Article: Acourate Digital Room and Loudspeaker Correction Software Walkthrough

[input username here]

I'm running digital room correction just fine for DSD at native rate...

 

I spoken to EMM Labs about this at length some time ago and they were emphatic that this is simply not possible. DSD is a differencing playback format. That is, the "sound wave" is either increasing or decreasing with each sample (e.g. 1 = increase, 0 = decrease--or the other way around, I cannot remember).

 

Unlike PCM, it does not, cannot, "describe" the wave within the sample because it does not decimate the data. This is likely the great benefit of DSD: it eliminates quantization error. HOWEVER it also means that the signal cannot be shaped via DSP (of course, the signal is "shaped" via a filter to remove hypersonic noise, but that's not the same thing).

 

DSD's lack of quantization leads to a number of problematic features for the medium (along with the one big feature: potentially higher sound quality). (1) It is difficult to edit multitrack DSD because it is very difficult to level-match the tracks without quantizing them first. DSD is really an archival format for analog, rather than a digital format the (easily) allows one to manipulate the signal in production (some, like the Sonoma, use finesse to minimize the "damage" but it's still a problem that affects all DSD mastering engines). This is why many mastering programs use DXD, a kludgy compromise between PCM (decimation) and DSD (ultra-high sample rate) (2) You cannot implement a digital domain volume control. Since you are differencing, the playback device has no idea how "loud" a particular sample is relative to any other (or to a LSB). Many have opined that Ed Meitner does not like digital volume controls because of they degrade sound quality. While this may be true (I don't know, honestly), it's true-r that he simply cannot "do" a digital volume control, given his preference for single-bit DA conversion. (3) Finally, you cannot use DSP with DSD because, again, you cannot shape or alter the samples with DSD because the samples are not described via a decimation process: no room correction with DSD is possible.

 

So Miska, you are not, in fact, "running digital room correction just fine for DSD at native rate." This is not to say you cannot/are not using DSP on DSD tracks, but rather that they must be decimated first--the DSP is running in PCM (or, much less likely though possibly, in DXD); it is possible that the "corrected" PCM file is the reconverted to DSD when presented to your DAC. But, in the latter case, once the sample is decimated and thereby subject to quantization error, even if it is subsequently converted back to DSD, the singular magic of the DSD format will be lost.

[dallasjustice]

Thanks Mitchco for grabbing hold of the third rail of audiophilia! This is the topic that most audiophile don't want to touch, but everyone needs to address; no matter how great the room. I will be adding a couple of subs on my front and back walls crossed over around 50 hz. I don't like the idea of correcting high frequencies. My past experience (albeit limited) with audiolense tells me correcting HF is usually detrimental to good sound.

 

Would I be able to set a partial (low frequency) target curve along with the crossover/delay in acourate? My NOS ladder dac already produces something similar to Bob's house "perceptually flat" response. :-O Does acourate work with all sample rates up to 192k? I know DIRAC is up to 96k and I believe audiolense can handle up to 192k.

Thanks again,

Michael.

[Miska]

Okay, but how are you doing it? If Acourate can't do it, some of us would be interested in what you are using as a substitute.

 

My (admittedly limited) knowledge of DRC programs is that they all work with PCM, and most will downsample 4X files to 24/96 in order to do the RC.

 

You can use DRC filters created by number of applications, like Acourate, with HQPlayer and it performs the necessary processing. IOW, it has convolution engine for DSD.

 

Any sampling rate used for creating the filter is fine, although I recommend using maximum supported sampling rate.

 

There are two separate things, 1) creating DRC filters, 2) applyingthe filter.

 

I've been using for example RoomEq Wizard for creating filters for use in HQPlayer to deal wit some room modes in bass region.

[Miska]

I spoken to EMM Labs about this at length some time ago and they were emphatic that this is simply not possible. DSD is a differencing playback format.

 

I've always loved the especially when someone says something I want being impossible. It makes the challenge especially interesting.

 

Of course I know exactly how my software algorithms perform this operation.

