Print

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

[dmitch77]

I've found an anomaly in the EQ section of the DSP studio. I have a sound file, linked below, which makes the EQ section behave in a manner I consider to be incorrect in some configurations.

The source audio is just "clicks" (created by wood blocks) along with reverb (room ambience). Here is a link to this file:

https://www.dropbox.com/s/jop2s9sh6f1es9o/EQ%20Test.flac?dl=0

It's 39 seconds of this audio, in FLAC format, 96 KHz 24 bit.

When EQ is enabled in the DSP studio and the 60 Hz band is boosted at 4.0 dB or more, an anomaly is clearly audible: there is a 60 Hz hum in the reverb tails of most of the "clicks". Turn off the EQ, or set the gain for the 60 Hz band to zero, and you don't hear that anomaly. I'm pretty sure that there is no 60 Hz hum in the audio file; listen to it good and loud with the EQ off. The reverb tails are just white noise.

This anomaly occurs on 2 out of three systems on which MC is running. (MC 25.0.81 on macOS 10.14.5 in all three cases.) The two systems on which the anomaly occurs have high quality DACs - 96 KHz 24 bit - one connected via USB, one via HDMI. The anomaly happens whether MC is configured to send the output directly to the outboard DAC, or to the default core audio output which is then directed to the DAC. The system on which the anomaly does not occur has a 44.1 KHz / 16 bit external DAC, and the anomaly does not occur regardless of how the audio is sent to the DAC (directly to the DAC by MC, or via default core audio device).

System 1, anomaly occurs
Late 2013 iMac, macOS 10.14.5
DAC is Focusrite Scarlett 18i20, 96 Khz 24 bit, connected via USB
MC 25.0.81
  - audio out directly to Focusrite : error occurs
  - audio out to default Core Audio, default is set to Focusrite: error occurs
  - audio out to default Core Audio, default is set to internal speaker: error does NOT occur

System 2, anomaly occurs
2014 Mac Mini, macOS 10.14.5
DAC is Marantz SR7009 AVR, 96 KHz 24 bit, connected via HDMI
MC 25.0.81
  - audio out directly to Marantz : error occurs
  - audio out to default Core Audio, default is set to Marantz: error occurs

System 3, anomaly does not occur
2017 27" iMac, macOS 10.14.5
DAC is low-cost 44.1 KHz 16 bit connected via USB
MC 25.0.81
Anomaly never occurs

[dmitch77]

Giving this a bump since it got no replies. It was actually intended as a bug report. JRiver folks - is there any chance that someone might get a chance to look at this one (and maybe fix it)? Let me know if there is any more info I can provide. It's 100% repeatable. Thanks.

[mattkhan]

 if the issues occurs only on certain downstream signal chains then how can it be a jriver bug?

It is a bit difficult to follow the description though, have you considered capturing the output stream so you can show the actual output Vs the expected clean output? This would show exactly what the distortion was

[dmitch77]

if the issues occurs only on certain downstream signal chains then how can it be a jriver bug?

It is a bit difficult to follow the description though, have you considered capturing the output stream so you can show the actual output Vs the expected clean output? This would show exactly what the distortion was

The problem happens when MC is outputting 24 bit 96 KHz audio, and it happens on two different platforms in this situation. It doesn't happen when MC is outputting 16 bit 44.1 KHz. The problem only happens using DSP Studio's EQ. (And not, for example, when using DSP Studio's Parametric EQ.) You don't see how this can be a JRiver bug?

I'm not sure I have the tools to capture the audio output to "show the actual output". What do you suggest for this? And what do you suggest for showing the "expected clean output"?

It's really easy to reproduce. Just play the audio I provided thru a hi-res DAC, boost the 60 Hz EQ by 4Hz, and listen. The problem is unmistakable. I'm pretty sure that if you try it, you'll hear it right away. Let me know if you need more info to repro.

Thanks!

