EEVblog #692 - Digilent Analog Discovery Review

[tggzzz]

With high Z the bandwidth of this device is about 20MHz (-3dB): (note the Digilent software does not permit exceeding 10MHz, but using the API you can extend that to 20MHz+)

How irritating. I always like to see the edge of my instuments' performance, so that then I can be sure that I'm not operating so close to the edge that I'm being misled (or misleading myself).

However, congrats to Digilent for releasing the API in the first place.

However with a standard 50R load the -3 dB bandwidth (function gen) is pretty much exactly 5MHz:

I'm content for the scope and AWG to have different bandwidths, providing I'm aware of it.

Thanks for the information.

[jaxbird]

Dave's complaint of the misaligned pinout text was corrected even before he did his review, I guess he had it around for so long several revisions were made in the meantime:

[jaxbird]

With high Z the bandwidth of this device is about 20MHz (-3dB): (note the Digilent software does not permit exceeding 10MHz, but using the API you can extend that to 20MHz+)

How irritating. I always like to see the edge of my instuments' performance, so that then I can be sure that I'm not operating so close to the edge that I'm being misled (or misleading myself).

However, congrats to Digilent for releasing the API in the first place.

However with a standard 50R load the -3 dB bandwidth (function gen) is pretty much exactly 5MHz:

I'm content for the scope and AWG to have different bandwidths, providing I'm aware of it.

Thanks for the information.

Same here, I would much prefer the software to allow up to e.g. 50MHz, even if the signal is 30-50 dB down, it just gives more flexibility and a better understanding of the usable limits. Using API you can get the Function Gen to output 20MHz+ using standard waveforms, but the signal is down.

[jaxbird]

And finally the spectrum analyzer peak at 39.4kHz (with input shorted) Dave complained about and attributed to possible switchmode supply design flaw or similar, is not present on my version:



I guess it could be external as the device isn't shielded at all, or perhaps it was fixed in later revisions. Mine is revision C.

[tggzzz]

ISTR reading somewhere that it was felt that the target market (beginners) would complain if they felt they has been mislead by finding the amplitude of a 5MHz signal was too low. Personally I don't agree with that, since they will need to understand what equipment's spec does and doesn't mean.

I believe Digilent is owned by NI. In that case there might be a less charitable motive for restricting the device, especially the spectrum/network analyser: to "differentiate" between the academic and professional markets and prevent the low-end professional market from being eroded by the academic tools. I carefully avoided using the word "crippling". I'm sure no marketeer would ever think of such a thing, let alone enact it.

I certainly don't know whether or not it has been done in this case.

[Syntax_Error]

Digilent was only very recently purchased by NI, and the AD and the EE board were both available for purchase (as in finished design already fielded for sale) before Digilent announced the company acquisition. Of course this doesn't mean that boardroom meetings weren't already going on for months; who knows. There was some worry online when the acquisition was first announced, as people were curious if the AD would infringe on NI's myRIO and myDAQ products. Since then, Digilent has announced that the AD and EE are not being dropped, and have also added the mentioned NI products to their sale lineup.

If being acquired by NI gives Digilent the stability they need to continue operating, then I'm happy for them. I know of no other company besides Digilent that does what they do in the market/educational spaces they operate in. If NI kills Digilent in the end, well, I will have to stick a few hundred pins in an NI voodoo doll and hope for the best, I suppose. 

[jaxbird]

ISTR reading somewhere that it was felt that the target market (beginners) would complain if they felt they has been mislead by finding the amplitude of a 5MHz signal was too low. Personally I don't agree with that, since they will need to understand what equipment's spec does and doesn't mean.

I believe Digilent is owned by NI. In that case there might be a less charitable motive for restricting the device, especially the spectrum/network analyser: to "differentiate" between the academic and professional markets and prevent the low-end professional market from being eroded by the academic tools. I carefully avoided using the word "crippling". I'm sure no marketeer would ever think of such a thing, let alone enact it.

I certainly don't know whether or not it has been done in this case.

If you are designing automated production tests, I believe this device is very handy, it's cheap, can do advanced waveforms x2 (+/-5V/5-20MHz/14bits), scope x2 (+/-20V/100MSPS/14bits), digital IO x16 (3.3V/100 MSPS) and PSU (+/-5V) plus you can stack multiple of these with trigger IO x2 for more advanced tests. The API is easy and accessible, device is very cheap, pretty much an ideal QC test device IMO.

