Macro 6 ?

User avatar
dsnyder
Posts: 21
Joined: 22 Oct 2019 13:52

Re: Macro 6 ?

Post by dsnyder » 28 Nov 2021 22:26

I've not given up, but I'm sad to report that I'm not getting closer to the results I desire.

I created a fresh project workspace folder with only my target.tgt file and original pulse recordings and started from scratch. My original pulse recordings were created using my Cross•Spectrum Labs EMM-6 mic and Focusrite 2i2 gen 2 interface. Microphone alignment was spot-on, and the playback level was at or under 75 dB. During the same session, I also recorded measurement sweeps in REW with exactly the same mic position and setup just to verify. Here's a link to my REW file (with and without subs engaged): dsnyder-REW.zip. As I've mentioned elsewhere, I find REW helpful to check my work, before and after corrections. I trust Acourate, but it's easy to make mistakes along the way. Results are only as good as the process followed…especially when recording the pulses!

This time, I tried an EPW of 6/0.9 at the suggestion of Ecwl. This enabled me to get by with less pre-ringing compensation, which I imagine is good. Here are my settings for Macro 4:

Image

I created the filter with no ICPA correction and loaded that in to Roon. Next, I tried compensating for only a single hump in the right channel at 69.45 Hz. Here's Macro 6:

Image

I've saved both filters as presets in Roon so that I can compare…not as seamlessly as Mitch's Hang Loose Convolver, but the the differences are sufficiently obvious that aural memory and precise level matching are not necessary.

Sadly, the listening results are poor, a bit worse even than with my previous attempt with EPW of 5.2/2 and more ambitious ICPA corrections. So far, the most coherent-sounding results are with 5.2/2 and no ICPA, but I'll keep experimenting. I have not ruled out the possibility that there's something wrong with my pulse recordings, but if that were the case, 5.2/2 with no ICPA would sound wrong as well, right?

My goal is to produce a correction with bass that is at least as coherent-sounding in phase as no correction at all (which is excellent) and with desired improvements in time-domain performance and overall tonal balance. This should be possible with V2, but the correct process continues to elude me, even after my improved understanding of ICPA compensation in Macro 6.

Heres a link to my new, cleaned up project folder: dsnyder_2021-11-28.zip

Once I have an opportunity to record new pulses (involves moving a chair and table out of the way), I'll share the results. In the meantime, I'd be grateful for references to general guidance on EPW selection. My limited understanding from Mitch's book is that we iterate until we have minimal or no significant "vecin" peaks above 100 Hz on the Phase plot and a Time plot with minimal or no pre-ringing, a nice high peak in the step, and parallel or matching behavior in both channels as time progresses. Generally, the high-frequency window width should be small to prevent over-correction. Higher values (above 4) for low-frequency window width tend to produce results that better conform to the target but can also produce unstable phase behavior. However, I do not have a good algorithm for arriving at optimal EPW values. Currently, I just randomly try low-frequency width values from 3 to 9 and high-frequency values from 0.9 to 2.2, stopping when I see results that produce a reasonable compromise between Time and Phase behavior. I do not know when, if ever, it would be a good idea to use different EPW values for the left and right channels, for example.

Thanks again for having a look. I'm hoping to get well-tuned with v2.0 eventually…
Acourate system: Roon -> Gustard X16 DAC -> Topping Pre90 -> Orchard Audio monoblocks -> Fritz Carrera 7 BE monitors + REL T7i subs
Ecwl
Posts: 48
Joined: 19 Feb 2021 01:05

Re: Macro 6 ?

Post by Ecwl » 29 Nov 2021 00:08

I think if you didn't get much improvements with ICPA then you should go without it? Hard to know why. I still wonder if you should just skip the corrections >100Hz for ICPA and see if you hear an improvement. But this time around, it's clear the ICPA settings were correct.

As for the 5.5/2 vs 6/0.9 setting, the differences would be that if you measure the peak of the step response, you're getting a much higher peak with 6/0.9 rather than 5.5/2 so the filter should sound a bit more dynamic with every drum strike or string pluck. The downside though is that you can see that in the RT60 measurements of the two filtered corrections that 6/0.9 offers less correction in the 100Hz range so you're getting more reverb in that area. It is possible that if you keep increasing the number to say 7/0.9, you'll get more correction and still be able to preserve the dynamics of the step response.

