Print

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

[altahomwtheatre]

Does anyone know how to import the REW EQ Text filter file into MC? 
Or does anyone know where these values are saved so I could possibly write directly to that location or file?

It is very cool to be able to put the REW filters into the parametric equalizer, but after running a number of tests - the re-enter is getting old.

Thanks for any help

[JimH]

Try a search here for "convolution".

[mojave]

I've requested it before but filter import hasn't been implemented. You can export "Filters impulse response as wav" in REW. This should be able to be used with JRiver's convolution engine, but I haven't figured it out how yet. This would save one from entering any filters and should be more accurate.

[arrowc]

You can export "Filters impulse response as wav" in REW.   That is what I do, I export each channel as a mono wav, then go to sound forge and copy paste into a 6 channel wave file. Load it into convolution, works well.

[altahomwtheatre]

I will have to do some reading about "convolution".  I dont have a clue what it is.

I did find the parametric EQ filter values in the registry.  I have created an Excel file that I can import the REW text file into and then automatically update the registry with the values.

It does a replace of values so I currently cannot use it with other EQ rules.  Maybe an update in the future.  For my current REW testing it seems to be working fine and saving a lot of time not having to re-type the 20 filters each time I make a change.

This program continues to amaze me with its capability and power.  If a pre-amp/reciever had this capability it would be the hottest seller available.  I just wish the JRiver documentation was better so that I did not have to work so hard to find and learn its capabilities.

[mwillems]

I will have to do some reading about "convolution".  I dont have a clue what it is.

I did find the parametric EQ filter values in the registry.  I have created an Excel file that I can import the REW text file into and then automatically update the registry with the values.

It does a replace of values so I currently cannot use it with other EQ rules.  Maybe an update in the future.  For my current REW testing it seems to be working fine and saving a lot of time not having to re-type the 20 filters each time I make a change.

This program continues to amaze me with its capability and power.  If a pre-amp/reciever had this capability it would be the hottest seller available.  I just wish the JRiver documentation was better so that I did not have to work so hard to find and learn its capabilities.

Convolution (in this context) applies a finite impulse response filter to a sound stream (instead of the minimum phase infinite impulse response filters applied in PEQ).  The main thing that this does that cannot be accomplished in an IIR filter is phase correction.  This allows for the possibility of applying a phase linear crossover, phase linear EQ, or "linearizing" the phase of existing minimum phase crossovers or EQ (whether in PEQ or in passive 2-way speakers).  In order to accomplish this phase magic it needs to work across a certain amount of signal time (that's the only way to adjust phase), so it winds up, in effect, delaying the signal.  The more intensive the phase manipulation, the more delay is required.  

There is an open question about the audibility of the phase distortion typically created by minimum phase filters.  It's certainly measurable, but it's unclear how audible it is.  The consensus (as I understand it) seems to be that it may be audible to some people, more so at low frequencies.  When I've done a good job with my filters, I think I can detect a difference, but I haven't ABXed it (which means I may be fooling myself).  However, I can definitely tell a difference between two filters with identical frequency response when I've goofed up the phase with a badly designed filter.  It can be tricky to get right, and is easy to get wrong (in my experience), in large part because small variations in position within a room can have very different phase relationships.  So with a badly designed filter you might get sweet sound for a 1 foot by 1 foot cube in your room, and anywhere else it sounds bizarre (pre-echo, etc.).  Intelligently designed convolution filters are less "localized."

But convolution can do frequency response correction too (it's just not a capability unique to it), and in audio, it's often used as the output of automated room correction software like Acourate or Audiolense (or the one you're already using REW).  The idea being that after taking appropriate measurements of your speakers in your room, the software then generates a frequency response and phase correction FIR filter which is applied in real time by a convolution engine.  It sounds like you're already working with the frequency response side of REW; convolution would just get you the phase correction as well.

