DIrect TMS9928 - GBS8200 YPbPr connection

Страница 4/5
1 | 2 | 3 | | 5

By DamnedAngel

Master (143)

Аватар пользователя DamnedAngel

25-09-2016, 16:14

hit9918 wrote:

the red picture looks like camera moiree, the bars change their angle, can I trust the pic.
I got plain vertical bars. they are too subtle for reliable counting.

Yes, you are right: The top lines are indeed artifact of the photo. The more spaced, bottom lines are the real ones (you can trust those). It is pretty hard to avoid the artifacts in digital photos!

By l_oliveira

Hero (529)

Аватар пользователя l_oliveira

25-09-2016, 23:12

Just mentioning Brazilian MSX computers use TMS9128 not TMS9928.

By maxis

Champion (512)

Аватар пользователя maxis

26-09-2016, 14:50

DamnedAngel wrote:

Yes, you are right: The top lines are indeed artifact of the photo.....

Finally here is the source of the aliasing. We found it Wink

By tvalenca

Paladin (728)

Аватар пользователя tvalenca

26-09-2016, 20:25

Hey DamnedAngel, now I think I understood your problem. Unfortunately, I'm convinced this is caused by the DRAM refresh, because it looks like the bars looks a little bit smaller than the pixels from the characters.

I can point 3 things you can check based on mere hypotesis, but all of them involve some degree of modifications on your Expert:

1) Maybe there's something wrong with your VDP. It's kinda long shot, but do you have another VDP to temporarly replace the one on your Expert? I never saw it happening with a TMS VDP, but I saw a handful of Color Computers with issues on colors that were solved only then had their VDCs (MC6847) replaced. Both TMS9128 and MC6847 have analog and digital parts on die and have YPbPr video outputs, so there's a small chance. Oh, I forgot to ask: Do your Expert have IC sockets to VDP and VRAMs?

2) you are having some kind of ripple on VDP power traces, and this is affecting the hability of VDP maintain video signals stable when refreshing the VRAMs. I would try adding a large capacitor near the VDP supply pins to check, and if it solves the problem, I would try find any issue on the PSU and traces on the digital board between PSU connector and VDP pins. Oxidization, bad solder joints, and if your Expert has sockets for VDP and VRAMs, check for oxidization on IC pins and sockets.

3) There are some ripple caused by bad layout (altough there's something else making it worse). Try adding some resistors like @maxis suggested. If your Expert has a IC socket for the VDP, you can lift the needed VDP pins from the socket, place the resistors on the socket and solder the other lead to the VDP pin. Or even try adding the resistors between two sockets and place those between the VDP and the Expert PCB (but don't use this socket-resistor-socket sandwich as a definitive mod, because it will generate a lot of issues due to bad contacts.

4) I can't think of anything else right now...

Unfortunately there's no secure way to check any of this before modding the board... maybe if you had an oscilloscope (and know how to use it).

By flyguille

Prophet (3029)

Аватар пользователя flyguille

27-09-2016, 20:04

DamnedAngel wrote:

Hi all (and specially tvalenca),

Quote:

HEY YOU HAVE AN EXPERT! you already have a good (if not great) quality RGB output.

Well, I just build the LM1881 circuit and image came out instantly. Pretty misadjusted, but I managed to torture my monitor into stretching the image into and acceptable box within the screen:


My big problem, however, are vertical stripes all over the image, which appear in every color but black and white:

Please note that the upper stripes are artifacts of the photo I took, but the lower stripes are real ones on the screen.

What is the cause of such stripes? Is my Expert's RGB output faulty? What can I do get rid of them?

Note: Please let me know if it would be adequate to start another topic, since it has nothing to do with the original subject (GBS8200/YPbPr).

TIA,

those vertical stripes is normal in VDP, is digital noise from the insides of the TMS/VDP chip, through the power line coupling in the analog output of the chip.

You can place a ceramic capacitor directly soldered from the V+ pin to ground, to minimize it, but it will not dissapear 100%.

By DamnedAngel

Master (143)

Аватар пользователя DamnedAngel

29-09-2016, 05:15

Okay, some more info (and no solution and no progress... :S):

1. L_oliveira is correct: My Expert has a 9128 and NOT a 9928.
2. Both my 9128 and its VRAMs have sockets.
3. Tried several capacitors in both 9128 and VRAMs (the board already has old 0.1uF ones), with no perceptible change.
4. I rebuilt my YPbPr setup (GBS8200 connected directly to 9128's YPbPr outputs) and, although the colors get wrong as I mentioned before (someone hinted my it's because wrong impedance matching), this configuration allowed me to confirm that the stripes remain without Expert's RGB circuitry. Thus, (unfortunately) it is not a problem in this module.

flyguille wrote:

those vertical stripes is normal in VDP, is digital noise from the insides of the TMS/VDP chip, through the power line coupling in the analog output of the chip.

Given my results, I am agreeing with you. But I really question if the problem is normally that acute... I really doubt one would be satisfied with such deep stripes in screen. I still think there has to be a way to enhance a LOT the image I am getting.

tvalenca wrote:

1) Maybe there's something wrong with your VDP. It's kinda long shot, but do you have another VDP to temporarly replace the one on your Expert?

I happen to have 2 Hotbits in my parent's house. Do you know if they use the 9128 as well - and, even better, if they are socketed?

tvalenca wrote:

2) you are having some kind of ripple on VDP power traces.

I will try to get my hands on an oscilloscope tomorrow to try to find out how my power lines go. Anyway, new capacitors did no good to the image...

tvalenca wrote:

3) There are some ripple caused by bad layout (altough there's something else making it worse). Try adding some resistors like @maxis suggested.

Could you or maxis explain better how (and where - would it go in all 8 address lines?) those resistors should be mounted? Did I get it right that they would go in serial with the lines? What resistance values should they have?

Thanks all for the patience to track the problem with me!!!

By DamnedAngel

Master (143)

Аватар пользователя DamnedAngel

29-09-2016, 13:49

Just another though I had during the night...

Would it be easy to somehow initialize and issue commands to 9128 to make its output somewhat colorful (i.e. 15,4,4 - or even 4,4,4) if I mounted it and its VRAMs in a breadboard?

Should there be an easy way to do such thing, I could test these three chips in a sandpit environment (since they are socketed), with alternative power sources, free of expert's noisy conditions to try to diagnose their health and try to build alternative configurations with resistors and capacitors...

best,

By hit9918

Prophet (2868)

Аватар пользователя hit9918

29-09-2016, 15:47

dip switch values to the port? Smile
a register write goes like this:
out &h99,value
out &h99,registernumber + 128

in BASIC it doesnt work because the interrupt handler does IN 99
which resets that flipflop telling whether VDP expects the first or the second byte
but on the board it should work

I got no idea in which mode it boots.
first candidate is to write register 7 to change the border color
but to get more than one color onscreen, you need to dipswitch some font to vram Tongue

By tvalenca

Paladin (728)

Аватар пользователя tvalenca

29-09-2016, 19:37

DamnedAngel wrote:

2. Both my 9128 and its VRAMs have sockets.
3. Tried several capacitors in both 9128 and VRAMs (the board already has old 0.1uF ones), with no perceptible change.
4. I rebuilt my YPbPr setup (GBS8200 connected directly to 9128's YPbPr outputs) and, although the colors get wrong as I mentioned before (someone hinted my it's because wrong impedance matching), this configuration allowed me to confirm that the stripes remain without Expert's RGB circuitry. Thus, (unfortunately) it is not a problem in this module.

2. Good!
3. That was kinda long shot also...
4. ok, we now know a little bit more of what we are dealing with.

DamnedAngel wrote:
flyguille wrote:

those vertical stripes is normal in VDP, is digital noise from the insides of the TMS/VDP chip, through the power line coupling in the analog output of the chip.

Given my results, I am agreeing with you. But I really question if the problem is normally that acute... I really doubt one would be satisfied with such deep stripes in screen. I still think there has to be a way to enhance a LOT the image I am getting.

He's absolutely right. I hope you are already considering that you may not get rid of these jailbars completely.

DamnedAngel wrote:
tvalenca wrote:

1) Maybe there's something wrong with your VDP. It's kinda long shot, but do you have another VDP to temporarly replace the one on your Expert?

I happen to have 2 Hotbits in my parent's house. Do you know if they use the 9128 as well - and, even better, if they are socketed?

Great, but I have bad news for you: Brazilian Sharp did not used sockets on Hotbits... if you plan to use those VDPs, you'll have to dessolder them. Anyway, you have to power up them to assure they're not producing jailbars... and Hotbit's analog board doesn't help that much (poor image quality).

DamnedAngel wrote:
tvalenca wrote:

2) you are having some kind of ripple on VDP power traces.

I will try to get my hands on an oscilloscope tomorrow to try to find out how my power lines go. Anyway, new capacitors did no good to the image...

Let's wait your results.

DamnedAngel wrote:
tvalenca wrote:

3) There are some ripple caused by bad layout (altough there's something else making it worse). Try adding some resistors like @maxis suggested.

Could you or maxis explain better how (and where - would it go in all 8 address lines?) those resistors should be mounted? Did I get it right that they would go in serial with the lines? What resistance values should they have?

I think he said in series, this means each resistor must be between the VDP and the VRAM. So, you need to break the connection and place the resistor, I think the safest way is to bend VDP terminals in a way they'll be parallel to the PCB, place the resistors on the socket, bend them to be parallel to the board, then bend the other lead back to connect the VDP terminal. You can see this on this sequence of photos:




DamnedAngel wrote:

Just another though I had during the night...

Would it be easy to somehow initialize and issue commands to 9128 to make its output somewhat colorful (i.e. 15,4,4 - or even 4,4,4) if I mounted it and its VRAMs in a breadboard?

Should there be an easy way to do such thing, I could test these three chips in a sandpit environment (since they are socketed), with alternative power sources, free of expert's noisy conditions to try to diagnose their health and try to build alternative configurations with resistors and capacitors...

it depends a little on how you want to do this. If you still want to use the MSX to send commands to this externally mounted VDP, you need to use A0-A7, D0-D7, /IORQ, /RD, /WR and GND lines, then use BASIC's INPs and OUTs do drive the VDP.

OR, you PROBABLY use a 8 position DIP SWITCH (or 8 small H-H switches) to represent DATA and one H-H and one push button to select the port (MODE pin) and pulse data in (/CSW pin) the way people used to program the IMSAI 8080. Never tried it thou and don't know if you could have any issue with /CSW signal timming. With this arrangement you can easily set video mode and backdrop color to anything you want, but to actually write something on screen you will have a (very) hard time.

hit9918 wrote:

in BASIC it doesnt work because the interrupt handler does IN 99
which resets that flipflop telling whether VDP expects the first or the second byte
but on the board it should work

What if you put the two OUTs on the same line? (OUT&H99,a:OUT&H99,b+128)

By hit9918

Prophet (2868)

Аватар пользователя hit9918

30-09-2016, 02:53

Quote:

What if you put the two OUTs on the same line?

it is no help, a line is not a disabled interrupt zone.
a second 9918 on another port could be coded with BASIC OUT. because it isnt touched by the interrupt handler.

Страница 4/5
1 | 2 | 3 | | 5