Author Topic: Scopes Built on FPGA's  (Read 1362 times)

0 Members and 1 Guest are viewing this topic.

Offline Old Printer

  • Frequent Contributor
  • **
  • Posts: 470
  • Country: us
Scopes Built on FPGA's
« on: May 03, 2018, 03:59:07 am »
Is the thinking that a scope built on an FPGA is more likely to be improved over time with updates, than one built with older technology. For instance, I have heard that Keysight is pretty well locked into their current ASIC because they have so much invested in it, or are minor upgrades possible without having to develop a completely new one.
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 2414
  • Country: us
Re: Scopes Built on FPGA's
« Reply #1 on: May 03, 2018, 11:00:16 am »
There are a number of Zynq based scopes.  If I quit fooling around with voltage references and buying HP test gear, I plan to write FOSS FW.  I've got all the dev tools collected.  The only thing lacking has been a long period without other distractions.
 

Offline hamster_nz

  • Super Contributor
  • ***
  • Posts: 1891
  • Country: nz
Re: Scopes Built on FPGA's
« Reply #2 on: May 03, 2018, 11:15:51 am »
Is the thinking that a scope built on an FPGA is more likely to be improved over time with updates, than one built with older technology. For instance, I have heard that Keysight is pretty well locked into their current ASIC because they have so much invested in it, or are minor upgrades possible without having to develop a completely new one.

(Mainly just to follow the thread, not really for useful input)

Red Pitaya might have beaten you to it. https://www.redpitaya.com/index2

User interface and features might improve, and bugs might get fixed by firmware, but specifications won't change much. The capabilities of the H/W are pretty much fixed at design time, and it would take a significant advance in DSP technology to squeeze a lot more out of an existing FPGA (i.e. a new algorithm that does twice as much with half as much resources).

FPGA chip technology advances over time, but sadly not for those already soldered to a PCB.  :(


 

Offline blueskull

  • Supporter
  • ****
  • Posts: 11540
  • Country: cn
  • Power Electronics Guy
Re: Scopes Built on FPGA's
« Reply #3 on: May 03, 2018, 11:16:51 am »
Every single (non crap, not the DSO138 shit) Chinese scope is based on FPGA. At this moment, no Chinese scope manufacturers can afford an ASIC (besides the new Rigol 6GHz monster), so FPGA is used instead.
« Last Edit: May 04, 2018, 08:45:00 am by blueskull »
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 2414
  • Country: us
Re: Scopes Built on FPGA's
« Reply #4 on: May 03, 2018, 10:59:29 pm »

User interface and features might improve, and bugs might get fixed by firmware, but specifications won't change much. The capabilities of the H/W are pretty much fixed at design time, and it would take a significant advance in DSP technology to squeeze a lot more out of an existing FPGA (i.e. a new algorithm that does twice as much with half as much resources).


Looking at the step response of my Instek scopes, I don't get the impression the person responsible was  particularly skilled at DSP.  I have yet to start my Zynq project work, but I shall be finding out what can be done.  In any case, the UI is so convoluted it's like a "maze of twisty passages, all alike".  So there is much room for improvement there.

I feel confident that with relatively modest effort some very useful features can be added.   I *wish* the OEMs were more cooperative and would open up the internal scripting system and apps to users.

My biggest interest is implementing compressive sensing.  Whether there is the compute power for that will be interesting.  A 5-10x increase in memory depth would be nice, especially on instruments with 140 Mpts.

The FPGA is what has made the cheap Chinese DSO possible along with a lot of other things.  Custom ASICs are fine for $300k instruments, but not for $300 ones.   The Zynq reduces the effort for reverse engineering an instrument for the purpose of writing new firmware very substantially.  Once you know the devices being used the number of possible permutations to be investigated becomes tractable.
 

Offline Old Printer

  • Frequent Contributor
  • **
  • Posts: 470
  • Country: us
Re: Scopes Built on FPGA's
« Reply #5 on: May 03, 2018, 11:17:32 pm »
Recently I was watching a 3 year old tear down by Dave of a Instek GDS 1000B. As he pried the heat-sink of the main processor he seemed delighted/impressed/surprised to find a XILINX ZYNQ fpga. Maybe mistakenly, I took this to mean that maybe on older DSO's there might be something other than a fpga under there. Maybe he was just impressed at the brand. I don't know.
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 2414
  • Country: us
Re: Scopes Built on FPGA's
« Reply #6 on: May 04, 2018, 01:22:33 am »
There's an FPGA in all of them, but the Zynq has a dual core ARM with two vector NEON FPUs.  From what I've seen the Zynq based scopes  are all  using 7010 devices, so at the price of some fiddly soldering, one could upgrade to a 7020 if one ran out of FGPA gates.  The Rigol DS1000E and DS1000Z series do not use Zynqs.  Most likely because of when they were designed.