I wouldn't be surprised if Digilent suddenly experience a peak in sales

[tggzzz]

ISTR reading somewhere that it was felt that the target market (beginners) would complain if they felt they has been mislead by finding the amplitude of a 5MHz signal was too low. Personally I don't agree with that, since they will need to understand what equipment's spec does and doesn't mean.

I believe Digilent is owned by NI. In that case there might be a less charitable motive for restricting the device, especially the spectrum/network analyser: to "differentiate" between the academic and professional markets and prevent the low-end professional market from being eroded by the academic tools. I carefully avoided using the word "crippling". I'm sure no marketeer would ever think of such a thing, let alone enact it.

I certainly don't know whether or not it has been done in this case.

If you are designing automated production tests, I believe this device is very handy, it's cheap, can do advanced waveforms x2 (+/-5V/5-20MHz/14bits), scope x2 (+/-20V/100MSPS/14bits), digital IO x16 (3.3V/100 MSPS) and PSU (+/-5V) plus you can stack multiple of these with trigger IO x2 for more advanced tests. The API is easy and accessible, device is very cheap, pretty much an ideal QC test device IMO.

I wouldn't be surprised if Digilent suddenly experience a peak in sales

I agree with all that, and that is is a good device for the money - so much so that Digilent appears to be not doing itself justice. That was the reason for my extremely jaundiced question.

I'm content that others, e.g. syntax error, can give reasons why it was too jaundiced.

[jaxbird]

I agree with all that, and that is is a good device for the money - so much so that Digilent appears to be not doing itself justice. That was the reason for my extremely jaundiced question.

I'm content that others, e.g. syntax error, can give reasons why it was too jaundiced.

Not my intention to pick on you in any way, I merely wanted to point out the general usability of the device in a production environment. I was probably a bit careless with my quoting.

Anyway I'm confident anyone pro or am investing in this will find satisfaction in its performance. Personally I enjoy being able to measure bandwidth in my (lowish frequency) opamp circuits.

 

[tggzzz]

I agree with all that, and that is is a good device for the money - so much so that Digilent appears to be not doing itself justice. That was the reason for my extremely jaundiced question.

I'm content that others, e.g. syntax error, can give reasons why it was too jaundiced.

Not my intention to pick on you in any way, I merely wanted to point out the general usability of the device in a production environment. I was probably a bit careless with my quoting.

Anyway I'm confident anyone pro or am investing in this will find satisfaction in its performance. Personally I enjoy being able to measure bandwidth in my (lowish frequency) opamp circuits.

You have no reason to apologise. if anything I should be the one apologising for posting unverified (and probably incorrect) thoughts.

[Howardlong]

I wouldn't be surprised if Digilent suddenly experience a peak in sales

I bought one off the back of this and consider it very good. Sure not a replacement for a proper 'scope, but it's impressive. The analogue bandwidth is definitely underplayed.

One thing I haven't manged to do is to properly compensate a pair of probes with the BNC adapter, the range of he probes I was using doesn't seem to be enough.

That said, for a portable piece of kit it is excellent. If I am to compare it with the Bitscope I've had for several years, it makes the BitScope look positively neandertal, particularly with respect to the UI. With USB tools, the software is key, and they seem to have done a pretty reasonable job on that on the Analog Discovery.

[Howardlong]

Hmm, although I am not sure it's such good value at the GBP200 or so Farnell have it at currently. Microchip Direct still showing GBP138 but not in stock until mid January.

[jaxbird]

Hmm, although I am not sure it's such good value at the GBP200 or so Farnell have it at currently. Microchip Direct still showing GBP138 but not in stock until mid January.

I got mine from MicrochipDirect, not sure how they can sell it at 2/3 the cost advertised by Digilent without requiring you to be a student for discounts. But, I guess we are all students of life

Anyway, a pretty decent offer if you are ordering stuff from Microchip anyway with their fixed per order, express shipping rate.

But yeah, as Dave pointed out, not meant to replace a dedicated scope for general measurements and circuit debugging. While the software is very good (compared to e.g. the Red Pitaya) I'd quickly get very frustrated not having any knobs to turn. But overall as a multiple purpose arb func gen, programmable scope and logic analyzer/signal gen it does represent high value at reasonable cost for me.