EDIT: Here's a link to a literature summary concerning the audibility of phase distortion.  It's kind of technical, but I think it's neat: http://www.audioholics.com/education/acoustics-principles/human-hearing-phase-distortion-audibility-part-2

[altahomwtheatre]

Thanks for all this information.  I have started to do some reading but will admit that this is an advanced topic for me.  I am only a home theatre guy who has played with Room EQ built into my pre-amps (Anthem ARC and Audessey MultiEQ) but never happy with the results.

I have a feeling that I will require some equipment upgrades.  I currently using a CM140 sound meter for measurements with my onboard Realtek soundcard.  I expect that if I want to make some improvements I will need to upgrade.

Any recommendations on mic and Soundcard? 

I was thinking of ECM8000 and am lost about the soundcard.  Is the soundcards primary purpose in my case simply to allow support for ASIO and Mic preamp?  I will likely continue to output through hdmi to Onkyo 5009 for processing.

[mwillems]

Thanks for all this information.  I have started to do some reading but will admit that this is an advanced topic for me.  I am only a home theatre guy who has played with Room EQ built into my pre-amps (Anthem ARC and Audessey MultiEQ) but never happy with the results.

I have a feeling that I will require some equipment upgrades.  I currently using a CM140 sound meter for measurements with my onboard Realtek soundcard.  I expect that if I want to make some improvements I will need to upgrade.

Any recommendations on mic and Soundcard? 

I was thinking of ECM8000 and am lost about the soundcard.  Is the soundcards primary purpose in my case simply to allow support for ASIO and Mic preamp?  I will likely continue to output through hdmi to Onkyo 5009 for processing.

I've used the ECM8000, and it's a perfectly usable mic.  One thing to be aware of: it needs phantom power, so make sure whatever soundcard you grab is capable of powering it.  If you're just planning to use the soundcard as an input, you can probably use any of a variety of USB sound cards designed for musicians (I used an M-audio fasttrack to power my Behringer).

My current solution is a calibrated USB mic (the OmniMic) from Parts Express that came packaged with a pretty decent software measurement suite.  It's not cheap, but I got it at a reduced price from a friend who upgraded his system.  I probably wouldn't buy it for what they're selling it for, but it's a perfectly good mic and it's nice that it's calibrated.  It also obviates the need for a separate "input" soundcard.

[sKiZo]

Going the hardware route here ... the MC import apparently still has "issues" and is way more complicated than I want to deal with.

Pulled the trigger on a miniDSP and I'll see how that goes. Supposed to be easy - do the REW thing, capture left and right data sets, then load them into the miniDSP via their plugin. Once the data is written, hook the box up to the stereo and yer jammin ...

Or so they say ... it always looks easier when someone else is doing it. <G>

[mojave]

I've done both and and you only save a couple minutes with the miniDSP and REW export vs manual entry in JRiver. The REW export and miniDSP import is nice, though. However, there are a bunch of downsides with miniDSP with trying to get proper gain setup, clipping issues, and proper delay. I've also found that JRiver more closely matches the predicted REW results with its filters.

Let us know how it goes.

[sKiZo]

Will do ...

One of my ongoing concerns has been all the little fiddlybits you've mentioned though. Proper gain setup and clipping are real issues here USING the MC features like leveling and room correction. I've currently got all of that stuff turned off and back to playing with the volume control. I also had high hopes for the new variable loudness feature, but that's also doing strange things here and remains bypassed with MC using the system volume setting. Needless to say, not a big fan of the software solution here.

PS ... Can't pass up yet another opportunity to whine about getting on-the-fly editing features unlocked on the ANALYZE AUDIO tables. Looks to me like that'd be a simple checkmark by the production team to make it happen. A good example is REPLAY GAIN - I can either accept what MC gives me, or change it to 0 ... neither of which gives me what I want and need. Just being able to click in a field and customize the setting would be a BIG improvement.

To each their own, I guess. What works for one, etc. ... All that said, MC is still one killer server ...

[sirkus]

Do I need to set speakers distance in MC Room Correction if I use convolution like Acourate or Audiolense?

If you are interested a guy wrote a cool import tool for Rew to ReaEq/Fabfilter Pro-Q:
http://sourceforge.net/projects/reweq2eqpreset/

[JimH]

Do I need to set speakers distance in MC Room Correction if I use convolution like Acourate or Audiolense?
http://sourceforge.net/projects/reweq2eqpreset/

No.

[sorepinky]

I ran individual left and right channel sweeps up to 200Hz with REW for my speakers (measuring at the seat) and used its EQ filter generator to create parametric filters (it optimised 8 filters for the left and 7 for the right).  It's easy enough to type these into JRiver's DSP Studio Parametric EQ feature and the results are subjectively fantastic.  The sound is seriously transformed for the better (at that seat anyway).

I have a question about the correct user-sequencing of the filters for application by the audio engine.

I have sequenced them in descending order of "area under each EQ filter curve" (REW lets you order them according to frequency, gain, Q, bandwidth, or area).  I put them in area ("integral") order thinking perhaps the "most energetic"/biggest ones ought to be processed first.  I put the left channel in this order followed by the right channel in this order (15 filters in total).  Does it matter?

Or do you cubic spline fit polynomials across the overall summed EQ curve and just calculate the gain to those in one hit?

Cheers.

[sKiZo]

Or do you cubic spline fit polynomials across the overall summed EQ curve and just calculate the gain to those in one hit?

What you do in the privacy of your own room is for you to know, and us to worry about. <G>

Interesting though ... I wasn't aware that I could type REW filters directly into MC's PQ. Have to study on that. I'm currently transferring the REW filter recommendations to hardware using a Technics SH9010 parametric. That only allows five filters per channel, but it does allow you to stack them on the problem frequencies.

But ... why stop with one set of filters? Target one band as you did and note the created filters, then target another problem area (say for instance, the upper mids which give me a bit of a problem), etc ... then compile and enter the filters to MC's PQ. Is there a limit on how many filters MC can handle?

I was thinking to keep the SH9010 for handling the "big lumps", run REW again with that engaged, then use MC to fine tune the digital side. That way I'd still get the benefits of the analog parametric eq for my vinyl and other sources ...

[sorepinky]

I do whatever I can as analogue DIY.  E.g. run a baffle step compensator across the midrange (shelving low pass with Fc around 500 Hz in my case - gives roughly a 5 or 6dB attenuation at around 1kHz c.f. 200Hz or so).  Also a phase misalignment compensator to match acoustic centres of tweeter with mid (I think 71 microseconds at 2.2kHz in my case) done around a coupe of opamps at the active crossover's tweeter output buffers.  It removes a "BBC dip" across the mid/tweeter transition where there is say near-quadrature misalignment that can't be fixed by simply reversing the tweeter wires.  For a single stereo DAC (i.e. just two channels used), MC can't manage that one of course (even with the most elaborate and expensive convolution plug-ins!) since any amplitude (and phase) changes will be presented to both drivers for nil effect after acoustic summing.  Any attempt to correct a crossover dip by EQ into stereo channels will cancel acoustically and therefore fail.  Acourate and Audiolense do not address this nor mention it.  DEQX can do it better than me by delaying the whole tweeter signal, but I'm not paying for that either.  MiniDSP could do it too at a much lower price than DEQX.

MC's DSP Studio might do the baffle step part, although I have not tried the shelving filters there.

Trying to do anything about the bananas across the upper midrange won't work.  Even if it did, move you head a bit (or the mic in REW) and you have a different bunch of bananas to contend with anyway.

If the JRiver software writers are at all interested in spline fitting polynomials, there is information in a 1985 publication about numerical methods (in FORTRAN77 unfortunately) by Borse.  I could get a proper reference.  It would seem quite elegant to replace a whole lot of overlapping (summing and in some cases substantially counteractive) filters with a splined set of equations articulated across the audible range - maybe with close spacing of bands down low (to enable bass EQ) and broader spacing from say 100Hz up, to cover baffle step compensation and perhaps more shelving to tame bright tweeters.

I might be talking a lot of old fashioned nonsense from what I learned long ago and this stuff may have no place in modern digital filtering so please excuse me if I'm off track.

It's interesting that you say you are retaining analogue EQ for your other sources sKiZo.  I am doing exactly the same (except despite my dotage I wouldn't use vinyl in a pink fit  ).  My recent fiddle with the DSP studio is just to hear the results prior to building an analogue version to mimic them.    It is super-interesting - your idea of using analogue for the main problem areas (to cover all sources) and JRiver for fine-tuning your stored library over the top of it.  This is very clever indeed and I am tempted to try it later if I can't get the "mere" fixed 8 band constant Q graphic to do the job.   I suspect that if the analogue is "close enough" the DSP overlay might be inaudible however.

Oh - your question about a limit to the number of filters.  I don't know the answer, however I did have twenty something running the other day before culling it to 15 by twiddling within REW.  One easy way to reduce the number is to ease the expectation on target flatness from 1dB to 2 or 3.  I managed to retain a 1dB prediction by fiddling with the end points and LF roll-off slope.

In the meantime, I'd still like to know about the correct user-sequencing of the filters for application by the audio engine as it stands.

[mwillems]

[snip]  For a single stereo DAC (i.e. just two channels used), MC can't manage that one of course (even with the most elaborate and expensive convolution plug-ins!) since any amplitude (and phase) changes will be presented to both drivers for nil effect after acoustic summing.  Any attempt to correct a crossover dip by EQ into stereo channels will cancel acoustically and therefore fail.  Acourate and Audiolense do not address this nor mention it.  DEQX can do it better than me by delaying the whole tweeter signal, but I'm not paying for that either.  MiniDSP could do it too at a much lower price than DEQX.

The easy solution, if you want to do digital filtering, is to use a DAC with more than two output channels, which is what most folks here with active setups do.  And JRiver's digital filtering can do things that would be difficult or impossible in analog (you can have an arbitrarily large number of filters, and can do phase adjustment through convolution). 

MC's DSP Studio might do the baffle step part, although I have not tried the shelving filters there.

As long as your baffle step shelf needs a Q of 1 or less, it works perfectly (I use a JRiver shelf for bafflestep).  I recently learned that the shelving filters don't support steeper slopes than 1 Q, but most baffle step compensation filters aren't that steep, so the available shelf filters should work for most folks.

In the meantime, I'd still like to know about the correct user-sequencing of the filters for application by the audio engine as it stands.

I haven't noticed any difference in measurements based on the order of operation within DSP studio, with a few exceptions.  The big exception is phase manipulation with convolution, it seems to make a big difference where in the chain it is.  I usually put it first and leave it there.  As to the order of operations within the PEQ modules, I haven't noticed any differences based on order of operations except for really obvious ones. For example, in a fully active 2-way setup you need four outputs, and you want to clone the two stereo channels as early in the chain as possible.  The way I structure it is:

1) convolution
2) clone channels
3) apply delay for driver alignment
4) crossovers
5) everything else (baffle step, eq, linkwitz transform, etc.).

Obviously reversing 2 and 3 would have a bad effect, but in my experience, reversing 4 and 5 had no effect that I could discern, and rearranging the different elements under 5 had no effect that I could measure (which doesn't rule out an effect, but I never saw it)

[sorepinky]

Thank you mwillems.  I don't quite see your point about convolution and its position in your sequence however.  My understanding is not high on the subject of convolution, but I thought "convolution" could include all manner of adjustments including delay, low pass, high pass etc., but you seem to be applying a specific meaning to it and I'm not quite sure what that is because you then refer to "delay" as a separate entity.  I guess you mean that you drag and drop the convolution check-box up in the DSP Studio's left panel list, but that would only affect the processing order so that a third party plug-in would process before say the Parametric Equaliser.  Maybe I have missed something.

Personally I haven't tried any convolution software (VST plug-ins nor library transformations) because I can't see through the marketing to an actual benefit that couldn't be achieved with REW, the DSP Studio's filters and/or a little DIY with a soldering iron and until I'm convinced that it does anything additional that is audible, I remain in the dark.  I know for one thing those products could not fix a driver misalignment through an ordinary stereo DAC.

This is looking like an interesting thread though.  

My stereo speakers are full-range and I run an analogue active LR24 3-way with stereo Linkwitz Transforms (analogue) in a horrible asymmetric room.  This is fed by a regular stereo DAC from a stereo USB-to-AES converter attached to the PC.  As far as MC knows, it might as well be feeding a stereo DAC in front of regular passively crossed speakers.  While I appreciate that six channels of DACs would be nice for an active stereo three-way, I can't afford DEQX (nor tolerate products with a serious lack of adequate gain for most power amps and speakers).  Next time, I'll probably go down the MiniDSP route.

Cloning of channels for amplitude might be a good idea too for full range stereo speakers, but only in a symmetric room (or any room with a mono sub + satellites).  I really like the capabability of MC's individual left/right filter entry.

Anyway, in the meantime this is how the left and right channels measured in the low frequencies at the favoured seat before applying the filters in MC's DSP Studio:



The REW-generated parametric EQ tables are of course quite different for left and right channels (and do not attempt to compensate for that 90Hz banana measured for the left channel output nor the next one up) so cloning amplitudes would not be right for this room apart from the upper midrange on up I suppose.

How can I verify that the new responses match the REW predictions?  I do not know how to interface REW with JRiver.

[sKiZo]

Attempting to stay on top of things here ... as you said, interesting thread.

I tried the MiniDSP route here and thought the results a bit sterile. I believe the application of an appropriate room curve could help, but it was about then I scored the Technics SH-9010 parametric eq. Once I figured out how to set the controls to match the results from REW, I was able to get very satisfying results with that and saw no reason to continue my experiments with the MiniDSP ... for now anyway. Come the winter, the walls will start closing in, and I'll be looking for projects to keep things interesting.

Couple points on the MiniDSP/REW combination. I did find that it was possible to "stack" filter sets from REW in the hardware. Run one set targeting a specific narrow band (e.g. 20-200hz) and import those filters to the parametric eq section BEFORE the crossover in the 2.1 Advanced Crossover plug-in.



Run REW again for a new set of curves with the MiniDSP in line, then target your next problem range. Get a new set of filters, and load those AFTER the crossover section in the plugin to Outputs 1 and 3 and be sure to copy those to Outputs 2 and 4 for stereo output. Not sure why (bug?) but I was getting noise without taking that step.

You also need to make sure the crossover section is bypassed and feeding a full spectrum signal to the end of the processor chain. In a perfect world, you'd be able to just click a bypass button, but in my version I had to zero out both the low and high pass presets to get a flat trace in that section ...

That was a while ago ... so take all that with the caviat ... "to the best of my recollection" ... 

Oh. Worth mentioning ... you'll want to go with a fairly strong power supply for the MiniDSP. I tried a smaller wall wart and was getting some dropouts - switching to a 15v 1A brick I had laying about made a world of difference.  Then again, that was the older hardware version ... it looks like they've since gone from a 5-24vdc recommendation to 12v only?

[mwillems]

Thank you mwillems.  I don't quite see your point about convolution and its position in your sequence however.  My understanding is not high on the subject of convolution, but I thought "convolution" could include all manner of adjustments including delay, low pass, high pass etc., but you seem to be applying a specific meaning to it and I'm not quite sure what that is because you then refer to "delay" as a separate entity.  I guess you mean that you drag and drop the convolution check-box up in the DSP Studio's left panel list, but that would only affect the processing order so that a third party plug-in would process before say the Parametric Equaliser.  Maybe I have missed something.

Convolution can do all of the things you described, but in my set up I'm using it only to linearize the phase in my crossover region.  I'm using JRiver's PEQ to time align the channels, do crossovers, EQ, etc.  I could do all those things in the convolution phase, but it would require a longer filter, which would add latency.  Generally, JRiver's built in IIR filters are much lower latency than attempting to accomplish the same thing in convolution through an FIR filter, so I only use convolution to do what JRiver's PEQ can't do, which is phase linearization.

