Author Topic: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth  (Read 4330 times)

0 Members and 1 Guest are viewing this topic.

Offline max_torqueTopic starter

  • Super Contributor
  • ***
  • Posts: 1282
  • Country: gb
    • bitdynamics
I have a =-10V signal.  It can float on a +-1000Vdc common mode voltage to earth.  I want to push that signal across an isolation barrier (1KV working, 3kV intermittent)  with 16bit resolution and around 500KHz bandwidth.

Two approaches possible:

1) stay in analogue domain, use some sort of analogue transfer and monitor/feedback

2) descretise into digital domain, with a high speed serial transfer across isolation, then back into analogue domain.  ie ARM or FPGA with ADC, pushing serial data across opto's to a DAC etc


Initially ii'm just looking for a single channel, but would be nice to expand to say 4 channels at a later date.  Absolute precision requirements tbc!


Any other options and opinions on "Best" method? 
 

Offline EmmanuelFaure

  • Regular Contributor
  • *
  • Posts: 154
  • Country: 00
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #1 on: December 01, 2016, 04:32:38 pm »
An analogue solution with IL300 analog isolator?
Is this for a commercial project?
 

Offline lukier

  • Supporter
  • ****
  • Posts: 634
  • Country: pl
    • Homepage
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #2 on: December 01, 2016, 04:38:48 pm »
I would use something like Tektronix A6902B - two isolated channels, up to 3kV (with the biggest scope probes I've ever seen), up to 450 kHz @ 3kV, more at lower voltages, up to 20 MHz.

But this is for oscilloscopes, thus I highly doubt you would get 16 bits of resolution.

I have A6902B at home, but I don't have any HV sources (> 120VDC, or a function gen that can be floated to such high voltages), so I don't think I can test how accurate it is.
 

Offline Gyro

  • Super Contributor
  • ***
  • Posts: 9504
  • Country: gb
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #3 on: December 01, 2016, 04:48:26 pm »
I don't think you are going to ba able to achieve 16 bit resolution (or at least linearity) using analogue transfer, nice though it would be. Even if you could, then it would be expensive in parts and development time.

You're probably best going with a floating ADC and micro and an optically coupled link. Some ADCs are self clocking so you might just get away with coupling the serial data stream across.

I'm not sure how easy it would be to obtain power on the high potential side, you could maybe consider a small transformer coupled supply across it isolation barrier to power the ADC.
Best Regards, Chris
 

Online T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21684
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #4 on: December 01, 2016, 07:10:16 pm »
ADC-DAC pair.  Seconded.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline max_torqueTopic starter

  • Super Contributor
  • ***
  • Posts: 1282
  • Country: gb
    • bitdynamics
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #5 on: December 01, 2016, 07:22:27 pm »
Happy to provide uC Vcc from usual low power isolated DC:DC across the barrier.  if i go digital, what micro / FPGA would people suggest?

( i have no direct FPGA experience, but this would be an ideal time to learn eh...   :-DD )

16b at say 500kS/s is 8Mb/s without any protocol overhead, so not too fast for a serial bus?

 

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 8172
  • Country: fi
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #6 on: December 01, 2016, 08:06:47 pm »
MCU or let alone an FPGA on the measurement side sounds an overkill when you just need a simple, bare ADC. ADC with serial output, SPI for example, isolate clock / other control signals, and the data output using fast optos or digital (transformer-based) isolators such as ADUM series. Then it will be trivial to connect this ADC to the MCU, just like you'd do it without the isolation.

If you only need analog output, you can do it with with ADC - DAC pair. Worst case, you'd need any simple low-performance microcontroller to provide some simple control signals for them.
« Last Edit: December 01, 2016, 08:09:42 pm by Siwastaja »
 

Offline max_torqueTopic starter

  • Super Contributor
  • ***
  • Posts: 1282
  • Country: gb
    • bitdynamics
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #7 on: December 01, 2016, 08:10:24 pm »
It doesn't really matter on which "side" the controller resides?  In either case i will need an isolated supply to the isolated side.

it either goes:

ADC -> Isolation -> uC -> DAC

or it goes

ADC -> uC -> Isolation -> DAC


Ok, so in case 2, there is a slightly higher power requirement on the isolated side but that's pretty small beer with any modern uC!

If i use a 16b ADC/ DAC, and realised a true >13b resolution i'd be happy   :-+
 

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 8172
  • Country: fi
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #8 on: December 01, 2016, 08:17:08 pm »
Oh, so your end product is analog, you are not going to digitize it afterwards, so you are not doing anything with the digital data? The 16 bit requirement gave me a wrong impression first :).

If you need an MCU to act as a master to toggle some command signals (like: initiate the conversion by pulling chip select low, then toggle clock line - it will be super simple, with no actual data line going through at all!), it indeed doesn't matter on which side it is; pick the side which seems to be easier to implement, depending on your design details.

So its either:

ADC -> Isolation (data) -> DAC
  ^
MCU -> Isolation (ctrl) ----^

Or:

ADC -> Isolation (data) -> DAC
                                        ^
  ^----- Isolation (ctrl) <--MCU


Are delta-sigma modulators good for BW resolution product this high (I don't know, haven't personally used)? If yes, it would be the easiest solution, with fewer signals to isolate.
« Last Edit: December 01, 2016, 08:24:35 pm by Siwastaja »
 

Online Zero999

  • Super Contributor
  • ***
  • Posts: 19520
  • Country: gb
  • 0999
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #9 on: December 01, 2016, 08:37:24 pm »
For the isolation part, you might want to consider using digital isolators from Silicon Labs. They're much faster than opto-isolators, cheap and available from most distributors.
http://www.silabs.com/products/power/isolators/Pages/default.aspx

If you opt for analogue then there's PWM but 16-bit accuracy might be difficult.
 

