Analog-to-digital converter

Go ahead.

Yeah, this is actually a triplet of Shield, but I wanted to pick the AD5593R ADC DAC.

How do they call it? The...

ADAC.

Yeah, this is a chip that does...

So now there's a shield.

Yeah, eight pins, and you can do ADC, DAC, or GPIO.

Simple.

family or search by whatever you want to search.

ADC, I2C, motor, driver, anything, just type here in search and we'll show you the results.

And last but not least here is the resources.

So we tried also to make it as a central place for the knowledge for the engineer that would work with our software.

So we have the special pieces of the software tools built in for that as well.

Then there's an interesting device in an analog pack, a product subfamily, SLG47011.

It has a 14-bit ADC and also an interesting plot called a memory table.

It's a huge 12-bit memory table with each word being 12 bits as well.

We will use silicon, but same physics, right? So we're not going to talk more about the principles of PIN because it's exactly the same as what was covered in LIDAR.

Then the other two components in every pixel of a CMOS APS is the readout circuitry And the readout circuitry, what it does is that it kind of sends the signal, the electrical signal that you get out of the PD or the photodiode, and it sends it to the ADCs to be digitized.

So that's the readout circuitry.

And it needs to be architected and designed such that you can read out every pixel in this APS array.

And then here, basically, this picture shows how we process one frame of data that we have collected.

So, a frame has M chirps and each chirp or actually the down-converted version of each chirp after, you know, mixing the received echo with the transmitted waveform and passing it through the ADC, each chirp would have N samples.

and complex samples.

So one frame of data overall, we can think of it or we can rearrange it as a m by n data matrix.

And this is all in DSP.

So that's a raw data matrix, one frame.

horizontal axis you can think of it as index n which is the ADC samples and vertically so each row is one down converted chirp.

So the row index gives you the chirp index and then the column index gives you the ADC sample number.

It can even be fully implemented in analog.

So you might not even, uh, need to use like a ADC and a DSP.

Uh, um, you can, you can do it much simpler.

But it comes with all, uh, uh, sorts of, uh, uh, you know, um, um, um, uh, caveats in terms of, uh, specifically the error in ranging has, has, has dependency on the amplitude of the received echo and also on the angle of where the target is.

So the signal has instantaneous frequency or c- it's actually just a constant frequency that is proportional to the time of flight, tau.

Right? So basically, at this point, uh, we have a sensor, right? Uh, because that block diagram that, that, uh, we showed, if you build that, the intermediate frequency that we get, if I just, uh, go back to the block diagram real quick so we can look at it together.

So this, uh, intermediate frequency signal that we have, uh, it's gonna get filtered and passed through an ADC.

But then the discretized version that we have, this signal has a frequency that is proportional to the time of flight.

And remember, this is the IF signal from one chirp, right? So you have down-converted one chirp You've downconverted that.

That's, that's the SIF that you get, and, uh, the ADC has, has sampled it, so you get endpoints or N samples in it.

So N samples corresponding to one downconverted chirp, then you do an endpoint FFT on it, and here I'm kind of loosely using the term FFT.

I mean, uh, we all know FFT is the actual, uh, algorithm, the implementation of a discrete Fourier transform.

So for example, in the, in the, in the case of the, what I showed you in the previous slide where we had three targets, if you do that, we'll see three peaks in the, in the FFT at indices K hat one, K hat two, and K hat three.

And then, uh, to actually find the range of, of, of the targets, uh, you, you first convert the indices to frequencies, right? Um, so the indices go from zero to N minus one because you are doing an endpoint, uh, FFT here.

Uh, to find the frequencies, you just, uh, do, uh, your index divided by N, which is the total number of points in the FFT, times the sampling rate F sub S. F sub S is your ADC sampling rate.

So that converts your index K to your delta F frequency, and then to, from delta F to range, that's another scaling, right, which, which we just looked at.