Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mattkhan]

I have a couple of questions on this hence the rather general subject for this thread

Channel Selection
am I right in thinking that the channels that should be selected depend on whether the PEQ stage is applied before or after room correction? e.g. given a 5.1 setup and a desire to cut a modal peak well beneath the XO; if room correction comes first then apply the filter to all main channels + SW, if PEQ comes first then apply it to the SW only

Filter modelling
Is there any mechanism to see what the filters selected should be doing to the response? For example, does jriver's approach to PEQ equate to any of the eq device types that REW supports?

How do linkwitz transform & shelving filters work in jriver?
I was playing around with these earlier. I understand how they work in general but I was having no success in applying these in jriver. Specifically neither filter seemed to be having any effect below about 40Hz. .

My setup was REW to default device routed through jriver via wasapi loopback measuring via an EMM-6 mic. Mixer (saffire mixcontrol) configured to output to L only. Room correction applied to "move bass to sub" at 80Hz. I then added a low shelf filter with Q=1, frequency = 30Hz, gain = +5dB and remeasured. The resulting trace was exactly the same. To check PEQ was working I added a peq filter (-5dB cut with q=5 at 40Hz) to trim a room mode and could see the cut in the resulting trace. I then shifted the LS filter to 60Hz and could see that the boost had been applied.

Therefore it really looks like there is some setting somewhere that was preventing me applying that sort of boost to the low end but I have no idea what it could be. I am comfortable that my setup can take it btw so I don't have any concern that I'm going to blow something up (excluding any fat finger errors!!).

so my Qs here are around shelving filters;

* what does Q mean in the context of a shelf filter? the order of the slope perhaps?
* what does the frequency correspond to? the +/- 3dB point of the filter?
* is there a setting that prevents what it perceives to be overboosting the low end?

Thanks
Matt

[mwillems]

I have a couple of questions on this hence the rather general subject for this thread

Channel Selection
am I right in thinking that the channels that should be selected depend on whether the PEQ stage is applied before or after room correction? e.g. given a 5.1 setup and a desire to cut a modal peak well beneath the XO; if room correction comes first then apply the filter to all main channels + SW, if PEQ comes first then apply it to the SW only

Yes, but just be aware that the crossovers aren't brickwalls, so if it's close to the XO frequency it might make sense to EQ all channels regardless of placement.  For example if you're trying to take down a 65Hz resonance and your XO is at 80Hz, unless you're using a pretty steep crossover slope your mains are still going to have meaningful amounts of program content at 65Hz.

Filter modelling
Is there any mechanism to see what the filters selected should be doing to the response? For example, does jriver's approach to PEQ equate to any of the eq device types that REW supports?

There's no in-program way to graphically see what JRiver's PEQ module is doing. The "adjust a frequency filter" is the same as REW's "PK" type filter, and the REW's low/high shelf and low/high pass can also be modelled in PEQ by using the filters with those names.  JRiver is generally more flexible than most of the equalizer types in REW, so I usually use "Generic Equalizer," which tends to just spit out a large number of PK type filters which are easy to feed into PEQ.

How do linkwitz transform & shelving filters work in jriver?
I was playing around with these earlier. I understand how they work in general but I was having no success in applying these in jriver. Specifically neither filter seemed to be having any effect below about 40Hz. .

The Linkwitz transform requires you to know your system Q and f3 (like the original linkwitz spreadsheet) and your target Q and f3.  Qz and Fz are the system Q and F3, Qp and Fp are the target.  Make sure you have the right channels selected to apply the transform (it's easy to miss the "channels" box at the bottom).  If you've done that and are still getting no effect below 40 Hz, could you post your settings and/or a screencap?

The shelving filters work as expected for me, even at low frequencies, but it's possible they may be interacting with another filter.  One thing to know is that the shelving filters don't currently support a Q steeper than 1 (even though it will let you put in a higher Q).  If you need a much steeper rise or cut, you might need to do it using the "adjust a frequency" filters instead.

