Strange spike with Focusrite ISA One

[kmnsed]

Hi!
I'm new to this forum, but I've read some interesting discussions here, and your technical knowledge and patience made me think I could get an answer for my problem.

I have a Focusrite ISA One preamp with the ADC card installed, and I've noticed a strange spike around 48 kHz when recording at 96 kHz sample rate.

I am planning to record some pretty delicate sounds (nature recordings in a Mid/Side setup), so I'm trying to avoid filtering and heavy post-processing.

The conversion in this device is handled by an AK5394 AVS multi bit delta-sigma chip. The transmitter is an AK4101 AVQ.
In my mobile setup the clock source is the ISA, digital audio is then fed to a Focusrite Saffire Pro 24, which is slaved to the ISA'clock. Firewire connects the computer and the Saffire. I assume the Saffire shouldn't alter the digital signal by any means, no sample rate conversion, no level change or mixing should occur.
I can't use word clock in this setup since the Saffire doesn't have WC input, so I must depend on the embedded clock. It seems solid though, no drop-outs or glitces there. I've tried all the formats the Saffire accepts, S/PDIF coax (via Klotz 75 ohm), S/PDIF optical, and ADAT optical, but this issue is always there. I've also tried eliminating the Firewire power from the computer (assuming it's noisy) and run the Saffire on an SLA battery, but the result was the same. It's worth mentioning that the cheap Saffire's converters don't produce this spike, so I assume it's not the computer or the Firewire interface that's guilty here.

I wouldn't believe that anyone could hear this spike, but I don't know how it would translate when playing back on an average (or below) DAC.
Should I be concerned?

 

[AnalogSteph]

I would try running a loopback test with RMAA (in various sample rates - 48, 96, 192 kHz) and see what that's got to say in terms of frequency response. Should be possible even if you're on a Mac, just as long you have access to a machine that can run RMAA - you can save a test file, play it back and then analyze the recorded version in RMAA again (make sure reference level is not below -10 dBFS, normalize/amplify if needed).

Looks like some sort of digital filter ringing / aliasing to me in any case. No, this doesn't look concerning, but I would certainly like to know where it's coming from, as it may indicate that you might have some extraneous resampling going on for one reason or another.

 

[kmnsed]

Thanks for your quick response.
Loopback testing could be a good idea, though only the Saffire could be tested this way. The ISA is ADC only. Maybe the Saffire is responsible for some kind of weird sample rate conversion in this setup, which seems rather bad news for me. I thought it would work well as a transparent interface between S/PDIF and Firewire.
I use Windows with genuine Nuendo daw and direct ASIO protocol, so software issues should be out of question.
I think my next task is testing the ISA with anothet interface. I have an E-MU 1820m system with a sync card that is capable of receiving word clock. I will post my results as soon as I can.
Now I am curious.

 

[daftcombo]

This looks like noise shaping. It shouldn't be audible at all, seing both the frequency and the level it reaches.

 

[kmnsed]

To be honest, that's the answer I've been looking for. My only concern is how this would translate during playback on an average DAC. I don't have experience using sample rates above 48k, where I've never ran to issues like this.
I don't know if I should filter this out. Using a FIR filter at 20k with a Q of 4 seems to eliminate the spike, but I feel like it's upsetting the spatial image especially with difficult sounds, like rain. And, given the nice image a M/S with an omni mid can deliver, I'd certainly like to avoid this.
In my weak understanding, this spike could reflect back to lower audible frequencies during playback, and that's what I'm afraid of.

 

[Blumlein 88]

I believe the ISA is capable of stand alone operation. Does the spike show up if you run it stand alone and feed the output to the Saffire ADC?
By this I mean the ISA analog output to the Saffire analog input. If the spike goes away in this configuration, then the low level spike is probably an idle tone of sorts in the ISA ADC card. Do you have the spike on both microphone and line level inputs?

 

[Guermantes]

Is this "spike" a constant tone with a fixed level (i.e. is it periodic in nature) or does it change in response to program level? If it is fixed at -90 dB then I would probably treat as part of the quiescent noise floor and maybe clean it up with an ultrasonic filter, though your pickings will be slim with VST plugins. Maybe Fabfilter Pro-Q -- I think it allows linear phase filters up to 30 kHz. Otherwise there is a free Chebyshev filter for Audacity that goes up to 48 kHz: https://wiki.audacityteam.org/wiki/Nyquist_Effect_Plug-ins#Chebyshev_Type_I_Filter

 

[kmnsed]

I've tested it with my trusted old 1820m and the spike is not there in the analog outputs, so it's not comig from the preamp or the phantom supply. One last thing I had in mind was using the AES digital out with a separate word clock - the spike was present there also. It must be the product of the ISA ADC then. No matter what inputs I use, it's there, even with shorted analog ins.
I think I have to accept it.

 

[kmnsed]

Is this "spike" a constant tone with a fixed level (i.e. is it periodic in nature) or does it change in response to program level? If it is fixed at -90 dB then I would probably treat as part of the quiescent noise floor and maybe clean it up with an ultrasonic filter, though your pickings will be slim with VST plugins. Maybe Fabfilter Pro-Q -- I think it allows linear phase filters up to 30 kHz. Otherwise there is a free Chebyshev filter for Audacity that goes up to 48 kHz: https://wiki.audacityteam.org/wiki/Nyquist_Effect_Plug-ins#Chebyshev_Type_I_Filter

It seems to be constant in level. I see some minor fluctuations, but it's not coherent with the input signal, so I assume it's the weakness of the analyzer I use.
Thanks for the tips, I will try these filters. Maybe I could implement a simple VST with Synthedit.