INTERACT FORUM
More => Old Versions => JRiver Media Center 19 for Windows => Topic started by: mattkhan on April 26, 2014, 05:58:43 pm
-
I have a 5.1 setup using acourate to deal with everything involved in bd (audio) playback except for the +10 LFE boost. My ears tell me something is not right with the SW (seems rather thin) but I can't see what I have done wrong. Therefore I thought I'd offer up my config for review, any & all comments much appreciated.
The basic setup is everything (room correction, delays, bass management) done in convolution except for the +10dB to the LFE channel which is done via jriver room correction.
The end result is great (main channel levels are matched nicely & sound v good) except LFE seems weak, there just isn't the impact I'd expect and hence this makes me think I have something wrong on the bass management side of things. This is not a missing +10dB for LFE problem, there is definitely +10 there.
jriver is configured according to the attached pics; output format is 5.1 with no mixing and everything resampled to 96kHz, room correction does nothing except add 10dB to the LFE & then convolution does the rest. The convolver cfg file is listed below, comments added to explain what each path is intended to do.
To make it clear what the WAVs are, a bit of background on (my use of) acourate .... the acourate UI is designed for stereo use but can be used for multichannell quite easily. This means I have 4 sets of filters; one each for L-R, L-C, L-SL and L-SR. I am running a 2 way (5.1) setup so each pair of channels produces 2 2 channels WAVs containing the correction filter. By default these are named as follows
Cor1S96.wav
Cor2S96.wav
so each wav has 2 channels (L & R for example) where Cor1 is the low pass to the sub and Cor2 is high pass to the main channel. I then copy them all into a single dir and append the channel names so I know what is what, i.e. I end up with;
Cor1S96_LR.wav
Cor1S96_LC.wav
Cor1S96_LSR.wav
Cor1S96_LSL.wav
Cor2S96_LR.wav
Cor2S96_LC.wav
Cor2S96_LSR.wav
Cor2S96_LSL.wav
I run a 120Hz XO so I just reuse 1 of the Cor1 filters for the LFE channel.
Finally the cfg file
96000 6 6 0
0 0 0 0 0 0
0 0 0 0 0 0
# low pass L to SW
C:\Users\Matt\Documents\Acourate\Filters\Cor1S96_LR.wav
0
0.0
3.0
# low pass R to SW
C:\Users\Matt\Documents\Acourate\Filters\Cor1S96_LR.wav
1
1.0
3.0
# low pass C to SW
C:\Users\Matt\Documents\Acourate\Filters\Cor1S96_LC.wav
1
2.0
3.0
# LFE
C:\Users\Matt\Documents\Acourate\Filters\Cor1S96_LSR.wav
1
3.0
3.0
# low pass SL to SW
C:\Users\Matt\Documents\Acourate\Filters\Cor1S96_LSL.wav
1
4.0
3.0
# low pass SR to SW
C:\Users\Matt\Documents\Acourate\Filters\Cor1S96_LSR.wav
1
5.0
3.0
# high pass L
C:\Users\Matt\Documents\Acourate\Filters\Cor2S96_LR.wav
0
0.0
0.0
# high pass R
C:\Users\Matt\Documents\Acourate\Filters\Cor2S96_LR.wav
1
1.0
1.0
# high pass C
C:\Users\Matt\Documents\Acourate\Filters\Cor2S96_LC.wav
1
2.0
2.0
# high pass SL
C:\Users\Matt\Documents\Acourate\Filters\Cor2S96_LSL.wav
1
4.0
4.0
# high pass SR
C:\Users\Matt\Documents\Acourate\Filters\Cor2S96_LSR.wav
1
5.0
5.0
Thanks
Matt
-
The configuration I can see looks good. It's hard to know without seeing what kind of filtering the convolution filter is doing on the sub, but what you're describing (thin bass despite correct measured volume) is, in my experience, usually a result of poor phase alignment in the crossover region (either due to crossover issues or failing to compensate for the extra delay resulting from the sub's position), and the best way to diagnose that is to run a measurement of the 2-way system (i.e. the sub and main interacting at a listening position).
Three troubleshooting questions:
1) How are you accounting for the sub's delay based on it's position in the room (time of flight delay)? It looks like you're reusing a filter for another speaker for the LFE channel, are they the same distance from the listening position? Or are all of that batch of filter set at the appropriate delay for the sub?
2) What type of crossover slope are you using for the sub and the mains (i.e. is it symmetrical, what orders, linear phase?, what's the delay like, etc.), and
3) How does the system measure? Since each of your speakers is effectively a 2-way with the sub, you should be able to run (say) a sweep on just the left channel and see how the sub is interacting with the left main. Does everything look good in the sub region through the crossover to the mains when you run a sweep on the left or right speaker?
It's probably some kind of phase alignment issue, an integrated measurement should show you something.
-
The configuration I can see looks good. It's hard to know without seeing what kind of filtering the convolution filter is doing on the sub, but what you're describing (thin bass despite correct measured volume) is, in my experience, usually a result of poor phase alignment in the crossover region (either due to crossover issues or failing to compensate for the extra delay resulting from the sub's position), and the best way to diagnose that is to run a measurement of the 2-way system (i.e. the sub and main interacting at a listening position).
thanks for looking at it. I'm don't think it's the punchiness I'm missing, music sounds really excellent. It's more a question of a lack of weight at the low end for films.
1) How are you accounting for the sub's delay based on it's position in the room (time of flight delay)? It looks like you're reusing a filter for another speaker for the LFE channel, are they the same distance from the listening position? Or are all of that batch of filter set at the appropriate delay for the sub?
the basic approach is as follows
- align the mic precisely between the L & R
- calculate the delay from the SW to the L & R, this shows the SW is about 40 samples further away so I apply the relevant delay to the low pass XO to bring them into line
- run mic alignment for the other channels against L, the other channels are also closer to the listening position by varying amounts so the high pass XO is shifted accordingly to align them with the L & R
i.e. the low pass XO is identical for each channel.
2) What type of crossover slope are you using for the sub and the mains (i.e. is it symmetrical, what orders, linear phase?, what's the delay like, etc.), and
2nd order linear phase Neville-Thiele at 120Hz
3) How does the system measure? Since each of your speakers is effectively a 2-way with the sub, you should be able to run (say) a sweep on just the left channel and see how the sub is interacting with the left main. Does everything look good in the sub region through the crossover to the mains when you run a sweep on the left or right speaker?
I can't be 100% certain as I haven't worked out a way to get actual timing data yet for individual measurements of each speaker using REW when convolution is involved. I don't think acourate has this feature either as it shifts the peaks to sample = 6000 automatically. I don't think there is an obvious discontinuity though, I've attached an unsmoothed measurement showing 15-300Hz. The dip at 125Hz seems to correspond to a room mode & otherwise it seems ok (given that I have only 1 sub).
-
I can't be 100% certain as I haven't worked out a way to get actual timing data yet for individual measurements of each speaker using REW when convolution is involved. I don't think acourate has this feature either as it shifts the peaks to sample = 6000 automatically.
I don't think there is an obvious discontinuity though, I've attached an unsmoothed measurement showing 15-300Hz. The dip at 125Hz seems to correspond to a room mode & otherwise it seems ok (given that I have only 1 sub).
Your methodology seems sound, and the graph generally looks quite good for a 1 sub setup, but that High Q dip at 120/125Hz looks a little suspicious given that's your crossover frequency. One way to confirm is to look at the phase.
Can you repost that measurement with the "phase" box checked? A room mode usually looks a little different than a crossover misalignment. I also have another phase-related hunch as to what could be happening, but it's hard to explain and the phase would confirm or disprove it.
P.S. - Am I reading that right that you have rising frequency response at 15Hz? Assuming that's not a measurement artifact, that's a pretty nice sub, what kind is it? ;D
-
Your methodology seems sound, and the graph generally looks quite good for a 1 sub setup, but that High Q dip at 120/125Hz looks a little suspicious given that's your crossover frequency. One way to confirm is to look at the phase.
Can you repost that measurement with the "phase" box checked? A room mode usually looks a little different than a crossover misalignment. I also have another phase-related hunch as to what could be happening, but it's hard to explain and the phase would confirm or disprove it.
I've attached 2 pics, one from REW (via the asio line in method) and one from acourate (showing L & R) using the correction filter directly.
P.S. - Am I reading that right that you have rising frequency response at 15Hz? Assuming that's not a measurement artifact, that's a pretty nice sub, what kind is it? ;D
yes that's right though the measurements through REW recently are a bit dodgy below about 30Hz, it really exaggerates the rise at the low end. You can see this by comparing against acourate, it is still rising but only a few dB as opposed to loads! Mind you now I look at it, the shapes of the 2 responses is a bit different anyway. This makes me suspect something is wrong in my jriver setup *or* that asio line in is doing something unexpected. The acourate graph suggests the sub is a few dB down on the mains.
I attached a couple more pics showing the full range measurements, one in REW and one from acourate. The former is 1/6 octave smoothed, the latter is run through acourate's macro 1 (psychoacoustic + frequency dependent windowing).
It's a sealed 65L enclosure with a down firing 15" Fi SP4 driver in it. I need to get a new amp though as my current one can't drive it properly (more money to spend!)
-
I've attached 2 pics, one from REW (via the asio line in method) and one from acourate (showing L & R) using the correction filter directly.
Ok your crossover looks fine (from what I can see), you're right, it's probably just a room effect.
yes that's right though the measurements through REW recently are a bit dodgy below about 30Hz, it really exaggerates the rise at the low end. You can see this by comparing against acourate, it is still rising but only a few dB as opposed to loads! Mind you now I look at it, the shapes of the 2 responses is a bit different anyway. This makes me suspect something is wrong in my jriver setup *or* that asio line in is doing something unexpected. The acourate graph suggests the sub is a few dB down on the mains.
It may also be that the two suites are doing different filtering. My experience is that REW's "unfiltered" response doesn't really look like the raw responses I get in any other measurement suite, which suggests to me that some kind of unusual filtering/smoothing is going on in REW (especially on the low end). I haven't played around with REW's gating settings, but that may be part of the story.
It's one of the reasons I stopped using it for measurement, I couldn't ever seem to get results that squared with other measurements I took. For example, raw measurements I take from a given position with the free measurement tool from Audiolense, look almost identical to the measurements I've taken in Holm, etc.
If the Acourate measurement is correct, the entire sub range looks about 3 or 4 dB down from the mains (more in some places), which would be very noticeable and could well explain your problem. You mentioned in another thread that you might try Holm, I'd be curious if it shows the same thing as Acourate. If so, then your issue is probably just levels.
I attached a couple more pics showing the full range measurements, one in REW and one from acourate. The former is 1/6 octave smoothed, the latter is run through acourate's macro 1 (psychoacoustic + frequency dependent windowing).
As noted, those Acourate measurements suggest that your levels are off, which (given that your phase looks good through the crossover) could well be the issue.
It's a sealed 65L enclosure with a down firing 15" Fi SP4 driver in it. I need to get a new amp though as my current one can't drive it properly (more money to spend!)
Interesting; I'm surprised that a sealed cabinet with more or less level FR has 360 degrees of phase wrap, but I haven't modeled your driver and enclosure so maybe that's expected in your setup? It might be worth trying to straighten out that phase wrap a little bit if it turns out it's not a levels issue.
That FR is darned impressive though; Do you like the sound of the driver? I'm potentially in the market for a new 15" for my frankensub.
-
Thanks for going over my results. I have downloaded holm so will give that a whirl when I next get time to measure. In the meantime I might just run the sub a bit hotter & see if it is just that.
Interesting; I'm surprised that a sealed cabinet with more or less level FR has 360 degrees of phase wrap, but I haven't modeled your driver and enclosure so maybe that's expected in your setup? It might be worth trying to straighten out that phase wrap a little bit if it turns out it's not a levels issue.
I've been experimenting with LTs recently & with correcting the near field response of the sub. For speed of experimentation I've been doing this via a number of IIR filters as it means I can do it all offline, if I embedded these into the acourate XO then it's quite slow going with all the measurements you need to take. The measurement I attached is one where I have a few filters in place to flatten the (presumably inductance based) hump at ~55Hz. I then add an LT onto that to reduce Q from ~0.711 to 0.500. I suppose the phase wrap comes from those filters? I have to say it's not something I've looked at given the basic problem I have with levels :)
That FR is darned impressive though; Do you like the sound of the driver? I'm potentially in the market for a new 15" for my frankensub.
I really like it. I've listened to a lot of music since installing it and it has lovely control which yields a really nice texture to the music. It drinks power for AV use though. I'm considering importing a speakerpower amp so I can really get it going as decent power amps in the UK are rather expensive unfortunately.
-
Thanks for going over my results. I have downloaded holm so will give that a whirl when I next get time to measure. In the meantime I might just run the sub a bit hotter & see if it is just that.
If that partially resolves it, you've probably found your culprit, which is good news.
I've been experimenting with LTs recently & with correcting the near field response of the sub. For speed of experimentation I've been doing this via a number of IIR filters as it means I can do it all offline, if I embedded these into the acourate XO then it's quite slow going with all the measurements you need to take. The measurement I attached is one where I have a few filters in place to flatten the (presumably inductance based) hump at ~55Hz. I then add an LT onto that to reduce Q from ~0.711 to 0.500. I suppose the phase wrap comes from those filters? I have to say it's not something I've looked at given the basic problem I have with levels :)
A linkwitz transform could definitely introduce that kind of phase shift so that may be it. Just FYI my experience has been that phase wrap at low frequencies that doesn't correspond to changes in frequency response can sound mighty weird, and lead to some undesirable bass consequences (ringing, muddling, loss of bass "authority", etc.) as the sound at 30Hz is nearly a full cycle behind the sound at 70Hz.
Obviously ported subs typically have at least that much wrap, but that's part of why sealed subs often have a better reputation for clean bass. Just something to think about if the levels turn out to not be the issue (or don't completely resolve the issue).
Phase distortion is not necessarily super audible, but studies (as well as my own anecdotal experience) suggests that it's most audible at low frequencies.
I really like it. I've listened to a lot of music since installing it and it has lovely control which yields a really nice texture to the music. It drinks power for AV use though. I'm considering importing a speakerpower amp so I can really get it going as decent power amps in the UK are rather expensive unfortunately.
Sweet, I'll see if I can arrange to hear one sometime. Thanks for the info.
-
If that partially resolves it, you've probably found your culprit, which is good news.
I think this was it. I measured in HolmImpulse and see a different low end to that in REW. I therefore upped the gain and all is well in the world again. I verified using REW RTA and the JRiver calibration tones and all looks good. Playback of some more intensive scenes verifies this. I'm glad it was a simple solution for once ;D
-
I think this was it. I measured in HolmImpulse and see a different low end to that in REW. I therefore upped the gain and all is well in the world again. I verified using REW RTA and the JRiver calibration tones and all looks good. Playback of some more intensive scenes verifies this. I'm glad it was a simple solution for once ;D
Great news! I'm always thankful when, after spending a lot of time on a tough problem, the solution only takes a few minutes and costs nothing ;D
I once spent a day and a half trying to figure out why my stereo imaging was poor in an early version of my bi-amp setup. The two speakers were identical and similarly placed, and had (a week before) measured almost identically, so I kept measuring the left speaker to check the crossover, check the delay, check everything. I couldn't figure it out. Until I finally re-measured the right speaker and realized that when I had taken everything apart two days prior I had wired the midbass module backwards so it was inverted with respect to the rest of the system :-[
But the good news was there wasn't a fatal flaw in my design, and it was a free fix that took 30 seconds.
-
Hi
Does the Holm measurement look the same as the one in acourate?
I also use acourate, and have finally worked (with a lot of help from Uli) out a procedure for integrating the subs
Basically you need to work out the delay, polarity and gain parameters of the sub that gives the flattest response. I used my sound card set to loopback and plogue bidule with vst plugins to adjust those parameters. I then used the RTA in REW to adjust them in realtime to get the best response. I think you could also use acourate-convolver for realtime parameter adjustment, or maybe the loopback driver in jriver
The next step is to generate the xover in acourate. Then apply the delay, gain and polarity settings to your xover filters. Then make a multiway wav file.
Apply this multiway wav file to the acourate recorder, to basically perform your sub integration measurement
Apply the room macros to your measured pulses, create filters and a convolver cfg file for jriver.
Now your sub will be integrated with your mains.
You can extend this procedure to multiple subs and active speakers
To generate the LFE channel, simply apply 10dB LESS attenuation compared to the filters above. If this is not possible, turn up the amp gain in the step above. Generating the LFE filter this way, instead of adding gain in jriver, will give more headroom in jriver
-
Does the Holm measurement look the same as the one in acourate?
Holm and Acourate seem quite similar except the acourate measurement rolls off sharply at ~15Hz, I've asked Uli if there is a reason to expect this.
I also use acourate, and have finally worked (with a lot of help from Uli) out a procedure for integrating the subs
Basically you need to work out the delay, polarity and gain parameters of the sub that gives the flattest response. I used my sound card set to loopback and plogue bidule with vst plugins to adjust those parameters. I then used the RTA in REW to adjust them in realtime to get the best response. I think you could also use acourate-convolver for realtime parameter adjustment, or maybe the loopback driver in jriver
The next step is to generate the xover in acourate. Then apply the delay, gain and polarity settings to your xover filters. Then make a multiway wav file.
Apply this multiway wav file to the acourate recorder, to basically perform your sub integration measurement
Apply the room macros to your measured pulses, create filters and a convolver cfg file for jriver.
Now your sub will be integrated with your mains.
You can extend this procedure to multiple subs and active speakers
To generate the LFE channel, simply apply 10dB LESS attenuation compared to the filters above. If this is not possible, turn up the amp gain in the step above. Generating the LFE filter this way, instead of adding gain in jriver, will give more headroom in jriver
Are you running multiple subs? ISTR a mail from you on the mailing list talking about correcting multiple subs. I think I have the sub aligned to each main channel correctly but the per channel delays are not working. I have dropped Uli a mail on this so am waiting to hear back from him. How are you handling main channel delays? via rotating the XO or some other way?
Thanks for the tip on the LFE channel.
-
Yep I use the rotation function to add delays to the main channel xover filters. I am using dual subs, but one at the front and one at the rear.
6dB down at 20kHz compared to 1kHz is pretty standard. Is your rolloff steeper than this?
I also found that pointing the mic at the front of the room (ie horizontal), produced less roff off than a vertical mic
-
Yep I use the rotation function to add delays to the main channel xover filters. I am using dual subs, but one at the front and one at the rear.
6dB down at 20kHz compared to 1kHz is pretty standard. Is your rolloff steeper than this?
I also found that pointing the mic at the front of the room (ie horizontal), produced less roff off than a vertical mic
I have the same slope hinged at 1kHz reducing 6dB by 20kHz.
ok so you rotate the XO2 for the C/SL/SR etc to bring them into alignment with the L and R? do you see that rotation in the peak of the resulting Cor2? not 100% sure if you're meant to see it there as a shift in the peak of the correction filter but, if not there, I don't know where a delay would be visible.
Do you have your approach written up btw? it would make interesting reading if you do.
-
just to add to this, I have verified experimentally that the delay applied to the XO2 is not appearing in the corresponding Cor2. Manually rotating this after the fact & using this in the convolution filter produces expected results when verifying delays in holmimpulse. Curiously it's interesting that the delays are being applied to the low pass. I will report back to Uli to see if this is a bug or some user error on my part.
It's interesting that I had noted the tonal shift (a certain boxiness to the sound) on the centre channel ever since I set this up.
-
To generate the LFE channel, simply apply 10dB LESS attenuation compared to the filters above. If this is not possible, turn up the amp gain in the step above. Generating the LFE filter this way, instead of adding gain in jriver, will give more headroom in jriver
I don't understand how this differs to adding 10dB to the LFE via jriver & appears to involve more work. If I add 10dB to the sub in room correction then, as I understand it, all it does is attenuate all the other channels by 10dB. If I go in and attenuate all the other low passes in acourate then I'm just doing the same thing but n times instead of once aren't I?
-
I don't understand how this differs to adding 10dB to the LFE via jriver & appears to involve more work. If I add 10dB to the sub in room correction then, as I understand it, all it does is attenuate all the other channels by 10dB. If I go in and attenuate all the other low passes in acourate then I'm just doing the same thing but n times instead of once aren't I?
Mattkhan, you're correct; adding 10dB to LFE in room correction just attenuates the other channels by 10dB. Adding 10dB in PEQ (instead of room correction) would not automatically self-offset in the same way, which may be part of the confusion. The only way there's an advantage to integrating the offset into your convolution filters is if your main channel filters already involve very substantial attenuation; in that case, you might be better off integrating the attenuation into the convolution filters (i.e. you might have less total attenuation that way). With an ordinary filter, you lose nothing by using room correction instead of offsetting the volume in convolution, and gain convenience.
Turning up the sub amp is still the preferable offset method (when possible) because that prevents the loss of 10dB worth of digital dynamic range in the other channels, but that's a separate issue.
-
just to add to this, I have verified experimentally that the delay applied to the XO2 is not appearing in the corresponding Cor2. Manually rotating this after the fact & using this in the convolution filter produces expected results when verifying delays in holmimpulse. Curiously it's interesting that the delays are being applied to the low pass. I will report back to Uli to see if this is a bug or some user error on my part.
It's interesting that I had noted the tonal shift (a certain boxiness to the sound) on the centre channel ever since I set this up.
Yes the issue is to avoid jriver attenuating all the other channels by 10dB.
As mwillems also noted, the preferred option if to turn up the gain on your sub amp.
Then you simply make the gain adjustment in your filters 10dB less for the LFE channel, compared to all the other low pass channels. This only involves generation of 2 different xover filters.
-
just to add to this, I have verified experimentally that the delay applied to the XO2 is not appearing in the corresponding Cor2. Manually rotating this after the fact & using this in the convolution filter produces expected results when verifying delays in holmimpulse. Curiously it's interesting that the delays are being applied to the low pass. I will report back to Uli to see if this is a bug or some user error on my part.
It's interesting that I had noted the tonal shift (a certain boxiness to the sound) on the centre channel ever since I set this up.
Are you using workspaces in acourate correctly?
You need to set the workspace in acourate to a given directory eg 'front channels'. You then generate the xover and multiwav file. Record the pulses, place the PulseL48 etc in the same directory.
Run all the room macros, and in macro 4 acourate will convolve the CorS48.dbl filters with the xover filters.
If you do not use the workspace in this manner, the convolution in the last step won't happen. You must not rename the xover files either
To me this sounds like your problem.
I can confirm that with this method I do get the delays applied to the Cor2 Cor3 etc. I have recently triamped and time aligned my front speakers. The difference was significant.
I have experienced some speakers sounding 'boxy'. The problem is either a poor measurement or the target curve. Repeating the measurement has often fixed this problem for me.
If that's not your problem, you can try moving the knee in the target curve higher in freq than 1kHz, and/or raising the -6dB @ 20kHz slightly.
-
Further to the boxiness:
The last time I had really poor sounding speakers after correction, the solution was to record the pulses at 48khz instead of 96kHz.
Uli also demonstrated how to avoid high freq issues by flattening the gain of the inverse file after 20kHz.
After macro3, if the inverse curve is rising after 20kHz, it may lead to problems.
The solution is to use the phase (?) function (can't remember exact name, but press F2 key)
Set the gain parameter to 0dB from the freq where it rises (21kHz in my case). Then select minimum phase
Doing this may also help with the boxiness, as if the inverse curve has too much gain in this high freq region, it may cause attenuation in other lower freq regions.
Note that recording at 48khz, minimises the high freq measurement
-
Thanks for the suggestions. I think I am doing all of those things correctly though.
I use a separate workspace for each pair of channels (named LC, LR, LSL, LSR)
I generate the XO in each workspace & don't rename them or otherwise mess with them
I record at 48kHz
I use phase extract to avoid boosting at the extremities of the frequency range (<14Hz and >20700Hz for me)
The problem is repeatable in my case. I can
create a new workspace
generate an XO
rotate of each of them by some different amount
run through the macros
compare the Cor with the XO and see that Cor1 is shifted but Cor2 is not
-
Hi
It seems you have something there.
I just had a look at my 5-way setup (2 subs and triamped mains)
There is a timing difference between the correction filter and the XO filter, but it only matters if it changes for each driver.
Here is the difference (in samples) for each driver (between XO and Cor):
sub1: 71
sub2: 71
Mains1: 459
Mains2: 457
Mains33: 457
So it is not maintaining the time delay I embedded for the subs.
My guess is that acourate thinks that it's getting a better correction by adjusting the timing.
But it's worth checking with Uli
-
thanks for checking your setup, at least now Uli has 2 clear bug reports to go on :)
It would be good if Uli added the ability to load a correction filter into the mic alignment tool just as you can in the log sweep recorder. It would make double checking mains alignment post filter a trivial task.
-
yes that's right though the measurements through REW recently are a bit dodgy below about 30Hz, it really exaggerates the rise at the low end. You can see this by comparing against acourate, it is still rising but only a few dB as opposed to loads!
Are you using same mic and soundcard calibration in REW as in Acourate? Some add a soundcard calibration in REW which changes the low frequency response.
I've been experimenting with LTs recently & with correcting the near field response of the sub. For speed of experimentation I've been doing this via a number of IIR filters as it means I can do it all offline, if I embedded these into the acourate XO then it's quite slow going with all the measurements you need to take.
I recently build four sealed subs that needed an LT. You should just be able to do it with your target. That is what I did in Audiolense. Convolution can make a perfect LT because it knows the exact amount of room gain you have to go along with the subwoofer's rolloff.
I'm considering importing a speakerpower amp so I can really get it going as decent power amps in the UK are rather expensive unfortunately.
You can get one from Mark Seaton at Seaton Sound cheaper than buying directly from SpeakerPower. He is a dealer and also uses SpeakerPower amps in his subs and active speakers.
I also found that pointing the mic at the front of the room (ie horizontal), produced less roff off than a vertical mic
The mic pointed at the speakers is a free field measurement and with it pointed vertical it is a diffuse field (or pressure response) measurement. The calibration for each can be quite different. My iSemcon microphone only varies by a few dB between free vs diffuse. I recently received some info from ACOPacific on their reference microphones and the same mic when pointed vertical is -9 dB at 20kHz and -5.1 dB at 10kHz. If you aren't using the proper calibration you can get very different results when trying to correct the full bandwidth.
I've read that a diffuse field measurement with the mic at about 80 degrees is recommend for room correction since it most represents what we hear. A free field measurement is more used for close mic/anechoic type measurements for designing speakers. However, many take room measurements with the mic pointed straight ahead.
Wow, the Acourate measurement and filter technique sounds rather laborious. This weekend I am helping setup a theater with 15 channels including 6 subwoofers, 3 mains, and 6 surrounds. I will be using Audiolense to take measurements at all 7 seating locations of all 15 channels and creating a filter. I only have to take 7 measurements with Audiolense vs 105 with Acourate (I think). Should be fun!
-
thanks for checking your setup, at least now Uli has 2 clear bug reports to go on :)
It would be good if Uli added the ability to load a correction filter into the mic alignment tool just as you can in the log sweep recorder. It would make double checking mains alignment post filter a trivial task.
As you have seen, Uli doesn't think that this behaviour is a bug, but it's the result of optimisation in macro 4.
I will check tonight the delays that a different correction produced, which had the mains' polarity reversed, to see if a different result was produced.
-
The mic pointed at the speakers is a free field measurement and with it pointed vertical it is a diffuse field (or pressure response) measurement. The calibration for each can be quite different. My iSemcon microphone only varies by a few dB between free vs diffuse. I recently received some info from ACOPacific on their reference microphones and the same mic when pointed vertical is -9 dB at 20kHz and -5.1 dB at 10kHz. If you aren't using the proper calibration you can get very different results when trying to correct the full bandwidth.
I've read that a diffuse field measurement with the mic at about 80 degrees is recommend for room correction since it most represents what we hear. A free field measurement is more used for close mic/anechoic type measurements for designing speakers. However, many take room measurements with the mic pointed straight ahead.
I have an earthworks M30, which only has calibration for free field, and also a calibrated ECM8000, which has calibrations for 3 different angles.
I actually think that the free field measurement is best for the stereo mains, and the diffuse field is better for surround channels.
Acourate is targeted more towards a stereo setup and making active speakers. It can do surround channels too, but the workflow for many channels takes a while. It also doesn't measure multiple locations, which Uli believes gives a more compromised result for stereo.
Audiolense is better suited to surround setups, but several users have compared the two, and audiolense doesn't give quite as good a result for a stereo setup.
So it really is horses for courses.
-
Are you using same mic and soundcard calibration in REW as in Acourate? Some add a soundcard calibration in REW which changes the low frequency response.
yes, it is exactly the same setup.
I recently build four sealed subs that needed an LT. You should just be able to do it with your target. That is what I did in Audiolense. Convolution can make a perfect LT because it knows the exact amount of room gain you have to go along with the subwoofer's rolloff.
I don't follow what you mean exactly, can you expand on that? My understanding is that the LT is applied to give a specific anechoic response which dictates the response characteristics of the sub itself & that impacts the step response (and hence the perception of a tight sub). You clearly need to set that to "agree" with your room but the room correction step has to be distinct to the LT step otherwise you're just correcting the room not the sub.
You can get one from Mark Seaton at Seaton Sound cheaper than buying directly from SpeakerPower. He is a dealer and also uses SpeakerPower amps in his subs and active speakers.
thanks, that's good to know.
I've read that a diffuse field measurement with the mic at about 80 degrees is recommend for room correction since it most represents what we hear. A free field measurement is more used for close mic/anechoic type measurements for designing speakers. However, many take room measurements with the mic pointed straight ahead.
the description I read was by Herb from CSL on HTS here (http://"http://www.hometheatershack.com/forums/spl-meters-mics-calibration-sound-cards/15951-cross-spectrum-microphone-calibration-service-usa.html#post146672"), I think the only correction system I'm aware of that uses that "multiple measurements at random incidence" approach is RoomPerfect (http://"http://www.steinwaylyngdorf.com/technology-and-innovation/roomperfect"). I use the mic at 80 degrees ish approach myself.
Wow, the Acourate measurement and filter technique sounds rather laborious. This weekend I am helping setup a theater with 15 channels including 6 subwoofers, 3 mains, and 6 surrounds. I will be using Audiolense to take measurements at all 7 seating locations of all 15 channels and creating a filter. I only have to take 7 measurements with Audiolense vs 105 with Acourate (I think). Should be fun!
yes it's not the most user friendly product for multichannel use but I don't think it's quite that bad :) an n.1 setup (where 1 refers to a mono sub not the no of subs) needs n-1 measurements for correction alone & then there are separate measurements for each n way speakers you are sorting out. If you need multi seat correction for a large theatre then I don't think it's not the right tool.
-
As you have seen, Uli doesn't think that this behaviour is a bug, but it's the result of optimisation in macro 4.
I will check tonight the delays that a different correction produced, which had the mains' polarity reversed, to see if a different result was produced.
I will be surprised if this covers my case as each channel (C, SL, SR) is reverted to 0 offset from the L XO. It's not impossible of course but it seems rather improbable.
-
It does seem like there is something else going on with your filters. Do you have ver 1.8.9 of acourate?
My system is maintaining the time alignment I program into the mains, it's just moving the sub-main time delay, presumably to get a smoother response. I need to measure to confirm this.
Could you post or email some screen grabs of your XO1, Cor1 and XO2, Cor2 impulses on the same plot? Just to confirm we are definitely on the same page
-
It does seem like there is something else going on with your filters. Do you have ver 1.8.9 of acourate?
I am using 1.8.9
Could you post or email some screen grabs of your XO1, Cor1 and XO2, Cor2 impulses on the same plot? Just to confirm we are definitely on the same page
sure, in this case acourate L is really my L and acourate R is really my C
XO1L vs Cor1L, peaks are at samples 32718 vs 32733 (seems like the sort of shift you saw?)
XO2L vs Cor2L, peaks are aligned at 32768
XO1R vs Cor2R, 32718 vs 32740
XO2R vs Cor2R, 32790 vs 32768
the last one seems a bit fishy to me
-
I think what your graphs show is that acourate aligns the tweeter to 32768 no matter what (as Uli said)
The sub/woofer is then aligned in time to give the best response.
So in summary, acourate is behaving as expected and there is no problem with your filters, but the room correction timing needs to be applied using jriver or something. ie if your centre is 1m further away than your front, you still need to apply 3ms delay to the front speaker using jriver.
The time delay built into XO filters can still be used to time align drivers within a single speaker
-
So in summary, acourate is behaving as expected and there is no problem with your filters, but the room correction timing needs to be applied using jriver or something.
The problem is this completely contradicts what he has advised me to do with respect to multichannel delays. This makes me think it is some sort of bug.
-
The problem is this completely contradicts what he has advised me to do with respect to multichannel delays. This makes me think it is some sort of bug.
OK so it's not a bug, it's a feature :) This issue is a reflection of acourate's stereo focus so, as Brad says, it aligns the tweeters and then shifts the woofers accordingly. This can be seen in my subwoofer correction for that channel which has been rotated to the left. In contrast, for a multichannel setup where I have a single sub, we want the woofer to stay in a fixed position and the tweeter to be rotated to the right. Therefore the fix is to manually rotate all CorR filters by (Cor1L-Cor1R) samples so that the Cor1R is aligned to the Cor1L and the Cor2R is in the right place given the theoretical delay (as captured by the XO) & the adjustments caused by phase correction.
-
I don't follow what you mean exactly, can you expand on that? My understanding is that the LT is applied to give a specific anechoic response which dictates the response characteristics of the sub itself & that impacts the step response (and hence the perception of a tight sub). You clearly need to set that to "agree" with your room but the room correction step has to be distinct to the LT step otherwise you're just correcting the room not the sub.
The perception of a tight sub is primarily due to frequency response at the listening position. Once your box size and driver are set, it is the frequency response that you listen to that matters. I built some sealed subs and first added an LT and one filter to them in JRiver's PEQ using a close mic measurement. I then added EQ so that they would be smooth at the listening position. This became one zone. Then I started over and just added an LT and EQ'd using measurements from the listening position. I ended up with fewer filters and the same frequency response. This became a second zone. Switching between zones and listening revealed them to be identical.
Adding filters through PEQ or convolution are cumulative and the final result will be the same regardless how you got there. It is all just math. Using JRiver's PEQ, if you add a dip at 60 Hz of -10 dB with a Q of 2 and then add a peak at 60 Hz of 10 dB with a Q of 2 what do you hear? You won't hear the dip or the peak. It is as if nothing happened. Same thing with adding an LT using a close mic and then pulling down room induced peaks with a filter later. It is as if the LT at the peak location never existed. Depending on how the convolution filters are added from Acourate with the LT, you could be losing digital headroom with the double filters.
The LT is always correcting the room and not the sub because the only way you know what LT parameters to use is to measure the room gain profile. The actual Fz and Qz of the sub stay the same regardless of where you measure, but you only know their actual value through a close mic measurement. However, the desired Fp and Qp are completely dependent on the room and how much gain the room adds and at what frequency the gain is starting to be increased.
-
The perception of a tight sub is primarily due to frequency response at the listening position. Once your box size and driver are set, it is the frequency response that you listen to that matters. I built some sealed subs and first added an LT and one filter to them in JRiver's PEQ using a close mic measurement. I then added EQ so that they would be smooth at the listening position. This became one zone. Then I started over and just added an LT and EQ'd using measurements from the listening position. I ended up with fewer filters and the same frequency response. This became a second zone. Switching between zones and listening revealed them to be identical.
Adding filters through PEQ or convolution are cumulative and the final result will be the same regardless how you got there. It is all just math. Using JRiver's PEQ, if you add a dip at 60 Hz of -10 dB with a Q of 2 and then add a peak at 60 Hz of 10 dB with a Q of 2 what do you hear? You won't hear the dip or the peak. It is as if nothing happened. Same thing with adding an LT using a close mic and then pulling down room induced peaks with a filter later. It is as if the LT at the peak location never existed. Depending on how the convolution filters are added from Acourate with the LT, you could be losing digital headroom with the double filters.
The LT is always correcting the room and not the sub because the only way you know what LT parameters to use is to measure the room gain profile. The actual Fz and Qz of the sub stay the same regardless of where you measure, but you only know their actual value through a close mic measurement. However, the desired Fp and Qp are completely dependent on the room and how much gain the room adds and at what frequency the gain is starting to be increased.
OK thanks for expanding on that. My approach has been to choose an LT that augments room gain to the extent I want (roughly) but that also reduces Q down to the value I'm after, I then let the room correction fine tune to the actual target curve.
Your post sounds like you don't believe the Q of the system itself is relevant to that perception of tightness, is that a fair interpretation? For example, this thread (http://"http://www.diyaudio.com/forums/subwoofers/8114-subwoofer-qtc-tightness.html#post88711") shows how the theoretical step response changes with different values for Q.
-
A subwoofer's step response can be derived from its impulse response. It is a way of looking at the time domain behavior. Its impulse response is changed through either a Linkwitz Transform or time domain convolution (such as Acourate). It is more accurate to change it with time domain convolution.
Your post sounds like you don't believe the Q of the system itself is relevant to that perception of tightness, is that a fair interpretation? For example, this thread shows how the theoretical step response changes with different values for Q.
From what I read, the thread deals with Qtc which is correctly described in the first post as "a complex mathematical equation derived from driver, electrical and enclosure parameters." Electrical parameters means the T-S parameters of the driver. The Qtc of a sub is set once you build it. Using PEQ filters, a Linkwitz Transform, or convolution affect the signal that you input into the subwoofer. Those changes to the signal will result in a changed system Q which absolutely is relevant to the "perception of tightness." However, there is no such thing a perfect system Q. It changes to what it needs to in order to get a flat frequency response. If you fix the frequency response, you are changing the impulse response. If you change the impulse response, you change the frequency response. They are both linked. So in the end, you just try to get the frequency response you desire and don't worry about what the system Q might be. You also don't worry about some intermediate system Q that happens after the Linkwitz Transform but before convolution. ;)
Whether you use an LT and then convolution or if you just use convolution, if the final frequency response is the same then the system Q is identical as well because the exact same change to the subwoofer's input signal was done.
-
For a minimum phase system, the impulse response is directly linked to the frequency (and phase) response.
However, for a non-minimum phase system that is not the case.
Hence with linear phase filters it is possible to improve the impulse response without changing the frequency response (in principle).
ie linear phase filters of acourate and the like do buy you more flexibility to correct problems
-
For a minimum phase system, the impulse response is directly linked to the frequency (and phase) response.
However, for a non-minimum phase system that is not the case.
Hence with linear phase filters it is possible to improve the impulse response without changing the frequency response (in principle).
ie linear phase filters of acourate and the like do buy you more flexibility to correct problems
I think mojave's point was that it's the final measured frequency response and phase that matter, regardless of whether you achieve it a) by doing a minimum phase correction (i.e. a conventional LT) followed by a linear phase correction (convolution) or b) just doing the whole thing in convolution from step one.
And with one caveat, I agree: the FR and phase results are what counts, not the intervening steps. The only caveat I'll offer (that is specifically applicable to the perception of "tightness" in subs) is that there is an additional measurable time-domain response parameter other than phase, namely, time decay. Decay is not readily visible in conventional FR and phase measurements, although the decay information is present in the impulse and can be "cooked out" of it. Decay is usually presented as a waterfall plot (which REW does quite nicely).
You can (at least theoretically) have two systems with nearly identical FR and phase measurements, that nonetheless have very different waterfall plots as a result of a lot of variables (size of the driver, the driver motor, box resonances, room resonances, a port, etc.). Some speaker elements are just "faster" than others (which you can easily see if you ever look at a set of speaker waterfall plots taken in free air, like Zaph does). But even with a "fast" driver, some boxes and some rooms will just keep ringing forever.
In my opinion, the waterfall (speed of the decay) of your system in the relevant frequency band is the single biggest predictor of the perceived tightness of bass. You can look at your own waterfall measurement and immediately "see" how tight the bass is or isn't at the measurement position.
That said, I think mojave's original point is correct: in terms of electronic speaker correction, ordinarily, the single best thing you can do for your waterfall is get a flat FR and flat/coherent phase (however you do it) as flattening FR will tend to minimize the excitation of resonances (by reducing the SPL at those places) and flattening/unwrapping phase will reduce certain types of ringing (like port ringing, for example). Once you've done that, every other step you can take to improve perceived "tightness" involves dealing with issues extrinsic to the audio signal (adding bracing or stuffing to your speaker box, treating your room, etc.).
-
Would it be correct to reduce your (mojave/mwillems) posts to "the transient response of a subwoofer is completely described by its frequency response & hence all subwoofers will sound the same if they are equalised to produce the same response in that room & don't produce audible amounts of distortion"?
-
Would it be correct to reduce your (mojave/mwillems) posts to "the transient response of a subwoofer is completely described by its frequency response & hence all subwoofers will sound the same if they are equalised to produce the same response in that room & don't produce audible amounts of distortion"?
I wouldn't say that's exactly my position. My position is that the transient response of the subwoofer is completely described by it's impulse response, which is informed by: frequency response, phase, and decay. I would say that two subwoofers that are equalized/convolved to produce identical FR and phase response (at comparable distortion) will "sound" as close as signal manipulation can make them, but still not necessarily identical, because of decay. If you ring two bells, they might have the same pitch (FR) and start ringing at the same split second (phase), but one might keep ringing for much longer after the strike than another based on what it's made of and mounted to (decay).
For those reasons, there will still be a residual difference in the sound between two subs with identical frequency response and phase response based on mechanical factors: the drivers, box construction, and position in the room. The driver, the sides of the box, or the walls of the room are like the bells, they keep ringing once excited for differing amounts of time. Those differences would appear in a waterfall plot, but not necessarily in a frequency response and phase measurement. The only ways to "correct" those residual differences is mechanical (room treatment, box damping, different drivers, etc.).
The bottom line is that, in my view, two speakers whose outputs have identical FR and phase measurements are as close as signal processing can get them, no matter how you got there, but they won't sound identical in part because of differing time-decay response.
To provide a sub-specific example: imagine two subs that both start the uphill journey of the transient at the same time, but drop off at different rates. One of the subs may "ring" at a given frequency for a millisecond after the time zero (the transient takes a millisecond to fully drop off), and the other sub may ring for two dozen milliseconds after the time zero. Those two subs might show identical frequency response and phase response because they both ramped up to the correct volume (FR) at the correct time/part of the sine wave cycle (i.e. phase), but the time it takes them to stop making noise (decay) isn't shown on a normal FR and phase plot. And that difference will be audible; the sub that's slower to stop ringing will sound more smeared than the other sub; the sub that decays faster will sound tighter.