My setup was REW to default device routed through jriver via wasapi loopback measuring via an EMM-6 mic. Mixer (saffire mixcontrol) configured to output to L only. Room correction applied to "move bass to sub" at 80Hz. I then added a low shelf filter with Q=1, frequency = 30Hz, gain = +5dB and remeasured.

The resulting trace was exactly the same. To check PEQ was working I added a peq filter (-5dB cut with q=5 at 40Hz) to trim a room mode and could see the cut in the resulting trace. I then shifted the LS filter to 60Hz and could see that the boost had been applied.

Therefore it really looks like there is some setting somewhere that was preventing me applying that sort of boost to the low end but I have no idea what it could be. I am comfortable that my setup can take it btw so I don't have any concern that I'm going to blow something up (excluding any fat finger errors!!).

The set frequency on a shelf is not the frequency where the rise is finished, the frequency at which the rise begins in earnest (like the corner of a crossover filter).  So a 1Q LS with 30Hz frequency wouldn't reach +5 dB until 10 Hz or so (well below the audio band). EDIT: the frequency is actually the center of the shelf transition band [see below].  I was going to suggest you try setting it at 60Hz, but it looks like you did that and that seems to have worked?  Just remember that a 1Q filter can take more than an octave to reach its full boost or cut depending on the scope of the rise or fall.

Also which channels did you have the shelf set to affect? Just the sub?

so my Qs here are around shelving filters;

* what does Q mean in the context of a shelf filter? the order of the slope perhaps?
* what does the frequency correspond to? the +/- 3dB point of the filter?
* is there a setting that prevents what it perceives to be overboosting the low end?

Thanks
Matt

I think I got the bulk of this above, but to be clear, Q means roughly the same thing it means in a normal parametric EQ filter.  It's a measure of how long it takes the shelf to finish rising or falling.  Higher Q's reach their maximum rise faster, lower Q's reach their maximum rise slower.  Shelving filters can have orders, but in this case the Q is finely adjustable so that doesn't really make sense here.  For your reference, I think a .5Q is equivalent to a "1st order" shelving filter, and a 1Q is a "2nd order" shelving filter

The frequency isn't a fixed +/-dB point because the total rise isn't fixed.  It marks the start of the rise or fall. center of the transition band [see below]

I'm not aware of any overboost protection on the low end.

[mattkhan]

Thanks for the info.

The Linkwitz transform requires you to know your system Q and f3 (like the original linkwitz spreadsheet) and your target Q and f3.  Qz and Fz are the system Q and F3, Qp and Fp are the target.  Make sure you have the right channels selected to apply the transform (it's easy to miss the "channels" box at the bottom).  If you've done that and are still getting no effect below 40 Hz, could you post your settings and/or a screencap?

I used the xls from the minidsp site for the calculations. I will doublecheck and remeasure later this week and post back with the results.

The shelving filters work as expected for me, even at low frequencies, but it's possible they may be interacting with another filter.  One thing to know is that the shelving filters don't currently support a Q steeper than 1 (even though it will let you put in a higher Q).  If you need a much steeper rise or cut, you might need to do it using the "adjust a frequency" filters instead.

The set frequency on a shelf is not the frequency where the rise is finished, it's the frequency at which the rise begins in earnest (like the corner of a crossover filter).  So a 1Q LS with 30Hz frequency wouldn't reach +5 dB until 10 Hz or so (well below the audio band).  I was going to suggest you try setting it at 60Hz, but it looks like you did that and that seems to have worked?  Just remember that a 1Q filter takes well more than an octave to reach its full boost or cut.

Also which channels did you have the shelf set to affect? Just the sub?

I had it set to affect the sub only with the PEQ stage listed after room correction. I applied the LS at 30Hz and saw no difference in the output at all. I then changed it to 60Hz and saw the expected lift clearly visible. I will repeat the experiment next time I measure and come back with screenshots. Tonight's rabbit hole is getting convolution working though so no doubt I'll have another thread up by midnight

[mojave]

I used the xls from the minidsp site for the calculations. I will doublecheck and remeasure later this week and post back with the results.

