exxos's Atari, Amiga & retro forum

I'm getting about 2V peak-to-peak noise on the +5V rail in the left socket and on the ground I'm getting bursts of what looks like 8MHz "packets" os sign waves which go way up, but most of the time it's in the 500mV range. Not nice at all.

That was a bad connection on the probe... forget that! :-)

I am getting pulses up to 2V though. I'll take a picture in a minute.

OK, here are the +5V and GND on the same screen, both 1V per division and triggered by a low excursion on the +5V.

As you can see it's pretty repetitive noise and the +5V and GND are in phase, so the differential in voltage is small. The problem is that the rest of the system doesn't see it that way. Some inductors on the line near the CPU should flatten this out a lot.

Stuffing a big old 470uF capacitor across +5V and GND to get rid of the high frequency stuff and it shows a VERY regular pattern.

I've found the source of the noise.. it's the MMU which sits directly under the board.

I taped a square of foil over the top of the MMU (between two lots of electrical tape) and practically all the high frequency noise disappeared and the other noise became lower frequency, though still as potent. Something in the MMU is generating a large, fluctuating magnetic field.

Instead of aluminium foil I need a lump of grounded soft iron on top of that flippin' thing!

P.S. The CPU switching is fully working, both CPUs are on the board with the external auxiliary power supply. Frontier is running normally in ST-RAM. I've not upgraded the firmware from the version @PhilC installed on the TF yet, so it's probably several version behind. Upgrading that will be the next project.

This is what GEMBENCH 6 is reporting (note that I've disabled the cache using the jumper): Oh, and the floppy isn't working in TF mode.

Well, with the shielding over the MMU and auxiliary power the machine is stable, though no floppy.

With maprom and fastram programs run the machine is snappy. Frontier runs almost as smoothly as it does on the TT, only held back by ST-RAM access speeds.

I've run GEMBENCH in this configuration, though I didn't know that it wasn't set to run in Fast-RAM by default so it's still running in ST-RAM, and it gives this result: Here's GEMBENCH run on a TT from ST-RAM as a comparison: One thing I've noted is that the floppy LED is flashing far faster than normal and at an irregular pace. Whatever is causing this is probably what's causing the floppy to not work. Switching to the 68000 without changing anything else and the floppy is operational.

I'm currently soak testing with Frontier.

So, electronically this project is a success, but physically it won't fit in the machine with the keyboard in place, so this part is so far not quite there.

But, I think that I can say that phase 1 is complete...

:girldance:

Very good troubleshooting work!

derkom wrote: 13 Oct 2020 22:46 :girldance:

Very good troubleshooting work!

It helped having the 157s and their capacitors in the middle of the board as I could see where the power line suddenly became very noisy. It was somewhere between the 68000 and the decoupling capacitors. In fact it could BE the decoupling capacitors picking up the signal from the MMU actually.

stephen_usher wrote: 13 Oct 2020 22:50 It helped having the 157s and their capacitors in the middle of the board as I could see where the power line suddenly became very noisy. It was somewhere between the 68000 and the decoupling capacitors. In fact it could BE the decoupling capacitors picking up the signal from the MMU actually.

My first attempt at the Stacy relocator (before it later became a switcher) moved the bus over from the left side to the right side of the motherboard, probably about the same distance as yours. The idea was that I could keep the old PSU (I hadn't yet decided to replace the screen) and build my upgrade stack where the SCSI controller used to live.

This proved very unstable and wouldn't run for more than a few minutes. After sticking a lot of impractical shielding between the PSU and my relocator, it got fairly stable with just the 68K, but still crashed with anything else on the bus. Probably a pullup problem at this point, as I hadn't replaced those. But the lesson learned from that attempt was that these are noisy computers and long traces make great antennas. :)

Looking at the signals on the STFM, the address and data lines were fine. No need to replace the pull-ups. However, /AS was terrible across the whole board. That pull-up on my board made the world of difference, similarly /RESET.