TF536 + 68000 relocator and ROM board project.

[stephen_usher]

Oh, monitoring the ground over the other side of the board, next to the ROM... Spikey when the ROM is being accessed! (Once a game which doesn't use the ROM starts the ground goes quiet.)

OK, so it looks like the majority of the problems are probably too small decoupling caps. I can test this properly after Tuesday.

[stephen_usher]

Seeing as I'm working from home today and had a little time this morning added 1uF tantallums on the current board.

The ground is a lot cleaner now, though still spikey directly next to the major chips, but maybe that's normal?

So, that's good isn't it?

Well, the strange thing is that the TF536 GB6 speeds have dropped significantly. e.g. the Integer Division score went from 1728% down to 1656% of a standard STFM.

[derkom]

Well, the strange thing is that the TF536 GB6 speeds have dropped significantly. e.g. the Integer Division score went from 1728% down to 1656% of a standard STFM.

What about the 68000? Still at the unusual 99%?

[stephen_usher]

Yeah, between 99% and 98% as usual.

[stephen_usher]

I've put a 10K pull-up on /DTACK as from the 'scope trace that rises quite slowly relative to the width of the signal and hence could give a timing issue.

That's improved things very slightly on the 68000 giving 99% and 100% for many of the tests where they were 98% and 99%.

It hasn't improved the speed of the TF though and I have noticed the display jump to the right, as described by Exxos as clock skew. (However, I can't see any skew at all when using the oscilloscope.)

I'll try a 2K2 resistor later but I can't do that this afternoon.

[exxos]

The skew only needs to be like 2ns to upset it. It's easily missed.

[stephen_usher]

Going to the highest sensitivity of my oscilloscope there seems to be a skew of around 0.25 to 0.5 ns at most across the board and onto the TF.

[stephen_usher]

I've found out why the benchmarks for the TF are lower than I had before... I was an idiot!

The previous benchmark was done with the cache enabled! D-oh!

Turn on the cache and the benchmarks are the same.

Still, I can't find why the 68000 benchmark is lower. Surely a clock skew of less than half a nanosecond can't really be the reason.

[exxos]

Clock skew is visible with pixel issues , like the screen shifts 16 pixels to the right and wraps around to the next line.. try something like 500R on ST DTACK see if that helps.

[stephen_usher]

As in pull-up?