Offline max_torqueTopic starter

  • Super Contributor
  • ***
  • Posts: 1282
  • Country: gb
    • bitdynamics
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #10 on: December 01, 2016, 10:23:03 pm »
Oh, so your end product is analog, you are not going to digitize it afterwards, so you are not doing anything with the digital data? The 16 bit requirement gave me a wrong impression first :).

If you need an MCU to act as a master to toggle some command signals (like: initiate the conversion by pulling chip select low, then toggle clock line - it will be super simple, with no actual data line going through at all!), it indeed doesn't matter on which side it is; pick the side which seems to be easier to implement, depending on your design details.

So its either:

ADC -> Isolation (data) -> DAC
  ^
MCU -> Isolation (ctrl) ----^

Or:

ADC -> Isolation (data) -> DAC
                                        ^
  ^----- Isolation (ctrl) <--MCU


Are delta-sigma modulators good for BW resolution product this high (I don't know, haven't personally used)? If yes, it would be the easiest solution, with fewer signals to isolate.

Assuming there is such a thing as a matched serial ADC and DAC, where the output and input data format is identical?  (so the MCU just provides the clock and control lines (CS etc), and data gets shifted out the ADC, straight into the DAC??)
 

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 8172
  • Country: fi
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #11 on: December 02, 2016, 05:52:55 am »
I'm 99.9% sure you will find a matching pair, because I have used several simple serial ADCs and DACs with bit-after-bit no-bullshit data format.
 

Offline Alex Nikitin

  • Super Contributor
  • ***
  • Posts: 1177
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #12 on: December 02, 2016, 09:12:05 am »
If you can get away with a somewhat relaxed requirements (200kHz BW, 0.03% non-linearity), than something like AMC1301 from TI may do the job.

Cheers

Alex
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13747
  • Country: gb
    • Mike's Electric Stuff
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #13 on: December 02, 2016, 09:43:25 am »
Definitely digitise it first and send it serially. There is no way you will get 16 bits with any analogue method.

There may be some scope to build a delta-sigma converter with isolation around the digital parts but unless it's a high volume thing it would be cheaper to simpley use a serial-output (e.g. SPI) ADC with isolators.
 
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Online 2N3055

  • Super Contributor
  • ***
  • Posts: 6632
  • Country: hr
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #14 on: December 02, 2016, 10:13:43 am »
Is there any reason why you couldn't have separate optical link for data...
1000 V is a lot, to make it reliably safe is very hard...

 

Online tszaboo

  • Super Contributor
  • ***
  • Posts: 7377
  • Country: nl
  • Current job: ATEX product design
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #15 on: December 02, 2016, 10:29:24 am »
I dont think there is a way to stay in analog. Your requirements are basically 1+MSPS ADC with 16 bit resolution, that is 16mbit/sec. You should use a digital isolator or fiber optics. Can you tell us, if this is a low quantity project, a one-off or price sensitive production? It is really a different answer for any of the three.
 

Offline Alex Eisenhut

  • Super Contributor
  • ***
  • Posts: 3338
  • Country: ca
  • Place text here.
Hoarder of 8-bit Commodore relics and 1960s Tektronix 500-series stuff. Unconventional interior decorator.
 

Offline max_torqueTopic starter

  • Super Contributor
  • ***
  • Posts: 1282
  • Country: gb
    • bitdynamics
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #17 on: December 02, 2016, 01:17:11 pm »
I have considered doing a "remote" isolated micro with ADC, and a fibre optic data link to a "mother board", and that mother board could have a some thing like ethernet on it for direct digital data output, and a DAC for an analogue output.  This would also potentially bring the total system cost down for a multi-channel arrangement.


it certainly wouldn't be a commercial product, more just a customisable and affordable way of allowing our existing data acquisition systems, which are designed for low voltage use, to interface safely to high voltage systems.
 

Offline snarkysparky

  • Frequent Contributor
  • **
  • Posts: 414
  • Country: us
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #18 on: December 02, 2016, 01:43:51 pm »
500khz bandwith means at least sampling 16 bits at 1 Mhz.  Actually more like 1.5 Mhz.   With 16 bits this is at least 24 Mhz.

Can you couple a digital signal that fast.  If so then digital is the way to go.
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13747
  • Country: gb
    • Mike's Electric Stuff
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #19 on: December 02, 2016, 02:13:58 pm »
500khz bandwith means at least sampling 16 bits at 1 Mhz.  Actually more like 1.5 Mhz.   With 16 bits this is at least 24 Mhz.

Can you couple a digital signal that fast.  If so then digital is the way to go.
110Mbits over 6kV good enough?
 multichannel so SPI with a single package
http://www.nve.com/v-series.php
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Online tszaboo

  • Super Contributor
  • ***
  • Posts: 7377
  • Country: nl
  • Current job: ATEX product design
Re: Transfer +-10V signal across 1KV (working) isolation at 500Khz bandwidth
« Reply #20 on: December 02, 2016, 02:53:29 pm »
500khz bandwith means at least sampling 16 bits at 1 Mhz.  Actually more like 1.5 Mhz.   With 16 bits this is at least 24 Mhz.

Can you couple a digital signal that fast.  If so then digital is the way to go.
110Mbits over 6kV good enough?
 multichannel so SPI with a single package
http://www.nve.com/v-series.php
You dont need to go that obscure. Texas instruments, analog devices and silicon labs makes digital isolators, which are some 3-5KV working volatge and 100 mbit.

One thing to keep in mind is the propagation delay. Digital isolators have significant delay,  which can mess up SPI reading, or generally something, where the clock and the data source are on different sides.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf