cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
Visitor
Visitor
548 Views
Registered: ‎09-03-2019

Arbitrary Function Generator

Hi all!The FPGA creates pulses that are allocated to time slots (corresponding to 1/4Ghz spacing or smaller). The two outputs are syncronized.The FPGA creates pulses that are allocated to time slots (corresponding to 1/4Ghz spacing or smaller). The two outputs are syncronized.

I'm real newbie when it comes to FPGAs! I wonder what kind of FPGA (would be this FPGA just fine: EK-U1-VCU118-G) to take and if an FPGA is suitable for my exercise at all.

Basically, I would like to create a pulse train, where the pulses are put in time slots of width 1/4Ghz or smaller. Both outputs of the FPGA should be syncronized with each other. Please see the figure for details. The first question goes like if an FPGA is able to create such a fast signal at all? I'm asking, because isn't the clock speed in the Mhz regime?

Another question would be about how to connect the FPGA to the intensity modulator and PC.

Thank you for any answers!

- Chris

 

 

 

 

0 Kudos
12 Replies
Highlighted
535 Views
Registered: ‎06-21-2017

Re: Arbitrary Function Generator

Let's start with a basic question.  1/4 GHz is 250MHz.  Is this what you mean or do you want a pulse of 1/4 nS?

0 Kudos
Highlighted
Visitor
Visitor
527 Views
Registered: ‎09-03-2019

Re: Arbitrary Function Generator

Hi Bruce, thanks for the quick answer!

Sorry, the time slot should have a width of 1/(4Ghz), that is 0.25ns or smaller.

0 Kudos
Highlighted
Teacher
Teacher
501 Views
Registered: ‎07-09-2009

Re: Arbitrary Function Generator

your not going to get an FPGA output to switch at sub ns rates ,

can you talk to your lecturer ?

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
Highlighted
Visitor
Visitor
494 Views
Registered: ‎09-03-2019

Re: Arbitrary Function Generator

Thank you for the answer, John!

What about the Tranceivers? Is there a way to get an output signal from there, since they operate at high speeds (28Gbps)? May I ask, what else could I do to get a high frequency pulse signal?

 

0 Kudos
Highlighted
Scholar
Scholar
492 Views
Registered: ‎09-16-2009

Re: Arbitrary Function Generator


@drjohnsmith wrote:

your not going to get an FPGA output to switch at sub ns rates ,


General purpose FPGA outputs no.  MGTs operate in the 10GHz+ ranges, however.

Your design requirements are NOT for the faint of heart, however - using the MGTs in such a fashion will be quite a bit outside the normal use cases.  Your specification of two "synchronous" outputs needs much more details - just PCB delays at those rates will have heavy skew impact.

This would be a fun (learning!) exercise - even for an experienced FPGA user.

Sounds like an intersting project - good luck!

Mark

 

Highlighted
Teacher
Teacher
464 Views
Registered: ‎07-09-2009

Re: Arbitrary Function Generator

Yes the GTx are fast
but they are not controllable like the normal IO
they are designed to accept parallel input, encode it and sending to serial, adding padding as needed. .
in this case, the encoding and padding is going to cripple your signal,

It might be possible to turn off all the encoding and padding, and use the GTx raw, but thats into the world of un supported ,

I'd suggest that even an experianced engineer would find that hard.

Also you need to get a good book on high speed design
at that sort of speed, the PCB is the biggest and most expensive part of the design,

Something like this I'd be wanting to use tools like hyper lynx to help design the tracking,

Possible yes, but a very long road to walk .

Good luck and I look forward to reading your paper when its published.


<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
Highlighted
Scholar
Scholar
453 Views
Registered: ‎09-16-2009

Re: Arbitrary Function Generator


@drjohnsmith wrote:

It might be possible to turn off all the encoding and padding, and use the GTx raw, but thats into the world of un supported ,


I wouldn't venture to call this mode of operating a GTx as "unsupported".  The TX direction is easier.  I don't think it would be all that much trouble to set the GtX up to bypass any line encoding, buffering, and idle insertions, and simply treat the TX GTx as a fast Parallel to Serial Converter.  (Chris' design must be able to produce 16, 32, or perhaps even 64 samples at a time).

The tricky bit is minimizing (and even just characterizing) the skew between two GTx transceivers.  That's the (very) tricky part.   (Along with the high speed board design problems noted).  Most higher protocols which bond multiple GTx together, are done in such a way to be (mostly) skew agnostic - or at least tolerant to higher skews - by the line protocol.  Chris, I very much think your problem may require some sort of at least initial closed-loop training of the independent delays.  It may very likely require dynamic training which tracks variant delays because of PVT variations.

Again, sounds like an interesting problem!

Regards,

Mark

Highlighted
406 Views
Registered: ‎01-08-2012

Re: Arbitrary Function Generator

Quite apart from the skew calibration problem, you also seem to have an issue with voltage levels.

The transceivers will output CML - a 100 ohm differential signal with a configurable differential amplitude of some hundreds of mV.  (It also has a configurable edge shape - see TXPOSTCURSOR and TXPRECURSOR attributes.)  There is a common mode DC offset.

You seem to want a 5V signal on a 50 ohm unbalanced cable + connector.  You will need an amplifier somewhere to achieve that.

0 Kudos
Highlighted
Teacher
Teacher
394 Views
Registered: ‎07-09-2009

Re: Arbitrary Function Generator

Im guessing this is a research project for a university,

  Im just wondering if you are in touch with CERN,

I know they have similar 'bits'.

 

 

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
Highlighted
Visitor
Visitor
313 Views
Registered: ‎09-03-2019

Re: Arbitrary Function Generator

Hi all!

 

Thanks for the many answers. Yes, the project is a kind of university research labwork.

Pulse pattern generators are existing with speeds of 32Gbit/s (from Anritsu for example), but they really cost a lot. I through, that I found a cheap solution with FPGAs and could get GHz speed, too. I'm an optics/photonics guy. That's the reason why I wanted to get a simple solution, since I don't know much about electronics. However, it looks like achieving Ghz signals from FPGAs is not straight forward, so I may just stick with MHz pulses at the clock rate of the FPGA.

0 Kudos
Highlighted
Teacher
Teacher
307 Views
Registered: ‎07-09-2009

Re: Arbitrary Function Generator

Yep , those arbs are astronomical priced,

   I use a couple at a ocmpany I'm doing work for, and they go faster than that,

Im guessing you only need an on / off pulse.  The optics I have been involved with is laser diodes, that need fast pulses, but at quiet high voltage relative to what the FPGA could do, like 30 volts, 2ns rise / fall time,  ( that took overlaping GAN fets, that turned on / off in sequence.)

 

An idea, if you want a small pulse, could be to use the phase shift of the MMCM , may be.

   My thought is to use two outputs of the FPGA , one slightly phase offset to the other.

      then externaly OR them together , using something like this

https://docs-emea.rs-online.com/webdocs/1478/0900766b81478d93.pdf

 

 

 

 

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
Visitor
Visitor
296 Views
Registered: ‎09-03-2019

Re: Arbitrary Function Generator

That's right. I just need on/off pulses basically.

Well, I need a small pulse and they should be spaced by 1/(4+Ghz). That's a cool idea, but the speed would increase (only) by a factor of number of outputs and the SMA outputs are limited. However, the approach sounds simple - thanks!

0 Kudos