That is a great way to determine what numbers to use. You have to take a close mic measurement of your subwoofer and match in the spreadsheet in order to get the proper system Q and F3.

[dvogel1]

I measured a couple of shelving filters to determine how the parameters affect the response. I think the frequency setting is the center frequency, i.e., the frequency at the midpoint of the transition.

Bandwidth and Q are not interchangeable. A Q of 0.5 does represent a 1st order response. Altering the Q stretches or shrinks the range of frequencies (bandwidth) that the filter impacts. They are inversely proportional.

If REW can measure electronic response it should be possible to measure your filter settings using the Loopback feature of MC.

The Parametric Equalizer function in MC is the most flexible and thorough DSP configuration utility I have encountered. It would be the best software I've ever used if it had a graphing function (not so subtle hint). That would help users immensely. Documentation would help too.

[mwillems]

I measured a couple of shelving filters to determine how the parameters affect the response. I think the frequency setting is the center frequency, i.e., the frequency at the midpoint of the transition.

Interesting.  My measurements suggested that the frequency was well before the midpoint.  What specific filter parameters did you test (I'll see if I can confirm on my end)?  It would be a heck of a lot easier to use if that were true.

Bandwidth and Q are not interchangeable. A Q of 0.5 does represent a 1st order response with a 6 dB/oct slope, Q = 1 has slope of 12 dB/oct. Altering the Q stretches or shrinks the range of frequencies (bandwidth) that the filter impacts. They are inversely proportional.

Thanks for that. I'm aware that Q and bandwidth vary inversely, but I didn't put it very well above.  I've edited it to be clearer.

The Parametric Equalizer function in MC is the most flexible and thorough DSP configuration utility I have encountered. It would be the best software I've ever used if it had a graphing function (not so subtle hint). That would help users immensely. Documentation would help too.

I agree, it's a really impressive DSP module, but a graphic representation would be very helpful.

[dvogel1]

I setup a High Shelf at 1 kHz, Q = .5, Gain = -10 dB. The result is as expected, the response is down 5 dB at 1 kHz and the slope is 6 dB/oct.

Please note that I measured the result. The Analyzer isn't useful to determine actual response but it is a valuable tool to spot test outputs.

[mwillems]

I setup a High Shelf at 1 kHz, Q = .5, Gain = -10 dB. The result is as expected, the response is down 5 dB at 1 kHz and the slope is 6 dB/oct.

Please note that I measured the result. The Analyzer isn't useful to determine actual response but it is a valuable tool to spot test outputs.

Thanks.  My previous tests with an external measurement suite (Holm) suggested that the frequency was before the midpoint, I'll try your settings and measure to confirm.

I know what you mean about analyzer not being a good indicator, I don't typically use it except for basic troubleshooting.  It's most useful to see where in the chain something isn't working correctly.

[mattkhan]

I got asio line in working (as per http://yabb.jriver.com/interact/index.php?topic=86792.0) so can share this measurement for the discussion

attached is a +10dB LS at 30Hz with Q=0.5 and Q=1.0 and the HS mentioned earlier (at Q=0.5 and 1)

I had no soundcard cal file loaded but seems like I need one.

[mwillems]

I got asio line in working (as per http://yabb.jriver.com/interact/index.php?topic=86792.0) so can share this measurement for the discussion

attached is a +10dB LS at 30Hz with Q=0.5 and Q=1.0 and the HS mentioned earlier (at Q=0.5 and 1)

I had no soundcard cal file loaded but seems like I need one.

Well that confirms that the frequency setting is the center of the transition band, and the HS filters are working correctly (i.e. the shape and slope). It does look like the sound card may need some calibration at the lower frequencies, not sure how to account for that measurement of the low shelf behavior otherwise (it's showing way more than 10 dB of boost on the .5Q shelf).  Is that an RTA trace or a sine sweep BTW?  I've seen results like that with sine sweeps when the measurement software isn't actually producing sound below 20Hz, but is just extrapolating from existing data.

