STE V1.5 32MHz BOOSTER - STATUS

[exxos]

Not sure why it would happen but can only assume oxydisation on the pin over the years, not that we can see it on the photo.

Does not appear to be any tarnishing or oxidisation from what I can see.

Could it be that the cpu and pcb flexed under the sudden application of heat? Reading the instructions on the paste I use talks about raising the temperature in a set procedure....

It was soldered in a oven originally.. Of course bad joints can happen if the PCB and the solder melts and the CPU pin isn't up to temperature and it does not stick.. But the whole thing is heated up equally so that should not happen.. Even so, I heat gunned this board taking the CPU off and on, and the pins resisted the re-soldering each time.. And I even plastered lots of flux over everything each time. So it's not soldering this amount of times is nearly impossible.. The only thing it could be as if the pins were bent such a way solder did not touch the pins.. But all the pins seem level to me..

[exxos]

So the current booster was running a couple of hours then died again I fluxed the board up and re-heated twice and it made no difference. The third time I heated it up , and pushed down on the CPU for a couple of minutes until it had cooled, and now it works!

So what I am going to do is see if I can get a thicker stencil made to get more solder on the pins. I will also see if I can paste up the pins directly on the CPU and get some solder on them before actually soldering it to the PCB. At least that way I can make sure solder is adhering to the CPU pins before actually soldering it to the PCB.

.. And here's me thinking that PLCC sockets were evil and the pain ass...

[exxos]

What I will try today is pre-soldering the CPU pins on the PCB as shown below.. Then I will paste the whole board and assemble.. This way the CPU pins have double the amount of solder on them..

pre-paste.jpg (204.89 KiB) Viewed 4416 times

[troed]

(The next update will be from Chris swearing over hard to find connections between pins )

[exxos]

(The next update will be from Chris swearing over hard to find connections between pins )

Most probably

[exxos]

Board is getting there

I just tested the 8MHz switch and while it works, it actually glitches during transition I think it is a problem with my wait state code.. So currently sidetracked while I investigate the issue...

IMG_3049.JPG (222.81 KiB) Viewed 4439 times

[exxos]

It does not look like a real-time 8MHz switch will be possible with this design. There simply isn't enough IO power in that PLD

Problem is I have to delay DTACK because the CPU is running faster. Though I also have to turn this delay off to get back to stock 8MHz speeds. The problem is when the switch happens the DTACK delay time basically becomes random and causes issues.

The fix would be to have some logic to allow the 8MHz mode wire synchronised to DTACK so the switch only happens when there is no bus activity at all. Unfortunately this simple GAL simply does not have enough logic elements to implement that code. It wouldn't be a problem with a second GAL, but I don't think it's really worth designing another add-on circuit just to do real-time switching. Frankly there are more important to be working on than that.

So while the booster will be switchable between 40MHz & stock 8MHz, it will likely cause a crash if switching with the power on.

EDIT:
I may make a quick GAL test and just upload the info here, so if anyone wants to adapt it design a small addon PCB, then they can add it themselves..

EDIT2:
OK my brain has gone to sleep..

Anyone think how to allow a single to only propagate when signal is HI ?

Logic would be regardless if SW (8MHz switch) is high or low, it will only allow signal to propagate when DTACK is HI..

I guess a clocked flip-flop would work if the clock input enable was based on DTACK.

EDIT3:
Seemed to have made some progress bodging the GAL code in getting the clock switching to work...

[PhilC]

(The next update will be from Chris swearing over hard to find connections between pins )

LMAO, was thinking along the same lines.

Solder sucker anyone?

[exxos]

So this one worked first time.. see how it goes.. got the new firmware in it...

It it seems OK, will sell it cheaper for someone to beta test it Will have a second one when I get more bits..

IMG_3054.JPG (152.82 KiB) Viewed 4469 times

[exxos]

So today I found a odd issue.. I have seen it before, but I put it down to the previous problems but this seems to be a new problem, it is difficult to tell if there is still a bad connection or not.. I'm not sure if the board I was testing yesterday and this issue or not, but I will retest later today to make sure.

At some random point generally over half an hour, GB6 locks up without bombs.. HALT on the CPU is going crazy I am not sure why..

I swapped out the 40MHz oscillator for a 38.4MHz one, and this did not change anything. So it is not like speed itself issue here, I think not anyway.

I don't think it is ROM related, as it does just the same if accessing ROM at 8MHz.

e2.jpg (89.56 KiB) Viewed 4447 times

e1.jpg (49.26 KiB) Viewed 4447 times

EDIT:
OK so it died pretty quick that time, I will try disabling the blitter to see if that helps..
I remember having some odd issues while ago where there was a couple of blitters which did not play nice with the booster, I don't know why, but out of 100s tested which worked fine, I just put it down to some bad blitter chips.. I also have had contact problems with the blitter causing issues in the past.. So I will investigate...

EDIT2:
One suspect thing is that BR has a heck of a lot of noise on it.. I guess it could be possible the CPU is seeing a bus request when there actually isn't one..

Or could be possible the slow rise time could falsely be triggering another BR on the CPU.. The CPU is running at around 25ns cycles, if we took a rough figure of BR is still below 1V around 50ns.. then likely as possible random points timings could just trip the CPU into another bus request..

br.png (3.86 KiB) Viewed 4436 times

.. and on x10..

br2.png (4.04 KiB) Viewed 4432 times

EDIT3:

Added 1K onto BR..

br3.png (4.08 KiB) Viewed 4430 times

Faster rise times, but also lot of oscilations on the low part..

It ran for about 90mins with blitter off, no issues, before with blitter on, was getting lockups after 10-20mins. So think I am looking in the right places here..

Now I am trying with blitter on and the 1K pull up on BR to see how long it runs for this time.. possible may need small pF capacitor to help with noise... but will see how it goes...