[UliBru]

I don't like the idea of correcting high frequencies. My past experience (albeit limited) with audiolense tells me correcting HF is usually detrimental to good sound.

 

Would I be able to set a partial (low frequency) target curve along with the crossover/delay in acourate?

I truly believe that 99.9% of Acourate users do not have a problem with high frequency correction. Personally I would not run without it.

But it is also possible to limit the correction, see the picture.

 

[hifitubes]

I've always loved the especially when someone says something I want being impossible. It makes the challenge especially interesting.

 

Of course I know exactly how my software algorithms perform this operation.

 

Miska - can you elaborate on how you are doing this. No one believes you.

[pacwin]

I spoken to EMM Labs about this at length some time ago and they were emphatic that this is simply not possible. DSD is a differencing playback format. That is, the "sound wave" is either increasing or decreasing with each sample (e.g. 1 = increase, 0 = decrease--or the other way around, I cannot remember).

 

Unlike PCM, it does not, cannot, "describe" the wave within the sample because it does not decimate the data. This is likely the great benefit of DSD: it eliminates quantization error. HOWEVER it also means that the signal cannot be shaped via DSP (of course, the signal is "shaped" via a filter to remove hypersonic noise, but that's not the same thing).

 

That doesn't make entirely sense. What then was the basis of the Genex Mix + engine based on Sony Oxford native DSD processing?

 

http://www.oxford-digital.com/pdf/GenexSmartODL_06_10_09.pdf

 

As far as I know this is still licensable technology from Oxford Digital

 

Technology Licensees - Licensing Audio Technology & Effects

 

Patents continue to be filed in the area of native DSD signal processing

 

Patent US7302303 - Mixing system for mixing oversampled digital audio signals - Google Patents

 

And there is prior research

 

AES E-Library » Research on Cascadable Filtering, Equalization, Gain Control, and Mixing of 1-Bit Signals for Professional Audio Applications (Research on Cascadable Filtering, Equalization, Gain Control, and Mixing of 1-Bit Signals for Professional Audio Applications, AES convention 102, 1997)

 

http://www.eecs.qmul.ac.uk/~josh/documents/ReissSandler-DAFX2004.pdf (DIGITAL AUDIO EFFECTS APPLIED DIRECTLY ON A DSD BITSTREAM, Proc. of the 7th Int. Conference on Digital Audio Effects (DAFX-04), Naples, Italy, October 5-8, 2004)

[Miska]

Miska - can you elaborate on how you are doing this. No one believes you.

 

Just like any other hardware or software manufacturer, I don't want to tell how my things exactly work. I've spent quite a lot of R&D effort getting where I am and I'm all the time working on new things and improving existing ones. So it would be just plain stupid. Most of the time I'm telling and talking much more than many other manufacturers who are dead silent apart from press releases produced by their respective marketing departments.

 

You can try it out yourself if you don't believe. I'm happy with it myself, especially compared to what else is available out there. If nobody believes me, oh well, too bad...

[Matt Ashland]

Is it possible to use this on a Windows computer to generate the correction and then use the results with a plug-in on a suitable Mac OS X player (like Audirvana)?

 

JRiver Media Center on Mac uses the same 64-bit floating point SSE accelerated convolution engine as the Windows version described in this article.

 

It should also supports automatic filter bank switching, just like the Windows side.

[bobkatz]

I'm running digital room correction just fine for DSD at native rate...

 

That's astounding! To be "pure DSD" from source to DAC you would either have to implement a volume control in the DSD domain or use a pure analog volume control post DAC. Miska, please tell us about your DSD drc solution. (So many abbreviations, so little time to learn them. :-)

[bobkatz]

Okay, but how are you doing it? If Acourate can't do it, some of us would be interested in what you are using as a substitute.

 

My (admittedly limited) knowledge of DRC programs is that they all work with PCM, and most will downsample 4X files to 24/96 in order to do the RC.

 

That's one distinguishing feature of Acourate. There is no up or downsampling and Acourate works at the incoming sample rate. This requires automatic filter switching but the reward is increased sonic transparency.

