Please help me to select an USB3 + FPGA board for my logic analyzer project

[DC1MC]

Dear digital design gurus, I'm had enough of the miserable and/or expensive logic analyzers available in sub 500EUR range, regretfully but are are crappy and/or obsolete designs.
 I want to have a sane board with an USB 3 interface from Cypress and a reasonably beefy XILINX or Altera FPGA with  at least 64MB DDR2 or preferable DDR3, if external  memory or more external memory can be added later is even better.

 Budget is maximum 400-450EUR (breakout board can be sold separately).

 Features in order of importance:

 - The USB3 interface has to be Cypress FX3, not the poor brother FX3S or, God forbid, some kind of pumped-up USB to serial interface.
 - The FPGA to USB controller should have a solid 32 bit bus, capable of maximum data transfer, I still see boards with FPGA capable to transfer Gigabits per pin !!! with pathetic PC interfaces, the POS with old FTDI serial to USB converter has the first place on special Olympics of bad design.
 - The FPGA should be minimum Spartan 6 series, LX45 preferred, but not mandatory.
 - The external RAM should be DDR2 or better DDR3, the MCB supports it and the price difference is not that big, and we are in 2018 not 2008, the minimum possible size 64MBit, better 256MBi, best 4GBit, doesn't matter if isn't soldered when sold, it has to have the routing and the pads.
 - The GPIO have to be sanely routed to the break-out connectors, I've seen a promising board with really bad routed gpios and even a mention that is "not suitable for high speed designs", then why the frak are you putting this fast FPGA on it, just to route the outputs miserably 
 - The board has to have all the power supplies on-board(single power supply), 5V perfect, any other value OK.
 - Due to volatility, the full design files should be available, even at a cost.
 - And finally, a high-speed breakout board will be nice, until I design a better analog FE for the analyzer, I want to also have some differential LVDS channels later, the single ended stuff doesn't cut it anymore.

 If you have encountered some of these boards, kindly please let me know.

 Best regards,
 DC1MC
 

[hli]

So you want to spend considerable time to implement a logic analyzer in a FPGA, and then want to spend even more time implementing the proper PC interface GUI to use it? What is it that you don't like at the existing logic analyzers? What are your targets in respect to sample rate, channels, sample depth, triggering and decoding capabilities?
I would encourage you to look at e.g. the Digital Discovery - it comes with a Spartan 6, 2GBit of DDR2 memory, and its properly routed to allow sample rates of up to 800MHz. And it comes with quite some capabilities in terms of triggering an decoding (e.g. complex triggers, or triggering on decoded protocols).

[ace1903]

Recently I have looked qmtech board named:
Cypress CYUSB3014 USB 3.0 Development Board with Xilinx Artix7 FPGA XC7A35T
Price is around 100eur on ebay but usb part is not so powerful(has 256mb ddr3).
Couldn't find mentioned github project.
If demo project and schematics is available, then for the price is good deal.
Would like to hear opinions for this board.

[asmi]

I use FT601 USB 3 to FIFO bridge IC in my board with Artix-7 FPGA and 256 MB of DDR3L x16 memory, and in my tests I was getting about 320-330 MBytes/s of "clean" bandwidth - meaning that was a real speed I was getting my data from FTDI's driver in my application, not "theoretical" bandwidth. It has 32bit bus to FPGA which runs up to 100 MHz. This IC is very easy to integrate and FIFO protocol is simple and flexible (IC can be configured for 1 IN, 1 OUT, 1 IN/1 OUT, 2 IN/2 OUT, or 4 IN/4 OUT bulk pipes), and vendor provides API libraries for C/C++/.NET which are very easy to use. IC also have 2 GPIO pins which could be used for out-of-band communication (including I2C via bit-banging) if you don't want to use dedicated USB OUT pipe for that.
FTDI sells breakout boards with this IC with hi-speed connectors compatible with both Altera/Intel and Xilinx devboards, so if you don't want to "roll your own" - you can just purchase a breakout and attach it to whatever board you have.

[DC1MC]

@hli:

"So you want to spend considerable time to implement a logic analyzer in a FPGA,"
Yes !!!

"and then want to spend even more time implementing the proper PC interface GUI to use it?"
No, there are reasonable existing programs for this, they just need to be fed with proper data.

"What is it that you don't like at the existing logic analyzers?"
The one that I can afford are more or less garbage, the one that are working properly, I can't afford .

"What are your targets in respect to sample rate, channels, sample depth, triggering and decoding capabilities?"
At least 100MHz in state mode, 500MHz in sample mode, 36 single ended channels, 18 LVDS channels because it's the $current year, sane RLE implementation, sampling gating (bonus gating on pattern), trigger on edge, trigger on pattern, adjustable number of  trigger occurrences, continuous streaming to the PC at (as close as possible to USB3) max speed using the on-device external memory for buffering of bursts.
Really sweet cherries on the top: simultaneous sampling/state acquisition (that is synchronized sampling in state mode using the external clock as synchronizer for the sampling clock), advanced protocol decoding(HDMI, USB, Ethernet), pattern generator synchronized with the internal or external clock.
 
