Centurbo 1 evolution 3b -> CT60 booster

[foft]

I upgraded the SDR and ABE today via JTAG to the latest 5E and 5M versions. Also the boot menu to 1.05.

Then I tried some ram experiments.
My Centram14 which worked fine with CT1(at 16MHz and 25MHz) does this with the CT60:
030 mode: pass, 060 non-boosted: fail, 060 boosted: fail
My Falcon wings does this with the CT60:
060 mode: pass, 060 non-boosted: pass, 060 boosted: fail (too slow, this is known to be the case for this ram...)

So, tried a little SIMM swapping in the Centram14 - and now its happy :D

[Rustynutt]

After working with a minimal invasive bus speeder (no wires) for the Afterburner, moved over to the CT60's to test the setup.

I run a 60MHz oscillator to the COMBEL, and 500KHz clock into the ACIA (KB).

It's quick. Found the CT60 and CT60e were more fussy, and would top out around 46-48MHz.

My CT63 has the accelerator circuit, but don't see the point in running wires across the board when a simple bus mod is all that's needed. It's stable at 106MHz CT clock.

The problem that came up with the Afterburner (it has many :) ), is DMA transfers. Regardless of clock patch, the only testing which results in 100% reliable SCSI/floppy is to down clock the bus to 8MHz for SCSI use.

With a 30MHz bus, DMA is an issue with the CT as well. Tried clocking it at 8MHz bus, but there's more going on there as it doesn't work like on an Afterburner.

Need to come up with a simple dual oscillator board think for the CT. Still, cleaner than running wires all over.

Nice thing is, using BSS HRTC, 640*400 TC mode is snappy. Haven't tried Daniel's video adapter and Blow Up with it yet, image at a minimum 640*480 TC is possible.

And Apex Media rocks with it in TC 640*400-640*480 virtual screen mode.

[foft]

Just checking I understand this. When you say no point in running wires, do you just put a local oscillator pcb next to both ACIAs and the COMBEL? i.e. plumbing them directly into ACIA pins 3/4 and L102?

I guess the DSP is still running at the original 32MHz clock?

The CT60 FPU and DSP solutions are pretty ... long wires for 32MHz and 50MHz. Over the top of other traces... One difference on the CT60 vs the CT1 is the addition of a GND for the clocks (except FPU), perhaps for a shielded wire? In the setup I have its just a ribbon cable with GND next to the signal wire.

I've just been trying the CT60 bus speeder in 030 mode. Connecting the middle pin of F30 boost to 3.3v enables it regardless of 030 or 060 mode, seems to work fine.

The normal bus speed is 16MHz after booting, so do you mean you disable any bus speeders then switch it back to 8MHz (via the COMBEL register) to do any SCSI transfers?

[mikro]

That would be fantastic Mikro. Please could you ask this:
In the Centurbo 1 Evolution 3b installation document it says that replacement GALs for U44 and U68 are provided. Do you still have the JED(or LCR/GAL) for the replacement debugged U44 and U68? If not, do you recollect what change was done from the original Atari version?

Actually, as it seems, they already have been published: https://mikrosk.github.io/ct60tos/rodol ... 0_GALs.zip.

[foft]

Thanks, I’ll extract that tonight and diff it vs the Atari version of that archive.

[Rustynutt]

Just checking I understand this. When you say no point in running wires, do you just put a local oscillator pcb next to both ACIAs and the COMBEL? i.e. plumbing them directly into ACIA pins 3/4 and L102?

I guess the DSP is still running at the original 32MHz clock?

The CT60 FPU and DSP solutions are pretty ... long wires for 32MHz and 50MHz. Over the top of other traces... One difference on the CT60 vs the CT1 is the addition of a GND for the clocks (except FPU), perhaps for a shielded wire? In the setup I have its just a ribbon cable with GND next to the signal wire.

I've just been trying the CT60 bus speeder in 030 mode. Connecting the middle pin of F30 boost to 3.3v enables it regardless of 030 or 060 mode, seems to work fine.

The normal bus speed is 16MHz after booting, so do you mean you disable any bus speeders then switch it back to 8MHz (via the COMBEL register) to do any SCSI transfers?

I'll try not to get crazy :)

The COMBEL (main) oscillator has a socket so I can experiment.

The ACIAs have their own 500KHz clock, with the traces (there are 2 from different COMBEL pins) from the COMBEL finely cut. Same as Nemesis, except the oscillator sets to the side of the ACIA (and powered from). And you don't have to cut and lift pins of the ACIAs.

The TOS ROM is edited to keep the Blitter at boot to 1/2 CPU clock. Allows getting around crashing the Blitter with too fast a clock when it is used to display the Atari logo. For CT mode, this doesn't matter.

I'm still experimenting with clock patches, but yes, without an accelerator, or the Afterburner, running the clock at 1/2 speed makes DMA reliable. With a 60 MHz clock, divided by 2 by the COMBEL, then at 1/2 speed, bus clock is 15.
The Afterburner has problems of its own, it runs at 46MHz max, so 11.5MHz, where the AB works well.

On the CT, with its own CPU clock, the main oscillator still runs the bus higher, but has no effect "on top of the expansion pins, ie, the CT is independent of the COMBEL CPU clock out.

The DSP has its own 46MHz oscillator, it's the only ceramic one I had to mount on the underside of the board.

Just started testing with the CT63. I tried editing CT TOS the same as the Atari ROM, but somewhere during boot it's changing. I did test the CT with a 16MHz oscillator, and it works, just haven't figured out how to set it with the CT installed. Mirko wrote the two programs that switch between 1/2 CPU and full clock. For some reason, the one that switches to 8MHz causes errors.

I have some benches I'll post later, it's worth it, and not having wires running everywhere is nice. I've had a Nemesis before, what a mess :)

I want to build a small PCB for two main oscillators soon.

That make sense?

[foft]

Thanks, I’ll extract that tonight and diff it vs the Atari version of that archive.

I ran a diff and this is the same as the Atari one, i.e. it's not the CT1 modified versions.

[foft]

That make sense?

Yep makes sense and very interesting. Thanks for writing it up.

Reusing the trace for the ACIAs makes it quite a lot cleaner, I guess the benefit of the wire to them is the motherboard itself is intact.

For the DSP also a local oscillator is much nicer than running that 50Mhz line across the board, crossing many traces.

[Rustynutt]

Exactly. The oscillator feeds the 500KHz clock backwards to the MIDI ACIA. Only requiring a couple pins to be jumped with a solder blob.
Take note, the KB ACIA power seems reversed from the schematic. Haven't quite figured out what is going on, but plus is minus and minus is plus on that IC. Guess I should at least look into it.
Mine are socketed as I'd replaced them due to cutting the pins on the original Nemesis install.

In time, will have a relocation board for the Afterburner to move it rearward. Once feature there will be to reverse feed the clock from the 500KHz expansion pin.

Another fun fact I just discovered yesterday. While the CT will boot without an NVRAM or TOS ROM in place, a standard Falcon will not, either without the IC present, or a botched repair job.

By chance, had left it out after removing the CT to test the standard setup and 60MHz COMBEL oscillator, and the Falcon boots normally even without the NVRAM present. I tested this down to about 48MHz from 60, and it boots fine until below 48 MHz (44.9 close as I had).

Why, have no clue.

[mikro]

Thanks, I’ll extract that tonight and diff it vs the Atari version of that archive.

I ran a diff and this is the same as the Atari one, i.e. its not the CT1 modified versions.

Hmmm and are we sure that they were modified? Because when I asked Rodolphe, he did respond and he sent me exactly those files (I binary-compared them to be sure).