[bobkatz]

Dallas, speaking of third rail, NOS dacs are the third rail of audioqphilia! Without getting into why I would never use a NOS DAC isn't it true that all current NOS Dacs are limited to 16 bit input? Well, Acourate Convolver (as well as the Convolver in JRiver) calculates at 64 bit floating point and dithers to 24 bit to feed the DAC. If you want to do drc, Dallas, properly you should use a DAC that will not truncate the high resolution information being sent to it. The issue with target roll off is the least of your worries, but yes, you can design any target. .

Thanks Mitchco for grabbing hold of the third rail of audioqphilia! This is the topic that most audiophile don't want to touch, but everyone needs to address; no matter how great the room. I will be adding a couple of subs on my front and back walls crossed over around 50 hz. I don't like the idea of correcting high frequencies. My past experience (albeit limited) with audiolense tells me correcting HF is usually detrimental to good sound.

 

Would I be able to set a partial (low frequency) target curve along with the crossover/delay in acourate? My NOS ladder dac already produces something similar to Bob's house "perceptually flat" response. :-O Does acourate work with all sample rates up to 192k? I know DIRAC is up to 96k and I believe audiolense can handle up to 192k.

Thanks again,

Michael.

[dallasjustice]

I previously PM'd you the details about my setup so you should know that my DAC fully resolves all bits and doesn't lack a reconstruction filter. I appreciate your contribution to DRC elsewhere. I don't appreciate your other rhetoric.

 

Dallas, speaking of third rail, NOS dacs are the third rail of audioqphilia! Without getting into why I would never use a NOS DAC isn't it true that all current NOS Dacs are limited to 16 bit input? Well, Acourate Convolver (as well as the Convolver in JRiver) calculates at 64 bit floating point and dithers to 24 bit to feed the DAC. If you want to do drc, Dallas, properly you should use a DAC that will not truncate the high resolution information being sent to it. The issue with target roll off is the least of your worries, but yes, you can design any target. .

[dallasjustice]

Mitch,

My mic calibration files are .frd, .csv and .xls. I got my mic from Herb and csacoustics. I am trying to use the logsweep and I can't figure out how to import my calibration files into it. It's looking for a .wav file.

[hifitubes]

That's astounding! To be "pure DSD" from source to DAC you would either have to implement a volume control in the DSD domain or use a pure analog volume control post DAC. Miska, please tell us about your DSD drc solution. (So many abbreviations, so little time to learn them. :-)

 

 

This is what I was told as a potential customer, as I am very interested in doing this, if it works, by purchasing HQPlayer. Hats off to Miska for working on this:

 

"If Acourate can export filters in WAV format you can use those. WAV needs to be mono per channel, but stereo WAVs are easy to split to mono ones.

 

Since PCM filter WAV is by necessity bandwidth limited, I have implemented "HF-expand" option for expanding the filter as flat above it's Nyquist frequency. However I recommend using as high as possible sampling rate for the filter WAV. If 352.8/384k is possible, then things would work pretty nicely without "HF-expand" too. And it's not bad with 176.4/192k either.

 

To enable DSD processing HQPlayer you need to un-check the "Direct SDM" box in "DSDIFF/DSF Settings" -dialog. Then you can perform DRC and also multichannel speaker distance and level adjustments plus main digital volume control. For any processing I recommend using max -3 dBFS main volume setting to avoid overloads. Overloads in DSD are more nasty than in PCM.

 

When source is DSD and output mode is DSD, DRC processing is performed at native source rate. You can combine it with per channel level and speaker distance adjustments plus global digital volume control, performed at native target rate. IOW, you can decide to resample. So usually when I play DSD64 content, I perform DRC and some level adjustments and output at DSD128 rate. DACs like exaSound E20 allow performing DSD64 -> DSD256 upsampling combined with processing. But you can also play DSD128 -> DSD64 if necessary.

 

If source is DSD and output mode is PCM, then DRC processing is performed in PCM at 1/16 rate after conversion."

