Resonant pole > 20Khz in digitised audio - is my interface broken?

[rivermAn]

When I record an analogue signal onto computer using my digital interface, the recorded sample always has a single resonant frequency between 20-30Khz. It's about -30dB in amplitude. Recorded audio generally sounds good - no noticeable distortion or jitter. The phenomenon is observed on every channel. I am using a Lynx Aurora 8 with AES card, have tried different sample rates/clock source/drivers etc.

Is this normal behaviour? My understanding is that such artefacts are undesirable in any recorded source. I'm worried something is wrong.

At the very least, I'd expect the analogue front-end of the Aurora to have a low-pass filter < 20Khz. Which makes me conjecture that the pole is induced by some problem with the rest of the signal chain.

 

[JohnRoberts]

Could be an artifact of sampling. Is it causing a problem?

JR

 

[ruffrecords]

Does the sample actually have any content above 20KHz?

Cheers

Ian

 

[rivermAn]

@ruffrecords The analogue source contains no Freq > 20Khz. The artefact is observable in all sources I've tried to record: mic preamp, DI, keyboard out. Even plugging a bass guitar straight into the Aurora mogami.

@JohnRoberts
To me, there's a lack of clarity in the high-end extension > 10Khz. Mixing sibilance or airy frequencies is challenging - particularly bringing out detail in the vocal. Pushing the higher end generally makes everything sound slightly distorted / crunchy and smeared. I avoid recording guitar amps with anything but ribbon mics, too hard to mix with drums/vox otherwise.

I am wondering if the pole is causing intermodulation distortion in the top end.

On the other hand, I don't trust myself enough to say that something is definitely wrong - particularly as listening to higher freqs is fatiguing, competes w tinnitus, sibilance often needs to be tamed anyway, distortion could be introduced by monitoring setup, etc etc

The pole was pointed out to me by a very experienced engineer - he said he hasn't seen it before

Will post a screenshot of the frequency amplitude spectrum

 

[Khron]

Electrolytics in the switchmode power supply aging, and disturbing the feedback loop compensation giving rise to oscillation? How old is that Aurora?

 

[living sounds]

Would be interesting to see what exactly you mean.

I've still got an Aurora16 somewhere. Never liked the sound much, despite very low THD. Muddy and a little crunchy. It sounds quite different between 44.1k and 48k, and the phase response looks quite different (and strange, not linear even in the midrange) between them. It does not use linear phase filters.

 

[JohnRoberts]

old story,,, back in the distant 70s I modified my heathkit SMPTE IMD analyzer (60Hz:7kHz) to instead operate at 19kHz:20kHz. This tester revealed easy to see and measure IMD products down at 1kHz.
===
It should be easy enough these days to make your own 19kHz/20kHz test signal and look for dirt at 1kHz...

enjoy...

JR

 

[rivermAn]

Originally, I said the pole was at -30db. It's much lower here. One preamp I tried pushed it to -30 last year.

Here are two spectrograms which illustrate the phenomenon.

Spectrogram 1.
Test: plug bass -> aurora ch5. No DI/Pre/whatever. Going straight into the converter. Volume is turned all the way up on the instrument, but not playing.

Pole is at 24Khz.



Spectrogram 2.
Test: plug bass -> API 312 DI -> aurora ch5. Less raw dog type situation. Volume is turned all the way up on the instrument, but not playing.

Pole is at 29Khz. Lower amplitude than first spectrogram. But still compared to other freq, pretty obvious.



I can replicate these results using gibson SG electric guitar instead of my p bass. Same pole frequencies.

Has anyone seen anything like this in their own recordings? Like it looks weird as. I've never seen anyone talk about it before.

Aurora is at least 10 years old, same with AES card. Bedroom studio use for the last 9 years. Not sure when it was first purchased. I actually don't know if it's always had this problem.

@JohnRoberts thanks - will look into that

I emailed Lynx, will report back

 

[ccaudle]

the recorded sample always has a single resonant frequency between 20-30Khz

Have you checked with nothing connected to the converter input? First step would be to determine if it is internal to the converter or picked up externally.

At the very least, I'd expect the analogue front-end of the Aurora to have a low-pass filter < 20Khz.

No way a device specified as +/- 0.1dB to 20kHz could have a low pass filter anywhere near 20kHz. Maybe 50kHz or 80kHz, although even 50kHz filter would be a little tricky to hold to -0.1dB at 20kHz. If you run the interface at 48kHz or 44.1kHz then the digital filter will remove anything above 24kHz or 42kHz respectively (give or take, most converter filters will actually alias for a few kHz above the Nyquist frequency), so I assume since you can see something at 29kHz that you are running at 96kHz sample rate.

Where is your instrument and converter situated relative to your recording computer? You used to see interference like that showing up from computer monitors, although that is a lot less of a problem with modern LCD monitors compared to the old CRT based monitors.