[JimH]

Can you take a look at this thread?

https://yabb.jriver.com/interact/index.php/topic,120675.0.html

[dmitch77]

Can you take a look at this thread?

https://yabb.jriver.com/interact/index.php/topic,120675.0.html

OK I looked. I don't see anything there related to anything I've reported. I don't have any dropouts. One of the machines I hear this problem on doesn't use Time Machine. The thread doesn't say anything about a problem only happening at 24 bit/96 KHz. The thread doesn't mention DSP Studio EQ. The thread doesn't mention 60 Hz hum.

What was it you were drawing my attention to?

[JimH]

Time Machine.

[dmitch77]

Time Machine.

Yes...? One of my machines that exhibit this problem does not use Time Machine.

[mattkhan]

by EQ you mean this?

[dmitch77]

by EQ you mean this?

Yes.

[blgentry]

I hear this too.

As the OP said, you have to play it kind of loud to hear it.  It also helps if you increase 60 Hz to more like +8dB.  I can also hear the effect at +4, it's just easier at +8.

Strangely, you can really hear it if you pause and unpause the track several times.  It's probably a matter of catching the leading edge of the waveform or something.  Several times when I unpaused, I got a very distinct 60Hz "hum".

So, I can hear it.  Does that mean it's a bug?  Frankly I have no idea.  Maybe an analog eq, pushed to +8dB at 60Hz would produce the exact same sound.  I don't have one to test with.

I tested with VLC on the same machine and didn't hear this.  But that's not saying much.  VLC's equalizer may or may not be doing exactly what it shows on screen.

As another test I set MC DSP Output format to convert 96kHz to 44.1kHz.  I still hear the exact same thing.

My gut feeling is that I'm simply hearing more of the 60Hz content that is present in that signal already when the EQ is engaged.  I'm not sure.

An analog EQ or another highly regarded digital player with a 60Hz boost would be good things to double check this against.

dmitch77:  It seems like you have several complete systems with different playback back ends.  I suspect that your system that does not show the error has different speakers with less bass response.  If it's not too hard, you might move the Mac from that system temporarily to plug in to one of the other systems where it's clearly audible.  Then see if you can hear the same thing on that one.

Brian.

[dmitch77]

I hear this too.

An analog EQ or another highly regarded digital player with a 60Hz boost would be good things to double check this against.

It's telling, I think, that using MC's Parametric Equalizer - also in DSP studio - to do more-or-less the same EQ tweak does not cause this hum. E.g., Low-shelf, 100 Hz, 6 dB gain. It raises the bass, as much as you want, and the hum never happens. That seems to indicate (to me, at least) that it's not in the source audio, it's not in the downstream hardware. It's in the EQ.

Moving Macs around at my location is not practical. The speakers on the Mac where this does *not* occur are pretty good and I can turn it up pretty loud. Thanks for looking at this.

[wer]

There would be an easy way to prove this, if the method works.

Grab a sine wave file, something like 1kHz.  It will have no 60Hz content.

Run it through both equalizer types, and look at the waveform output of each, and see if a hum is being manufactured where none could have existed.

If this doesn't produce a hum, it proves nothing, because it might only happen when there's signal in that frequency range to process...

[dmitch77]

There would be an easy way to prove this, if the method works.

Grab a sine wave file, something like 1kHz.  It will have no 60Hz content.

Run it through both equalizer types, and look at the waveform output of each, and see if a hum is being manufactured where none could have existed.

If this doesn't produce a hum, it proves nothing, because it might only happen when there's signal in that frequency range to process...

Good idea. I tried with 1k sine wave. No anomaly. Tried with pink noise, tried with silent audio. No anomaly in either case. There is something in that audio file I supplied that triggers the bug.

[wer]

So when Brian heard this, it was out of the same file?  Is that the one and only file that produces the anomaly?

I would then try this: convert the flac to wav and try again.

If there's no hum, it's the file itself and not the actual waveform data.

If it DOES hum, try applying an equalizer to the wav in Audacity, or some other high quality editor.

Remember, 60Hz humming could be a ground loop hum present in the actual recording.

But given the fact the peq doesn't reveal it, I think not.  It would be interesting if some strange combination of data triggered a mathematical error in the geq code.  That would only be indicated if there's hum in MC on wav but no hum from audacity.  In that case, Matt would have to look...

