Thank you! I added that to the list of hardware we may use. Amazing that such an innocuous device could have such a high sampling rate!
Have you tested the audio I/O bandwidth?
Yes, I have tested devices purchased a few years back in Germany and they work indeed with 192 kHz sample rate and have an input bandwidth up to 96 kHz. Below is some info how I use it as a low-cost ultrasonic microphone for bat recordings. Will try to get me a recent one here in Australia to check if they still work like before (the specification at least has not changed).
Edit: I have uploaded a corrected schematic (some of the lines had moved while copying from the original)
That’s not possible. The audio bandwidth must be less than half the sample rate.
One quick and easy way to test is to make a loopback cable, then play a sweep signal 0 to 96kHz through the device, looping back into the input, and record.
Steve,
I’d like to comment on this, not because I want to have an argument about it, but to clarify things for those who might come across this thread and are not deeply involved in signal processing theory.
So, a quote from: Analog Devices, Walt Kester, What the Nyquist Criterion Means to Your Sampled Data System Design:
Simply stated, the Nyquist criterion requires that the sampling frequency be at least twice the highest frequency contained in the signal, or information about the signal will be lost. If the sampling frequency is less than twice the maximum analog signal frequency, a phenomenon known as aliasing will occur
This means in theory (everything being ideal) that 96 kHz can be sampled correctly using a sampling rate of 192 kHz. Having said that, we know that nothing in the technical world is ideal, so it is advisable to stay away from the theoretical limit in practical systems, and I believe that’s what you were referring to.
I should have better said: The Ugreen device works close to 96 kHz, which is the theoretical limit when sampling with 192 kHz. There will be some loss in sensitivity when approaching this limit.
All of this isn’t relevant when using the device as an ADC for a bat recording system, because the sensitivity variation across the frequency band of interest due to the characteristic of the MEMS microphone itself is much greater.
Now, how have I tested this?
1. I created a test file using Audacity. Not a sweep as you have suggested, but a sequence of discrete frequencies 10 kHz apart.
2. I have stored this file on my FIIO M3K player, which can output data with a sample rate of 192 kHz.
3. I have then played that file from the M3K through a wide bandwidth amplifier, which was connected to drive a high bandwidth tweeter speaker.
4. I have then recorded this transmitted test signal through my DIY microphone, which was mounted side by side with a professional Pettersson M384 ultrasonic microphone and I have compared the recorded results.
5. This comparison to a high-quality professional microphone showed clearly, that my DIY microphone was suitable for a lot of cases when recording bat calls.
6. I have repeated this in situations with real bats calling. Also, these tests showed that the quality of the results was satisfying for the bat call analysis process.
Now, based on our recent interaction on the Audacity forum I have run some more “technical” testing. I have simply fed the MK3 out put to the Ugreen sound adapter and found the following when comparing the results (top test file, bottom recorded data):
In summary there is a loss of sensitivity when looking at the higher end of the frequency scale and the 95 kHz signal doesn’t look extremely good, but overall good enough for my application. If I find the time, I’ll check what is causing the loss at high frequencies, is it the MK3, the Ugreen adapter or both? This is not separable with this test.
The data sheet of the sound chip used for the Ugreen adapter, which is an ALC4042, says this:
This is specified for the 48 kHz sample rate. The ADC -1dB point at the high end is specified as 0.454 *sample rate frequency. If I transfer this to 192 kHz sample rate the upper -1dB point should be at ~ 87.2 kHz. It still records the 95 kHz tone though.
Exactly that.
It is very common for high quality audio devices to limit the analog range to <30kHz even at high sample rates. This has multiple benefits for normal audio: It helps to reduce interference, provides a more gentle high frequency cut-off slope with less ripple in the pass band, and can help reduce IM distortion and improve cross channel separation.
While that little Ugreen device might cut a lot of corners regarding sound quality in the audio range, your tests show that it is a great fit for a low cost, high bandwidth device for analog frequencies up to ~90kHz. Nice find.