"I would encourage you to look at e.g. the Digital Discovery - it comes with a Spartan 6, 2GBit of DDR2 memory, and its properly routed to allow sample rates of up to 800MHz. And it comes with quite some capabilities in terms of triggering an decoding (e.g. complex triggers, or triggering on decoded protocols)."
I was encouraged to look on it some while ago when I was looking for an LA, I've looked carefully on it, at it's price and capabilities, and IMHO it's a rotating overpriced turd, that was created specially to gouge students and universities of a lot of cash (the prices in Germany are outrageous) for mandatory projects, the obsolete FPGA platform, the miserable architecture and design, not to mention the useless software, convinced me that along with the brother turd, the Analog discovery, it's a project to teach the students to discover that 99% of everything it's crap (Sturgeon was an optimist) and they just have dropped around 500EUR for a piece of it. At least they can proudly show the 56EUR "high-speed" adapter, it could be that there is written with very small letters: "suckeeers" . Also not in a 1000yrs this thing will sample reliable anything close to 800MHz, ever. And the linked review was kind of lukewarm, really the guy was mentioning sniffing I2C and UARTS ?!?! with this thing, 200+ EUR for it ?!?!, well at lest I've seen that they dropped the price for the "high-speed" adapter, just 46EUR 
===============

@asmi Good to know, maybe Cypress will drop the prices for FX3, personally fact that I can't control the USB personality and firmware and make it a USB class device, the need of special drivers and libraries instead of using the OS optimized class drivers and no libraries and DLLs, turned me from buying an otherwise promising board. I may be biased here, because I know very well the FX3, but even when trying to be as impartial as possible, I don't see any kind of advantage of this chip, maybe it's cheaper. Also FTDI is on my black list for the USB-serial fake chips disaster, they even had the gale to submit a Linux patch to destroy the chips as well !!!
===============

@ace1903 Actually I've found the Artix board of the same seller: https://www.ebay.com/itm/Cypress-CYUSB3014-USB-3-0-Development-Board-with-Xilinx-Artix7-FPGA-XC7A35T/173113636330 and if by the end of the week I don't get a better suggestion I may give it a try.
Thanks anyway for the suggestion, this was welcome and what I've asked for.

 Best regards and thanks for all the advices, they are welcome but some may be a bit harsh criticized, all the critics are for the object and NEVER to the person.

 Cheers,
 DC1MC
 


 

[asmi]

@asmi Good to know, maybe Cypress will drop the prices for FX3, personally fact that I can't control the USB personality and firmware and make it a USB class device, the need of special drivers and libraries instead of using the OS optimized class drivers and no libraries and DLLs, turned me from buying an otherwise promising board.

Drivers are part of Windows and don't need to be installed separately - OS will do it automatically (at least on my Win10 Pro).

I may be biased here, because I know very well the FX3, but even when trying to be as impartial as possible, I don't see any kind of advantage of this chip, maybe it's cheaper.

It's 4 times cheaper

Also FTDI is on my black list for the USB-serial fake chips disaster, they even had the gale to submit a Linux patch to destroy the chips as well !!!

As someone who also was on receiving end of piracy I totally understand their motivation, even if I don't approve their actions.
At the end of the day, I'm just a hobbyist, so whatever I do will not matter a single bit, so I'm taking advantage from very good solution.

[KrudyZ]

Why don't you just get a used Tektronix or Aglient logic analyzer.
These have dropped in value to the point where they can't give them away.
In addition, there is no way that your home brew kit would get anywhere near the signal integrity of the commercial  probes.
Of course if you want to do this just for fun, then there is little point in getting something that just works...

[thm_w]

"I would encourage you to look at e.g. the Digital Discovery - it comes with a Spartan 6, 2GBit of DDR2 memory, and its properly routed to allow sample rates of up to 800MHz. And it comes with quite some capabilities in terms of triggering an decoding (e.g. complex triggers, or triggering on decoded protocols)."
I was encouraged to look on it some while ago when I was looking for an LA, I've looked carefully on it, at it's price and capabilities, and IMHO it's a rotating overpriced turd, that was created specially to gouge students and universities of a lot of cash (the prices in Germany are outrageous) for mandatory projects, the obsolete FPGA platform, the miserable architecture and design, not to mention the useless software, convinced me that along with the brother turd, the Analog discovery, it's a project to teach the students to discover that 99% of everything it's crap (Sturgeon was an optimist) and they just have dropped around 500EUR for a piece of it. At least they can proudly show the 56EUR "high-speed" adapter, it could be that there is written with very small letters: "suckeeers" . Also not in a 1000yrs this thing will sample reliable anything close to 800MHz, ever. And the linked review was kind of lukewarm, really the guy was mentioning sniffing I2C and UARTS ?!?! with this thing, 200+ EUR for it ?!?!, well at lest I've seen that they dropped the price for the "high-speed" adapter, just 46EUR 
===============

Student price for analog discovery is 260 and digital discovery is $200 USD for everyone. It used to be much cheaper in the past, I think 199 or less.
Shipping is $43 to Germany.

