96kHz is my interface defective?

[conklin]

So I've had a FirePod for a number of years now and it's been rock solid with Sonar/Cakewalk and it survived a few Windows versions as well (Vista, 8 and now 10).   All at 24 bit 44.1 or 48kHz.

But I did the odd song or two at 24/96  and wasn't all that impressed, and have been working on a project that would be 24/96 and something really isn't right but couldn't put my finger on it.

I downloaded RightMark Audio Analyzer and got this for a frequency response at 24/96 :

 

 

Yikes!  Looks like it's rolling off at around 6kHz  and what is all that garbage from 25.5ish on up? 

I went ahead and loaded a 1kHz test tone in Cakewalk and looked at the graph in SPAN and on playback all looks like it I would expect:

HOWEVER, everything I've recorded at 24/96 has this extra junk in the 25k area going on up to the limit:

 

Also happens with 88.2kHz as well.   

The rest of the test results from RMAA were ok (what I expected) for 44.1/48 but were really bad at 96.  

So is my FirePod defective?   It's on all 8 inputs, mic and line  Even patched into the 2 returns on the back of the unit - same results.  I assume that it's something being added during the A-D stage as the test files I loaded up playback without the extra "stuff".

Tried with ASIO and WDM drivers and there's no difference.  I loaded up a song from 2013  and sure enough, same junk in the upper end - and no, I don't really like the overall sound of that recording.  Couldn't quite put my finger on it then and never gave it a thought to look at the response. 

I'm on the latest updated Win10, but I don't really think it's an os problem...

 

 

[Jon L. Jacobi]

All you capture with 96kHz is garbage that no one can hear and has to be filtered out anyway.  That stuff you see may be dithering to frequencies that no one can hear or noise in your interface or from somewhere else. If you want reduced latency, you can get it with 96 kHz if you can run a small enough buffer but it puts a lot more strain on the CPU. 

I've heard people argue that those frequencies can affect the lower frequencies and therefore should be captured, ignoring the fact that they already affected the lower frequencies before capture. 

This guy knows it better than I do.

https://xiph.org/video/ 

[Robert Bone]

I just never ever record beyond 48k.  Couldn't that stuff get filtered off, if you were wanting to record at 96k?

Bob Bone

[conklin]

That second video at xiph.org - wow that takes me back to audio lab from many years ago?  Explained the same way too!

Probably could filter it out/off but I may just go ahead and redo everything at a lower rate.  I guess that all that extra ultra sonic "wonderful stuff"(sarcasm)  is causing problems in the audio range thus the really bad results in the RMAA and the odd roll off.  The THD test came in at 5% - yes five percent not .00xx%

I would have thought that there would have been a filter already applied before any audio got digitized - or maybe there is and the FirePod just isn't all that great?

Anyway, thanks for the replies guys.  Looks like I'll be sticking with 44.1

 

 

[azslow3]

5 hours ago, conklin said:

So I've had a FirePod for a number of years now and it's been rock solid with Sonar/Cakewalk and it survived a few Windows versions as well (Vista, 8 and now 10).   All at 24 bit 44.1 or 48kHz.

But I did the odd song or two at 24/96  and wasn't all that impressed, and have been working on a project that would be 24/96 and something really isn't right but couldn't put my finger on it.

I downloaded RightMark Audio Analyzer and got this for a frequency response at 24/96 :

 

 

This graph shows that your interface, at the moment of the test, physically works in 48kHz mode. Check its own control panel (if it has one).

At least check that RMMA is in ASIO. In MME it WILL NOT switch the interface, so the interface continue to work in the last mode it was asked to work.

From where the garbage comes I have no idea, on my interfaces the part up to 24kHz has the same general shape,  but the upper part is zero (I mean when hardware is in 48kHZ, RMMA in MME with 96kHz).

Can it be some "windows audio improvement" or some other software "effect"? (I do not have FirePod, but when I had SoundBlaster it has tried to "improve" my sound internally). I ask because even up to 20kHz part is horrible. Sure, I do not expect it is as flat as for top current interfaces. But even my M-audio Fireware (without pre-amps)  looks way better (it falls after 20kHz in 48kHz mode).

If that garbage has found its way into your recording, there is nothing you can improve there. But at least it should be possible to switch your interface into real 96kHz for future recordings.

 

[conklin]

21 minutes ago, azslow3 said:

This graph shows that your interface, at the moment of the test, physically works in 48kHz mode. Check its own control panel (if it has one).

At least check that RMMA is in ASIO. In MME it WILL NOT switch the interface, so the interface continue to work in the last mode it was asked to work.

From where the garbage comes I have no idea, on my interfaces the part up to 24kHz has the same general shape,  but the upper part is zero (I mean when hardware is in 48kHZ, RMMA in MME with 96kHz).

Can it be some "windows audio improvement" or some other software "effect"? (I do not have FirePod, but when I had SoundBlaster it has tried to "improve" my sound internally). I ask because even up to 20kHz part is horrible. Sure, I do not expect it is as flat as for top current interfaces. But even my M-audio Fireware (without pre-amps)  looks way better (it falls after 20kHz in 48kHz mode).