[dmitch77]

I would then try this: convert the flac to wav and try again.

If there's no hum, it's the file itself and not the actual waveform data.

If it DOES hum, try applying an equalizer to the wav in Audacity, or some other high quality editor.

Same problem with WAV. It's not a problem with the FLAC file.

I don't see what testing with Audacity would show that testing with the MC parametric EQ doesn't. I don't have Audacity here, it's too late tonight to install it and learn how to use it. Maybe someone else could give it a shot?

[wer]

Someone tested with the EQ in VLC, the quality/reliability of which is suspect.  Audacity is a good quality editor, and should behave reliably.  The idea is to make a more persuasive argument.

[mattkhan]

I tried it on windows, it seems like an obvious bug in the equaliser to me. It's not limited to 96kHz though you need need to bump the equaliser a bit higher at 48kHz to make it audible. It's definitely not due to the content of the track either (which you will be able to see via a spectral analysis of the track) as the fact that entering the same shape filter via PEQ does not produce the same artifact.

The artifact itself is also clearly visible in the analyser view btw.

It would probably be nice if they just got rid of the equaliser screen and converted it into a pretty front end for PEQ

@OP the equaliser is an octave band graphic equaliser (I believe) so using a PEQ with Q=1.414 at the specified frequency should have the same effect (but without the distortion)

[blgentry]

Thanks mattkhan for saving me the effort of trying to replicate the 60 Hz EQ in the parametric.     I suspected that it would produce the same results, but you say it does *not*.  That definitely seems like a subtle bug.

Brian.

[dmitch77]

@OP the equaliser is an octave band graphic equaliser (I believe) so using a PEQ with Q=1.414 at the specified frequency should have the same effect (but without the distortion)

Good to know. Thanks for the info and for looking into this.

[dmitch77]

A few more observations.

1. Adding max boost at 60 Hz in Audacity to the same source does not result in the anomaly.
2. Adding max bass boost in an analog EQ (post DAC) does not result in the anomaly.
3. JRiver's EQ only exhibits the anomaly when the volume of JRiver itself is fairly high (at least 40% or -30dB).

Thanks again for the folks who looked into this.

[Matt]

I'm trying to reproduce this, but not sure I'm having much luck.

I'm playing the 96k file to my Mytek DAC.

When I enable the Equalizer DSP and boost 60 Hz, it still seems pretty good on my HD 650 headphones.

Maybe try reducing the level with the "Pre Amp" slider (or internal volume) so you're not prone to clipping?

Thanks.

[dmitch77]

Maybe try reducing the level with the "Pre Amp" slider (or internal volume) so you're not prone to clipping?

I'm 100% sure that the problem I'm hearing is not clipping. I know what digital clipping sounds like, and this ain't that. Plus, the source signal is so quiet when this happens that clipping can't possibly be an issue.

[mattkhan]

I'm trying to reproduce this, but not sure I'm having much luck.

I'm playing the 96k file to my Mytek DAC.

When I enable the Equalizer DSP and boost 60 Hz, it still seems pretty good on my HD 650 headphones.

Maybe try reducing the level with the "Pre Amp" slider (or internal volume) so you're not prone to clipping?

Thanks.

I did the following

- turn off all DSP in MC inc output format
- turn on the equaliser
- set to flat
- move 60Hz to +4dB
- reduce internal volume to -32dB
- play the supplied test track
- check analyser, no distortion around 60Hz
- change internal volume to -31dB
- large 60Hz spike appears

this is nothing to do with the output device, it is purely inside MC so you don't even need to listen to anything (just monitor the content coming out of MC)
spectral analysis of the track shows it is at about -100dB in the 60Hz range, i.e. no content
there is no clipping in the waveform of the source track with an equivalent +4dB PEQ applied

if you still can't reproduce using the above approach then I have no idea what is going on.

same behaviour on linux (and when outputting to a null device)

[blgentry]

That's really bizarre.  With internal volume at 0dB I can just barely see a blip at 60Hz on one or two of the impulses.  But sure enough, when you get in the range of -20 to -31 it shows up in spades.  It's much more pronounced with internal volume reduced.  How strange.

