Open source 14MHz 68000 accelerator

[Jope]

Expansion.lib configures the big chunks to addresses beginning from $200000. The IO side of things gets autoconfigured beginning from $E90000

So yeah, you could potentially map yours from the top downwards.

[opivy224]

Hello, I put this together and it worked once.. but then stupid me cleaned flux off with 90% IPA and may have reinstalled it before it had all dried under the CPU socket. This time I got a black screen and now it won't work at all. What components should I start troubleshooting? Replace CPU maybe?

[Mathesar]

Originally Posted by opivy224

Hello, I put this together and it worked once.. but then stupid me cleaned flux off with 90% IPA and may have reinstalled it before it had all dried under the CPU socket. This time I got a black screen and now it won't work at all. What components should I start troubleshooting? Replace CPU maybe?

Sorry to hear that! I don't think something is really broken. I would rather think that some conducting stuff got stuck underneath chips, between pins, etc. The circuitry around the the 74HC(T)4046 is the most sensitive to that. I would recommend a second, thorough cleaning. Let it dry and try again!

[opivy224]

Originally Posted by Mathesar

Sorry to hear that! I don't think something is really broken. I would rather think that some conducting stuff got stuck underneath chips, between pins, etc. The circuitry around the the 74HC(T)4046 is the most sensitive to that. I would recommend a second, thorough cleaning. Let it dry and try again!

Thank you for the reply. I haven't cleaned it yet, but I noticed if I press down on it while in the socket it works somewhat but will eventually freeze with screen artifacting. I will give it a good cleaning tonight and try again.

[opivy224]

Cleaned it and same issue. If I press down on the CPU socket it works but will eventually crash or display artifacts on the screen. I bought this as a kit from someone on ebay. Should I try a new CPU socket?

[Mathesar]

Ok, that definately sounds like some contact issues. Where exactly do you push down?
If you push down on one of the logic IC's it could also be a bad solder joint on the accelerator board that only makes contact when you push on it. I wouldn't immediately suspect the motherboard CPU socket.

[byteoftce]

I have now also built 2 pieces of the Card.
Great work, but I had a lot of trouble getting the card to work.
R3 as a fixed resistor didn't work at all. Although the measuring voltage at J2 was 2.52V (4.7KOhm), the card did not start and there was a green or black startup screen.

I also tried it with 4.3K and 5.1K resistor. no chance :-(

Only after I used a 10K rotary resistor does the card work now. (set to 4.95K ohms (J2 2.94V).
If I only adjust the rotary resistor by 50 ohms, the card does not work.

To rule out other problems, I also tested different CPUs.
I tried MC68000FN16 / MC68HC000FN16 / MC68HC000FN20 / MC68010FN16 / MC68000FN12 (for the endurance test). The same problem with all CPUs. Since then, both cards have been running stably, only if the Amiga was off for 1 day do I have to start it 4-5x before the Amiga starts, after that it runs perfectly again.

As it seems it is very much due to the ICs used and their normal tolerances. Maybe this will help one or the other if he builds the card.

[Mathesar]

Originally Posted by byteoftce

I have now also built 2 pieces of the Card.
Great work, but I had a lot of trouble getting the card to work.
R3 as a fixed resistor didn't work at all. Although the measuring voltage at J2 was 2.52V (4.7KOhm), the card did not start and there was a green or black startup screen.

I also tried it with 4.3K and 5.1K resistor. no chance :-(

Only after I used a 10K rotary resistor does the card work now. (set to 4.95K ohms (J2 2.94V).
If I only adjust the rotary resistor by 50 ohms, the card does not work.

To rule out other problems, I also tested different CPUs.
I tried MC68000FN16 / MC68HC000FN16 / MC68HC000FN20 / MC68010FN16 / MC68000FN12 (for the endurance test). The same problem with all CPUs. Since then, both cards have been running stably, only if the Amiga was off for 1 day do I have to start it 4-5x before the Amiga starts, after that it runs perfectly again.

As it seems it is very much due to the ICs used and their normal tolerances. Maybe this will help one or the other if he builds the card.

It is true that the timing of the clock doubler is very dependent on the tolerance of the VCO in the 4046 chip. The VCO needs to run at 7MHz when the VCO control voltage is around half the supply voltage to get a 14MHz clock with 50% dutycycle. That is why the test point should read around 2.5V. However, I had my boards running at voltages between 2..3V indicating a duty cycle between 40%..60%. That your boards are so sensitive is very weird. I would love to see some scope shots of the clock (is it jittery?). What brand/source of 4046 chip did you use? Are you sure all the othercomponents of the loop filter are correct?

[byteoftce]

Originally Posted by Mathesar

It is true that the timing of the clock doubler is very dependent on the tolerance of the VCO in the 4046 chip. The VCO needs to run at 7MHz when the VCO control voltage is around half the supply voltage to get a 14MHz clock with 50% dutycycle. That is why the test point should read around 2.5V. However, I had my boards running at voltages between 2..3V indicating a duty cycle between 40%..60%. That your boards are so sensitive is very weird. I would love to see some scope shots of the clock (is it jittery?). What brand/source of 4046 chip did you use? Are you sure all the othercomponents of the loop filter are correct?

For further tests I have ordered an Oszi, but unfortunately it will
take some time from China until it arrives here. I took a picture of a PCB.
All components are ordered via Digikey and are working properly so far.
I have followed the recommendations in the project files as best as possible.

I've also considered whether it makes sense to tap the clock generation externally via CDAC (I think that this can be done via Gary's Pin26. There was no information about this in the documentation for the project.

