DIrect TMS9928 - GBS8200 YPbPr connection

Страница 3/5
1 | 2 | | 4 | 5

By DamnedAngel

Master (143)

Аватар пользователя DamnedAngel

22-09-2016, 01:57

Hi Maxis,

Thank you for your interest in helping me.

Anyway, what I got from your reply (I have to admit that the content was more technical than I could handle without further research, so please correct me if I got anything wrong) is that either the video output from my MSX is far from good or that its levels/formats are not that suitable for the monitor I am using.

I can understand that, but I think these arguments are pretty incompatible with tvalenca's previous considerations:

Quote:

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

(...)

The GBS8200 isn't a bad solution at all, but its far from being the best image quality you can get, any and every other option I gave you above are way better than it.

(...)

...the RGB matrix is good and internal amplifier is pretty decent) I wouldn't dare modding the analog board on a Expert computer, because it is a miracle it is working and have such a good RGB output in the first place.

TValenca left me the impression that I would be able to get a pretty nice image right out my Expert's output + LM1881 Sync buffering. I would not mind not getting the best image in the world, but what I got is very far from my expectations.

So my question is: If it is expected that an Expert provides a decent RGB output but it is not quite being able to do so, do you consider possible that I should have to renew my RGB output components?

Thanks a lot!

By tvalenca

Paladin (728)

Аватар пользователя tvalenca

22-09-2016, 03:36

Hey DamnedAngel!

I see you got a Samsung 710n, which is an OK monitor to hook on Expert’s RGB output. It doesn’t have the best scaler (and you need it, because TMS9128 outputs a total of 240 lines on active video (192 plus borders) but 710n is meant to display 1024x768 pixels. On the other hand it seems the monitor could lock perfectly the image, as you don’t have blurry pixel borders.

Actually, I’m not sure if I can see anything on the photos you have posted... but I know that this isn’t your fault; LCD monitors are tricky to take pictures of and, most importantly, most vintage devices have some degree of what we call jail bars. I'm really curious to see what you are getting. You shouldn't get anything moving over a still image, neither shaded/faded patterns over solid colors. Unfortunately, those jail bars are nasty to get rid. I sold my two Experts recently, so I can’t test it right now (I have a Samsung 510n which is basically the same monitor with a smaller LCD panel) but I know for a fact that you wouldn’t get a jailbar-free image getting signals directly from VDP. There are some ways to minimize them thou.

The most common cause to jail bars on MSX is the VDP itself, or the fact that it handles extreme power-hungry components, called DRAMs. The DRAM refresh operation generates electromagnetic interference. Also on Expert you ALWAYS HAVE TO BLAME THE POWER SUPPLY, so you may be having noise on the 5V rail itself or only around the VDP, but you need an oscilloscope and some specific knowledge to diagnose this. You may have an dried up capacitor (or some of them) on the analog board (the board where the PSU, Audio and video circuits are mounted). Considering how poor quality the components were here in Brazil when Gradiente produced Experts, I would begin looking for leaked and dried up capacitors to replace.

Oh, and Maxis, DamnedAngel’s MSX already have a similar circuit inside. I really suspect a mix of dried up capacitors and EMI.

By maxis

Champion (512)

Аватар пользователя maxis

22-09-2016, 11:49

tvalenca wrote:

Hey DamnedAngel!

The most common cause to jail bars on MSX is the VDP itself, or the fact that it handles extreme power-hungry components, called DRAMs.
...
Also on Expert you ALWAYS HAVE TO BLAME THE POWER SUPPLY
...
I really suspect a mix of dried up capacitors and EMI.

Hmm, IMHO, there are 2 different types of vertical lines or patterns (as you call those JAILBARS). Some are very regular, and there as you have mentioned (I agree with you) it is a result of interference with the pixel clock.

The other JAILBARS are not regular, but still synchronized to the VSYNC pattern. So, this could be the result of the beating of the two different frequencies. One is the MSX pixel clock, but the other could come from the switched mode power supply or any other independent oscillator (something could actually oscillate, but the transistor matrixing in Gradiente seem to be all DC coupled). They do synchronize via the VSYNC, since the power peak is repeatable on each field of frame. Could it be the switched mode power supply of the RGB->VGA converter?
Another IMHO, the Gradiente schematics, you have sent, has the linear power supply. These regulators do not oscillate. So it can't provide any contribution to these irregular large bars on the screen.

