exxos's Atari, Amiga & retro forum

stephen_usher wrote: 12 Oct 2020 22:48 Hmm.. that would mean going up to a four layer board, which increases the costs hugely.

Yep, welcome to my world ;)

Thinking about this, the signals I measured last night show that the biggest concern is the /RESET line which is picking up something, 60Hz on the short trace and 230Hz on the longer one.

There are a few things to try. Firstly a terminating resistor, which acts as a damper at the end of the wire stopping reflections of the noise and countering resonance, so I can try a 10K pull-up. Alternatively, as it's a "slow" signal, I could put an inductor on the wire. Again, this would dampen any high frequency noise.

The noise on the power and ground are significant but a lot smaller than any signal. Again, adding inductors would smooth out the high frequencies here. Just have to make sure that they don't have values which will resonate with the 0.01uF capacitors at RF frequencies. We don't want Radio 2 on the power line.

exxos wrote: 12 Oct 2020 22:16 You need to use power planes for this stuff. If your just routing a power "track" its going to be nothing but headaches.

stephen_usher wrote: 12 Oct 2020 22:48 Hmm.. that would mean going up to a four layer board, which increases the costs hugely.

I think @exxos is right for a complicated board where you need power all over the place, but I don't see the need on something simple like this. If you really find that you need to keep power out of the board, you can always just inject power near the sockets, and avoid going up to a four-layer.

But I think a decently fat power trace, and supplemental power onto the TF536, should work fine. It certainly seems to be fine on my board, but then again, mine can't double as a serving tray. :lol:

I put 1K on reset as a pull up to stop issues. The reset circuits sux on original machines anyway, its why I went with a modern solution on the H4.

derkom wrote: 13 Oct 2020 11:36 I think @exxos is right for a complicated board where you need power all over the place, but I don't see the need on something simple like this. If you really find that you need to keep power out of the board, you can always just inject power near the sockets, and avoid going up to a four-layer.

This is basically going with what Atari did and why the ST's have so many crazy issues because of bad power rails. So "works" yes, works "ideally" no. If people want to do things properly, then power planes is a must. running separate power cable to the board will only help with voltage or current draw issues, not noise.

Yeah, though there are mitigation measures for high frequency noise, such as inductors. Also narrow, long, straight traces act as better antennae.

Don't forget, the bus etc is being pulled up via 10K (or whatever) resistor. "power" in each signal is hardly anything. However, capacitive coupling "loading" as said before, can cause very slow rising waveforms which is where problems start.

The only real power in the signal traces is when switching to 0V, which is why you need a good ground. You have all sorts of frequencies at play up to 50MHz (plus higher harmonics) all randomly going on, its a huge mess.

So you decrease the resistance pullups to compat loading (mostly capacitive), have a solid ground to stop said mess, avoid right angles in traces as much as possible as it increases inductance which compounds said problems. Straight traces are the best compromises, thinner traces give higher impedances in the MHz ranges so act as a "natural" few ohms resistance which can actually help with ringing on signals.

The address and data signals are clear, just some low amplitude high frequency noise and good levels. /AS need to be looked at as it's messy and only gets up to about 3V, so a pull-up on that is probably advisable.

The noise on the +5V rail and GND is probably in the region of 0.2V, if that.

The /reset, however has sharp, high frequency spikes with the oscillations attenuating in a few microseconds. On one socket these appear every 60th of a second and on the other about once every 230th of a second. The amplitude is about 2.5 volts!

@stephen_usher if it helps I can probably post some waveforms later on, will look at the /reset for starters.

Ok, so ive just checked /reset and /halt and with the scope I measure the ripple at the motherboard socket as 200mv on ground and /reset but thats then 500-600mv on the left socket and 280 to 360mv on the right hand socket.

Well, I've fitted 10K pull-ups to /AS and /RESET and they're looking a LOT healthier. /AS is going above 4V now. /RESET is as noisy as +5V.

I've also fixed the E2 line so the right-hand socket works properly. It was my completely useless surface mount skills which was causing the problem.

I've not tried with the TF yet.