[UliBru]

My mic calibration files are .frd, .csv and .xls. I got my mic from Herb and csacoustics. I am trying to use the logsweep and I can't figure out how to import my calibration files into it. It's looking for a .wav file.

 

Mitch has written in his article:

I also copied my microphone calibration file into the subfolder I am using for this article. Note that the mic calibration file may have a “.cal” file extension. To make it easy to import into Acourate, change the file extension to “.txt”.

Your .frd file is also a text file with ascending pairs of frequency and amplitude in each line. Thus simply rename the file extension to .txt and follow the article.

[bobkatz]

I previously PM'd you the details about my setup so you should know that my DAC fully resolves all bits and doesn't lack a reconstruction filter. I appreciate your contribution to DRC elsewhere. I don't appreciate your other rhetoric.

 

Dear Dallas: Sorry to rile you. We got out of sync. I read your PM after writing my reply in this thread! You have supplied some important information about your NOS DAC. I'm still suspicious because anything that doesn't measure flat has to be colored. Many if not most loudspeakers measure a high frequency rolloff and so target design is often to mirror the slope of the loudspeaker's native response but get rid of the major up and down shifts. So I think even with a flat target and your rolled-off DAC you're going to have some issues unless your loudspeakers are flat on axis and possibly flat in the power response. Which ordinarily (without an NOS DAC) would be EXTREMELY bright!

 

Anyway, let's please agree to agree that there are a number of controversial (third rail) subjects in the audiophile world, among them:

 

1) NOS DACs vs. Oversampling DACs

 

2) Linear phase versus minimum phase and Apodizing anti-image filters

 

3) and now, "To DRC or not to DRC". As a card-carrying audiophile up until about a year ago I was thoroughly in the "analog domain crossovers and analog domain EQ" camp. I built my own analog crossovers and even some filtering to deal with remnant low frequency artifacts. Out of necessity, because every DRC I had ever tried had some form of compromise or veiling. It was (and still is) the rule that DRC is not an audiophile-acceptable product. It's going to be very hard to convince most audiophiles that there is now even a system (Acourate, for example) that is the sonic exception that breaks the rule.

[dallasjustice]

I noticed the thread is acting weird for some reason. It's like the posts appear and then disappear. I don't know what's going on.

 

Yes, you are correct there are some incendiary topics in this world and you are correct that I managed to hit a couple of them right off the bat. I wanted to use some language to get this thread going since I find that there's too much discussion on things that don't really matter in the grand scheme of things and too little discussion concerning the things that DO matter. I believe this topic is of utmost importance and I look forward to learning as much as possible from folks like you, Uli and Mitch. I only hope that my ridicule inducing remarks can draw more attention to room acoustics. I am a criminal defense lawyer, so I am used to taking a beating for someone else's mistakes. :-)

 

Dear Dallas: Sorry to rile you. We got out of sync. I read your PM after writing my reply in this thread! You have supplied some important information about your NOS DAC. I'm still suspicious because anything that doesn't measure flat has to be colored. Many if not most loudspeakers measure a high frequency rolloff and so target design is often to mirror the slope of the loudspeaker's native response but get rid of the major up and down shifts. So I think even with a flat target and your rolled-off DAC you're going to have some issues unless your loudspeakers are flat on axis and possibly flat in the power response. Which ordinarily (without an NOS DAC) would be EXTREMELY bright!

 

Anyway, let's please agree to agree that there are a number of controversial (third rail) subjects in the audiophile world, among them:

 

1) NOS DACs vs. Oversampling DACs

 

2) Linear phase versus minimum phase and Apodizing anti-image filters

 

3) and now, "To DRC or not to DRC". As a card-carrying audiophile up until about a year ago I was thoroughly in the "analog domain crossovers and analog domain EQ" camp. I built my own analog crossovers and even some filtering to deal with remnant low frequency artifacts. Out of necessity, because every DRC I had ever tried had some form of compromise or veiling. It was (and still is) the rule that DRC is not an audiophile-acceptable product. It's going to be very hard to convince most audiophiles that there is now even a system (Acourate, for example) that is the sonic exception that breaks the rule.

