How to get higher speed UPDI more reliable?

[mechatroniks-git]

with 2.3.2, I
57.6k working all the time.
230.4k rarely works, I think I got it once or twice
460.8K, never.

I'm using LeoNerd's UDPI programmer, the device (CP2105) claims to support up to 921.6K. It has built in 4.7K resistor. My PCB I've installed a 0-ohm. Running at 5V. 6" Jumper to pcb from programmer.

Otherwise, how can I get 115.2K working?

[SpenceKonde]

I'm unfamiliar withthe design of his device - If you have an oscilloscope that you can generate a trace of an unsuccessful programming attempt with I can tell you what's wrong electrically.

I dropped support for 115.2 because I wasn't aware of any devices that needed that, but didn't work at other settings. You can reenable it by editing programmers.txt - copy-paste the 2230400 baud version, change the name

Are you using the 12v pulse? we probably do things that are interfering with that.

I can't really comment further without seeing a schematic. The 4.7k resistor is markedly inferior in terms of robustness of the connection compared to a 470-2.2k resistor small signal schottky diode

I think me and LeoNerd need to get in contact so we can sort out whether there is a way that we can make sure SerialUPDI is compatible with it. This needs to happen FAST because there's already prototype hardware that will demand certain behavior for RTS and DTR by SerialUPDI being developed by BOTH primary developers of SerialUPDI. And once both of us are shipping hardware, we're not going to break compatibility with our products so that our competitors products work ;-)
Are you

[mechatroniks-git]

Board schematic is here

Leonerd schematic is here

I've replaced the on-board 4.7K (R3) on the board with a zero-ohm, as the 4.7K exists inside the LeoNerd. I'm not using the 12V pulse.

Tried to grab a waveform, any tips on how to get a good grab?

[SpenceKonde]

When we spoke yesterday, it was revealed that I guess there are a few versions of that board from LeoNerd as he's been trying to find the ideal thing to put between TX and RX. I hope I convinced him of the Right Way of a 1k resistor in series with a schottky diode..... I cant see a damned thing in that trace... I use a Siglent too and I would suggest three things: 1. instead of using a camera, stick a USB thumb drive tinto one of those USB ports and take screencaps that way. 2.use "single" trigger mode. You want to see a trace where the lowest point shows two different voltages (corresponding to whether the low was caused by programmer or target driving the line low) - that's the most useful. Or if you just see a few bursts of activity right at the start then nothing and only one voltage for the valleys, tbat imoplies that the target isn't answering. 3. If you have the "decode" module, you can set it toi decode UART, the baud rate you're using, even parity and 2 stop bits.

…

On Mon, May 24, 2021 at 4:03 PM Mechatroniks ***@***.***> wrote: Schematic is here: https://github.com/tetrabiodistributed/PAPRA-PCB/blob/main/pcb/PaprControlPCB/PCB%20Design%20Files/Papra%20v0.4%20Schematic.pdf I've replaced the on-board 4.7K (R3) with a zero-ohm, as the 4.7K exists inside the LeoNerd. I'm not using the 12V pulse. Tried to grab a waveform, any tips on how to get a good grab? [image: image] <https://user-images.githubusercontent.com/57600622/119401166-fd3e5d00-bc8f-11eb-8979-d494c9ac7ff7.png> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#436 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABTXEW74NFYMGHPBPYLLDZLTPKWIDANCNFSM45H2DZ4Q> .

--

____________ Spence Konde Azzy’S Electronics New products! Check them out at tindie.com/stores/DrAzzy GitHub: github.com/SpenceKonde ATTinyCore <https://github.com/SpenceKonde/ATTinyCore>: Arduino support for all pre-2016 tinyAVR with >2k flash! megaTinyCore <https://github.com/SpenceKonde/megaTinyCore>: Arduino support for all post-2016 tinyAVR parts! DxCore <https://github.com/SpenceKonde/DxCore>: Arduino support for the AVR Dx-series parts, the latest and greatest from Microchip! Contact: ***@***.***

[SpenceKonde]

One thing I'm not sure I made clear - 921600 baud is probably not realistic - I couldn't make it work at least. that's fast enough where physical configuration starts to be relevant (ie, long wires are bad) Is the power supply rail noisy? askin part becasue the UPDI trace is also going under a power supply selection. that has some pretty wide traces, If large and chang and there are severely diminishing returns at that point Anyway. against tinyAVR even if you can make it work, based on my analysis of 460800 MHz performance, the write time was dominated by USB latency (initially, overwhelmingly dominated by, - final set of improvements were made a priority when I discovered that 17k/s write speed from Dx fell below that of jtag2updi on tiny, which led to the final round of performance improvements that reduced USB transactions, But with the short pages and extra command because of the page write and different nvmctrl left me with like 50% USB latency, so even at infinite baud rate, theeffective rate would be twice what's currently possible. That USB latency is a killer

• the tiny write performance that screamed on the Dx-series was dismal on tiny, because of smaller pages and more separate commands.
  Any luck with those traces? If you could ever catch the failure in the act ( and capture it on the scope) , I could tell what was wrong from the trace. IT also crosses my mind that it might have the same problem as the HT42B5234 on tinys. I would also try just a plain old serial adapter, instead of some purpose made device ) add a diode to most serial adapters and there's your IUPDI programmer. Looking at my hardware. I see that one of the CP2102's andthe HT42b354 have a 10k resistor across the diode too it effectively acts as a pullup when the USB serial adapter isn't trying to drive it low, and it seemed like it was having trouble driving it low enough with 4.7, but with just the diode, at 460800 the rise time was too long, so I added that 10k "pullup" to tx ( when USB adapter is transmitting, that resistor is no longer acting like a pullup, of course, but at that point, that's not a bad thing either.

[SpenceKonde]

I give up. Browser has eaten edits to above message several times with multiple paragraphs l;ost. Capture a trace at like 10 or 20ms/division in "single" trigger mode , plus a similar capture from the middle of a write at lower baud where it's working., and I can tell from that what is wrong electrically. By now I can tell a lot from a trace that looks like this ( a successful write to a tiny - looks like recent version and 128b pages, at 230400 baud, and nothing in the trace fgives me any doubt about the system electronically.

[SpenceKonde]

Also please post the failure message! There is also a software bug that can cause programming failures against tinyAVR devices at higher baud rates - though I could never make it manifest except on the HT42B534, which is pretty exotic.