MatriX Keys go wild

bfe9cd11c7ddee62

Now when days get warmer, my matrix is getting into trouble.
When I press a key, other LEDs are activated as well. Sometimes 4 LEDs, sometimes a lot more. I guess a defect IC? Does someone know, what could be the root cause?

17703227d9536da1

Hi Robert, very strange indeed!

Maybe you could check again in test mode to see what keys trigger on the button presses?

Ideas to check:

what PSU is used? Noisy PSUs could cause issues
is it button triggering or corrupt LED data?
are flux residues present around the shift register ICs? Could be worth to reflow the pins?
are BAT54S diodes installed on the boards? This might help to keep the signal 0V more consistent across neo_ boards
are all cables still okay? Sometimes a “Wackelkontakt” can cause strange behaviour.

One other tip we have seen is to wire between the clock signals of the 165 shift registers directly on the chips, especially if the lower boards (closer to the extra row) become more corrupted. This appeared to reduce signal ringing.

Best regards,
Andy

048668344ed25ef3

I didn’t place few 10k resistors and had similar issue.
Maybe there are some resistor soldered only from one side or LED legs or IC legs unsoldered. It’s getting contact, but heat expansion might cause problems.
I had one IC side unsoldered. Soldered corners and then drag soldered only one side

bfe9cd11c7ddee62

Thanks Andy and Adrian,
I have checked, as you suggested but still not found the culprit.

Measuring the serial bus J89 on axes PCB with a scope, I see the signal with lots of overshoots. In the moment, when the bus is touched by any wire, everythings works correctly and the button LED (<->) starts lighting. Well, I could permanently attach this antenna wire, but that is not solving the root cause…

I think, there is an issue with the bus termination or the driver. It would be good to have the circuit diagram about the bus.

17703227d9536da1

Hi Robert, any further insight from the questions I posed before?

The other user fixed his issue by wiring the clock signals of the shift registers together. It seems to only happen occasionally though.

bfe9cd11c7ddee62

Hi Andy,
Yes, I’ve checked your 5 points, except the BAT54S, which I had to order.
I guess, wiring the pins 2 of both 1650 has the same effect as my antenna wire. Having this wire in places fixes the issue, possibly because of the additional small capacitance that smoothes the transients a bit. I will insert the diodes later and then test without the wire.

I am wondering why you made the floating grounds with the diodes and did not connect all grounds directly to the chassis?

17703227d9536da1

Hi Robert,

Another idea is to try RC termination on the end of the chain to reduce reflections. This has been used in MIDIbox before. You could use the empty J89A socket here.

For wiring the 165 parts together, you could also argue that the impedance on the clock line is reduced with additional wires.

For not connecting 0V between boards, the idea is to have a common “star” point that is at the power entry, and to use thicker-guage wires with connectors designed to handle more current. Otherwise significant current would flow through the thin ribbon wires and be transmitted through all boards. I am not sure what is best and I am not sure if it really makes a difference, but for instance it is not recommended to use “flying bussboards” for Eurorack modular. One could also simply bridge the SOT-23 pads with solder to re-make the connection.

Are you using the Leicke power supply or something else?

Best,
Andy

bfe9cd11c7ddee62

Hi Andy,
Yes, I am using the recommended Leicke.
Could you also post the circuit and dimensions of the RC termination of the MIDIbox? I think, this is the best idea.

17703227d9536da1

I found a few old posts here:
http://midibox.org/forums/topic/14439-large-shift-register-chains-with-mios32/
http://midibox.org/forums/blogs/entry/33-signal-integrity-for-long-srio-chains-mios32/
http://midibox.org/forums/topic/18510-rc-termination/
http://www.ucapps.de/midibox_seq/mbseq_v4_dout.pdf

The last one has a suggested circuit, so maybe a 100R resistor plus a capacitor of 100pF (maybe up to 100nF max??) is worth a try. You could even build that into a micromatch connector or ribbon?

Hope it could help!

Best,
Andy

bfe9cd11c7ddee62

Hi Andy,
thanks for your hints, this has solved the issues! I only needed to terminate one bus, then the keys worked all the time.
Robert

17703227d9536da1

Great solution! Glad you could translate this to the axes board, but of course we’re using the same pinheader throughout.

We could also add such termination to a future PCB revision in case it’s needed.

Enjoy!