It's one of the reasons that vented speakers get a bad rap; a port introduces significant phase wrap and also ringing around the port frequency. In that case, I've found that fixing the phase can fix most of the port ringing, and, more generally, phase manipulation can definitely improve decay measurements to some extent. But some kinds of ringing or decay problems (i.e. box vibration, "slow" drivers), can't always be fully "resolved" through signal manipulation.
Does that make sense?
Check out this article for an interesting case study in why time decay can make a big difference in the sound of a speaker: http://www.soundonsound.com/sos/sep08/articles/yamahans10.htm
And check out this page over at Zaph audio: http://www.zaphaudio.com/5.5test/compare.html . He's tested dozens of drivers, and if you compare the Frequency Response graphs with the "CSD" graphs (cumulative spectral decay, aka waterfall plot) you can see how differently different speakers perform on those tests even with fairly similar FR.
I can't speak for mojave's view (although I'm interested to hear his thoughts).
-
I can't speak for mojave's view
I'll agree with your view so I don't have to type as much. :)
If you take a close mic measurement of a subwoofer/enclosure/amplifier, you can look at a waterfall chart to see the decay. You can never improve the native decay. The faster and cleaner the decay, the better the subwoofer can track the signal. It will sound different because there are more nuances that are smeared in a frequency response measurement. I think the main driver parameters that affect decay are motor strength (BL). mass (Mms) and suspension compliance (Cms). You can also affect it with box size and alignment. Finally, if you don't have sufficient power for transient bursts, then the subwoofer will sound less dynamic all things else being the same.
-
Yes this all makes sense, thanks.
If we go back a few posts to something I said earlier
I don't follow what you mean exactly, can you expand on that? My understanding is that the LT is applied to give a specific anechoic response which dictates the response characteristics of the sub itself & that impacts the step response (and hence the perception of a tight sub).
I was referring to that temporal aspect of how long & strong the ringing is after the initial impulse. It seems to me that this information is adequately described by the step response & a spectral decay/waterfall view (which AIUI is created by shifting a window forward over the IR and transforming into a frequency response for that window) is only needed if you want to drill down into how specific frequencies decay (e.g. to look for offensive resonances). Here's an article that seems to make the same point - http://audioxpress.com/files/2008/10/dappolito2960.pdf
To give an example, I have attached the near field frequency & step response of my sub before and after I cut out that (presumably inductance driven) hump. The flat version is clearly more like the theoretical step response as it strength and length of the post step ringing is clearly reduced.
The issue I've been mulling over though is the one mojave pointed out around any intermediate response being irrelevant to what you actually hear. Once I shift the mic to the LP then obviously the response changes and hence further correction is then required to trim the modal ringing down to produce an acceptable frequency response. I was therefore wondering whether that intermediate step is a worthwhile step or not. I haven't got round to a near field measurement of the sub with the final correction filter in place yet. I guess that could be informative (to compare the 2 approaches to correction; room only & room + near field).
-
I was referring to that temporal aspect of how long & strong the ringing is after the initial impulse. It seems to me that this information is adequately described by the step response & a spectral decay/waterfall view (which AIUI is created by shifting a window forward over the IR and transforming into a frequency response for that window) is only needed if you want to drill down into how specific frequencies decay (e.g. to look for offensive resonances). Here's an article that seems to make the same point - http://audioxpress.com/files/2008/10/dappolito2960.pdf
They can be useful for a lot of tasks besides resonance hunting (which is plenty useful), like comparing different drivers/speakers, or comparing different room positions for your existing speakers. Also, waterfalls make it much easier to visualize certain aspects of the response. For example, I can't tell from looking at an impulse how many milliseconds it takes before the FR output is -40dB across the board. On a waterfall, that's easy to see.
You're absolutely right that all the information in a waterfall is in the step response/impulse (the waterfall is derived from it, as you say), but FR and phase are derived from the impulse too. For me, anyway, it's much easier to see all those things when they're extracted from the impulse.
To give an example, I have attached the near field frequency & step response of my sub before and after I cut out that (presumably inductance driven) hump. The flat version is clearly more like the theoretical step response as it strength and length of the post step ringing is clearly reduced.
The issue I've been mulling over though is the one mojave pointed out around any intermediate response being irrelevant to what you actually hear. Once I shift the mic to the LP then obviously the response changes and hence further correction is then required to trim the modal ringing down to produce an acceptable frequency response. I was therefore wondering whether that intermediate step is a worthwhile step or not. I haven't got round to a near field measurement of the sub with the final correction filter in place yet. I guess that could be informative (to compare the 2 approaches to correction; room only & room + near field).
I think comparing measurements will ultimately answer your question. If the correction from both methods is the same, and they measure the same, (for our purposes) they are the same. The only reason to do a two-step process is if you get different (and better) results doing it that way than doing the one step process. In my speaker correction guide, I advised correcting the speaker first and then moving onto the room as a second step, because I've gotten objectively better correction doing it that way using the software I used. In that case the two-step method produced measurably different results than the one step method.
It may be that Audiolense and Acourate are sophisticated enough that the one-step method is every bit as good (or better) than the two step method. Your comparison plan is the best way to find out ;)
I'll agree with your view so I don't have to type as much. :)
;D
-
It may be that Audiolense and Acourate are sophisticated enough that the one-step method is every bit as good (or better) than the two step method. Your comparison plan is the best way to find out ;)
acourate is as sophisticated as the operator, this might be where the problems start ;D
-
Just to correct a mathematical point:
The complex amplitude response as a function of frequency is related to the complex amplitude response as a function of time via the Fourier Transform.
That is the Freq and phase plots have the same information as the impulse plot. Delay is contained in both.
It is just that some plots such as waterfall, ETC, group delay process the information to make the delay easier to see.
Useful information about how to treat a room or speaker can be gained from all the different plots. In fact, it is good to look at the different plots to make sure you haven't been fooled by a certain representation.
Some people will argue at length on various fora that their approach is best and you don't need (insert a plot).
-
acourate is as sophisticated as the operator, this might be where the problems start ;D
I find it generally the case that the more flexible a solution, the more ways it can go wrong!
However, the flexibility of acourate does allow you to produce a better solution than the more automatic products.
Further, the macros in acourate do produce a pretty good result for a stereo system with the basic settings.
But there is a big learning curve to go further
-
I was therefore wondering whether that intermediate step is a worthwhile step or not. I haven't got round to a near field measurement of the sub with the final correction filter in place yet. I guess that could be informative (to compare the 2 approaches to correction; room only & room + near field).
From what I understand, the 2-step process is useful if you use minimum phase filters in step 1. Then you will benefits from the EQ at more than just the one listening position
-
That is the Freq and phase plots have the same information as the impulse plot. Delay is contained in both.
Just to make sure we're not talking past each other: I was talking about using waterfalls to observe decay, not delay. Delay is definitely shown in a conventional phase graph (that's in large part what the graph is showing), but conventional FR and phase graphs don't show spectral decay, which is primarily what a waterfall or CSD shows.
Of course, all three graphs are derivable from the impulse (the impulse has all the information in it), and all three graphs are important to speaker analysis.
Useful information about how to treat a room or speaker can be gained from all the different plots. In fact, it is good to look at the different plots to make sure you haven't been fooled by a certain representation.
I agree completely.
-
one last thing to doublecheck on this, am I right in thinking that the following output format settings are the right ones to use for a 5.1 setup
channels: 5.1 channels
mixing: JRSS mixing
subwoofer: silent
i.e. use jrss to mix to 5 main channels where the source channels != 5
or to put it another way, the wiki (http://"http://wiki.jriver.com/index.php/Mixing") has option 3 "If you want JRiver to upmix and let the receiver handle speaker setup and bass management, then . . ." which could be written as "and let the [receiver|convolution engine] handle speaker setup and bass management" right?
-
one last thing to doublecheck on this, am I right in thinking that the following output format settings are the right ones to use for a 5.1 setup
channels: 5.1 channels
mixing: JRSS mixing
subwoofer: silent
i.e. use jrss to mix to 5 main channels where the source channels != 5
That's correct, assuming that you're still doing your bass management downstream in convolution. The settings you've chosen will result in 5.1 remaining 5.1, and everything else will get mixed to 5.1 or 5.0 (including stereo audio). My understanding is that if there's no LFE in the source, JRSS will not create one with those settings, but if there is an LFE channel (e.g. 7.1 source) it will retain it in the mix.
If you don't want stereo audio to get mixed to 5.0 (or 5.1), make sure the box labelled "for stereo sources only mix to 2.1" is checked. That will, with your other settings, result in stereo sources remaining stereo (because you have subwoofer set to silent).
-
That's correct, assuming that you're still doing your bass management downstream in convolution. The settings you've chosen will result in 5.1 remaining 5.1, and everything else will get mixed to 5.1 or 5.0 (including stereo audio). My understanding is that if there's no LFE in the source, JRSS will not create one with those settings, but if there is an LFE channel (e.g. 7.1 source) it will retain it in the mix.
If you don't want stereo audio to get mixed to 5.0 (or 5.1), make sure the box labelled "for stereo sources only mix to 2.1" is checked. That will, with your other settings, result in stereo sources remaining stereo (because you have subwoofer set to silent).
OK thanks. I find that 2.0 upmixed to 5.0 just sounds weird (quite thin & unnatural) so I will keep that "mix to 2.1" option checked.
FWIW, and going back a few posts to the discussion about whether to correct the near field response of the sub first, I got round to measuring this today. The measurements are of 2 sweeps with the correction filter applied, the red line is the end result of acourate correction with a linearised sub (i.e. in my case, the "inductance hump" removed) and the green line is without that and just correcting the room response alone. There are minor differences in the precise shape of the correction filter and the step responses are almost identical, you could argue the linearised sub is *marginally* "quicker" but whether that is audible is beyond me. I think this shows that, for me, it's not worth the effort involved in the extra measurements as it looks like the need to correct the room response wins (i.e. the final correction filter is barely any different as the room rings more than the sub does).
For completeness it would be interesting to embed that near field response change within the sub and then repeat the room correction and compare the outcome. However I don't get paid for this, it's just idle curiosity :D
-
Hi
I think what I said earlier was that the driver linearisation can correct problems (due to the driver) all around the room.
So to compare, you would want to see which approach gives a better result in multiple locations.
Of course, then deciding which is actually better is another thing.
-
Hi
I think what I said earlier was that the driver linearisation can correct problems (due to the driver) all around the room.
So to compare, you would want to see which approach gives a better result in multiple locations.
Of course, then deciding which is actually better is another thing.
Exactly; that's the main reason to deal with them separately (that and it's easier to get reproducible results).
But it's all relative; as the wavelengths get exceedingly large compared to the room, localization becomes less and less relevant. A quarter wavelength at 20 Hz is 13 feet; unless you have a very large room or very widely scattered listening positions, correction in one spot is probably about as good as any other relatively nearby spot at 20 Hz. Most domestic listening spaces aren't even large enough to have room modes at 20Hz.