Thanks for those measurements, though. I hadn't had enough time at home to get my rig set up. I'll edit my post above to make it clear for anyone following after.

[dvogel1]

I performed the same test. At Fc = 30 Hz the response doesn't have enough low end to rise a full 10 dB. The cursor is at 30 Hz (5 dB).

Fc = 30 Hz, Gain = +10 dB
green - Q = 0.5
blue - Q = 1.0

[mwillems]

I performed the same test. At Fc = 30 Hz the response doesn't have enough low end to rise a full 10 dB. The cursor is at 30 Hz (5 dB).

Fc = 30 Hz, Gain = +10 dB
green - Q = 0.5
blue - Q = 1.0

That looks a little more like it; those look like textbook filters.  

I offered my own guesses on this above, but do you have any thoughts on why mattkhan's LS loopback measurements might look so odd?

Regardless, thanks for taking the extra time to post the measurement!

[mattkhan]

my measurements do look odd and I'm not sure the lack of a soundcard cal is to blame. I don't believe the response of the card varies that badly, I don't think I see that variation in other measurements vs a different interface I have a cal file for. I will need to doublecheck this obviously to be sure.

@dvogel1 - how did you take your measurements?

Is that an RTA trace or a sine sweep BTW?  I've seen results like that with sine sweeps when the measurement software isn't actually producing sound below 20Hz, but is just extrapolating from existing data.

It was a normal REW sweep, 15Hz-24kHz 256k length and levels were normal (no clipping) through the sweep.

[dvogel1]

I use ARTA software and my setup is hardwired. It took me several attempts to understand how the loopback works and tie it all together. Pay attention to how Windows is setting up the recording device and be sure the loopback input is not "playing through" the output (test signal) device.

How does REW work for electronic measurements? It's unlikely that you need to calibrate your soundcard (cal is generally for mics only). REW should allow you to do two channel measurements where one channel is the reference (test) signal, but that shouldn't be necessary. I'm using a single channel setup.

I've attached an overview of my setup. Write your setup down once you get it working. I'll try REW once you figure it out...

[mattkhan]

My setup in this case is using asio line in and saffire mixcontrol to handle routing through its loopback channels

Set REW output to DAW7
Set DAW7 -> Loopback1 in mixcontrol
Open JRiver Asio Line In for 1 channel offset 14
Set DAW1 -> Loopback2 in mixcontrol
Set REW input to Loopback2

This is my 1st attempt at using asio line in so it could be the problem.

[mattkhan]

I think I have the problem pinned down now, there were 2 causes

1) the response of the virtual loopback
2) I had mic cal file loaded 

I've attached a comparison of the response from the physical vs virtual loopback with no soundcard cal file. I then created a soundcard cal file for the virtual loopback & repeated the measurements of the PEQ which now shows a valid measurement.

The confusing bit is that the soundcard cal file for the virtual loopback is ruler flat so I have no idea which magic switch I have thrown to make it work  ?

[mwillems]

I think I have the problem pinned down now, there were 2 causes

1) the response of the virtual loopback
2) I had mic cal file loaded 

I've attached a comparison of the response from the physical vs virtual loopback with no soundcard cal file. I then created a soundcard cal file for the virtual loopback & repeated the measurements of the PEQ which now shows a valid measurement.

The confusing bit is that the soundcard cal file for the virtual loopback is ruler flat so I have no idea which magic switch I have thrown to make it work  ?

Great news 

Not sure about the mystery of the ruler-flat soundcard .cal file.  The un-calibrated software loopback response does look a lot like a mic calibration, now that you mention it; is it possible the mic cal was still on with that measurement?

[dvogel1]

Please note that I edited my reply on Jan 4 (reply #4) and removed the comments regarding fixed filter slopes. Filter order doesn't imply a fixed response slope.