[Frex]

Hello all,

I ordered mine on MicrochipDirect, and i receive it in start January.
I think too that the device seem very promising but in my own opinion
there is a concern with the connexions.
Of course, only 10cm of  wires cost about nothing and it's a convenient way
to use some fast connect prototype board (student use), but in more serious
lab, good and stable connexions are a requirement.
I have seen also that a small extension board with 4 BNC is available for sale,
but i think that could be better.

So, in order to get that, i will probably build one fit to lab needs.
Does anyone has already think that before me ?
Somebody would be interested ? (i could organize a PCB group buy
like i've already do in DIYaudio forum).
Is there users could help to define precisely "the interface board" requirements ?

For me, the minimum is :
-------------------------------------
4 x female BNC for Analog inputs
2 x toggle switchs for selecting single ended/differential
2 x female BNC for Analog Outputs
1 x HE10 connector for logic probe

Some parts will be added to protect I/O from overvoltage.
The AnalogDiscovery board and the interfec board will be mounted
together in small and low cost Hammond aluminium enclosure, to give
a better EMC shield and also a more pro tool and look and feel.
Does anyone can give me size of the internal board to choose an enclosure ?
Of course, this add-on won't cost more that the tool itself ! (i hope).
Some suggestions ?

Frex
 
 

[softone]

Hello all,

I ordered mine on MicrochipDirect, and i receive it in start January.
I think too that the device seem very promising but in my own opinion
there is a concern with the connexions.
Of course, only 10cm of  wires cost about nothing and it's a convenient way
to use some fast connect prototype board (student use), but in more serious
lab, good and stable connexions are a requirement.
I have seen also that a small extension board with 4 BNC is available for sale,
but i think that could be better.

So, in order to get that, i will probably build one fit to lab needs.
Does anyone has already think that before me ?
Somebody would be interested ? (i could organize a PCB group buy
like i've already do in DIYaudio forum).
Is there users could help to define precisely "the interface board" requirements ?

For me, the minimum is :
-------------------------------------
4 x female BNC for Analog inputs
2 x toggle switchs for selecting single ended/differential
2 x female BNC for Analog Outputs
1 x HE10 connector for logic probe

Some parts will be added to protect I/O from overvoltage.
The AnalogDiscovery board and the interfec board will be mounted
together in small and low cost Hammond aluminium enclosure, to give
a better EMC shield and also a more pro tool and look and feel.
Does anyone can give me size of the internal board to choose an enclosure ?
Of course, this add-on won't cost more that the tool itself ! (i hope).
Some suggestions ?

Frex
 

I have already sold the product that looks alike to your demand in Japan.
Please refer to the following page.
http://softone.a.la9.jp/english/FrontBox/FrontBox1.htm

Shoichi Yoshimoto

[jaxbird]

Hello all,

I ordered mine on MicrochipDirect, and i receive it in start January.
I think too that the device seem very promising but in my own opinion
there is a concern with the connexions.
Of course, only 10cm of  wires cost about nothing and it's a convenient way
to use some fast connect prototype board (student use), but in more serious
lab, good and stable connexions are a requirement.
I have seen also that a small extension board with 4 BNC is available for sale,
but i think that could be better.

So, in order to get that, i will probably build one fit to lab needs.
Does anyone has already think that before me ?
Somebody would be interested ? (i could organize a PCB group buy
like i've already do in DIYaudio forum).
Is there users could help to define precisely "the interface board" requirements ?

For me, the minimum is :
-------------------------------------
4 x female BNC for Analog inputs
2 x toggle switchs for selecting single ended/differential
2 x female BNC for Analog Outputs
1 x HE10 connector for logic probe

Some parts will be added to protect I/O from overvoltage.
The AnalogDiscovery board and the interfec board will be mounted
together in small and low cost Hammond aluminium enclosure, to give
a better EMC shield and also a more pro tool and look and feel.
Does anyone can give me size of the internal board to choose an enclosure ?
Of course, this add-on won't cost more that the tool itself ! (i hope).
Some suggestions ?

Frex
 

You are pretty much describing the bnc addon board, it's only ~$20 just buy it if you cannot work with the pinouts.

