Ralph Doncaster

> > > +1 > If you push the hex files to your repo, I can make some independent and intermediate tests for your t85_default and t85_aggressive versions. As you...

I removed more dead code from usbdrv.c, and got past the next page boundary for t85_default: 1398 bytes.

> > > > I suspect most USB hub controllers don't implement a PLL with a timing range of +/- 1.5% for low-speed devices as required by the spec. Tim,...

Today I started converting the usbdrv.c ProcessRx code to asm. t85_aggressive is now below 19 pages, at 1212 bytes. https://github.com/nerdralph/micronucleus-firmware/blob/asm/firmware/releases/t85_aggressive20200828.hex next I'll work on BuildTx.

I found some easy size optimizations in the C code by using fixed registers for usbMsgPtr and usbMsgLen. Those two changes save a total of 36 bytes, so now t85_default...

I finished converting usbBuildTxBlock to asm, and now t85_aggressive is 1146 bytes. https://github.com/nerdralph/micronucleus-firmware/blob/asm/firmware/releases/t85_aggressive20200830.hex

Today I decided to do some investigation into the reliability problems I'm having with micronucleus on a t85 at 16/16.5MHz. While it's well known OSCCAL tuning on the t85 doesn't...

The resolution of the high/low count was only 16 cycles as I was only sampling 1 phase with a 8-cycle sampling loop and then multiplying by 2. I updated it...

I just noticed something that *might* explain some of the unreliability I'm seeing while running on the internal RC oscillator: https://github.com/micronucleus/micronucleus/blob/master/firmware/osccalASM.S#L153 I haven't been testing with a dedicated bus; there's...

> > These collision are not as much of an issue as one might think. Only old (1.1?) USB hubs forward all traffic to every port. Anything more modern should...