That's much less true at 80Hz where the quarter wavelength is around 3.5 feet. Then you're going to get much more variation. And (IMO) above about 100Hz, correcting the speaker separately from the room and/or multi-seat correction becomes more important for good results if you have more than one listening position.
But this is a long way to saying that I'm not necessarily surprised that it doesn't seem to make a huge difference which way you come at it with deep bass because positioning becomes less important at very low frequencies.
I don't have tons of hands on experience with Acourate and other similar suites; they may have some secret sauce that let's them distinguish between the speaker and the room from a single measurement (some folks have more or less said as much about audiolense). But the physics of rooms dictate that what's good for one spot can't really be just as good for another, so that has to be addressed somehow (at least at frequencies with wavelengths short enough to have room modes in your room).
-
Hi
I think what I said earlier was that the driver linearisation can correct problems (due to the driver) all around the room.
So to compare, you would want to see which approach gives a better result in multiple locations.
Of course, then deciding which is actually better is another thing.
I don't see how that applies in the modal region, for higher frequencies yes but I don't see how it could low down. Besides which there is essentially no measurable difference in the near field response anyway.
-
I don't see how that applies in the modal region, for higher frequencies yes but I don't see how it could low down.
It definitely doesn't apply "below" the modal region (in the bass region below which your room has no modes), but it potentially has application in the modal region. A modal resonance is like any other resonance it can be excited to varying degrees depending on SPL. An anti-mode on the other hand, can often just be an unfillable suck out due to cancellation. So you can still address the mode maxima by trimming speaker FR peaks to avoid over-excitation of resonant modes. In those contexts, room positioning can be relevant.
An example: imagine three horizontally arranged seats in a large room. Modes are periodic, so let's say the room has a 60Hz antimode in the center and resonant modes equidistant to the left and right. So lets say the center seat is smack in the middle of an anti-mode at 60Hz, and the left and right seats are in the middle of the 60 Hz resonant modes. Let's also stipulate that the subwoofer itself has a 4dB FR peak at 60 Hz.
If one takes a single correction measurement at the center seat, unsophisticated software would see a huge suckout at 60Hz and try to correct it by boosting (which is exactly wrong). More sophisticated software would recognize (possibly based on phase/group delay info) that the suckout was a room effect and would leave the FR alone at 60Hz. Meanwhile, even in the "no correction" hypothetical folks in the left and right seat are getting blasted out of their seats anytime a 60Hz note is played because it's not just 4dB louder from the speaker, it's probably between 7 and 10dB louder (depending on how serious the mode is).
In our example, if one (instead) did speaker correction ironing out the speaker's 4dB peak (or took multiple measurements and corrected based on that), the left and right seats would still hear a too loud 60Hz, but it would be significantly less "too loud." The center seat would likely see minimal differences because it would still have a gigantic 60Hz suck out (that may or may not actually be deeper given how room cancellations can be at step one). Additional seats not in a 60Hz mode or anti-mode would experience better response too.
Besides which there is essentially no measurable difference in the near field response anyway.
And that's the real answer ;D. Theory is nice, but you took measurements at different locations and found it didn't make a difference in the relevant frequency band in your room, which is all you really need to know.
[Above edited to hopefully make it clearer what my point was]
-
It definitely doesn't apply "below" the modal region (in the bass region below which your room has no modes), but it potentially has application in the modal region. A modal resonance is like any other resonance it can be excited to varying degrees depending on SPL. An anti-mode on the other hand, can often just be an unfillable suck out due to cancellation. So you can still address the mode maxima by trimming speaker FR peaks to avoid over-excitation of resonant modes. In those contexts, room positioning can be relevant.
An example: imagine three horizontally arranged seats in a medium large room. Modes are periodic, so let's say the room has a 60Hz antimode in the center and resonant modes equidistant to the left and right. So lets say the center seat is smack in the middle of an anti-mode at 60Hz, and the left and right seats are in the middle of the 60 Hz resonant modes. Let's also stipulate that the subwoofer has a 4dB FR peak at 60 Hz.
If one takes a single correction measurement at the center seat, unsophisticated software would see a huge suckout at 60Hz and try to correct it by boosting (which is exactly wrong). More sophisticated software would recognize (possibly based on phase/group delay info) that the suckout was a room effect and would leave the FR alone at 60Hz. Meanwhile, even in the "no correction" hypothetical folks in the left and right seat are getting blasted out of their seats anytime a 60Hz note is played because it's not just 4dB louder from the speaker, it's probably between 7 and 10dB louder (depending on how serious the mode is).
In our example, if one (instead) corrected the speaker's 4dB peak (or took multiple measurements and corrected based on that), the left and right seats would still hear a too loud 60Hz, but it would be significantly less "too loud." The center seat would likely see minimal differences because it would still have a gigantic 60Hz suck out (that may or may not actually be deeper given how room cancellations can be at step one). Additional seats not in a 60Hz mode or anti-mode would experience better response too.
I don't disagree but we seem to be talking at cross purposes. I read what you've written as a description of room correction not speaker correction whereas I'm saying I'm not seeing an obvious case for speaker correction in the modal region. I am not arguing for no correction at all in that region.
In the room correction case, we're specifically moving the speaker (sub) away from a nice theoretically on target response in order to not excite those modal frequencies so much so that we experience an on target response at the listening position.
In contrast in the speaker correction case, we are trying to get the speaker itself to produce the on target frequency response irrespective of what the room might do to it.
-
I don't disagree but we seem to be talking at cross purposes. I read what you've written as a description of room correction not speaker correction whereas I'm saying I'm not seeing an obvious case for speaker correction in the modal region. I am not arguing for no correction at all in that region.
In the room correction case, we're specifically moving the speaker (sub) away from a nice theoretically on target response in order to not excite those modal frequencies so much so that we experience an on target response at the listening position.
In contrast in the speaker correction case, we are trying to get the speaker itself to produce the on target frequency response irrespective of what the room might do to it.
My example (I had hoped) was trying to make the case for speaker correction in the modal region (but I guess I didn't do a very good job ;D). The point of my example was that an attempt to do room correction based on a single listening position measurement could result in objectively worse results than speaker correction would have (i.e. just ironing out the 4dB lump in the speakers response). That is to say, if, in my example, one did room correction based solely on the center seat, the center would be no better off at 60Hz, and the two satellite seats would be much worse off. If you did speaker correction instead, the center seat would be no worse off and the two satellite seats would be much better off. So in that case, speaker correction would "win" (at least at 60Hz).
But the point of speaker measurement and correction is not to produce an on target frequency response and then stop completely; it's to provide a stable first step in total correction (speaker and room) by disentangling which phenomena can be corrected and which can't. You can theoretically achieve a similar effect in the modal region by measuring in multiple listening positions, but it can be hard to sort out what's causing the issues. For example, just measuring out in the room how do you know what's a dip in speaker FR (correctable) and what's a dip from a web of anti-modes (uncorrectable)? Sophisticated software might be able to tell the difference, it's true, but hard to know in any given case without doing the extra measurements to find out.
Think of speaker measurement and correction as a diagnostic or methodological step, not an end product. It may be that it's irrelevant in a specific frequency range, with a specific speaker, in a specific room (as in your case), but my example was intended to show that in the right room (or the wrong room depending on how you think of it), speaker measurement and correction could help resolve an issue that might be hard to diagnose just doing room measurements and correction. That's why speaker correction is valuable in the modal region, because without doing it before or alongside room correction, it can be very challenging to get optimal correction for multiple seats.
If you're only trying to get single seat correction, I agree that high quality room correction will get you just about everything you want in the modal region.
I hope I haven't confused the issue further :-[
-
It seems to be a question of terminology.
I don't know why you refer to that as speaker correction when you are correcting the effects of the room. Whether that correction is based on a single seat or many seems by the by to me (in terminology terms that is). At least that is my experience anyway, perhaps your experience is different.
I do agree that having the info from a near field measurement is valuable info though even if you don't do anything to it, both for the reason you give and for understanding the magnitude of room gain.
I agree with what you say though and I am not arguing for a mutually exclusive approach to any of these things. To be precise, when I said the modal region I mean the lower end of that where modes are more widely spaced. I think as you approach the transition frequency then things get less clear cut so coming at it from multiple angles may give better results. It is not something I have managed to do though as I have not found a good way to take a near field measurement of my mains (they have multiple woofers and multiple tweeters in close proximity).
-
It seems to be a question of terminology.
I don't know why you refer to that as speaker correction when you are correcting the effects of the room. Whether that correction is based on a single seat or many seems by the by to me (in terminology terms that is). I agree that having the info from a near field measurement is valuable info though even if you don't do anything to it, both for the reason you give and for understanding the magnitude of room gain.
I was hypothesizing that the speaker itself had a 4dB peak at 60 Hz, not the room. The point was that a room measurement taken in an anti-mode would not detect the speaker's 4dB peak. A 4dB response peak in the speaker is undesirable by itself (anywhere but an anti-mode), and would have the additional bad effect of making the 60 Hz resonant modes at neighboring seats go extra crazy. The point was that just correcting the 4dB non-linearity in the speaker itself would be an improvement at 60 Hz over what the single seat room correction would have generated. Does that make sense?
I agree that the room's reaction to the speaker's non-linearity is part of the issue, but even just correcting the speaker so it was flat at 60Hz with no attempt at room correction would have produced superior results at 60 Hz in the hypothetical.
You may be right, we may just have definitional issues; but I think we're understanding each other's gist at this point ;D
I agree with everything you say though and I am not arguing for a mutually exclusive approach to any of these things. To be precise, when I said the modal region I mean the lower end of that where modes are more widely spaced.
If that's what you mean I think we're on the same page. I was referring to the range from 30 Hz or so to 200 Hz where most domestic rooms have strong modes as the "modal region." If you just mean the range where a room is approaching the "no modes at all" region, I think I'm with you.
I think as you approach the transition frequency then things get less clear cut so coming at it from multiple angles may give better results. It is not something I have managed to do though as I have not found a good way to take a near field measurement of my mains (they have multiple woofers and multiple tweeters in close proximity).
We may have talked about this over in the other thread, but what kind of arrangement of drivers do you have? Does measuring at the acoustic center not produce good results? I have a 2x2 array of woofers and got good results measuring at the acoustic center (even measuring very close). It may be tougher with more widely or irregularly spaced drivers though.
-
I was hypothesizing that the speaker itself had a 4dB peak at 60 Hz, not the room. The point was that a room measurement taken in an anti-mode would not detect the speaker's 4dB peak. A 4dB response peak in the speaker is undesirable by itself (anywhere but an anti-mode), and would have the additional bad effect of making the 60 Hz resonant modes at neighboring seats go extra crazy. The point was that just correcting the 4dB non-linearity in the speaker itself would be an improvement at 60 Hz over what the single seat room correction would have generated. Does that make sense?
ah ok, I didn't get that was what you were saying. Interestingly that is roughly (I forget the absolute magnitude) my situation in that the sub has a hump centred on about 60Hz. Reducing this makes next to no difference to the in room response at the listening position, the modal response just swamps it completely. In another room, perhaps especially one that has significant treatment to deal with such modal ringing, then i imagine the result could be quite different. I have no experience of that though.