[Superdad]

Audirvana and Pure Music can host Mac OS X AU plugins.

 

Two convolvers in AU plugin format are:

 

Lernvall LAConvolver - free:

Lernvall Audio

 

One issue is that you need a different filter impulse response for each sample rate of music you will play. Lernvall performs SRC on the impulse response in real time to match the sample rate of the music. Alternatively, you can configure Audirvana or Pure Music to upsample the music to a fixed sample rate, such as 192 Kbs, and then use a 192 Kbs impulse response from Acourate.

 

Hi Bob: I downloaded and looked at LAConvolver's settings from within Audirvana. I don't yet have a filter file to run (will reboot into windows tonight to try Acourate--oh, that's right, their trial does not generate filters!), but I do have a question:

 

At the top of LAConvlover's window, it says "Plug-in Sample Rate: 44,100," and the User Guide page says "If the sample rate of the file doesn't match that of the plug-in the impulse is sample rate converted as it is loaded."

That would seem to imply that even if I generate a filter file at say 176.4kHz (which is what I like Audirvana/iZotope to real-time upsample all my Redbook tunes to), LAConvolver is going to downsample my filter file. Is that the case? And what does that mean for the accuracy of the filter and for my actual music files which I want upsampled?

 

Perhaps LAConvolver wont be the best choice for those of us who use A+ to upsample? Please educate me further if you will be so kind.

Thanks,

ALEX

 

P.S. I also just remembered that the Edirol UA-5 which I can use as a mic-pre/USB ADC captures only to 96kHz, and that's with their drivers which may not be ASIO-compliant. Darn, I wish Mr. Brüggemann would allow for an actual trial product so I could see if I could sort all this out.

[Bob Stern]

Lernvall LAConvolver

 

Alex/Superdad: Sorry for the terrible lead. I find the same thing: LAConvolver seems to permit only 44.1 sample rate.

 

Bob Katz's recommendation for Reverberate seems a good prospect because the product is updated frequently and is supposed to be compatible with Mountain Lion. The cheaper "Core" version, which is what you want to run an impulse response generated by another program, is about $50.

LiquidSonics - Reverberate Core | Convolution Reverb for VST, RTAS and AU

[Superdad]

Lernvall LAConvolver

 

Alex/Superdad: Sorry for the terrible lead. I find the same thing: LAConvolver seems to permit only 44.1 sample rate.

 

Bob Katz's recommendation for Reverberate seems a good prospect because the product is updated frequently and is supposed to be compatible with Mountain Lion. The cheaper "Core" version, which is what you want to run an impulse response generated by another program, is about $50.

LiquidSonics - Reverberate Core | Convolution Reverb for VST, RTAS and AU

 

No problem. Thanks Bob. Any thoughts about the LiquidSonics plug-in versus the other one from your first recommendation, the MellowMuse? I am too tired tonight (and just finished a long heart-to-heart with our A+ senior student daughter about colleges--while having having had a bit to much to drink--me not her!) to figure out the difference between them. Both are $50 bucks. I just want whichever will take various resolution Acourate filter files and apply them as an AU within the wonderful Audirvana Plus while I upsample Redbook to 176.4kHz. (Wow, thank the OS X speller for keeping this whole paragraph from being unintelligible!)

 

Good night,

ALEX

[UliBru]

I also just remembered that the Edirol UA-5 which I can use as a mic-pre/USB ADC captures only to 96kHz, and that's with their drivers which may not be ASIO-compliant. Darn, I wish Mr. Brüggemann would allow for an actual trial product so I could see if I could sort all this out.

1) According to the website of Roland there is an ASIO driver for the UA-5 soundcard. Even if a soundcard is without ASIO driver it can be used by the Asio4All driver, which simulates an ASIO interface.

 

2) In most cases I use 48 kHz for recording. A samplerate of 96 kHz allows to record up to 48 kHz = fs/2. But most users do not have a microphone and a speaker good for this frequency range. How to correct something that you cannot measure very well?

 

3) there is a free logsweep recorder AcourateLSR2 on the Acourate website. You can also download the Acourate trial, it allows a recording too.