I took another try at making loopback measurements on my MC PC itself, instead of using a second computer to take measurements with. It's painfully easy. The measurement software (ARTA) is setup to output the test signal to the motherboard default output and record data from the discrete soundcard line input. I setup MC to listen to the motherboard default audio (WASAPI). MC then processes the input from the default audio through the DSP engine to the specified Audio Device (HT Omega soundcard/ASIO). A single cable is needed to connect the soundcard line input to the soundcard output to be tested (like a second loopback but for the measurement). My setup is 8 channel output.

Did I mention how impressive, logical, and otherwise excellent JRiver MC is? It's beautiful software.

[mwillems]

Please note that I edited my reply on Jan 4 (reply #4) and removed the comments regarding fixed filter slopes. Filter order doesn't imply a fixed response slope.

I wondered about that when I originally read it.  None of my measurements suggested a fixed slope, but I was starting to question my earlier measurements because I obviously managed to get the frequency issue wrong    

Thanks again for helping us get all this sorted out, we're all working together in the spirit of inquiry.

I took another try at making loopback measurements on my MC PC itself, instead of using a second computer to take measurements with. It's painfully easy. The measurement software (ARTA) is setup to output the test signal to the motherboard default output and record data from the discrete soundcard line input. I setup MC to listen to the motherboard default audio (WASAPI). MC then processes the input from the default audio through the DSP engine to the specified Audio Device (HT Omega soundcard/ASIO). A single cable is needed to connect the soundcard line input to the soundcard output to be tested (like a second loopback but for the measurement). My setup is 8 channel output.

Did I mention how impressive, logical, and otherwise excellent JRiver MC is? It's beautiful software.

I know what you mean.  When I started using JRiver I had this motley assortment of sound devices that I used for measurement (cables, adapters, etc.) and for playback with my active speakers (a crossover in a box, DSP, etc.).  Now, I've got about four or five fewer "boxes" in my living room because JRiver can do such an incredible amount of heavy lifting, and can listen anywhere in the house into the bargain.  

JRiver changed almost everything about how I listen to music (except the chair and the guy in it)  

[SteveR]

The Linkwitz transform requires you to know your system Q and f3 (like the original linkwitz spreadsheet) and your target Q and f3.  Qz and Fz are the system Q and F3, Qp and Fp are the target.  Make sure you have the right channels selected to apply the transform (it's easy to miss the "channels" box at the bottom).  If you've done that and are still getting no effect below 40 Hz, could you post your settings and/or a screencap?

I'm experimenting with the PEQ to add some LF boost ot offset the natural roll of of my 2 channel stereo speaker set up. I'm using the Linkwitz spreadsheet (link from the wiki), and I'm confused as to which F value to use in the DSP. You're suggesting using F3, but the spreadsheet model calculates Fsc and F3 from the box data, and it is Fsc, the system resonance that is carried forwards onto the Transform Calculator tab, where you have f(0) and Q(0) as the system data, and Q(p) and f(p) as the target figures, which I presume are Q/f3 and Qz/Fz respectively.

Which value do I need to enter into the DSP for Fp? System Resonance, or Lower Cut Off Frequency?

Thanks.

[mwillems]

You don't need to enter system resonance anywhere in the JRiver DSP module.

Qz is your current unmodified system Q.  Fz is your current unmodified -3dB point.  You'll need to measure those.

Qp is your desired system Q.  Fp is your desired -3dB point (not the resonance).

So you need to choose the Qp and Fp based on your system needs and what the calculator tells you your excursion limits will allow.   

Qz and Fz are the same as Q(0) and F(0) as shown in this link:
http://www.linkwitzlab.com/pz-eql.xls

[SteveR]

Thanks for the quick reply. I'm not able to measure them in room, as have not yet the capability to do that, so am using the theoretical values from the spreadsheets (True Audios Linkwitz Transformer Circuit Design Spreadsheet). The input box parameters are given to me by the speaker manufacturer, so they should be correct. To be honest, I'm trying to use the PEQ as a more scientific bass adjustment, rather than simply using the EQ sliders and using my ears.