The high speed adapter though is certainly overpriced.

[iMo]

800MHz sampling rate with a 46Euro adapter is an illusion, imho. For such an sampling rate you need special construction techniques, impedance matching, sub-ns comparators at the input, etc. Such a probe costs x-times more than the digilent digital discovery analyzer itself.
Btw - another similar and simpler design -http://www.dreamsourcelab.com/hardware.html - also lacking a "high-speed" probe, I guess..

[DC1MC]

@thm_w Let's not confuse American prices with German prices, the Deutsche Adler guards proudly the borders, and (via Zoll interception) extracts VAT and sometimes custom tax out of EVERYTHING. But even on the American prices, it's overpriced.

@imo The DS it's the only design that has a grain of sanity and resonable, too bad that it's USB2 and the FPGA is dated, I've found nice cheap FPGA boards with a newer Artix, large DDR3  and USB3 via Cypress FX, my digital designer friend had a quick look at the routing and says that if the manufacturer didn't  screw anything major we can push the sampling rate up to 600MHz with the SERDES pins, and stream via USB3 close to maximum possible speed.
The Analog Front End it's a real important issue, people kind of neglect it, and I'll have to consider it carefully, eventually dissecting some professional LA pods.


 Cheers,
 DC1MC

[mac.6]

@ace1903 Actually I've found the Artix board of the same seller: https://www.ebay.com/itm/Cypress-CYUSB3014-USB-3-0-Development-Board-with-Xilinx-Artix7-FPGA-XC7A35T/173113636330 and if by the end of the week I don't get a better suggestion I may give it a try.
Thanks anyway for the suggestion, this was welcome and what I've asked for.

Just got the artix 7+SDRAM board from this seller (with simple vga/usb daughterboard). Seems fine (just compiled some example) and shipping was really fast from china (no custom).

[rhb]

I happen to be pursuing the same goal, but chose to swat flies with a hammer and use a Zybo Z7-20 board, mostly for the 1 GB DRAM.  That choice was reinforced by the widespread use of the Zynq series for DSOs like the Instek GDS-2000E line.

I find the notion that the Digilent Discovery is an LA rather astonishing.  It *could* be if it had fast comparators to buffer the inputs.

Very few, if any,  logic analyzers will collect long traces on a fast, bursty bus.

I'm still digging through Zynq documentation trying to sort out how fast the GPIO will really go and what the problem areas are.

[DC1MC]

Well, SERDES pins are going to 1.2Gbit/s, but the talent is doing the RLE and clock gating and pushing the stuff in memory organized as FIFO, not to mention the whole triggering wizardry.

[senso]

Do like of of those Dangerous Prototypes gizmos and use the DRAM as inputs and sample directly into it.

Slap some DDR4 at some 3MT's and pray it all doesn't end being a mess of RFI.

Regarding VAT, welcome to europe.

[rhb]

Well, SERDES pins are going to 1.2Gbit/s, but the talent is doing the RLE and clock gating and pushing the stuff in memory organized as FIFO, not to mention the whole triggering wizardry.

As far as I'm concerned, the first task is to implement sampling from input to DRAM on the FPGA.  Likely the easy way to get deep logging capability is to use Linux to offload data files from a DRAM buffer via the  LAN and USB ports.  With the FPGA writing the data to memory and the ARM cores taking over from there.

1 channel of 8 bits  at 1 GS/S
1-2 channels of 8 bits at 500 MS/S
16 bits of comparator state at 
sampling the ADC and LA  at random intervals for compressive sensing

Storing the sample intervals doubles the data volume, but with 5-10x gains in effective sample rate/resolution, it's probably still a good deal  A lot depends upon the FP capability of the ARM cores, but a DSO application might get 3-5x gains in memory depth for free.

Digilent have one of those "we can not deliver to PO box" rules not caught by their website.  But they claim my order is "awaiting shipment".   So hopefully in a few days I'll be able to actually work on this rather than just read manuals.  And read manuals.  And read manuals.

[MegaVolt]

Cypress CYUSB3014 USB 3.0 Development Board with Xilinx Artix7 FPGA XC7A35T
Couldn't find mentioned github project.

https://github.com/AndreZheng/QM_XC7A35T_DDR3
https://github.com/AndreZheng/QM_CYUSB3014_USB3.0

[casinada]