08:06 <bouliiii> Hello, I am trying to solve a strange issue with my icoboard. I am not sure this is a Yosys question even though I am using the icestorm toolchain. Basically, some pins on the board seem defective and output nothings (signal stays low). Repro case is trivial and consists just in trying to output anything on one given pin. Some work fine and some

08:06 <bouliiii> not at all. Would you get a suggestion to tackle this issue? I am a bit lost and unsure if I should buy a new board.

08:10 <daveshah> Is there anything special about the pins that aren't working?

08:10 <daveshah> All in one bank for example?

08:10 <bouliiii> Is there a resource for that?

08:11 <bouliiii> I was looking yesterday at the db files in icebox but I am not sure this is the right file.

08:12 <daveshah> Either the schematic or if that isn't clear, then the pinout xls from Lattice

08:15 <bouliiii> OK. I am going to. Also, for any of those faulty pins, I don't see any trace on the back or the front of the board. Does it mean that they go to the same inner layer?

08:17 <daveshah> Yes, although it might not be the same layer each time, I don't know the stackup

08:33 <bouliiii> I looked at it quickly (but not exhaustively). There is no clear pattern unfortunately, The pins I am looking at are mostly bank 0 and bank 1. For each bank, I can find _both_ faulty ones and working ones.

08:35 <daveshah> https://github.com/cliffordwolf/icotools/tree/master/examples/pinout

08:35 <tpb> Title: icotools/examples/pinout at master · cliffordwolf/icotools · GitHub (at github.com)

08:35 <daveshah> Should print the pin name via a 9600 baud UART to each pin

08:35 <daveshah> Probably a good test, if that fails then the board is pretty much definitely trash

08:37 <daveshah> I expect Trenz will be good about returns though, you should just be able to send it back and not need to buy another

08:46 <bouliiii> OK. With the logic analyser, I see the correct strings for the OK pins and still flatlines for the faulty ones

08:49 <daveshah> My guess would be a badly soldered BGA

08:52 <bouliiii> OK. I am going to contact Trenz and get a new one. Thanks *a lot* for the help!

12:33 <ZipCPU> bouliiii: Beware, there were issues at one point with the IcoBoard's pinouts not quite being accurate. You might want to check for outputs on the wrong pins as well.

12:36 <cr1901_modern> daveshah: Could you please review my PR when you get the chance today?

12:38 <daveshah> cr1901_modern: LGTM, merged

12:39 <cr1901_modern> tyvm for the quick response :P. Now I can merge back into my branch and keep going.

12:39 <cr1901_modern> (If I can stay awake lol)

12:41 <cr1901_modern> >I think splitting them up will be a *net* win overall

12:41 <cr1901_modern> That was a missed opportunity for a pun

12:41 <daveshah> lol

12:59 <bouliiii> ZipCPU: we tried the broadcast approach (using latest icestorm and latest db) and I checked them one by one. Still broken.

13:01 <bouliiii> ZipCPU: Also the chip is running hot. I am wondering if this is really expected. I am going to buy a new one anyway. Let's hope I did destroy this one by mistake. At least, I'd like to know how and not repeat it... :-(

13:14 <bouliiii> ZipCPU: actually, is there a possibility that the icebox db is incorrect and the pins are incorrectly assigned ? I have a gamma icoboard but it looks like the pin assignment only depends on the package (ice40-hx8k-ct256). Broken pin assignment would mean that all boards using ice40-hx8k-ct256 would suffer from this issue. I guess this cannot be

13:14 <bouliiii> true.

13:15 <daveshah> Given that the icoboard was mostly tested with the open source tools, that seems very unlikely

13:16 <ZipCPU> I'm going to agree with daveshah here. I have an icoboard that I then tested with open source tools.

13:18 <bouliiii> ZipCpu: for example, C13 and A15 output nothing for me. Whatever I tried. One PMOD (B11, B10...) is also always low. If you have any of this working, I guess the board is broken. I bought it long time ago though and only got the change to use it recently.

13:19 <bouliiii> The board itself is amazing. It is such a pleasure to work with.

14:25 <ZipCPU> bouliiii: Those two pins are not pins I have worked with

14:26 <ZipCPU> I do know that claire had at one time built a design around outputting a repeating (but different) character out every single pin using a serial port. That allowed him to probe and double check all of the pinouts on the board after I had initially found problems with them

14:27 <ZipCPU> I'm sure if you look hard enough through the demo designs you might find that one

14:27 <ZipCPU> Most of my work was with PMods, hence why I didn't check those values

14:43 <bouliiii> ZipCPU: this sample has been mentioned by daveshah. I tried it and all OK pins output the right string and all KO pins output _nothing_ (that is not even an incorrect string, just nothing). I just bought a new board. The mystery will be soon solved (or not... if the new board shows the same issue :D)

14:44 <ZipCPU> If that example had problems, then my guess is that there's a hardware issue with your board

14:44 <ZipCPU> I look forward to hearing how well the new board meets your requirements ;)