Personally I haven't tried any convolution software (VST plug-ins nor library transformations) because I can't see through the marketing to an actual benefit that couldn't be achieved with REW, the DSP Studio's filters and/or a little DIY with a soldering iron and until I'm convinced that it does anything additional that is audible, I remain in the dark.  I know for one thing those products could not fix a driver misalignment through an ordinary stereo DAC.

I use RePhase to generate my convolution filters; it's a free tool.  I've never used any of the pay convolution filter generators.  Like I said, I only use it to linearize the phase of my crossovers, which is (to me) barely audible.  I think I can hear the difference, and it definitely measures differently/better. It's free, so why not.  

You're right that convolution can't fix driver alignment with a stereo DAC, and that's why most folks here with active speakers use four, eight, twelve or sixteen channel DACs.  That way you have much more control over your channels and EQ.  You can do lots of things beyond time alignment, including asymmetric crossover topographies that can't be easily achieved in analogue, etc.

My stereo speakers are full-range and I run an analogue active LR24 3-way with stereo Linkwitz Transforms (analogue) in a horrible asymmetric room.  This is fed by a regular stereo DAC from a stereo USB-to-AES converter attached to the PC.  As far as MC knows, it might as well be feeding a stereo DAC in front of regular passively crossed speakers.  While I appreciate that six channels of DACs would be nice for an active stereo three-way, I can't afford DEQX (nor tolerate products with a serious lack of adequate gain for most power amps and speakers).  Next time, I'll probably go down the MiniDSP route.

There are a lot of other multi-channel DACs on the market besides DEQX and MiniDSP, and with JRiver you really don't need any of the software/firmware built into those boxes.  You can do almost all of your processing on the PC with JRiver, either through PEQ or free convolution tools.  You really just need a high-quality six or eight channel DAC, and there are a large number of those, check out these threads for info:

http://yabb.jriver.com/interact/index.php?topic=77508.0
http://yabb.jriver.com/interact/index.php?topic=55607.0

There are at least five or six six-to-eight channel DACs that are less than a fifth of the price of the DEQX and in the same range as MiniDSP, but with better distortion specs.  For example, I'm currently using a Steingberg UR824, which I got for around $650, but if you wait for a sale, you might be able to get it cheaper.  It's got 8 channels out and sounds great.  There are even multi-channel internal computer soundcards that have pretty good specs, like the Asus Essence ST with the H6 daughterboard, that cost about $200.  I used and ST/H6 combo for a while  before I got the Steinberg (although you then have to deal with the issues attendant upon having your DAC inside a PC).  

I used to have an analog active crossover, and the first time I hooked up a multichannel DAC and let JRiver do the crossover digitally, I was so blown away I swore I'd never go back to analog.  It's not quite magic, but it's close.

Cloning of channels for amplitude might be a good idea too for full range stereo speakers, but only in a symmetric room (or any room with a mono sub + satellites).  I really like the capabability of MC's individual left/right filter entry.

I'm cloning channels to split my High Frequency and Low Frequency, so room symmetry isn't an issue.  I copy the left channel to an unused channel and then high pass one and low pass the other, likewise with the right.  Sorry if I didn't explain that very well.

How can I verify that the new responses match the REW predictions?  I do not know how to interface REW with JRiver.

Have you experimented with JRiver's loopback function?  I don't use REW very much, but, for example, with Holm Impulse you can just drop JRiver into loopback and it'll feed the measurement signal right through JRiver and you can measure that way.  

[sorepinky]

Thank you again mwillems,

Now I see your cloning point.  Very clever.  Your wealth of knowledge seems vast and I thank you for sharing some of it here.  I'm sure many people will gain from the links that you provided too.  I'm off to investigate them now.

I remember trying to run REW through the Berkeley Alpha USB ASIO driver to the sound system and there were strange aberrations in the measurements brought about (I think) by the Berkeley's clock being out of sync with the PC's clock (I think others here have seen similar things with REW), so I guess JRiver's loop-back might have the same problems, but I'll certainly look at the feature.  Thanks for that.

Cheers.