 

[living sounds]

I see, you are recording at >48 khz and see a spike between 20 and 30 khz.

No, my Aurora never did that. It does produce progressively rising out-of-band noise at higher sample rates thanks to noiseshaping, but this looks very even. In fact, the Aurora has always been a perfectly clean converter in terms of spuria (as noticeable patterns vs. just evenly distributed noise).

Could this be down to interference from a strong RF source? What happens without anything connected? I think this might be coming from the bass picking up a signal, could be am LCD monitor as ccaudle suggested.

 

[FIX]

Most likely the analyzer in the computer. I have a FFT spectrum display that is on my iPad Pro and it has the same peak up there.

 

[thomasdf]

I have a pair of DBX 902 deesser and when inserted on my interfaces which use impedance balacing, they go completely crazy because of some ultra high frequencies. When I use them after a tape recorder, no problem. It drives me mad because I like them first in the signal chain... but I avoid using them because of that.

 

[abbey road d enfer]

I have a pair of DBX 902 deesser and when inserted on my interfaces which use impedance balacing, they go completely crazy

Not clear. I don't think your interface has impedance-balanced inputs. The 902 is connected before the interface, right?

 

[ruffrecords]

I think it would be useful to determine if the interference is generated internally or injected from outside. So I would suggest doing a test with the input shorted. Do you still see the artifact? if you do it must be generated internally. if it goes then it looks like it is getting into the input somehow.

Cheers

Ian

 

[Dan Elleson]

Hey mate. That's down in the noise, and it's not unusual to see harmonics at those levels high up in the spectrum that are due to interference from power supplies, internally or externally, for a connected device, nearby lighting, a connected device, even if that device isn't connected.

I used Lynx Auroras in a great commercial studio for years where I was also the tech, and they were great. I would often use analysers there and I never noticed any issue like that, that I can remember.

You probably don't have the tools you need to properly test and isolate this, but one thing you can do is get an analyser plugin that will show you more noise so you can see what's really down there. Pro Q isn't really the tool for the job. Voxengo Span is free and usable in this scenario. Get a good look at the whole noise floor, run the unit at 96kHz. Test it with nothing connected to the interface like the other dude said, and if you're using a laptop, try running off battery while you do it. With modern systems, you can get interference like that creeping in from so many different sources. Turn off or unplug as many nearby devices as you can do without, including lights. See if you can make that harmonic vanish.

BTW that's not a pole. A pole is a mathematical concept in filter design. What you have there is a harmonic. It may or may not be due to a resonance in some part of the system, but that kind of peak in the spectrum, with input that has no energy up there, it's not a resonance, unless it's an electrical one that is causing an oscillation, which it won't be.

 

[thomasdf]

Not clear. I don't think your interface has impedance-balanced inputs. The 902 is connected before the interface, right?

After ! Inserted in Protools so : interface out > DBX902 > vocal chain (1176, or vintage tube limiter, or anything else) > interface in back into Protools. The deesser was crazy, randomly compressing even when Protools is not playing. When unplugged, it stays quiet and normal (no comp happening). You told me it may be due to the DAC generating UHF content, back then. And it most certainly looks like it indeed !
I changed the DBX902 input filtering caps to cut above 22Khz, they act almost normal now, but I can clearly hear they eat the air out of my vocals now. Sometimes it's great... sometimes not
But I don't want to hijack the OP's thread

 

[abbey road d enfer]

After ! Inserted in Protools so : interface out > DBX902 > vocal chain (1176, or vintage tube limiter, or anything else) > interface in back into Protools. The deesser was crazy, randomly compressing even when Protools is not playing. When unplugged, it stays quiet and normal (no comp happening). You told me it may be due to the DAC generating UHF content, back then. And it most certainly looks like it indeed !
I changed the DBX902 input filtering caps to cut above 22Khz, they act almost normal now, but I can clearly here they eat the air out of my vocals now. Sometimes it's great... sometimes not
But I don't want to hijack the OP's thread

OK, so yes, the interface shifts the noise spectrum towards ultrasonics. I had the same problem with a Native Instruments soundcard I used for measurements, but would do crazies with compressors. I had to add a very sharp LPF (5-stage 10th order) to make it half usable.

 

[zamproject]

I changed the DBX902 input filtering caps to cut above 22Khz, they act almost normal now, but I can clearly here they eat the air out of my vocals now. Sometimes it's great... sometimes not

-3 @ 22k is more than -1 in the 10k/15k region, that's a lot of change
Maybe yo can shift a little higher ? what is the freq of parasitic content ?

 

[thomasdf]

no clue ! I cannot hear or even monitor it with my spectrum analyzers... but the sidechain clearly reacts to it.
Yes, I should try fine tuning the cutting point...

 

[zamproject]

which still strange as a deesser probably have a band pass in the sidechain ?!?