HB-F9S v9938 partially dead?

Pagina 5/6
1 | 2 | 3 | 4 | | 6

Van Grauw

Ascended (10064)

afbeelding van Grauw

02-01-2016, 22:21

.

Van flyguille

Prophet (3028)

afbeelding van flyguille

02-01-2016, 22:18

blackmsx wrote:

Yes, the firmware is the one that came with the machine, which I belive is the original. The IC 13 was missing and so I used the socket to install the 512k ram.

As I said before, the scroll problem is present ONLY when I plug a Sunrise IDE cartridge I borrough. Can the Sunrise cause somehow the clock problems?

Here are some oscilloscope captures. Upper channel is /WAIT signal, lower M1 signal:

This is the msx with nothing plugged in:

and with the Sunrise plugged but no CF card inserted:

But with when the card gets detected things look quite different:

See how the /WAIT signal has changed.

At other thread some people suggested me that can be a power supply problem... but the wierd thing is that the problem only appears when the CF card is detected....

Yes!, you found the problem, that stream of low pulses is the generated by the s1985?... no other device uses it , right?.

trace the WAIT/ the circuitry and draw it, I needs to see it.

Then, open the sunrise cartridge, and measure the WAIT/ pin against Ground and +5v, using resistence scale, not diode. It must be infinite unless there is a pull resistence.

Now pay attention, to the traces, here you has a impedance problem. The theory, is, a trace maybe is malfunction and is generating resistence, unless the sunrise cart does a short circuit to +5v when CF card is on (like in the socket, or something), wich you needs to measure, trace even if the /WAIT signal reach the CF socket for some reason, etc.

The other possibility, is the connection from the s1985 to z80 /WAIT is not as good, again, a TRACE doing resistence, etc. Why it can be? because if it is doing resistence, it only will fail if other device is sharing the line, but if it is ok, is to say, from s1985 pin (the pin, not the trace), up to z80 /wait pin is 0ohm resistence, it is ok), but again, maybe it is not connected directly, maybe is an TTL AND logic in the middle, for merging different sources of /WAIT into one, that is the reason why I asks you to draw the circuit you found.

If not trace malfunction in the msx pcb, if the AND logic is ok if has one, then what is bugged is the cartridge itself, but what exactly in it?, you needs to inspect all the /WAIT path in that cart, and see what it does. Maybe in the cart it is just buffered, so you will see that it connects to soma glue logic only. In any case, if no reason for a fail in the mother, it is the cartridge.

Van blackmsx

Expert (77)

afbeelding van blackmsx

02-01-2016, 22:25

I think I solved the problem!

I open the Sunrise cartridge and cut a trace going from the /WAIT on the slot connector to pin 27 on the IDE connector. The scroll problem is now gone, and the IDE interface seems to work fine without this signal. I'll do further tests to make sure.

Thanks to everyone for their help!

Van flyguille

Prophet (3028)

afbeelding van flyguille

02-01-2016, 22:29

Another possibility, is the "cheap" circuitry, that it has not any TTL AND logic for merging the /Wait from differents sources, and it is all directly, and with that cartridge in that computer it is just a compatibility problem. Like the cartridge biasses more up the signal, making the s1985 impossible to place a LOWER status because it don't reach the threshold.... this kind of bias problems (also known as impedance problems).

If so, the fix is to modify it. Place a AND ttl logic in the middle, cut traces, piggying the ttl chip.

Van flyguille

Prophet (3028)

afbeelding van flyguille

02-01-2016, 22:30

blackmsx wrote:

I think I solved the problem!

I open the Sunrise cartridge and cut a trace going from the /WAIT on the slot connector to pin 27 on the IDE connector. The scroll problem is now gone, and the IDE interface seems to work fine without this signal. I'll do further tests to make sure.

Thanks to everyone for their help!

Most probabbly as you uses a CF adapter, that the adapter or the CF card is short circuiting that line.... more like the adapter

BAD idea cut it in the sunrise cartridge.

Also can be as simple as a IDE connector pin folded.

Van blackmsx

Expert (77)

afbeelding van blackmsx

02-01-2016, 23:01

flyguille wrote:

Most probabbly as you uses a CF adapter, that the adapter or the CF card is short circuiting that line.... more like the adapter

BAD idea cut it in the sunrise cartridge.

Also can be as simple as a IDE connector pin folded.

Yes, not a very subtle solution. Well, I can patch the track back with a little solder later, after I find a better solution. I've also checked the IDE-CF adaptor and I couldnt find any shorts....

Anyway, so far I've loaded some games and is working fine, and the scroll bug is completely gone. CHKSPD now reports the right CPU clock.

Your suggestion is to use the AND gate to buffer the /WAIT signal? How exactly will you connect it?

Van flyguille

Prophet (3028)

afbeelding van flyguille

02-01-2016, 23:03

blackmsx wrote:
flyguille wrote:

Most probabbly as you uses a CF adapter, that the adapter or the CF card is short circuiting that line.... more like the adapter

BAD idea cut it in the sunrise cartridge.

Also can be as simple as a IDE connector pin folded.

Yes, not a very subtle solution. Well, I can patch the track back with a little solder later, after I find a better solution. I've also checked the IDE-CF adaptor and I couldnt find any shorts....

Anyway, so far I've loaded some games and is working fine, and the scroll bug is completely gone. CHKSPD now reports the right CPU clock.

Your suggestion is to use the AND gate to buffer the /WAIT signal? How exactly will you connect it?

first you checked that in the MSX PCB is directly connected the slot /WAIT directly to s1985 and z80?, maybe with some diode in the middle? or it already has a ttl AND logic in the middle?

Checks, that please.

Van flyguille

Prophet (3028)

afbeelding van flyguille

02-01-2016, 23:07

in my MSX that is also based on s1985, it has a ttl AND logic.

where a single AND gate is used, and the input that goes to the slot is pulled up with a 4k7 resistor.

the other input is connect to s1985 IIRC no pull up, and the output to z80 /WAIT, no pull up there.

any of the two input that goes LOW, the output will LOW aswell, so that way separating circuits, and eliminating any impedance problems.

Van blackmsx

Expert (77)

afbeelding van blackmsx

02-01-2016, 23:30

On the HB-F9s the /WAIT line is connected directly. No diode or TTL logic. Only the 1K pull-up resistor (R121)

By the way, in the cartridge the /WAIT signal goes directly from the slot connector to pin 27 on the IDE connector. From there goes to the CF pin 42. I tested the resistance from the slot connector to pin 42 on the CF card and gives 0 ohms

Van flyguille

Prophet (3028)

afbeelding van flyguille

03-01-2016, 00:42

so, the internal impedance of the CF card itself, don't meets the MSX standard requirements.

And see that CHEAP CHEAP circuitry of the msx MB, that don't buffers ANYTHING....

It is all tied together, the first device pulling the line down triggers the WAIT, but as soon as you inserts anything else on the line that biases it UP, the s1985 simply don't has enought internal LOW impedance as to pull the line down.

You can try to change IIRC you mentioned the MB has a 1K resistor to pull up it... you can change it with a 4.7K resistor. so that will weak the line enough that the s1985 can pull it down include with the CF inserted.

Pagina 5/6
1 | 2 | 3 | 4 | | 6