[byteoftce]

A new intermediate...
I have now replaced the HCT4046AMG4 from TI with a 74HCT4046AD from NXP as a test. Unfortunately, the problem has gotten worse. Now I have to switch the Amiga on several times until it boots and the dependency of the resistor R3 is even more difficult to set (10-20 ohms). I also replaced the capacitor C9 (100p) with a different batch as a test. When the Amiga boots everything runs stable, otherwise a green screen appears when switching on. It becomes problematic if I install the card in a second Amiga. There the resistance must be set differently again until the card runs. So it also depends on the Amiga itself (tested with 4 Amiga 500 Rev.6).
I'm still wondering if it makes more sense to tap the clock at Gary (CDAC pin 26), but I'm not sure if that's correct. More precise results will probably only bring the oszi which I'm still waiting for at the moment....

[pandy71]

Originally Posted by byteoftce

A new intermediate...
I have now replaced the HCT4046AMG4 from TI with a 74HCT4046AD from NXP as a test. Unfortunately, the problem has gotten worse.

Verify DS for both, also you can give a chance to 74lv4046 from TI it is way faster than regular HCT and work with 5V.
Most of 4046 are quite slow - perhaps PLL can utilize NE564 with better stability.

[Mathesar]

Originally Posted by byteoftce

A new intermediate...
I have now replaced the HCT4046AMG4 from TI with a 74HCT4046AD from NXP as a test. Unfortunately, the problem has gotten worse. Now I have to switch the Amiga on several times until it boots and the dependency of the resistor R3 is even more difficult to set (10-20 ohms). I also replaced the capacitor C9 (100p) with a different batch as a test. When the Amiga boots everything runs stable, otherwise a green screen appears when switching on. It becomes problematic if I install the card in a second Amiga. There the resistance must be set differently again until the card runs. So it also depends on the Amiga itself (tested with 4 Amiga 500 Rev.6).
I'm still wondering if it makes more sense to tap the clock at Gary (CDAC pin 26), but I'm not sure if that's correct. More precise results will probably only bring the oszi which I'm still waiting for at the moment....

I still suspect some other component having the wrong value. C10 should be 100n (C10 is not very critical though), C11 and C15 should both be 22p (pico!).
The NXP parts are generally high quality. It shouldn't be that sensitive.

But yeah, give CDAC a try. I put that option in as a fallback when the PLL is not working. So far I haven't heard from anyone that they had severe problems with the PLL but as you do have them: try the CDAC option.

[byteoftce]

Your last post might contain the solution. But I prefer to ask again. Are you sure the C11 and C15 must have 22Pico? According to the BOM list (Github), C15 is specified as 100 pico....

edit:

I think I found the mistake. The reference to C11 was entirely correct. I love our Chinese, who seem to be up for a joke from time to time. Instead of 22pico, they labeled a capacitor strip with 220pico capacitors as 22pico. Someone probably thought binary 0 or 1 (220 or 221 ...). Oh, that's right ^^.....
Now that I've replaced C11, it also works with fixed resistors. R3 with 4.7K -> J2 2.6V- Almost perfect....

[Mathesar]

Originally Posted by byteoftce

Your last post might contain the solution. But I prefer to ask again. Are you sure the C11 and C15 must have 22Pico? According to the BOM list (Github), C15 is specified as 100 pico....

edit:

I think I found the mistake. The reference to C11 was entirely correct. I love our Chinese, who seem to be up for a joke from time to time. Instead of 22pico, they labeled a capacitor strip with 220pico capacitors as 22pico. Someone probably thought binary 0 or 1 (220 or 221 ...). Oh, that's right ^^.....
Now that I've replaced C11, it also works with fixed resistors. R3 with 4.7K -> J2 2.6V- Almost perfect....

You fixed it! Yeah, 220pf is definitely too much and will make the PLL loop unstable. I am very happy to hear that your card now works as it should.

Also, you're right, C15 must be 100pf. Sorry for the confusion. It probably also works with 22pf but 100pf is best. My first prototype had no C15 at all and that worked not so well. Due to the weak signal level of the Amiga's 7Mhz clock the phase comparator produced a very jittery output. With capacitive coupling the phase comparator biases itself and the output is then (well, almost ) jitter free. PLL's are always tricky and I spend quite some time to make make it work proper.

[TymoDEV]

Great job. Looks very nice on the new accelerator card. I am considering to buy it. I already knew the simple turbo x2 system before.
Is it possible to work with 68HC000FN20 at turbo x4 [28 MHZ]?
I am dreamming about turbo x4 version. If it would be possible?

[Samurai_Crow]

Only with Fast RAM. Otherwise there won't be enough bandwidth.

[TymoDEV]

I don't understand: WHY. If CPU works with CHIP RAM, takeovers all BUSes. The clock multiplied by 4 connected to the processor will tell it to work 4 times faster even with CHIP RAM. All access times for prcessor can be multipled by 4, too, just like turbo x2 [14 MHz]
Only bad thing is revision of CPU 68K. If can to work at 28 MHz.
Plastic packages can be good at that clock like MC68030RP33.
We need only wright revision of 68K.
Long time ago Supra released Supra Turbo 28 with 68HC000 at 28 MHz. If somebody can tell us something about this turbo card? Who can tells anything, how it has been tested?

[Jope]

The Supra Turbo28 has a little bit of cache RAM that helps it access Zorro 2 memory quicker. Only 4MB can be cached.

[TymoDEV]

Most important thing is if 68HC000 really works stable at 28 MHz.

[Samurai_Crow]

Originally Posted by TymoDEV

Most important thing is if 68HC000 really works stable at 28 MHz.

Incorrect. If you can only get bandwidth for 7-14 MHz worth of performance, that's the amount of performance you will get even if the processor is solid at 900 MHz. It's like the old saying goes: "A chain is only as strong as its weakest link."