I bought a Zybo Z7-20  and a BeagleBoard X-15 for dev work.  The documentation for the Zybo is pretty mediocre, so I got a MicroZed to use while I work through all the tutorial examples for it.  I've got an Instek GDS-2027E which I shall use for testing once I get that far.  This will take a while as I am prone to playing with other shiny baubles that catch my eye.

I'm an old hand at DSP having worked in reflection seismology my whole career.  I can think of a lot of tricks to teach a computer that can sample at 1-2 GS/S and implement real-time FIR and IIR filters and FFTs.  I haven't quite reached the $1k mark for the project , but as the saying goes, "The chicken has an interest in breakfast, but the pig is committed." 

I'm chaffing a bit at the moment because @cellularmitosis sucked me into volt-nut land.  I'm working on a small Peltier chamber today built around an old CPU heatsink and fan, a couple of TEC1-12706s and a 4" surface mount steel electrical box.  This is an experimental build before I build a larger unit.  Mostly motivated by I bought 10 but needed 8 for the big unit plus 4 high temperature units.  So a more limited range box seemed a good use for the two extra devices.
 

Online SMB784

  • Regular Contributor
  • *
  • Posts: 166
  • Country: us
Re: Scopes Built on FPGA's
« Reply #7 on: May 04, 2018, 02:23:17 am »
From what I've seen the Zynq based scopes  are all  using 7010 devices, so at the price of some fiddly soldering, one could upgrade to a 7020 if one ran out of FGPA gates.

I'm not sure this is possible because I don't think the 7010 is pin compatible with the 7020.  The 7010 has 100 IOs whereas the 7020 has 200, which should equate to way more pins on the 7020.  Does anyone know this for sure?
"Anything will lase if you hit it hard enough."

-Arthur L. Schawlow
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 2414
  • Country: us
Re: Scopes Built on FPGA's
« Reply #8 on: May 04, 2018, 03:11:15 am »
I *think* the scopes are using the CLG400 BGA package  which is an option for many of the low end devices.  The Zybo Z7 is available with either the 70110 or 7020.  However, it *is* a BGA swap, so I was not really very serious about the remark.
 

Offline Retep

  • Regular Contributor
  • *
  • Posts: 78
Re: Scopes Built on FPGA's
« Reply #9 on: May 04, 2018, 03:32:44 am »
The FPGA is what has made the cheap Chinese DSO possible along with a lot of other things.  Custom ASICs are fine for $300k instruments, but not for $300 ones.   

$300k instruments are pretty much by definition low volume products where NRE dominates and the unit cost is less of a concern. In those kind of products it makes perfect sense to use standard components (FPGA's) were possible, and only when there no other alternative (to meet the desired specs) use some custom silicon.

Products around the $300 mark are likely to be high volume, which means the NRE can be spread over many products and the unit costs is a more important concern. If the volumes are high enough an ASIC may be more cost effective than a relatively expensive FPGA. Of course there are other considerations, like time-to-market, field updates (bug fixes)...etc. which may make FPGA's a more attractive choice even for cheap/relatively high volume products.
 

Offline Keysight DanielBogdanoff

  • Frequent Contributor
  • **
  • Posts: 675
  • Country: us
  • ALL THE SCOPES!
    • Keysight Scopes YouTube channel
Re: Scopes Built on FPGA's
« Reply #10 on: May 04, 2018, 05:46:31 am »
On a $300k instrument, the ASIC is 100% about performance. On a $300 oscilloscope it's about price performance. That is, what's it going to cost to get to the level of performance we want/need? That can include things like GUI and responsiveness.

We actually talk about it a bit with one of our ASIC planners on the EEs Talk Tech podcast:

https://eestalktech.com/2017/05/25/all-about-asics/

https://youtu.be/jGKY29t-uQs?t=1m1s
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf