8-bit acorn hardware • Re: BBC Micro NMI spamming

Worth checking that your network/termination is in good order before plugging in this machine - with it idle measure D+ and D- on the socket with a meter and they should be somewhere around +1.5V with the D+ a bit (0.2-0.5V) above D-.

Yes, there is around 1.990V and 2.001V on D+/D- and the same on C+/C- (on a working Beeb I get 1.982V and 1.972V respectively) the difference only seems to be 10mV though. I am measuring the voltage at the DIN connector on the back of the machine.

That isn't right (not enough difference). Not your current problem(*), but you should really fix your termination. If using the common phantom-powered termination, open-circuit GND connection between the clock and the terminators can cause this effect.

(*) I'm beginning to think you have more than one fault here...

Plugging in the station shouldn't have any effect - if significantly pulled up or down, can only really be the 75159 (or random solder shorts etc).

Yeah, plugging the station in and voltages are now all over the place. C- = 2.042V, C+ = 0.761V, D+ = 1.327V, D- = 1.071V

Removing the 75159 makes no difference to the voltages measured on the pins when connected to the network.

The whole of the rest of the circuit is only connected to the line through 100K resistors and so really shouldn't be able to have material effect on the network.

However, those 'plugged-in' readings are not necessarily bad. As measured with a DC meter, the asymmetric waveform of the clock will give just those sort of readings even though the scope will show them to be the same size: C- is high for longer than C+ for typical clock settings (though for a short network there's no advantage in an asymmetric clock).

And those revised D+/D- readings look closer to correct than the version with the station unplugged - I'm wondering if plugging in the station fixes an open-circuit GND in the wiring to the terminator (potentially via mains earth connections of various stations plugged in).

I did notice something very strange on the scope... the yellow trace should be flat when there's no data (and there isn't) Looking at IC94.jpg I suspect you're right about the solder bridge somewhere, but I've looked with the microscope and I can't see any on the underside, so that leaves me with it being under a socket somewhere. Maybe this trace may give a hint at where the short might be. I've checked all the resistor values and they tally with the other Beeb.

That scope shot does look like classic crosstalk from clock to data. On Issue 3, (and even more so on Atom) this could potentially occur between adjacent vertical resistors depending how you bend them, but yours is an Iss 4 which had the layout re-done to fix that.

The potential for the crosstalk is that you've got IC94p7, R44, R52 which all have the clock as a low-impedance full-5V signal, and they necessarily end up rather close to IC94 pins 4,5 which are the data inputs at very low amplitude and high impedance.

You've got an inversion between clock and the unwanted pulse on the data, so probably the clock output (pin7) coupling to something on the negative side of the other comparator (pin 5, R35/R40).
And only a short pulse rather than solidly reproducing the clock, so maybe more of a capacitative coupling than a dead short, or perhaps an open-circuit somewhere on the data side making that node more receptive to coupling.

BTW: my pin naming convention is looking at the socket from the back, the PCB side, with D+ on the left next to the PSU, followed by D-, GND, C+ and C- on the right (I'm not sure if this is the correct pinout, I used the polarity of the comparator inputs as a guide)

I think you have it correct.

Statistics: Posted by arg — Thu Feb 15, 2024 8:52 am

---