If that's what you mean I think we're on the same page. I was referring to the range from 30 Hz or so to 200 Hz where most domestic rooms have strong modes as the "modal region." If you just mean the range where a room is approaching the "no modes at all" region, I think I'm with you.
In a normal sized room, at least a normal sized room for the UK, I am pretty much talking about less than 80-90Hz. This covers the 1st 2 axial modes but not much more hence the modes are easy to spot.
We may have talked about this over in the other thread, but what kind of arrangement of drivers do you have? Does measuring at the acoustic center not produce good results? I have a 2x2 array of woofers and got good results measuring at the acoustic center (even measuring very close). It may be tougher with more widely or irregularly spaced drivers though.
my front 3 are MK MP150 so 2 woofers next to 3 tweeters. I've tried measuring at 20-30cm and at 1m, either way seemed a bit messy (each time aiming at the centre of the speaker) which I put down to being off axis from all of them, not sure really.
-
my front 3 are MK MP150 so 2 woofers next to 3 tweeters. I've tried measuring at 20-30cm and at 1m, either way seemed a bit messy (each time aiming at the centre of the speaker) which I put down to being off axis from all of them, not sure really.
If you're game to try again, I'd advise measuring the woofers and tweeters separately in the near field, by aiming at the acoustic center of whichever stage you're measuring rather than the acoustic center of the speaker as a whole.
So when near-field measuring the mid-bass section, I'd suggest positioning the mic exactly in between the two woofers. Similarly, I'd advise measuring the tweeters by positioning the mic directly centered on the center tweeter. Then stitch the two measurements together at the crossover. That way you can create a composite near-field measurement that's on axis for both stages, if that makes sense.
At 1 meter, you should probably be on axis with the center tweeter rather than pointed at the center of the speaker. Bass frequencies will find their way to the mic, Treble won't always.
-
If you're game to try again, I'd advise measuring the woofers and tweeters separately in the near field, by aiming at the acoustic center of whichever stage you're measuring rather than the acoustic center of the speaker as a whole.
So when near-field measuring the mid-bass section, I'd suggest positioning the mic exactly in between the two woofers. Similarly, I'd advise measuring the tweeters by positioning the mic directly centered on the center tweeter. Then stitch the two measurements together at the crossover. That way you can create a composite near-field measurement that's on axis for both stages, if that makes sense.
At 1 meter, you should probably be on axis with the center tweeter rather than pointed at the center of the speaker. Bass frequencies will find their way to the mic, Treble won't always.
I'll give that a try next time I get a chance to spend some time on it as I'm in the middle of getting a 3D LUT going atm, audio seemed sufficiently done to move on :)
What distance would you advise measuring at?
-
I'll give that a try next time I get a chance to spend some time on it as I'm in the middle of getting a 3D LUT going atm, audio seemed sufficiently done to move on :)
That sounds exciting, good luck!
What distance would you advise measuring at?
In the guide I suggest taking 3 inch measurements, 1 foot measurements, and a single 1 meter measurement on axis with the tweeter, and then averaging them, and I still think that's good advice based on my own tests.
If you're only going to take measurements at one distance: I think the three inch measurements are the most important, especially if you're going to go on to do room correction later on. If you get anomalous results, try moving out a little bit until things start to make sense.
-
That sounds exciting, good luck!
it was surprisingly easy; mash a few buttons, leave it going for a few hours, mash a few more buttons... and out pops a nicely calibrated projector :)
In the guide I suggest taking 3 inch measurements, 1 foot measurements, and a single 1 meter measurement on axis with the tweeter, and then averaging them, and I still think that's good advice based on my own tests.
If you're only going to take measurements at one distance: I think the three inch measurements are the most important, especially if you're going to go on to do room correction later on. If you get anomalous results, try moving out a little bit until things start to make sense.
OK. I've looked into the equivalent functions in acourate and all the tools seem to be available so I think I'm ready to give that a go when I get a chance to measure.
-
. . .I'm in the middle of getting a 3D LUT going atm, audio seemed sufficiently done to move on :)
Can you give me a link to how you did the 3D LUT?
-
Nevermind, I found it.
-
In case anyone else looks for it, I used this thread from AVS -
http://www.avsforum.com/t/1471169/madvr-argyllcms
Before and after measurements in this thread - http://www.avforums.com/threads/calibrating-an-htpc.1833478/#post-20728150
I was surprised how far my PJ had drifted in 18 months tbh but that is projectors for you I guess.
-
If you get anomalous results, try moving out a little bit until things start to make sense.
I took some measurements of the R speaker & am not sure what to make of them. I dumped them in my gdrive (http://"https://drive.google.com/folderview?id=0BxdmSMpV-t3GSTlvUmNLRl9UZHM&usp=sharing") in case you were interested to look at the raw data.
I took 3" measurements for each individual driver, 12" for the woofers as a whole and tweeters as a whole and 1m for the speaker as a whole.
As far as I know, the crossover is at 1500Hz & the crossover is different for each tweeter so that the middle tweeter behaves differently to the top and bottom tweeters in order to control the dispersion pattern in some way (I've never seen explicit details of how this works so that's as much as I know.
All graphs are smoothed using acourate's psychoacoustic & FDW functions (basically what it provides as macro 1).
tweeters_3inch.jpg; red = top tweeter, green = middle, brown = bottom
woofers_3inch.jpg; red = top woofer, green = bottom
12_inch.jpg; red = woofer, green = tweeter
correction.jpg; turquoise = sub + r at listening position, blue = speaker at 1m, black = correction filter
FWIW I tried importing these as wav's into HolmImpulse but holm crashes every time I do that, not sure what format it is after (I tried the 3 different ways acourate can export it; 24bit pcm, 32bit, 64bit).
My thoughts;
Tweeters
- the individual tweeters do appear to behave quite differently, the 3inch measurements may well be useless as a result
- the 12in tweeter measurement shows rising frequency response above 10kHz which isn't in the 3in measurements
- the 12in measurement is about 3dB down from 2kHz to 10kHz, my target curve is the b&k curve (6dB down from 1kHz to 20kHz) which is also about 3dB down between 2-10kHz
- the 1m measurement does *not* show that same downward tilt >2kHz
My gut feel is that I should leave the tweeters alone as their natural response appears to be approximately the shape I'm after anyway. The mystery is the spike above 10kHz though this is not found in a measurement at the listening position.
Woofers
- the individual 3in measurements match quite closely; they both show elevated levels between ~130-300Hz, a dip centred on ~400Hz and then rolling off from 1kHz
- the 12in measurement accentuates the 400Hz dip but moderates the <300Hz levels though there is still a peak centred at ~150Hz
- the 1m measurement still shows that ~150Hz peak
The question here is whether I should do something about the 150Hz peak & also whether the 400Hz dip should be compensated for.
Looking at the correction filter I have I can see it is cutting <200Hz so this is being "fixed" without speaker correction.
The speaker is ~280mm wide which I think corresponds to baffle step issues hitting at ~410Hz. Is this coincidence? Again the room correction filter is seeing this as the 400Hz region is at 0dB in the filter so it is trimming around that region.
Overall it looks like acourate is handling this pretty well without needing me to correct the speaker. The 1m measurement (once smoothed) seems v similar to the listening position measurement >300Hz and you could argue that <300Hz, in an untreated room, is just going to result in correcting one way then the other for questionable benefit.
I would be really interested in other views though. All this is getting me more interested in building my own speakers :)
-
I then followed your guide for stitching together multiple measurements by doing the following
- splitnjoin at 1500Hz {sum of 3 3" tweeter measurements} with {sum of 2 3" woofer measurements} to produce composite 3" measurement
- splitnjoin at 1500Hz {12" tweeter} with {12" woofer} to produce composite 12" measurement
- average these 2 with the 1m measurement
- normalised gain of this average trace with my listening position measurement
this produces the attached graph (brown is my listening position measurement, green is the correction filter, blue is the average), I guess the question is whether the final correction is really going to change much if I were to correct the speaker in the 150-800Hz range.
-
As far as I know, the crossover is at 1500Hz & the crossover is different for each tweeter so that the middle tweeter behaves differently to the top and bottom tweeters in order to control the dispersion pattern in some way (I've never seen explicit details of how this works so that's as much as I know.
Yes it definitely looks like they're using the tweeters differently to control directivity, I would base any correction of the tweeters on more distant measurements (1 foot or 1 meter).
FWIW I tried importing these as wav's into HolmImpulse but holm crashes every time I do that, not sure what format it is after (I tried the 3 different ways acourate can export it; 24bit pcm, 32bit, 64bit).
I think it's expecting 16 bit mono wav or a text file.
Tweeters
- the individual tweeters do appear to behave quite differently, the 3inch measurements may well be useless as a result
I think I agree.
My gut feel is that I should leave the tweeters alone as their natural response appears to be approximately the shape I'm after anyway. The mystery is the spike above 10kHz though this is not found in a measurement at the listening position.
All of your measurements (except at your listening position) show a differential rise above 10KHz, and all of them (including your listening position) show a 3dB-ish lump at 17 or 18 KHz.
Sound becomes very highly directional above 10K, so even if your speaker has a rise at 10KHz at almost any real room position (we're talking a few inches off axis) you'll tend to get a roll off like you're seeing at your listening position, but if you look closely at your listening position measurement, there's still that same 17 or 18KHz lump. That lump is probably a resonance in the tweeter (there's no other reason for a sudden high Q rise in FR up there), and even though sound that high up isn't super audible, it can cause intermod down below.
It looks like the Acourate filter is trying to correct it, but if it still shows up in your "corrected" measurement, I would EQ it more aggressively. I had a compression driver with a spike like that at 18.5KHz, and EQing it out was one of the best things I ever did for my highs. I was skeptical that it would make a difference, but it improved my distortion measurements (in Holm) and (on a subjective note) it dramatically reduced listener fatigue. I can't hear 18.5KHz unless it's very, very loud, but apparently it can still hurt my ears ;D
The question here is whether I should do something about the 150Hz peak & also whether the 400Hz dip should be compensated for.
It looks like Acourate is more or less doing exactly that. I think your correction is probably doing a pretty good job with those already. You could try generating correction based on the 1 meter measurement and see how it compares, but you're pretty close already.
The speaker is ~280mm wide which I think corresponds to baffle step issues hitting at ~410Hz. Is this coincidence? Again the room correction filter is seeing this as the 400Hz region is at 0dB in the filter so it is trimming around that region.
I think the baffle step is a coincidence, you'd expect to see a continued downward trajectory if it were the baffle step (rather than a dip and then a rise).
