Build a TDC using FPGA

[rjulianomartins]

Hello,

We're implement a fast scanning LiDAR system in our Lab and we want to move into FPGA module for fast acquisition/processing.

But I have a few questions about performance in ToF calculations using FPGA:

(1) What requirements should I look to build a sub-ns resolution data acquistion ? Looking into the products page I saw that typical clocks on FPGA system are 40/80 MHz, does that mean that the minimum time that can be evaluate is 1/40Mhz or 1/80MHz ?

(2) I have quite a lot of experience in LabView programming but I've never used the FPGA module, how user-friendly it is ? Do I need  a specific formation in FPGA programming ?

(3) Can I use PCI FPGA cards to do it? in that case what should be a good price/performance card to look for ?

Thanks!

Renato

[James_McN]

Hi Renato,

Depending on the card it is possible to derive higher frequency clocks to improve resolution. 100-200MHz is typically possible - 250MHz is difficult but if the logic is simple enough it might be possible.

The bigger limitation will be the IO - what format does the signal come in at? You will need IO fast enough. The PCIe cards are generally limited to about 80MHz on digital (I think, from memory) and about 1MS/s on analog.

The flexRIO products have much faster front ends available but are generally only available in PXI or embedded controller format.

[UMASO]

Hi Renato, 

I have some experiences of NI FlexRIO products, and let me briefly answer your questions as far as I can.  

(1) In the pavilion at last NI-Week 2019, there was a Demo of fast scanning LiDAR.  The core of the Demo was the latest FlexRIO device; PCIe-5775.  Its sampling rate is 6.4GSps, 12bit, 1.25Vp-p.  It also has Digital input/output to synchronize with external accessories such as scanning devices.  

The such fast 6.4GSps data is handled 32 samples in parallel per FPGA clock.  If the sampling rate is around 200MHz (such as traditional NI 5761/5734), only a single sample is handled per FPGA clock.  It is more difficult to handle multiple samples per single clock cycle, and any devices with >200MHz sampling rate usually require to handle multiple samples per clock.  

(2) In my opinion, learning curve for LabVIEW FPGA is much shorter than other languages.  However, it still requires a certain amount of efforts.  In addition, there are not so many training materials/courses provided for LabVIEW FPGA.  As far as I know, below is the only material, which provides information deeper than what you can learn from regular LabVIEW FPGA training course.  

The NI LabVIEW High-Performance FPGA Developer's Guide

To maximize the performance of PCIe-5775 (and same device family such as 577x / 576x) for what you want to accomplish such as LiDAR, it may require multi-months to years experiences.  Contacting NI partner is another option, and if I were you, I would carefully choose the best FPGA-specialized NI partner and ask them to build the first prototype for the starting point for both your FPGA study and your future system.  

The caveat here is that many partners insist "yes, we can do LabVIEW FPGA", but that does not tell you HOW MUCH they are skilled in LabVIEW FPGA.  So, please somehow carefully scrutinize the best partner for the best outcome for your system development.  

(3) As aforementioned, PCIe-5775 would be the best for the latest NI's FPGA device family.  

Best regards,

[rjulianomartins]

Thanks for the responses.

The signal  is a output (300mV to 1V) 15 ns-width pulse  from a photodetector and initially I have to setup a simple ToF setup. Then I'll use it synchronized with external devices to form the image.

I was already looking for the PCIe-5775 as a solution.

Regarding the PCIe and PXIe connections; does it makes a difference ? Looking onto datasheets it seems to be pretty much the same.

We're acquiring also a new workstation and we're making sure to have the x8PCIe-3rd generation to be fully compattible with the card.

best,

Renato

[UMASO]

It's great that you have already considered PCIe-5775.  

The main advantage of PXIe form factor is its capability of synchronization with other modules and data transfer between modules and PXIe controller. 

Although 5775 has its MGTs available for fast data movement with other modules for both PCIe and PXIe, P2P inside PXIe chassis is still powerful and easy to use only available for PXIe.  Also, if high-throughput data streaming is required during your system development or for the final product, PXIe provides better solution for it.  For example, there are high throughput storage available for PXIe, such as HDD-8261, etc.  Also, recent line-up of PXIe controller such as PXIe-8861 supports NVMe interface to SSD.