Analog Discovery 3

[Guest]

Hi all,

I just got the mail about the upcoming Analog Discovery 3. Interesting.

I'm mostly interested in the raw performance of this thing...

• What sample rate can it sustain when pushing data to a single DAC channel via USB, in playback mode? What can it do to two synchronized DAC channels, assuming a PC that is not the bottleneck?
• What sample rate can it sustain when pulling data from a single ADC channel via USB, in record mode? What can it sustain when pulling data from both ADC channels simultaneously, assuming a PC that is not the bottleneck?

I hope that these numbers are significantly improved when compared to the performance of the AD2.

I would also be interested in the different configurations it supports (in terms of memory allocated to the different instruments).

Lastly, how difficult is it to open the enclosure and taking out the PCB?

In the AD2, it was just a matter of unscrewing a few screws but the new enclosure doesn't show any. Since the connection to the external BNC breakout is rather clunky/flimsy, I have gotten into the habit of 3D-printing an enclosure that can take a joint AD2 and BNC breakout board as shown below:
 

 

 

 

[oltechie]

Don't know the answers to your questions, but that sure is a cool way to set up the AD2! 

I am hoping the USB C connection will deal with the flaky USB issues of the AD2. With the usual USB C docking station style hubs available these days the AD2 has a very great tendency to not have enough power to operate or suddenly disconnect. And the small cheap Windows tablets in my test stands no longer have USB2 or 3 slots at all! The only way I am able to use one with USB C is to get an ancient USB2 powered hub and use very short cables. Oddly, using an external power supply does not seem to help. 

If someone from Digilent checks in, here are my AD3 questions:
1. JTAG or SWD protocol? Does that mean we can program chips? Probably not, but that would make the AD3 my dream machine. I have to drop $500 on every test stand for hardware that is really just a serial port to SWD translator to program the chip. 

2. Will it work with the PI finally? All the Raspberry PI derivatives I have tried do not seem to be able to find the AD2 device or keep it connected. Neither the USB3 or USB2 slots work. Believe me, I have tried all the tricks.

3. More informative winky light that you can actually see? The AD2 LED starts blinking as soon as you connect, and then goes on blinking until you power off. Not helpful when you are fighting disconnection issues from flaky hubs and cables. The cover is no longer partly translucent but there is something that could be an LED next to the USB C inlet. Hope it is not just power. 

Guess I'll find out when my units arrive.

[attila]

Hi @reddish

The Scope Record is a bit improved due to higher device buffer/fifo and other optimizations, 10MHz for one channel and 5MHz for two channels.
I haven't tested the Wavegen Play yet. Maybe we can optimize this a bit too.
The enclosure is held by 4 screws.

[attila]

Hi @reddish

After optimizations, the Wavegen Play works up to about 5MHz.

[Guest]

Okay thanks for the info. Unfortunately you guys didn't go for proper USB3 support, that could have helped a bunch....

There's so much untapped potential performance in the hardware itself when it comes to streaming, but as far as I can tell from the outside, it seems that the low-level communication protocol choices are an unpleasant bottleneck that we'll have to live with forever.

[attila]

Hi @reddish

The current implementation transfers chunks of data so each independently running instrument (or channel) in the device gets transfers slots on demand.
Theoretically, the maximum achievable streaming could be double of the current rate 10/5MHz, but this would require non-blocking transfers, new dedicated functions...

 

[Xavier Jiang]

On 6/5/2023 at 7:06 PM, attila said:

10MHz for one channel and 5MHz for two channels.

Correct me if I am wrong, but I have noticed the behavior (regarding wavegen) that instead of storing the waveform (and associated settings) in the AD buffer, the waveform is being continually streamed from the PC, even though the wave is repeating.

I know the "play" function of wavegen streams chunk of the file to the AD, which is because of the small buffer on the AD, but I seem to be able to see the same behavior even when I am not streaming media (e.g., generating a square wave)

When testing scope record on the Analog 2, I keep getting weird results. Eventually I am able to dial in the settings, but for some reason I can stream the 100MS/s measurement (although not continuously, I guess).

And lastly, is the Analog Discovery 3 USB 3.0 capable? That will unlock a lot more bandwidth. Although from what I am seeing we have not maxed out the USB 2.0 speeds yet.

 

One of the other problems I have found when using the Analog 2 is that the device gets fairly warm during use. None of the components are alarmingly hot, but it leave me wondering if any additional measurements are taken to address the heat issue on the AD3. Although a custom 3D-printed case with a fan would easily address this "issue".

Edited June 21, 2023 by Xavier Jiang

[attila]

Hi @Xavier Jiang

Could you provide more detail about the problem ?

The devices can capture and generate waveforms which fit in the device buffer (4,8,16,32k .. 128M) up to maximum rate (100/125MHz).
The Record and Play uses data streaming. It is intended for larger waveforms and the device buffer is used as fifo.
The AD2 can Record up to about 3MHz on 1 or 2 channels.
The AD3 is still USB FS (2.0 480Mbps). The Scope buffer is shared between the channels. It can record at about 10MSps, 10MHz for 1 channels and 5MHz for 2 channels.

It is normal for the devices to heat up, the AD2 consumes at least 1.5W and the AD3 3W.

[Xavier Jiang]

4 hours ago, attila said:

the problem ?

I was trying to capture the transient (a.k.a. "bounce) of some buttons, so I want the highest amount of samples (a.k.a. 100M) possible from the Analog 2.
However the Analog was complaining about "data loss" and "reduce sample rate", even if I set the amount of samples to be beneath 8Ki (the default allocation size). In the lab.
I am unable to produce that problem, and now I can get it to do *live* at 100M, so I don't know.
I even get it to do 25M waves, although its very apparent that the DAC is not fast enough, but that's fine -- I'm just trying to see what will cause it to fail.

I can only imagine that by the time they get to play with Analog 3 they get to do even more crazy stuff.

Edited June 21, 2023 by Xavier Jiang

[attila]

Hi @Xavier Jiang

You probably used Record mode to capture multiple samples, which uses data streaming and has limited rate, ~3MHz for AD2 and ~10MSps for AD3.