It's either a non-linearity in the driver, or (more likely) its a room boundary cancellation effect (rear wall bounce is a likely culprit). How far are your speakers from the rear wall? If the distance from the front of the cabinet to any wall is about .75ft (i.e. the speaker back is up against a wall or they're in corners), its very likely a boundary effect.
I've also seen wide 400 Hz-ish modal dips in fairly small rooms (one I can remember that had the issue was about 8ft by 9ft). It could also be a box resonance if your box is about .75ft deep (which you can test by placing something heavy on top of it and seeing if it changes)
How well does your acourate correction resolve that issue? Is there still a dip there in your measurements with the correction on? That will help you diagnose the source.
Overall it looks like acourate is handling this pretty well without needing me to correct the speaker. The 1m measurement (once smoothed) seems v similar to the listening position measurement >300Hz and you could argue that <300Hz, in an untreated room, is just going to result in correcting one way then the other for questionable benefit.
I agree that acourate is doing a very good job, and that may be (in part) because the differences between your 1 meter measurement and your listening position measurement are not very significant. I've measured a few dozen speakers in a half dozen rooms and that kind of agreement between the 1 meter measurement and the listening position in an untreated room is a pearl of great price.
You either have very a nice room (acoustically speaking), or acourate's filtering is even more incredibly sophisticated than I already thought it was ;D
I would be really interested in other views though. All this is getting me more interested in building my own speakers :)
It's a lot of fun if you enjoy tinkering, and you can get sound quality that far exceeds anything you could buy for the same amount of money. The time investment on the other hand... well, it's a hobby, right? ;D
Give a shout if you want to discuss any designs, I'm not an expert on speaker design, but I've built about four sets of speakers and a few subs, and assisted on several more. So at the very least, I can tell you about some of the "pitfalls" we encountered and how we resolved them.
[Continued below]
-
this produces the attached graph (brown is my listening position measurement, green is the correction filter, blue is the average), I guess the question is whether the final correction is really going to change much if I were to correct the speaker in the 150-800Hz range.
Given the relatively close resemblance of your 1 meter and your listening position measurements, I doubt it would make a huge change, although the listening position measurement does look quite a bit more jagged than the averaged response you attached.
One thing that I haven't seen is how your system measures with the correction filter applied (and if you have more than one listening position, how the other positions look). If you have one listening position and the corrected measurement looks nice and flat across the band, then it probably wouldn't help much to do the speaker correction first.
But it may be the case that the correction filter is adding boost in places where it won't help at all, and you'll still have a dip there but with a goofy tonal balance due to all the boost, or it may be that the cut applied by the filter isn't enough to EQ out a particularly stubborn resonance, etc. Or it may be that the correction filter creates a really weird measurement at another listening position (if you have one). There's just no way to know what's going to happen with any given correction without doing a corrected measurement.
If any of those are the case, those would be reasons to do the speaker correction separately, if you see what I mean.
-
Thanks for your comments.
Here's another graph, I've spread out the measurements so they are easy to see
green; listening position, uncorrected
blue; predicted result of convolution
red; listening position, corrected
brown; 1m measurement
turquoise; correction filter
It looks like the correction has spotted & fixed that 17-18kHz peak. It also looks like it might be overcorrecting a bit >1kHz, I might try tweaking the FDW parameters to see if that helps. There is also an anomalous behaviour around the XO (~1250-1750Hz) that could be looked at. The woofer seems to have been handled ok, the 250Hz peak is perhaps a room effect?
I probably need to go and check the L and C now. It's never done this thing :D
-
here's another graph, I've spread out the measurements so they are easy to see
green; listening position, uncorrected
blue; predicted result of convolution
red; listening position, corrected
brown; 1m measurement
It looks to me like the 400Hz piece is probably room related and the correction you've got is about as good as it's likely to get. It looks like the correction did a good job of resolving that 17KHz resonance as well.
But the corrected measurement at 1.6 KHz and around 230Hz looks non-optimal and correctable. The 1.6KHz in particular looks particularly avoidable (that huge peak only exists in the corrected trace). If I had to guess, since your crossover is right there, the application of boost and/or phase correction in that region is having unexpected consequences due to the interaction of the drivers at the crossover.
Have you tried correcting the phase around the crossover (or is that a built in part of the correction)? If not, the crossover region is one of the primary areas where phase manipulation can really pay dividends, and it's also an area where speaker correction can be helpful (because the phase relationship at the crossover can be hard to interpret/correct when off-axis).
Looking at your averaged trace of speaker measurements above, I'm not sure whether generating correction based on that would necessarily produce better results. It might, but it might not. It looks like it would result in less boost being applied at 1.6KHz at least, and if it has more "representative" phase measurements it may result in better correction around the crossover for that reason as well.
If you can manually dial down the boost at those two regions (or increase the cut), that would probably get you 90% of the way there (i.e. the correction otherwise looks pretty good). That said, since you already have the averaged speaker measurement, you could try correcting based on that and see what happens (since you've already done the hard part ;D)
In all seriousness, though, that's a nice looking corrected response overall. +/- 4dB around the target across the band is a big deal, even in a treated room.
-
It's either a non-linearity in the driver, or (more likely) its a room boundary cancellation effect (rear wall bounce is a likely culprit). How far are your speakers from the rear wall? If the distance from the front of the cabinet to any wall is about .75ft (i.e. the speaker back is up against a wall or they're in corners), its very likely a boundary effect.
I've also seen wide 400 Hz-ish modal dips in fairly small rooms (one I can remember that had the issue was about 8ft by 9ft). It could also be a box resonance if your box is about .75ft deep (which you can test by placing something heavy on top of it and seeing if it changes)
they are on wall speakers, about 5-6" deep, and my room is 13.5' x 12' x 10'
Give a shout if you want to discuss any designs, I'm not an expert on speaker design, but I've built about four sets of speakers and a few subs, and assisted on several more. So at the very least, I can tell you about some of the "pitfalls" we encountered and how we resolved them.
cool, thanks for the offer. It is just the germ of an idea atm but I'm quite tempted to just try and knock up some small monitors just for experimentation purposes really. I'm planning an extension atm which will contain a games room, it would nice to be able to build my own speakers & sub for that room. This gives me ~2yrs to experiment :)
It looks to me like the 400Hz piece is probably room related and the correction you've got is about as good as it's likely to get. It looks like the correction did a good job of resolving that 17KHz resonance as well.
But the corrected measurement at 1.6 KHz and around 230Hz looks non-optimal and correctable. The 1.6KHz in particular looks particularly avoidable (that huge peak only exists in the corrected trace). If I had to guess, since your crossover is right there, the application of boost and/or phase correction in that region is having unexpected consequences due to the interaction of the drivers at the crossover.
Have you tried correcting the phase around the crossover (or is that a built in part of the correction)? If not, the crossover region is one of the primary areas where phase manipulation can really pay dividends, and it's also an area where speaker correction can be helpful (because the phase relationship at the crossover can be hard to interpret/correct when off-axis).
acourate deals with phase correction "automatically", you can influence what it does by tweaking the frequency dependent window params & the strength of the pre ringing compensation but you can't control it directly (as far as I know). If you're interested in what acourate does then this site (http://"http://digitalroomcorrection.hk/http___www.digitalroomcorrection.hk_/Room_Correction.html") is really useful.
The basic method is that speaker correction is expressed through the XO files, this deals with linearising the speaker/driver & time alignment and results in the sweep you use to measure at the listening position. This measurement is used to create an amplitude correction for the room. Acourate then combines this with the XO to generate the final correction filters, this stage is generally an iterative process as you work through various combinations of FDW & PRC params until you get a stable outcome (things to consider include group delay discontinuities, pre ringing & IACC).
I think I'll take the Q around the effect on the XO to the acourate user group.
In all seriousness, though, that's a nice looking corrected response overall. +/- 4dB around the target across the band is a big deal, even in a treated room.
thanks, Acourate is a lot of work but the results really are great.
-
I took a few more measurements just to see if I could see more of what was going on.
Firstly I checked our 2nd seat just to see what the effect of the correction was on there. Acourate targets a single seat only but you can precisely control the frequency range over which correction is applied & you can apply arbitrary IIR filters on top. This means I could, for example, manually EQ the bass region based on a spatial average if I wanted to (I don't bother though atm).
seat2_L.jpg; red = uncorrected, green = corrected (left speaker)
seat2_R.jpg; brown = uncorrected, blue = corrected (right speaker)
I then look at the L speaker (previous graphs have been the R speaker) to compare the 1m against the LP, this is in 1m_LP_corrected_comparison_L.jpg
red; listening position, uncorrected
green; predicted result of convolution
brown; listening position, corrected
blue; 1m measurement
turquoise; correction filter
the same oscillation around the 1.5kHz XO is present here too though it's not really any worse then the actual measurement in the same range. The L speaker is +/-3 throughout apart from that XO region though so it's difficult to complain too much.
Finally I compared my LCR (all same model of speaker) at 1m, I thought this might say something about the room effects
red; left
green; centre
brown; right
I'm not sure what to make of this one.
-
It seems the anomalies must just have been a result of moving the mic between before & after measurements. I took the Q to the acourate mailing list and was asked to repeat the measurements more methodically. From this I can see I was doing acourate an injustice before as it is really producing +/- 1.5dB at the measurement position.
1-0 to acourate by the looks of it :-[
-
Sorry about that, I just noticed this thread had some replies I hadn't seen/processed. It looks like you sorted the issue out, though? Regardless, that's a heck of a response curve. I assume when you say "more methodically" you mean without moving the mic in between measurements?
-
Sorry about that, I just noticed this thread had some replies I hadn't seen/processed. It looks like you sorted the issue out, though? Regardless, that's a heck of a response curve. I assume when you say "more methodically" you mean without moving the mic in between measurements?
yes basically. There were a few other tweaks around how I worked through the workflow & organised data to make it more obvious what was going on. I have to say the surround soundstage for films is quite amazing now.
-
yes basically. There were a few other tweaks around how I worked through the workflow & organised data to make it more obvious what was going on. I have to say the surround soundstage for films is quite amazing now.
I bet!
Surround is the next big frontier for me; I pretty much wore out my welcome (and my wife's patience) building my washing-machine-sized mains, so I'm waiting a few years before introducing several additional speakers ;D
-
I still need to calibrate my system... something to do once I'm back home, just need to get a mic.. :)
Calibrating a full surround system sounds like a lot of fine tuning.
-
I bet!
Surround is the next big frontier for me; I pretty much wore out my welcome (and my wife's patience) building my washing-machine-sized mains, so I'm waiting a few years before introducing several additional speakers ;D
I get a free pass to some extent as she uses the system more than I do. This didn't stop her rolling eyes when I was suggesting building a new sub with 2-3 18" drivers the other day mind you ? ;D
I still need to calibrate my system... something to do once I'm back home, just need to get a mic.. :)
Calibrating a full surround system sounds like a lot of fine tuning.
worth it though :)