High speed ADC multiplexing.

[Paptrack]

Hello.
I'm working on a design where I need a 200Msps ADC. I'll be sampling 4 transducers, but not simultaneously. Each of the transducers will be sampled at 5-minute intervals. At the output of the ADC, I would like to connect an MCU, which is why I'm considering ADCs with parallel CMOS output (such as the ADS58C48, for example).
My question is: do you think it's feasible to multiplex the 4 parallel digital outputs of the ADC to connect to the MCU? Or would it be more logical to select a single-channel ADC and multiplex the analog part?

Thank you in advance.

[Andree Henkel]

In your application case (very slow multiplexing of the Channels)
I would use ADC with just one analog input channel, thus you only have one digital output bus.

then I would multiplex-switch the 4 analog signals into the one input channel.

[Andree Henkel]

the ADC you mention has parrallel DDR (double data rate) LVDS (low voltage differential swing) outputs.

good luck to find µCU with differential inputs, you wll likely additionally need FPGA, which converts LVDS to CMOS

say you are sampling one channel with 14 Bit at 200MHz, LVDS gooes into FPGA, 2*14=28 line CMOS Bus runing at 100MHz goes from FPGA to µCU - this would be feasible

[mikerj]

Input muxing seems like the obvious solution, with such slow switching things like charge injection become a non-issue.  That said at 200 Msps there are presumably some fairly high frequencies on the outputs of the transducers so channel isolation is likely to be of more concern.  You could even use signal relays to give high isolation at such low switching speeds.

[S. Petrukhin]

I think multiplexing the digital part is always easier and more reliable than invading the analog input circuits.

[max_torque]

i agree with the last ^^ poster tbh.  AT 200MSPS you'll be wanting a decent analogue front end / signal path, so doing the mplex'ing in the digital domain sounds like the way to go. As you are not requiring fast change to the mutiplex addressing then your multiplexers can be  switched and left, the high speed digital data sent/recieved, and you can then swap to a new m/plexer setting etc

Switching digital logic is also really cheap, so a basic digital buffer with an enable pin can be used to select the digital bus to the correct ADC with very minimal cost

[Paptrack]

Thank you all for your comments.

Just to mention that the ADS58C48 has also the option to configure the digital output as parallel CMOS interface, proper to interface with an MCU.

Coming back to the multiplexing, I have the following doubt. I will be sampling 4 ultrasound transducers of different frequencies: 05KHz, 2MHz, 4MHz and 20MHz. One transducer per channel. The signal from the transducer needs to be amplified, BP filtered and matched previous to ADC. So I am afraid a mux device (althougt it won´t be switching) could affect the analog signal.

So, despite although muxing the output will be the most expensive solution, as I will need and amplifier and BP filter per channel, I think it will be the most reliable solution. Do you agree?

Thank you.

[S. Petrukhin]

Coming back to the multiplexing, I have the following doubt. I will be sampling 4 ultrasound transducers of different frequencies: 05KHz, 2MHz, 4MHz and 20MHz. One transducer per channel. The signal from the transducer needs to be amplified, BP filtered and matched previous to ADC. So I am afraid a mux device (althougt it won´t be switching) could affect the analog signal.

So, despite although muxing the output will be the most expensive solution, as I will need and amplifier and BP filter per channel, I think it will be the most reliable solution. Do you agree?

Do you want to make the analog path (gain and filter) common? For different frequencies? What will it look like?
If this is a feasible task, yes, you can switch the "microphone", taking care of the emerging requirements not to disrupt the signal, take into account the normalization time, switching clicks, etc..
It's possible, no one says it's a forbidden option.
Muxing of the parallel digital output is very simple: from having CS pin in the ADC (it can support a common bus, perhaps) and even a separate MCU with a large number of pins. 

[Paptrack]

Thank you 

Yes, having an MCU with a lot of input pins could be a solution. In this case, I would need an MCU with at least 4 DMA (one per channel) to transfer the information to memory.

[S. Petrukhin]

Yes, having an MCU with a lot of input pins could be a solution. In this case, I would need an MCU with at least 4 DMA (one per channel) to transfer the information to memory.

No, you will need one DMA channel to which you will change the source and destination addresses.
You promised to take turns interviewing the ADC. 

[S. Petrukhin]

But not every MCU can capture 200Msps.
Were told about FPGA have reason.
Presence of LVDS outputs hints about serious flow. 

[David Hess]

Either method of multiplexing, analog or digital, could be used.

On the analog side, operation to 20 MHz and higher is relatively easy, but most people are not familiar with the methods used at higher frequencies.  Analog oscilloscopes did it all the time by switching the current outputs of transconductance stages.  A better method in your case involves operational amplifiers which have an output disable, so their outputs may be connected in parallel with only one amplifier active at a time.

[Paptrack]

Ups, sorry  . You are right, I can change the origin and destination of the DMA. What a slip-up!

Thank you.

[Paptrack]

A better method in your case involves operational amplifiers which have an output disable, so their outputs may be connected in parallel with only one amplifier active at a time.

I'm sorry David Hess. I'm afraid I don't understand. Could you please explain a bit more? Thanks.

[S. Petrukhin]

I'm sorry David Hess. I'm afraid I don't understand. Could you please explain a bit more? Thanks.

In addition to the expensive ADC, you will also need a unique op-amp with CS control. 

[Paptrack]

I'm sorry, but I still don't understand. What would be the op amp for? How should I use it? Should it be needed input multiplexing or output multiplexing? Excuse my ignorance

[S. Petrukhin]

I'm sorry, but I still don't understand. What would be the op amp for? How should I use it? Should it be needed input multiplexing or output multiplexing? Excuse my ignorance

You talked about filtering and amplification, and probably op-amp will be used in this case.
It is quite reasonable to use the op-amp in the filter or amplifier before the ADC, which obeys the output control.

[Paptrack]

Thanks. I will try to investigate a bit more.

[David Hess]

A better method in your case involves operational amplifiers which have an output disable, so their outputs may be connected in parallel with only one amplifier active at a time.

I'm sorry David Hess. I'm afraid I don't understand. Could you please explain a bit more? Thanks.

https://www.analog.com/en/resources/analog-dialogue/articles/amplifier-disable-function.html

[Paptrack]

Thanks a lot. Now I understand. And why multiplexing using op amps is better than using mux devices? Cost? Efficiency?

[xvr]

And why multiplexing using op amps is better than using mux devices? Cost? Efficiency?

Because you will use opamp at any case (for amplifuer/filters/etc) and with OE enabled opamps you will get Mux for free (no separate IC needed)

[David Hess]

Thanks a lot. Now I understand. And why multiplexing using op amps is better than using mux devices? Cost? Efficiency?

It is a low cost high performance solution when only a limited number of signals must be multiplexed.

FET based analog switches have worse bandwidth and feedthrough, and have to deal with series resistance so would need to be buffered anyway.  Transconductance based diode or bipolar switches are more complex.