One other thing I noted, when I switch in the PEQ the overall track volumes drops massively, rather than an increase in bass volume. What is the reason for that? I don't have any form of volume levelling checked.

Thanks again.

[mwillems]

Thanks for the quick reply. I'm not able to measure them in room, as have not yet the capability to do that, so am using the theoretical values from the spreadsheets (True Audios Linkwitz Transformer Circuit Design Spreadsheet). The input box parameters are given to me by the speaker manufacturer, so they should be correct. To be honest, I'm trying to use the PEQ as a more scientific bass adjustment, rather than simply using the EQ sliders and using my ears.

Attempting a Linkwitz Transform without taking measurements can be risky, especially if you'll be applying significant boost.  The manufacturer's measurements are a good starting place, but there are variances.  If you're not applying a whole lot of boost, it won't be a big issue, but it may also not lead to correct results.

Also (forgive me for asking if you've already thought of this), but your speakers aren't vented/ported, right?  If they are, you're almost certainly wasting your time with a linkwitz transform (unless they have an unusually low port tuning). I only ask because most commercially produced speakers seem to be vented these days, and an LT is not usually a good idea for vented designs.

One other thing I noted, when I switch in the PEQ the overall track volumes drops massively, rather than an increase in bass volume. What is the reason for that? I don't have any form of volume levelling checked.

Thanks again.

I don't have MC in front of me right now, but there's a check box at the bottom of the linkwitz transform options that reads something like "normalize volume" or something like that.  It reduces the output by 10dB, to account for boost the linkwitz transform is applying and prevent clipping.  If you're sure you've got enough digital headroom you can uncheck it to get that 10dB back, but don't uncheck unless you're sure you've got enough headroom.

A linkwitz transform often applies significant boost, and if you have your digital volume maximized this will cause clipping.  So you need to make sure you have enough digital headroom to account for the boost of your specific transform (which may be more or less than 10dB).  If you use internal volume and don't typically set it above 80%, you may not need the check box; just make sure to set a maximum volume that allows enough headroom.  And if your transform is applying a lot of boost, the 10dB may not be enough, and you'll need to further offset it.  

[mojave]

The miniDSP spreadsheet is very helpful for setting up a Linkwitz Transform after you have taken a close-mic measurement of your subwoofer, provided it is sealed as mentioned by mwillems.

[SteveR]

Attempting a Linkwitz Transform without taking measurements can be risky, especially if you'll be applying significant boost.  The manufacturer's measurements are a good starting place, but there are variances.  If you're not applying a whole lot of boost, it won't be a big issue, but it may also not lead to correct results.

Also (forgive me for asking if you've already thought of this), but your speakers aren't vented/ported, right?  



A linkwitz transform often applies significant boost, and if you have your digital volume maximized this will cause clipping.  So you need to make sure you have enough digital headroom to account for the boost of your specific transform (which may be more or less than 10dB).  If you use internal volume and don't typically set it above 80%, you may not need the check box; just make sure to set a maximum volume that allows enough headroom.  And if your transform is applying a lot of boost, the 10dB may not be enough, and you'll need to further offset it.  

Thanks. I'm applying around 3-5dB boost at most, I'm moving the -3dB roll off point from around 55Hz to 45Hz, so less than half an octave. The speakers are indeed sealed, a pair of ATC SCM40 Mk2, which also have plenty of excursion and power handling. The whole experiment has been done at low SPLs anyway to ensure that the analogue circuits are all within their rated capacity. I hadn't thought about digital clipping, I'll check what headroom I've got.

The miniDSP spreadsheet is very helpful for setting up a Linkwitz Transform after you have taken a close-mic measurement of your subwoofer, provided it is sealed as mentioned by mwillems.

I'm not using a subwoofer, I was just trying to apply some relatively modest bass lift to the stereo speakers. Ine the long term, I'm going to add a sub and probably something like the DSpeaker antimode so that all sources are corrected, not just J River MC, I use Spotify a lot as well.