 

4) an efficient test of different plugins can be carried out by using test filters. I prefer a test filter that simply creates an echo. Thus you can test the filter data format and also the function. I can create such filters for you. Just tell me what format you need.

[bobkatz]

Lernvall LAConvolver

 

Alex/Superdad: Sorry for the terrible lead. I find the same thing: LAConvolver seems to permit only 44.1 sample rate.

 

Bob Katz's recommendation for Reverberate seems a good prospect because the product is updated frequently and is supposed to be compatible with Mountain Lion. The cheaper "Core" version, which is what you want to run an impulse response generated by another program, is about $50.

LiquidSonics - Reverberate Core | Convolution Reverb for VST, RTAS and AU

 

 

That wasn't my recommendation, actually. I haven't used any Mac convolvers yet. Let me tell you a bit about the Sourceforge Convolver config format, which you can also read about at the Sourceforge Convolver website.

 

1) The measuring program has to create a filter file for each sample rate. These are usually in WAV format.

 

2) Some measuring programs also will attempt to create a file in .cfg format to conform with the Sourceforge spec. The .cfg files essentially tell the convolver which wav file to use for which sample rate. And there needs to be a separate .cfg file for each sample rate as well. cfg files can be edited with a text editor and it may take you a few hours to get one done. If your convolver conforms to the Sourceforge format, I have some example .cfg files that you can use to start with.

 

3) The convolver you use should include a log or a status file to show you that it is using the correct filter file and that it is switching properly with the sample rate. it's always good to check this file until you are comfortable your convolver is working properly. Never turn your back on computers :-). If not, yes, the upsampling option is a good choice and it may even be the better sounding choice, depending on which DAC you are using and how it performs at different sample rates. My newest Avocet DAC module internally upsamples to close to 8x 192 and has a very nice (chip-based ASRC) filter and is very well designed so I'm not sure I hear much of an advantage upsampling in front of the DAC or not. There was a time when upsampling in front of the DAC was definitely advantageous in my system.

 

The differences you hear could be caused by:

 

a) the quality of the filtering in the DAC's digital filter

b) The quality of the filtering in the analog filter if there are different analog filters for each rate

c) jitter

or just d) Your imagination !!!!!!

 

Hope this helps!

[Bob Stern]

Any thoughts about the LiquidSonics plug-in versus the other one from your first recommendation, the MellowMuse?

 

Congrats on your daughter's academic achievement!

 

I backtracked and discovered that the recommendation for LiquidSonics Reverberate Core came from both Uli and *progear in the following thread. However, I think they tried only the Windows versions.

http://www.computeraudiophile.com/f11-software/drc-arc-vs-acourate-vs-xx-10324/

 

I haven't implemented convolution filters at all. I only jumped in because I'm a long-time Mac user and it looked like no one had any suggestions for a convolver for Mac.

 

Alan Jordan tried several convolvers for Mac 3 years ago and had success only with Mellowmuse IR1A. But he did not try LiquidSonics Reverberate Core.

RE: Feedback in case you convolve on a Mac - aljordan - Computer Audio Asylum

[UliBru]

I backtracked and discovered that the recommendation for LiquidSonics Reverberate Core came from both Uli and *progear in the following thread. However, I think they tried only the Windows versions.

I have even never used Reverberate :-). So I have just brought Reverberate on the list because I got a report by another Acourate user with a Mac:

On the front of the convolution players, I have two final contenders:

- Liquid Sonic's Reverberate, that fills the bill and creates a nearly identical reproduction of your convoluted file;

- SIR2, that seems a little brighter and "tighter" (nothing like ARC, which now sounds really brittle and thin in comparison)

 

Reverberate is my final choice, not only for economic reasons, but also to "save" my ears, in view of long days of work...

He also gave a hint to pay some attention on the setting:

I found out that Reverberate had a tiny switch hidden in a corner, that was doing normalization of the convolution files, "on" by default!

JRiver Mediacenter also has a normalization checkbox (on by default). By normalization bit accuracy may get lost.