Just keep in mind, while high performance, this is not meant to replace a proper arb fuction gen, scope or digital analyzer in a proper lab environment.

[Engineer1]

Dave,

Thanks for this superb review, which I've just found. It was well worth waiting the two years (I'm the person who, back in the day, asked Digilent to send one to you!) I think the unit you have is probably a couple of board revisions back (although I don't know more detail on this - all my units are of the same vintage), so hopefully some of the niggles you saw are improved in the current version.

Anyway, I still use, and love, the AD. I'm glad you like it too.

If anyone wants to do any background reading, there's another thread with some links to useful information through it:

https://www.eevblog.com/forum/chat/new-analog-devicesdigilent-analog-discovery-board/msg170932/#msg170932

I remain evangelical about this magical box.

Cheers.

Steve.

[Floyo]

I am thinking of getting this device as a complement to my other equipment, and mostly for taking on the road to always have the "basics" at hand. And the provided software looks sufficient for this. Then there is the API to make custom measurements a possibility.

One thing I am wondering though, which the docs do not mention anything about, How hackable/reprogrammable would the on-chip FPGA be, especially with regards to real-time continuous streaming of data to the pc, for audio/SDR applications. Some quick guesses show that not all the analogue signals are going to be able to be transferred across the USB 2.0 bus at their full 125Msps.

The documentation http://www.digilentinc.com/Data/Products/ANALOG-DISCOVERY/Discovery_TRM_RevB_1.pdf  (page 33) shows transfer speeds of up to 20MB/s are possible and that an FPGA bit file is loaded at each boot up of the device.

Programming seems to be done with the following software http://www.digilentinc.com/Products/Detail.cfm?Prod=ADEPT2

The Xilinx XC6SLX16-1L FPGA has ~14K and 72K of ram (if I am reading the specs right).

So I would like to know everybody's thoughts on the usage of this device as a potential mixed signal FPGA dev board, Since I do not yet have enough experience with programmable logic to say what can and can't be done with this hardware.

[tggzzz]

I am thinking of getting this device as a complement to my other equipment, and mostly for taking on the road to always have the "basics" at hand. And the provided software looks sufficient for this. Then there is the API to make custom measurements a possibility.

...

So I would like to know everybody's thoughts on the usage of this device as a potential mixed signal FPGA dev board, Since I do not yet have enough experience with programmable logic to say what can and can't be done with this hardware.

If that is important to you, have a look at the Red Pitaya. More expensive, faster, based on a Zynq, ethernet, linux, less mature than the Analog Disovery - but designed to be hackable at the FPGA and software level. Examples are given and encouraged.

The Zynq is faster than the XC6, contains a dual-core ARM-A9 processor which can run linux on both cores, or can have an RTOS on one core and linux on the other. Host software runs on win/linux/osx.

Care with the using the extension connector; the GNDs aren't ideally placed (!) and only 3V3 interface families can be used. AFAIK the hardware is closed source.

http://uk.rs-online.com/web/c/?sra=oss&r=t&searchTerm=red+pitaya

[Floyo]

I already had a look at the Red Pitaya, but to me it has quite some issues. First being the gigE interface, this is going to be relatively hard to hook up in a cooperate network, probably requiring one to carry a router. Then there is the price, quite a bit higher than the 159$ student price for the Analog Discovery.

That said, the hardware choices otherwise look better than the Analog Discovery, having SMAs for the probes, the auxiliary ADC/DACs, and the onboard processing grunt needed to do some beefy processing. But, looking at their website and Mjlortons review, the software seems to be a definite weak spot compared the the Analog Discovery, it looks rather immature and simple in places.

All of this wouldn't be such a big deal if the documentation on how to get started coding for the whole Zynq soc would be easy and accessible, because I imagine programming the Zyng as a novice at FPGA's would be quite a challenge, And in this case I don't want to spend  too much time figuring out and coding the measuring equipment.

So unless more docs become available on a beginner level, or the software becomes better I personally see the Red Pitaya as a bit of a gamble, and would still like to see if the Analog Discovery could perform in a similar role, albeit at reduced performance, just to get started.

[tggzzz]

I already had a look at the Red Pitaya, but to me it has quite some issues. First being the gigE interface, this is going to be relatively hard to hook up in a cooperate network, probably requiring one to carry a router. Then there is the price, quite a bit higher than the 159$ student price for the Analog Discovery.

