• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Interface Mystery

Joakim

Member
Joined
May 6, 2020
Messages
28
Likes
16
Inspired by ASR, I've started looking into the performance of various bits of equipment around the house and have come across a bit of a mystery. A certain 'audio recorder' (which shall go nameless for now in case this is all a misunderstanding on my part) is said by the manufacturer to be a 44.1K/16 bit device. As far as I can tell, though, it isn't behaving like one. I've done various recordings and feel the answer is in there if I knew how to interpret the data. If anyone knows how to do that and is up for it, I'd appreciate your take!

What you see in the attached images is a sine sweep captured by the mystery device as well as recordings of a high-pitched sine tone around 12 kHz using the same box and an RME ADI-2 FS interface. I haven't included a sine sweep recorded by the RME interface as it does exactly what you'd expect it to: continue up to Nyquist without drama or aliasing and dropping off once it hits the lowpass filter.

Oh, and I've also been in touch with the manufacturer who swears the sample rate is 44.1K and that there is an anti-aliasing filter in place around 20 kHz.

The attached images show:

1. A sine sweep recorded by the Mystery Device
2. A 12K sine tone recorded through an RME ADI-2 FS set to 44.1K (Interface A)
3. A 12K sine tone recorded by the Mystery Device, supposedly also at 44.1K (Interface B)
 

Attachments

  • Interface-B_441000_SineSweep.jpg
    Interface-B_441000_SineSweep.jpg
    702.9 KB · Views: 197
  • Interface-A_44100_Audition.png
    Interface-A_44100_Audition.png
    545.7 KB · Views: 200
  • Interface-B_44100_Audition.png
    Interface-B_44100_Audition.png
    588.9 KB · Views: 189
First off are you using Windows and does the mystery device use Asio?
 
Hi Blumlein, thanks for the reply. According to the manufacturer, "[e]verything is programmed in assembly language", so I don't think there is a normal operating system. I myself am on a Mac and I've gotten the recorded files off the device using a USB stick.
 
That 3rd image of the mystery interface looks like what you get with 12khz and full imaging from no filter back down to the 10,050 hz frequency.

Create a white noise signal at -4 db at 44,100 hz. Play it on the mystery interface and record at 96 or 192 khz. You'll be able to see what the filter is doing past 22,050 hz. It should look something like this.

1588793848849.png


If you don't have a way to do that, play a tone of 17,050 hz at - 6db thru your mystery device. and record it. If you get a clear full level tone at 5 khz along with 17,050 hz then you'll know there is no filter on the mystery device. Some of your postings so far certainly look like that is the case.

BTW, just noticed you are a new member. So welcome to ASR.
 
Aha, I think you're right! My labeling of the frequencies isn't perfect, but what you describe in your second example is exactly what you see in the series of screenshots of the sine sweep. If I play a sine tone at 17,050 Hz, I get a more or less identical (and equally loud) tone at 44,100 / 2 - 17,050 = 5,000 Hz. This was confusing to me as I thought you wouldn't get aliasing until you go above the Nyquist frequrency? As the sequence shows, it starts right away at 50 Hz and continues until the signal drops off at 22,050 Hz. I also find it interesting that it does drop off at that point in an identical way to my RME interface (see screenshots). It actually looks like the device has a sample rate of 44.1K and a lowpass filter somewhere around 20K, were it not for that reflected copy of itself.

I'm going to make a white noise recording as you suggest and see what it does. Based on my tests with sine tones, playback is perfect, with no weird effects that I can detect. It's only on the way in that something unusual seems to happen.
 

Attachments

  • RME-Near-Nyquist.png
    RME-Near-Nyquist.png
    379 KB · Views: 134
  • MysteryDevice-Near-Nyquist.png
    MysteryDevice-Near-Nyquist.png
    420 KB · Views: 130
And thanks for the kind welcome. I've been a reader for a while. It's a nice change from some of the other audio forums out there. :)
 
What is the source you are using to record? If I've not mixed it up, you are recording with the RME and mystery device. The mystery device is showing these other frequencies that shouldn't be there during recording.

What I'm trying to figure out is whether your mystery device is suffering from aliasing upon recording or some other device has lots of imaging upon playback.
 
I've mostly been sending signals from the outputs of my RME interface and recording them using either the mystery device or back in through the inputs of the interface. I have tried other sources, though, and it doesn't seem to make a difference. Anything that passes through the inputs of the mystery device gets affected the same way — lots of aliasing regardless of frequency — but anything that's played back by it seems to come out ok. I have attached screenshots of frequency analyses of the white noise files you suggested, recorded and played back by the mystery device (Image 1) and transferred to it via USB and then played back by it (Image 2). If I'm reading this correctly, the upper frequencies have been affected by recording in a way that's consistent with what I've seen from sine tones.
 

Attachments

  • Recorded_Noise.png
    Recorded_Noise.png
    319 KB · Views: 119
  • Original_Noise.png
    Original_Noise.png
    275.2 KB · Views: 125
I've mostly been sending signals from the outputs of my RME interface and recording them using either the mystery device or back in through the inputs of the interface. I have tried other sources, though, and it doesn't seem to make a difference. Anything that passes through the inputs of the mystery device gets affected the same way — lots of aliasing regardless of frequency — but anything that's played back by it seems to come out ok. I have attached screenshots of frequency analyses of the white noise files you suggested, recorded and played back by the mystery device (Image 1) and transferred to it via USB and then played back by it (Image 2). If I'm reading this correctly, the upper frequencies have been affected by recording in a way that's consistent with what I've seen from sine tones.
That really looks okay either way. You'd like for anything 22 khz or above to be down 60 db or more. Looks like the device is using a half band filter which is common and usually with record and playback is alright. The filter might be a bit less good than wanted, but not dramatically so. It probably is consistent with what you are seeing. Lots of imagining back down below 1 khz, because the filter could be steeper ideally.
 
Yeah, I'm not too worried about the playback side at all and it's hard to hear a difference between the noise files. Having an equally loud tone or (more likely) harmonic being recorded at 10,050 Hz whenever there's one at 12K is more of a worry.

I'm starting to think this is so unusual that's it's hard to find a good explanation of it outside of technical papers. I've found references to lower and upper images, and I think that has something to do with it, but mostly it's mentioned in passing as something that's no big deal to get rid of. If it's not too much to ask, can you think of a good way of explaining why this is more of a problem than just having frequency content above half the sampling rate being reflected back? I think it might be possible for the manufacturer to fix it if I could just explain it better, and I don't want to make a big deal of it in public as it's a one-person operation that makes really unique and useful products. Sorry, this really is a lot to ask, coming from a first-time poster!
 
Okay, it isn't too hard to test this. @Joakim

At 192 khz make a white noise file, and also make a 20 hz to 80 khz sweep. Play that file with the RME and record it with the mystery device at 44.1 khz sampling. That should make it clear what the aliasing filter is doing in the mystery device.

Next make a 44.1 khz file of white noise, and also make one with a 20 hz to 20 khz sweep. Play that with the mystery device, and record it with the RME with the RME running at 192 khz sampling. That should tell you what the reconstruction or anti-imaging filter in the mystery device is doing.

Also check which filters you were using with the RME. It has Sharp and Slow filters to choose from. It could have been the source of these spurious signals if set to Slow filters.

In any case the above tests should define what the filtering of the mystery device is both for recording and playback.
 
Thank you, this is all so useful. The attached images show the following:

1. White noise at 44.1 kHz played by the Mystery Device and recorded by the RME
2. Sine sweep at 44.1 kHz played by the Mystery Device and recorded by the RME
3. White noise at 192 kHz played by the RME and recorded by the Mystery Device
4. Sine sweep at 192 kHz played by the RME and recorded by the Mystery Device
5. High-contrast version of 4 to show what going on above 22.05 kHz

The 44.1 kHz files were created in Adobe Audition and transferred to the device via USB. The 192 kHz files were played in Audition and out through the RME outputs. I should also mention that I'm using an ADI-2 FS (not the Pro version) and with that one you can't choose filters.
 

Attachments

  • WhiteNoise_44100_Recorded.jpg
    WhiteNoise_44100_Recorded.jpg
    790.3 KB · Views: 122
  • SineSweep_44100_Recorded.jpg
    SineSweep_44100_Recorded.jpg
    402.8 KB · Views: 131
  • WhiteNoise_192kHz_Recorded.jpg
    WhiteNoise_192kHz_Recorded.jpg
    767 KB · Views: 128
  • SineSweep_192kHz_Recorded.jpg
    SineSweep_192kHz_Recorded.jpg
    415.5 KB · Views: 111
  • SineSweep_192kHz_HighContrast.jpg
    SineSweep_192kHz_HighContrast.jpg
    309.9 KB · Views: 113
Last edited:
Is the mysterious device a standalone/portable recorder? I can show you my Realtek ALC892's line in performance, and it is on a $62 motherboard.

Original 96k signal and spectrogram settings:
original.PNG


Played through my Creative soundcard and captured by the Realtek using 48k sample rate:
48k rec.PNG

So yes, halfband filter with rather strong aliasing above 21k, but quickly fades to -80dB below that.

Additional test. Some people tend to believe that Windows' built-in resampler is always ridiculously bad, so I set the Realtek's recording sample rate to 96k, but record at 48k in order to activate the Windows resampler, and here is the result. Actually it is cleaner LOL.
96k rec win src 48k.PNG
 
Interesting results. The basic filters on ADC and DAC of the mystery device look alright. But then on those sweeps you get what looks like both aliasing and imaging back into the recorded band. You are beyond my competence and knowledge on this in detail. I've seen some filters that have odd behavior well beyond nyquist. It might filter out the ultrasonics properly to 192 khz but let an image get reflected at very high frequencies above that and it can alias back down strongly. Something to do with the modulators in the digital filter. Something I don't know much about.

Maybe @DonH56 can take a minute and tell us what this looks like to him. Hopefully his regular job isn't too harried for him at the moment. He actually knows what he is doing.
 
Hi Blumlein, thanks for the reply. According to the manufacturer, "[e]verything is programmed in assembly language", so I don't think there is a normal operating system. I myself am on a Mac and I've gotten the recorded files off the device using a USB stick.

No damn way. In Assembly?! This must be a very old device, or very limited in terms of UI or something.. I really wanna see what this is lol. I can't imagine anyone actually programming in assembly in the modern day anything substantial outside of calculators or something like recorders using non-LCD screens from 2 decades ago or something..
 
The Mystery Device DAC is OK. Not great, but I've seen worse too. The ADC is beyond dreadful. It looks like it's running at half rate (22.05 kHz) with the worst anti-aliasing filter ever, possibly none, and then upsampled to 44.1 kHz, again with poor or no filter.
 
Interesting results. The basic filters on ADC and DAC of the mystery device look alright. But then on those sweeps you get what looks like both aliasing and imaging back into the recorded band. You are beyond my competence and knowledge on this in detail. I've seen some filters that have odd behavior well beyond nyquist. It might filter out the ultrasonics properly to 192 khz but let an image get reflected at very high frequencies above that and it can alias back down strongly. Something to do with the modulators in the digital filter. Something I don't know much about.

Maybe @DonH56 can take a minute and tell us what this looks like to him. Hopefully his regular job isn't too harried for him at the moment. He actually knows what he is doing.

Thanks for the kind words though filter theory is not my main thing (yes, have done quite a bit of it in the past, mostly analog or sampled-analog, some DSP). My workweek has been 60-70+ hours lately thus limited (and probably terse, sorry!) posts. I had yesterday afternoon off and spent it playing with my new toy (SDP-75). I'll try to loop back on this later today.

Many filters (analog and digital) have stop-band ripple of various degree, and sampled (DSP or sampled-analog) do have images around the sampling frequency, requiring additional filters to suppress those images and prevent out-of-band noise (and aliasing, if before an ADC). Sigma-delta ADCs can use a pretty low-order analog anti-alias filter before the ADC, since the sampling rate is so high, then high-order (digital) filters after the converter (usually on-chip). Similarly, sigma-delta DACs still require an output image filter, but again can be pretty low-order due to the high sampling rate.

There are architectures using Hadamard sequences and other magic to reduce the sampling rate using parallel converters, but they are complicated, and aligning the response across multiple samplers even on chip is challenging (understatement).
 
Thanks for the replies! I really appreciate everyone's take on this.

It is starting to seem like the device is doing something unusual. You will always have (mostly subtle) differences between sample rates and implementations of anti-aliasing filters, but this seems different as it's aliasing like crazy without any frequency content anywhere near Nyquist. To sum it up as succinctly as I can:

WhatTheHeck.jpg
 
Back
Top Bottom