To me, it's what you like and what you don't like since it's your system.

I've attached the VBA filter for your system because you said on Audiophile Style that your room length is 15.5 feet and your resonance frequency would be 36.3Hz so the filter is built to crossover at 50Hz with a gain of 0.7 to take away most of that resonance but not all. You can always try add that prefilter (same for both channels) to your current favorite correction (your preferred target response, 5.5/2 and no ICPA) and see if having that pre-filter creates a better convolution filter overall. Once again, if you don't like it, you don't have to use it. The main downside to the prefilter is that it would decrease the gain by up to 4dB which can affect the soundstage depth or depth of volume of vocals and instruments. Like I said, if you have time, you should try it.
Attachments
VBA50Hz7gain.zip
(493.34 KiB) Downloaded 79 times
User avatar
UliBru
Posts: 445
Joined: 19 Oct 2019 13:58

Re: Macro 6 ?

Post by UliBru » 29 Nov 2021 09:51

Hello David,

now I have checked your files more deeply. See my proposed correction here. You can copy the files into a workspace folder, then edit the file Acourate.ini and replace all pathnames C:\AcourateProjects\DavidSnyder by the workspace pathname. Then run Acourate, select the workspace and it remembers all the settings.

Some remarks:
1. in one of your projects I have detected prefilters of gain > 4 dB. If you create prefilters normalize them to 0 dB gain.
2. I have created a different target. Please test it also by listening. Of course you can modify it to your needs
3. my EPW approach in macro4 goes like this: usually I start with 5/5, 5/5. Then I check the result by the test convolution. Before I apply PRC I reduce the EPW values in iterative steps. So I do down to 2/2, 2/2. If the peal in the group delay display get quickly narrow I select the EPV value where the peak has disappeared. Sometimes also low EPW values do not help. Then I increase the EPV values to avoid narrow peaks and finally apply PRC.
In your case I have found that EPW 2.3/2.3, 2.3/2.3 is ok for me. In nearly all cases I use identical values. So I rarely apply different values for low/high frequencies or left/right channel.
With EPW 2.3 the step response already looks pretty nice.
4. Finally I have checked ICPA once again. With my prefilters + target I had to slightly change the ICPA values. But there is no preringing yet in the step response.
5. If you like to verify Acourate filters by a measurement then you can do it directly, REW is not required. Select the samplerate 48000 in Acourate. Load Cor1L48 and Cor1R48 into curve 1+2. Save as stereo WAV. Apply the wav file in the Acourate logsweep recorder.

I hope this helps.
Stay well tuned
Uli

Moderator
Acourate system: JRiver/Roon -> AcourateConvolver -> miniDSP U-DIO8 -> TacT M/S2150 amps -> DIY horn speakers
User avatar
dsnyder
Posts: 21
Joined: 22 Oct 2019 13:52

Re: Macro 6 ?

Post by dsnyder » 30 Nov 2021 17:21

Hi Uli,

Thanks so much for taking a closer look. I will give your changes a try and let you know soon.

On the pre-filter, that was a good catch. The one I was trying with a gain of > 4 dB was a failed experiment. The pre-filter normalized to 0 dB was reducing the gain in low frequencies too much, so I tried a non-zero normalization. However, I was not thinking clearly and forgot that after inversion, gain would be reduced even more! LOL. For my next experiment, I may try normalizing to -4 dB instead, which I think will be closer to the desired results.

Your explanation of EPW approach was helpful. I'm looking forward to trying 2.3/2.3. I'll go through the same process you followed, starting with 5/5 and working down so that I can observe changes in group delay peaks, etc. Thanks again.

-- David
Acourate system: Roon -> Gustard X16 DAC -> Topping Pre90 -> Orchard Audio monoblocks -> Fritz Carrera 7 BE monitors + REL T7i subs
Ecwl
Posts: 48
Joined: 19 Feb 2021 01:05

Re: Macro 6 ?

