Samplerates of multiple audio inputs of the Audio Expander

giuliomoro

yes

Bethanychm

giuliomoro It looks so wired, so as IN2. Is this the contact issue of caplete and cape?

giuliomoro

Can you show a picture of the boards from the side? I suspect J10's pins are not making proper contact with the socket underneath. Also make sure you run with 8 analog channels enabled when running these tests.

Bethanychm

giuliomoro Yes, I run with 8 analog channels enabled.

Bethanychm

giuliomoro Got!
Here are the new results for 3.3V：

Blow are the mics results with capelet unabled:

Bethanychm

giuliomoro It looks that the two boards are plugged in tightly.

giuliomoro

Oh wait another thing: for this last test, you need to disable the audio expander checkbox in the IDE for the selected channels, otherwise any DC voltage will be filtered digitally.

giuliomoro

Right, so there we see that in https://forum.bela.io/d/3201-samplerates-of-multiple-audio-inputs-of-the-audio-expander/24 we get equal readings from the 3 channels for a 3.3V input when connected directly to the cape, but
then in https://forum.bela.io/d/3201-samplerates-of-multiple-audio-inputs-of-the-audio-expander/32 we get different readings when plugged through J10 on the capelet, even though J10 is a passthrough.

In the last set of pictures, the scale and offset of the channel enabled in each picture is the same, correct? Just double checking.

Do a couple more things for me:

with the capelet connected, plug 3.3V into channel 0 of j10, look at the signal on the scope. Then, place your thumb on the free part of J10, as close as possible to the jumper wire, and the rest of your fingers underneath the BBB. Squeeze them tight while you look at the scope: do you see changes? Does the channel ever reach the expected value of approx 0.8?
Can you show the capelet unplugged from the cape? Are the pins on J10 as long as those on P8 and P9 ?

I see that you are using a rev B cape, where J10 is shorter than P8/P9, so that could be the reason why things don't make great contact.

Bethanychm

giuliomoro Thanks for your double checking, the scale and offset of the channel enabled in each picture is the same.
Then

Yes, I can see the value changing to approx 0.8 when I tightly squeeze them.
J10 looks as long as P8/P9:

So if I need use 4 mics to measure AoA with sample rate of 44.1KHz, this contact problem will be a trouble, right?

giuliomoro

Right. Well, so it is indeed a case of bad contact because of the older rev B cape underneath it... depending on how your skillset and tools look like, you have two options:

the fastest -with a desoldering station and good soldering skills, but potentially destructive- is to desolders J10 from the Bela cape and put it back on with the sockets sitting slightly raised on the PCB, to bring them closer to the capelet
the slowest - requires good angled cutters - is to trim the 46+46 pins of P8 and P9 on the capelet - leaving the ones on the single-row headers unchanged. Aim for cutting off about 2mm off and make sure you don't overdo it. Ultimately very few pins of the capelet are actually needed to make it work, so chances to damage it irreparably are pretty low. Still, you'll want to do a pretty even cut, using a jig of some sort to make sure the resulting length is even.

Bethanychm

giuliomoro It sounds like the second solution is safer. I have another question: If I use audio inputs 0/1 (L/R) of the cape and ananlog IN2/3 (enable analog IN0-3 as audio inputs, sample rate is same with audio sample rate)of the capelet to receive signal of 4 mics, is it a possible alternative solution to use 4-mic array to mearsure AoA? If it is possible, what issues do I need to pay attention to? For example, how can I synchronize these 4-mic signal inputs?

Here is the simple code for FMCW audio output and 4-mic inputs with real time logging, is it feasible?

WriteFile file;

//FMCW
float gFrequencyLow = 18000.0;	// Frequency of the sine wave in Hz
float gBandwidth = 4000.0;
float gT_chirp = 0.04;
float gPhase = 0;
float shift = 0;
int gAudioPerAnalog;

bool setup(BelaContext *context, void *userData)
{	
    printf("Audio Sample Rate: %f\n", context->audioSampleRate);
    printf("Audio Frames: %d\n", context->audioFrames);
    printf("Audio Output Channels: %d\n", context->audioOutChannels);
    printf("Audio Input Channels: %d\n", context->audioInChannels);
  
    printf("Analog Sample Rate: %f\n", context->analogSampleRate);
    printf("Analog Frames: %d\n", context->analogFrames);
    printf("Analog Output Channels: %d\n", context->analogOutChannels);
    printf("Analog Input Channels: %d\n", context->analogInChannels);
    
    gAudioPerAnalog = context->audioFrames/context->analogFrames;
    printf("Audio/Analog Frames: %d\n", gAudioPerAnalog);
    
    file.setup("trans_and_rece.bin");
    file.setFileType(kBinary);
    file.setFormat("binary: %.4f %.4f\n");

    return true;
}

void render(BelaContext *context, void *userData)
{
    for(unsigned int n = 0; n < context->audioFrames; n++) {
                 float out = cos(gPhase);
		float gFrequency = 1/gT_chirp * shift;
		gFrequency = gFrequencyLow + gFrequency;
		gPhase += 2.0 * M_PI * gFrequency / context->audioSampleRate; 
		if(gPhase > M_PI)
			gPhase -= 2.0 * M_PI;
		
		shift += gBandwidth / context->audioSampleRate;
		if(shift >= gBandwidth * gT_chirp)
			shift -= gBandwidth * gT_chirp;

                 file.log(out);   // log out
                 audioWrite(context, n, 0, out); %audio oupt channel:0

                  // first 2 mics inputs from audio IN0-1
                 for(unsigned int ch = 0; ch < context->audioInChannels; ch++){
    		       float val =  audioRead(context, n, ch);
    		       file.log(val);
    	          }

                   // then 2 mics inputs from analog IN2-3 of capelet
                    float input3 = analogRead(context, n/gAudioPerAnalog, 2);  
    	            file.log(input3);
    	            float input4 = analogRead(context, n/gAudioPerAnalog, 3);
    	            file.log(input4);
     }
}

void cleanup(BelaContext *context, void *userData)
{
}

giuliomoro

What's AoA? You can mix and match analog and audio channels but be mindful of their phase difference. The audio ADC has approx 20 frames group delay due to its internal filter. For the analog ADC that's more probably below 2 all included, however analog inputs are using a single ADC with multiplexes inputs , so its channels will all have a different phase. If your application is heavily phase-dependent, you'll have to compensate for these differences

Bethanychm

giuliomoro I am sorry for writing wrong analog channels in https://forum.bela.io/d/3201-samplerates-of-multiple-audio-inputs-of-the-audio-expander/37, now I have modified it to analog IN2-3. I want to use 4-mic array to measure the arrival of angle (AoA) for the reflected signals. It is indeed heavily phase-dependent. I understand that analog multiplexes inputs introduce different phases between IN0-3. How about the interval when swtiching from audio IN0/1 to analog IN2-3 (actually relates to IN0 I think?)? Can you give me some methods to compensate for these differences?

giuliomoro

Bethanychm

Bethanychm . How about the interval when swtiching from audio IN0/1 to analog IN2-3 (actually relates to IN0 I think?)?

The audio inputs are sampled synchronously, so they will not have a phase difference between them. They will be about 18 samples later than the analogs. For sub-sample phase adjustment, things are complicated ... you'd need a fractional delay line with decent interpolation, or a resampler where you can set the group delay arbitratrily. Could start with a four-point interpolator like the one in Pd's [tabread4~], or look at what they are using for [vd~]. For scientific accuracy, howeverm you may need a FIR linear-phase filter. Libresample can probably do what you need, but it may be overkill. You'd need to find a sufficiently short set of FIR coefficients that do the phase shift you require and then can apply them with the Convolver class to apply them to the signal.

Bethanychm

giuliomoro Sincerely thank you for your answer! It seems to be more straightforward to solve the contact problem between the cape and the capelet. How can I compensate the phase for analog In0-3 when they are used to measure the arrival of angle (AoA)?

giuliomoro

Ultimately, if you wanted to fix this in hardware, a better option is to only use audio inputs that are all of the same type and synced. We have the Bela Mini Multichannel Audio Expander which provides 6 audio inputs. That has microphone inputs and it's cheap, however, that's designed for the BelaMini, which is now discontinued so, unless you already have one of those, you'd need to make a PCB adapter to connect it to the BeagleBone Black you already have.

In the BBB form factor, there's the CTAG FACE, which you could even use in combination with your Bela cape's analog inputs and outputs (not the audio ones). The inputs of the CTAG are not designed for microphones though and they don't have a preamp (unlike the ones you've been using so far) and they have full-scale 0.9Vrms (I think) input, so the signal may be a bit quiet, but it may work.

If you can wait a few months, have a look at the Bela Gem Multi (with 10 mic inputs): https://bela.io/gem

Bethanychm

giuliomoro Thank you for your valuable suggestions. I need to use them urgently recently, so I cannot wait for a long time.

Bethanychm

giuliomoro I want to know if I purchase a new product, how long does it usually take to ship to Asia?

giuliomoro

you'd need a fractional delay line or resampling filter where you can specify a fractional delay as I mentioned above

giuliomoro

Bethanychm we ship on Wednesdays and if you pick DHL as an option, it will take 4 or 5 days I think. If the DHL option doesn't show up for your country, email orders at bela.io to add it.

Bethanychm

giuliomoro So the latest ship date is next Wednesday?

« Previous Page Next Page »