When I did my listening tests a few days ago, I had internal volume at 0dB and had Volume Leveling and Adaptive Volume both engaged.  So I know MC was applying some amount of volume reduction in that case.  I guess that's how I heard it.

Very strange stuff!

Brian.

[wer]

Just wondering if this bug is getting any attention, since it's been reproduced on Mac and Windows?

[JimH]

Matt commented above.

[wer]

You must have thought I couldn't see that post.  Yes, I saw that he said he couldn't reproduce it.  Since then more people did reproduce it, on Mac and Win. And so did I.  Matt wondered about clipping. Mattkhan proved it wasn't, and gave a step by step.  Since then no word from Matt.

So since everyone seems to be reproducing it except Matt and some time had passed since he was provided with new info, I wondered if it's going to be looked at some more.  I'll stop wondering.

[dmitch77]

So since everyone seems to be reproducing it except Matt and some time had passed since he was provided with new info, I wondered if it's going to be looked at some more.  I'll stop wondering.

I'm still wondering.... :-) The status is not clear to me at this point.

[dmitch77]

Seriously, what is the status of this one from JRiver's perspective? Is it an ongoing investigation?

Thanks.

[Awesome Donkey]

It's probably the same as Matt said above, they can't reproduce it, more than likely.

[Matt]

It's probably the same as Matt said above, they can't reproduce it, more than likely.

I haven't been able to reproduce it.  I asked Hendrik to take a look too.

[blgentry]

Please try the steps outlined in Matt Khan's post above:

https://yabb.jriver.com/interact/index.php/topic,121775.msg843181.html#msg843181

You should be able to visually see the anomaly in the DSP analyzer using these steps.

Brian.

[Matt]

I can't see any energy, except in the bands where the frequency actually is, at any internal volume levels.

[wer]

Brian or Mattkhan, could you make a video showing it?

[mattkhan]

https://youtu.be/XUFE7k2Dn_A

[Matt]

I think we're going to migrate our equalizer to use our high-end parametric equalizer code.  We'll be busy for a few days, but I think the results will be awesome.

[mojave]

I was able to reproduce with Windows 10 and see the same as in the video.

[wer]

Matt, since this was such a strange phenomenon, could you or Hendrik explain a bit about the nature of what you found?

[Matt]

Matt, since this was such a strange phenomenon, could you or Hendrik explain a bit about the nature of what you found?

I didn't reproduce the issue, but I did see a little bump at 60 Hz if I maxed the slider.

However, switching to our PEQ code should be a really clear win.

[Matt]

You can test on Mac now:
https://yabb.jriver.com/interact/index.php/topic,122034.msg844296.html#new

[dmitch77]

You can test on Mac now:
https://yabb.jriver.com/interact/index.php/topic,122034.msg844296.html#new

Verified here; sounds (and looks, in the Analyzer) great.

Thanks!

Doug

[Matt]

Verified here; sounds (and looks, in the Analyzer) great.

Thanks!

Doug

Very happy.

Thanks.

[wer]

Matt, does the new parametric code replicate the Q from the old equalizer exactly?  (Which was what, by the way?)

I'm wondering if we should expect to actually hear any difference with a particular set of EQ sliders, or is it just being rendered with higher precision?

Thanks for this improvement!

[Matt]

The new sliders have a Q of 1.0 and slope of 12.0.

[wer]

But what was the Q/slope for the OLD sliders?  The same?

[mattkhan]

I don't follow, what sort of filter are you using that has a slope?

Q=1 is wider than an octave btw, it is about 1.4 octaves

[Matt]

The slope is not used for the equalizer filters.  I just reported it because it was staring at me!

The old equalizer is a mystery.  It was written around the time I was starting at JRiver.  I don't think we know any of the parameters about it.

[wer]

Fair enough.  I'm guessing we might expect to hear some noticeable differences in how the new EQ affects the sound then, compared to the curves we were already using.