Post by Ecwl » 30 Nov 2021 22:45

It was a pleasure to be able to see Uli’s workflow compared to what I normally do. I learnt quite a lot from it.
I was previously told somewhere else that Group Delays <100Hz and preringing corrections are not audible so I should feel free to crank up the EPW settings. Trying Uli’s recommendations and reflecting on my own personal workflow, I realized, aggressive bass correction with Group Delays requiring preringing correction is audible to me as some bass notes has a wobbly sound to them that disappears when you decrease the correction level and stopped preringing correction.
The challenge is that I do find that setting the low frequency EPW correction to 4-6 tend to create a more coherent bass that is very appealing (unless you start noticing those wobbly sustained bass notes). Certainly higher settings seems to at least lower the RT60 in the bass region. Moreover, depending on the speakers and space, some systems, like DSnyder’s cannot have a large low frequency EPW setting without Group Delays. So instead of using 2.3/2.3 or 2.2/2.2 for his system, the most I can get without group delay would be 2.7/1.2 for the left channel and 3.8/1.2 on the right channel. That may or may not give DSnyder satisfying bass correction (but probably not).
I have to admit, I find myself to be exquisitely sensitive to loss of transient accuracy and dynamics which seems to correlate well with the amplitude of the first peak of the corrected step response. And when the low and high frequency EPW is set the same like 2.2/2.2, the peak often is not as high as when the low frequency setting is higher than the high frequency setting. Using Uli’s files and DSnyder’s data set, just looking at his left channel, using Uli’s rule of not letting there be any group delays, I measured the peaks in macro 5 for example with maximum EPW settings like this:
2.6/1.1 2.03936
2.7/1.2 2.06973
2.7/1.3 1.99789
2.7/1.4 1.95615
2.6/1.5 1.92949
2.2/2.2 1.82845
Once I have the values for both channels, I try to pick a set where the left & right channel combo have the highest IACC.
And I would often choose the EPW’s with the highest step response peak. Perhaps this method also introduces its own artifacts but it sounds the best to my ears. I guess all corrections are just about trade offs. Like I prefer VBA prefilter over Alternative to VBA prefilter.
Anyway, I’m getting off topic from Macro 6.
User avatar
dsnyder
Posts: 21
Joined: 22 Oct 2019 13:52

Re: Macro 6 ?

Post by dsnyder » 01 Dec 2021 07:05

Hi Uli,

The new target looks reasonable. Like you, I prefer for it to reflect the natural response of the speakers in the room rather than fighting them too much. I'm not sure what I think about the slight bump at 60 Hz, but I'll give it a listen.

Image

Looks like EPW in Acourate.ini is 2.2 across the board (Macro 4). From test convolution (Macro 5), the phase looks quite relaxed compared to what I'm used to seeing:

Image

Step response is not terrible without ICPA:

Image

But, it's definitely better with ICPA:

Image

This was the case with my previous attempts too, but the sound was worse. Hoping that's not the case here. I'll definitely let you know, but I have faith. :)

For Macro 6, it looks like you just had to compensate for the two non-parallel bumps below 100 Hz in each channels and the two parallel bumps at around 180 Hz. All relatively low-Q fixes.

Image

It's a bit late this evening for me to pull out my mic and confirm these results with REW, but I do have time to listen. Will start with that in a few minutes.

The reason I prefer to verify with REW vs. Acourate is that, as far as I can tell, the method you proposed plays the sweeps with no modifications and then does a convolution of the recordings with the correction filters. To me, this is not the same as energizing the room with the convolved signal and recording it with no corrections (apart from a mic correction curve). I realize that, mathematically speaking, they should be identical, but until my brain is able to run apply convolution filters to what I hear, I feel better about doing what seems like a proper end-to-end test. However, please let me know if I'm not understanding correctly how LogSweepRecorder works.

Edit: Sadly, the filter as designed, with ICPA corrections sounds super phasey in the bass. In fact, bass lines in acoustic recordings almost sound like they are coming from behind me with the filter enabled. I tried a number of recordings with natural bass that should appear to come from behind the loudspeakers, and it sounds like it's coming from behind my left ear. Craziness.

Much better sound with exactly the same filters but ICPA disabled. I may tweak the target curve slightly to taste or try your EPW settings in my project. I'm surprised that I'm the only person who is experiencing crazy, phasey bass with any amount of ICPA enabled, but perhaps it's something odd with my room or setup. I may try another pass with the subs powered off to see if that makes a difference. Failing that, there's plenty in Acourate to enjoy without ICPA. Thanks again.

Edit #2: I guess what's most frustrating about this, besides not being able to use this cool new feature in Acourate, is that, until now, I felt that I could trust the Acourate simulation results and what I see in REW as confirmation that corrections are moving in the right direction. A strong correlation between textbook-looking step response and good sound was reassuring. This experience has left my faith in these measurements shattered. I guess the microphone only has one "ear" while I have two.
Acourate system: Roon -> Gustard X16 DAC -> Topping Pre90 -> Orchard Audio monoblocks -> Fritz Carrera 7 BE monitors + REL T7i subs
User avatar
UliBru
Posts: 445
Joined: 19 Oct 2019 13:58

Re: Macro 6 ?

Post by UliBru » 01 Dec 2021 16:42

Hello David,

I'm a bit sad that you are not successful with ICPA.
But I can observe something that is not so common. The step response pictures show some low frequency oscillation after time 0.21
It looks a bit like the subwoofer are quite late and not well integrated.
Now there is of course some XO frequency between subs and main speakers. Is this frequeny now in the range 60 to 70 Hz where ICPA shows peaks?
So indeed it should make sense to test with subs powered off.

Target curve: the target you have sent me had a much bigger bump (at a lower frequency). I have assumed that you have tried to achieve more bass. IMO a slight bump around 60 Hz is good for a added warmth.
Stay well tuned
Uli

Moderator
Acourate system: JRiver/Roon -> AcourateConvolver -> miniDSP U-DIO8 -> TacT M/S2150 amps -> DIY horn speakers
User avatar
dsnyder
Posts: 21
Joined: 22 Oct 2019 13:52

Re: Macro 6 ?

Post by dsnyder » 23 Dec 2021 20:43

Yesterday was my last day "in the office" in 2021, so I will have some time during the holidays to revisit the apparent disconnect between what I'm hearing and measuring with ICPA. I'll start with the simple case with subwoofers powered off.

With ICPA enabled, what I'm hearing is out-of-phase behavior around 100 Hz. For example, runs on an upright acoustic bass in a good recording appear to come from a location in the soundstage that's behind the loudspeakers and a bit to the right without ICPA. However, with ICPA engaged, those same bass notes sound diffuse or even seem to come from behind my left ear. The sound is similar to wiring the speakers out of phase, but it only affects lower frequencies. For this comparison, I had built two filters with the only difference being toggling the "ICPA" setting on Macro 4.

The differences in sound are so obvious that I feel there must be a way to visualize what I'm hearing by analyzing the left and right impulse responses after correction. I think what I want is a way to plot phase differences between the two channels, but even at low frequencies and accounting for propagation delays, raw phase response is challenging to analyze.

What's frustrating is that I know the setup is good. With no correction, the soundstage is stable and remains behind the loudspeakers with excellent to astonishing depth on good acoustic recordings, as it should. Phasey electronic recordings project sounds all over the room, as expected. The same is true, although to a slightly reduced degree, with standard Acourate filters (no ICPA). But, at low frequencies, the soundstage is all over the place with ICPA engaged.

If I can find a way to measure and visualize the differences, perhaps we can find a solution. Since the setup sounds very good via Roon with convolution disabled, I don't think this is a setup issue, but I'm always game for learning and improving there. I have not ruled out the possibility of there being something off with my measurement protocol, but impulse responses from LSR via my Focusrite Scarlett 2i2 interface and Cross•Spectrum EMM6 mic seem to correlate well to REW with my Cross•Spectrum UMIK-2 mic when measuring pre-recorded sweeps played back via Roon and my usual DAC (Gustard X16).

I'll keep at it. It's important for me to be able to trust both what I'm hearing and measuring and for them to, more-or-less, agree!

If you have tips on how to meaningfully plot phase differences between channels, at least below 200 Hz, please let me know!
Acourate system: Roon -> Gustard X16 DAC -> Topping Pre90 -> Orchard Audio monoblocks -> Fritz Carrera 7 BE monitors + REL T7i subs
Ecwl
Posts: 48
Joined: 19 Feb 2021 01:05

Re: Macro 6 ?

Post by Ecwl » 23 Dec 2021 21:36

I just re-measured my system and decided to measure at a slightly louder volume this time around. The LSR3/Acourate logsweep recording volume bar never got above -10dB. My previous measurements, the loudest it got was maybe -20dB and most of the time was staying at -30dB.

What I found was that there was an excess phase shift at the lower volume measurement that I previously corrected but on the higher volume measurement, that single excess phase shift turns out to be part of a more complicated ringing excessive phase shift in a broader range of frequencies that I should not have corrected.

So the only thing I can think of for dsnyder is that maybe measuring at 75dB MAX is possibly too low a volume to measure? As a result, maybe you're missing out on a resonance that is not adequately captured at 75dB MAX whereas it might be captured at 80-83dB MAX?

Otherwise, maybe your ears are actually correct and your system simply doesn't need or won't benefit from ICPA corrections.
User avatar
dsnyder
Posts: 21
Joined: 22 Oct 2019 13:52

Re: Macro 6 ?

Post by dsnyder » 23 Dec 2021 22:33

Ah. That's a good thought. I'll record a number of sweeps at different levels to see what differences I can observe in Acourate and REW.

Looking at my previous recording of the main speakers + subs, here's what I see in Acourate for unwrapped phase from 20 to 100 Hz. This is just the original pulse recordings before any corrections or smoothing.

Image

I exported the pulses as a stereo WAV file and imported the impulse response into REW just to see how it is represented there, and I got this plot:

Image

I don't understand the units, and the differences in REW seem to be greatly exaggerated vs what I see in Acourate for the same data. The divergence from 52 Hz to 72 Hz is interesting. The difference in phase, as reported by REW, appears to be about 165º over that frequency range, which is almost completely out of phase! This is likely the region where I'm hearing bass notes go out of phase after ICPA correction and points to a possible issue with my pulse measurement affecting only the right channel. I sure hope it's that simple! :geek:

It's true what they say…garbage in, garbage out!

Edit: During that original session, I wanted to test the offline measurement feature in REW. So, I used Roon to play saved sweeps that included timing references. I then recorded the sweeps using my EMM6 microphone into a Zoom H4n Pro handheld recorder which happens to supply 48V phantom power for the mic. Finally, I imported the WAV file from the Zoom recording into REW, and I see the same results. In this case, the signal path for recording is completely different, but the plots are basically the same. One used my Focusrite Scarlett 2i2 to play and record the sweeps while the other used my Gustard X16 DAC and the ADC in the Zoom H4n Pro recorder.

Image

I also looked at sweeps that I recorded without the subs, and I see the same phase anomaly from 52 to 72 Hz in the right channel, although the magnitude of error is a little less at around 135º. This is quite a puzzle. If I get brave, I may try physically swapping my speakers to see if the problem moves, indicating something is not quite right with the speakers.

Edit #2: before I go moving speakers around (it was a pain to get them precisely located where they are) I may try swapping channels for bits further upstream, including the monoblock amps. Still, my ears tell me that this must a measurements issue rather than something wrong with the system or setup. :shock:

At least I feel like I'm getting closer to an answer. If ICPA added ~165º of correction that was not actually needed from 52 to 72 Hz, that would explain the phasey behavior I'm hearing in that range with ICPA engaged.

Edit #3: One last plot, also from the same session. This one was recorded in REW using my UMIK-2 with Roon playing the sweeps from my Gustard X16 DAC. Virtually identical to the other unwrapped phase plots even though I had to physically swap the mics.

Image

This pretty much eliminates the possibility of an issue with the ADC or microphone. Variables I'm left to experiment with include mic placement and everything else in the room.
Acourate system: Roon -> Gustard X16 DAC -> Topping Pre90 -> Orchard Audio monoblocks -> Fritz Carrera 7 BE monitors + REL T7i subs
Post Reply