If that garbage has found its way into your recording, there is nothing you can improve there. But at least it should be possible to switch your interface into real 96kHz for future recordings.

 

Good point about the drivers.  I ran the test again and RMAA is indeed in ASIO.   The PreSonus audio panel is set at 96 with internal clock.   Same results.  I even went into all the windows settings I could think of, and there's nothing else turned on.  Even the GameBar I have disabled.   Yeah sometimes the updates flick stuff back on but nope.  I'm on a wired connection and so there's no wireless anything in the box.    The FirePod plugs into the firewire card that PreSonus recommended.  Again everything was all great at 44.1 but I just had to kick it up to 96 LOL!

I even ran a test this morning right after turning it on thinking that it may somehow be heat related but that didn't change the results either.

Here's the response of 24/44.1:

That was what I was expecting to see - flat enough - not gonna loose any sleep over .37dB?

but man that graph for the 96k - whoah!  That explains some of the EQ choices ...

I did try using the PC EQ to filter that out and it did lower the garbage a lot.   Might be able to salvage some of the stuff I did record.

[azslow3]

I do not know how Presonus drivers work, but your first graph is definitively not from 96kHz sampling hardware. Can it be it shows 96kHz, while in fact doing that is software?

Not sure it is possible with Presonus (possible with my M-Audio), does it work with ASIO and WDM at the same time? I mean check that Windows is not configured to use it for something, in 44.1/48kHz mode (the only way I know to convince Windows it to have another interface for it... when there is only one, Windows try to grab it). My interfaces report correctly - when the interface is used for Windows / other app, its ASIO frequency can not be changed. But I do not have Presonus, they f.e. could "trick" by letting frequency change in software while still locking hardware.

Other possible way is checking Windows settings. In the Control Panel (old one) / Hardware and Sound / Sound. In playback and recording, right-click on each device, Properties / Advanced. Check everywhere is 96kHz, at least for all IO of FirePod.

[Gswitz]

I have seen this before with my Audient ASP 880 at double rates. 

The noise above 20kHz was proportional to the signal below 20kHz. It was really interesting. If you play guitar while you record you could see the huge noise above 20kHz.

Audient had me send it back to them... The first time they failed to fix it, but the second time was the charm. It no longer has that problem.

It also doesn't have a low-pass filter removing everything above 20kHz. If I want to record a dog whistle, it will work just fine. My RME never had the problem.

When you are recording at 44.1 then you want to filter out most sound above 20kHz so that you don't get weird artifacts. But if you are recording at 88.2 or 96, you want to filter out sounds kinda below 40kHz (or remove a little more than half the sampling rate in kHz).

It looks like your interface is applying the low pass filter at 20kHz when it should be applied at 40kHz. Then there is this weird bloom of strangeness above 20kHz sort of mirroring the signal in a strange distorted pattern.

You can record it and pitch shift it down to hear it if you want using Melodyne or audio stretching.

So, if you don't want to get it fixed or upgrade to an interface that doesn't do it, you might want to apply your own filter to your recording to remove the noise above 20kHz so artifacts don't sneak into your music. I'm sure it will bug you immensely because your device supports recording up to 96 but isn't moving the filter appropriately. But, just because you know it is doing it doesn't mean you can't keep using your interface to make great recordings. You will still get increased fidelity/head room by using 96. You just have to remove the detritus above 20 yourself when, if your interface wasn't misapplying the filter, you could record sounds above 20 and remove them only when required for export to a 44.1 file for example.

I'm guessing lots of interfaces have the same issue but the users don't know it because their spectral analyzers don't go up above 20kHz so they can't see the noise.

Whenever you work at double or quad rates, you should use an analyzer that shows the whole recorded spectrum.

Btw, for the Audient, the fix required a firmware update not a driver update, and the Audient asp880 has no method for consumers to upgrade the firmware.

You may wonder how i feel about the Audient after this trouble. I like it just fine. It is a decent interface and i don't strive to avoid it if i can.  In other words, i may use inputs on the Audient when i still have open slots on the rme ucx.

Edited November 16, 2019 by Gswitz

[Starship Krupa]

Really, those higher sample rates that they put on interfaces have no bearing on any kind of real-life situation. The storage media companies pay the interface manufacturers to put them there so that people will use up more drive space.   An analysis showing that it's freaking 2dB down by the time it hits 20KHz, then looks like Dr. Frankenstein's lightning storm above 25K isn't surprising when you try to record at sample rates higher than 48K.

Just kidding!

I own a pair of Firepods, actually Firepods with their firmware upgraded to the FP10 specification. After you posted this I decided to download the RightMark Analyzer and try it on my own system.

No problem, looks great, at 96/24 my system is only down about 0.1 dB at 30KHz and about 0.6 dB at 20Hz, with none of that craziness above 20K.