That said, the hardware choices otherwise look better than the Analog Discovery, having SMAs for the probes, the auxiliary ADC/DACs, and the onboard processing grunt needed to do some beefy processing. But, looking at their website and Mjlortons review, the software seems to be a definite weak spot compared the the Analog Discovery, it looks rather immature and simple in places.

I wouldn't disagree with any of that.

All of this wouldn't be such a big deal if the documentation on how to get started coding for the whole Zynq soc would be easy and accessible, because I imagine programming the Zyng as a novice at FPGA's would be quite a challenge, And in this case I don't want to spend  too much time figuring out and coding the measuring equipment.

So unless more docs become available on a beginner level, or the software becomes better I personally see the Red Pitaya as a bit of a gamble, and would still like to see if the Analog Discovery could perform in a similar role, albeit at reduced performance, just to get started.

Now there I differ: I would regard the Red Pitaya as less of a gamble w.r.t. fiddling with the contents of the FPGA. There are understandable Red Pitaya examples available, and the ecosystem is designed to enable it. Neither of those are true for the Analog Discovery.

Programming any FPGA is non-trivial at any level. The toolsets themselves are extremely complicated, as are the devices, as is the Verilog/VHDL language and coding styles, as is fitting the design into the FPGA within the timing constraints. Getting up to speed is vaguely equivalent to learning Java, learning an IDE, learning a enterprise frameworks, and deploying an application that meets the speed requirements. All do-able, all fun, but not to be underestimated.

OTOH, if you want instruments you can use, I too would choose the Analog Discovery (and have - hopefully mine materialises tomorrow ).

[jaxbird]

I had a look at some of the Red Pitaya code, and if it's correct it translated into that the device needs something like 1000ms to change frequency on the analog output, that makes applications like a network analyzer pretty useless, or you'll need a lot of patience for the scan to finish with e.g. a 500 point plot.

Can anyone confirm that my interpretation of the code is indeed a limitation of the Red Pitaya?

[tggzzz]

I had a look at some of the Red Pitaya code, and if it's correct it translated into that the device needs something like 1000ms to change frequency on the analog output, that makes applications like a network analyzer pretty useless, or you'll need a lot of patience for the scan to finish with e.g. a 500 point plot.

Can anyone confirm that my interpretation of the code is indeed a limitation of the Red Pitaya?

By "code" do you mean "C" or "VHDL"?

That effect (i.e. 1000ms) would be significant. What causes you to believe that figure might be right? What do you think the root cause might be?

[jaxbird]

I had a look at some of the Red Pitaya code, and if it's correct it translated into that the device needs something like 1000ms to change frequency on the analog output, that makes applications like a network analyzer pretty useless, or you'll need a lot of patience for the scan to finish with e.g. a 500 point plot.

Can anyone confirm that my interpretation of the code is indeed a limitation of the Red Pitaya?

By "code" do you mean "C" or "VHDL"?

That effect (i.e. 1000ms) would be significant. What causes you to believe that figure might be right? What do you think the root cause might be?

I'm referring to the C code, I have not looked seriously into the VHDL code.

And of worth notice they have not released any Network Analyzer app for the Red Pitaya yet.

[tggzzz]

I had a look at some of the Red Pitaya code, and if it's correct it translated into that the device needs something like 1000ms to change frequency on the analog output, that makes applications like a network analyzer pretty useless, or you'll need a lot of patience for the scan to finish with e.g. a 500 point plot.

Can anyone confirm that my interpretation of the code is indeed a limitation of the Red Pitaya?

By "code" do you mean "C" or "VHDL"?

That effect (i.e. 1000ms) would be significant. What causes you to believe that figure might be right? What do you think the root cause might be?

I'm referring to the C code, I have not looked seriously into the VHDL code.

And of worth notice they have not released any Network Analyzer app for the Red Pitaya yet.

I suspect they are hoping that third parties will do something and contribute it to the bazaar. One starting point mighe be http://blog.redpitaya.com/?p=501

But again I ask: that effect (i.e. 1000ms) would be significant. What causes you to believe that figure might be right? What do you think the root cause might be?

(You won't be crucified for making an honest mistake! I know: I've made enough in my time )