On the other hand, these irregular large JAILBARS can be the result of the scaler circuit aliasing (Gradiente TMS9928->RGB transistor matrices can go well beyond 100 MHz, i.e. beyond Nyquist of the TV/VGA scaler and produce high order frequency components, normally filtered by the old CRT TVs).

@DamnedAngel
Hence, these are the questions:
1. Do you connect your MSX to the monitor via the RGB->VGA converter? If yes, then how your RGB->VGA converter is powered?
2. When you see these large irregular bars and you change the brightness/contrast on the monitor do these bars move left/right?

By tvalenca

Paladin (728)

Аватар пользователя tvalenca

22-09-2016, 21:44

@maxis, I'm not sure if we explained this correctly, but there's no RGB-VGA converter on this setup. The VGA monitor he is using supports 15kHz resolutions without any kind of conversion. he is using just plain wires to connect R, G and B signals directly, and CSYNC signal goes through a LM1881 just because these monitors requires TTL level SYNC signals and MSX RGB output doesn't match perfectly. (He could have used a 74LS00 instead with same results) And another plain wire for GND, of course.

There are some monitors, discovered by trial and error (because this is an undocumented feature) that supports tv resolutions on the VGA input. These monitors are often offered in some markets as TV-Monitors or industrial-automation monitors. The Samsung 710n DamnedAngel owns *probably* has a related model that used to sold with industrial-automation equipment (CNC, Embroider/needlework machines, any industrial machinery with automated procedures) or as a replacement for RGB CRT monitors on industrial plants.

By maxis

Champion (512)

Аватар пользователя maxis

23-09-2016, 16:27

@tvalenca,

Oh, I see. So from the signal chain point of view it can't be any better. Direct YPrPb to the digitizer of LCD panel...
And LM1881 can't create such the artifacts.
Now, it seems that the question asked remains relevant:
- If we change the brightness/contrast on the monitor would it change/move the "large" bars position on the screen?

Also all my 9938 based Yamahas (and a couple upgraded with 9958) do produce the regular vertical bars pattern (well visible via SCART connection).
Yamahas don't have the line termination resistors on the VDP VRAM address bus.
However, the later models, like Panasonic FS-A1ST and Toshiba FS-TM1 (MEI for Toshiba) have the series termination resistors on the V99x8 address lines - no bars at all!
IMHO, the worst case is CX7M-128, where there are 2 banks of 4164 64Kx1 connected to the poor V9938 (16 chips in total). I have the motherboard, but actually use 4x 4464 instead of 16x 4164.

By tvalenca

Paladin (728)

Аватар пользователя tvalenca

23-09-2016, 20:22

maxis wrote:

@tvalenca,

Oh, I see. So from the signal chain point of view it can't be any better. Direct YPrPb to the digitizer of LCD panel...

Almost: The first two Gradiente models has internal YPbPr to RGB matrix conversion.

maxis wrote:

And LM1881 can't create such the artifacts.

Agreed.

maxis wrote:

Now, it seems that the question asked remains relevant:
- If we change the brightness/contrast on the monitor would it change/move the "large" bars position on the screen?

I don't see why it should, but as I sold my expert a few weeks ago, I can't answer that. Lets wait DamnedAngel answer that.

maxis wrote:

Also all my 9938 based Yamahas (and a couple upgraded with 9958) do produce the regular vertical bars pattern (well visible via SCART connection).
Yamahas don't have the line termination resistors on the VDP VRAM address bus.
However, the later models, like Panasonic FS-A1ST and Toshiba FS-TM1 (MEI for Toshiba) have the series termination resistors on the V99x8 address lines - no bars at all!
IMHO, the worst case is CX7M-128, where there are 2 banks of 4164 64Kx1 connected to the poor V9938 (16 chips in total). I have the motherboard, but actually use 4x 4464 instead of 16x 4164.

Well, definetely there's something about the DRAM refresh, so I belive you when you say that some termination resistors should help. But which resistors and how we should hook them? That would not be an easy task fitting them. Expert uses a TMS9128 VDP, so VRAMs are two 4416.

Curious that 16x 4164 VRAM bank. didn't thought anyone would use this way.

By hit9918

Prophet (2868)

Аватар пользователя hit9918

24-09-2016, 00:33

I got the same bars on 9929 composite. like every seond pixel a vram acess.

the color changes one pixel more early than the brightness. is the PAL encoder doing this?
for example color 6,7 . red cursor on blue sky. left of the cursor is a column that has the brightness of color 7 but the red of color 6.
the color changes before brightness.

when I switch my TV to s-video then I get a black and white picture and can see the overlaid wave that make the PAL colors!
I can see how those waves change ahead of the brightness.
like the early colors wasnt an interpretation of the TV but came this way out of the composite port.

By maxis

Champion (512)

Аватар пользователя maxis

24-09-2016, 13:49

tvalenca wrote:
maxis wrote:

Now, it seems that the question asked remains relevant:
- If we change the brightness/contrast on the monitor would it change/move the "large" bars position on the screen?

I don't see why it should, but as I sold my expert a few weeks ago, I can't answer that. Lets wait DamnedAngel answer that.

@tvalenca

For the large irregular bars, when going up/down constrast/brightness, it will change the backlight color of the LCD monitor. Hence the power profile of the built-in switched mode power supply. If the power supply output voltage ripple is the source of the digitizer artifacts, then the pattern will change/shift right left. If the large irregular bars will stay the same after altering brightness, then this is not the LCD power supply, but the result of the digitizer aliasing.

The termination resistor installation is simple. Cut the trace right at the VDP pin. Put the series resistor of 22-33Ohm of the 0603 size.

hit9918 wrote:

... I got the same bars on 9929 composite. like every seond pixel a vram acess.
I can see how those waves change ahead of the brightness.
like the early colors wasnt an interpretation of the TV but came this way out of the composite port.

@hit9918

Small regular vertical bars - a known problem. However, the chroma misaligned with luma suggests that the color modulator likely missing the delay line.

By DamnedAngel

Master (143)

Аватар пользователя DamnedAngel

24-09-2016, 18:53

Hi tvalenca, maxis and hit9918,

Thanks a lot for your contributions!

Sorry for the delay, but I was on business trip and just returned.

So here they go:

tvalenca wrote:

Actually, I’m not sure if I can see anything on the photos you have posted...

I think the following photo shows it clearly. It does not contain photo artifacts, All the vertical lines are the stripes I mentioned before.

The stripes happen also on R and G not only with B, EXACTLY at the same locations.

tvalenca wrote:

You shouldn't get anything moving over a still image, neither shaded/faded patterns over solid colors.

You are right. The stripes do not move. They stick to the image even I pan the image using the monitors controls to center the image.

tvalenca wrote:

Also on Expert you ALWAYS HAVE TO BLAME THE POWER SUPPLY

maxis wrote:

how your RGB->VGA converter is powered?

tvalenca was right on this: I am not messing with the standard RGB signal from Expert's output, only using an LM1881 to separate sinc signals. For the information, I am indeed using the 12v pin of Expert's RGA output, however:
1. I am using a 7805 regulator before feeding the LM1881. I do not have an oscilloscope to check how much ripple I still have after the 7805, though.
2. I tried a 12v switched font of an old TPLink router, and result was MUCH WORSE! With this font, the image got shaky.

maxis wrote:

Some are very regular, and there as you have mentioned (I agree with you) it is a result of interference with the pixel clock.
The other JAILBARS are not regular, but still synchronized to the VSYNC pattern.

Well, with my limited MSX-VGA experience, I easily regard the lines I am getting as pretty regular. So, according to what you stated, it may be related to pixel clock... however...

hit9918 wrote:

I got the same bars on 9929 composite. like every seond pixel a vram acess.

The stripes do NOT seem to align with pixels in my case. In screen 0, width 40, I counted 124 stripes in a line of 40 characters - a ratio of 3.1 stripes per character:

Additionally, when I change from screen 0 to screen 1, the stripes do not move a bit. In this mode, the 32 character line contains 131 stripes (not counting the borders), roughly 4.1 stripes every 8 pixel character:

So... I admit I had never heard of clock pixel interference, but... shouldn't we expect that such interference present itself in integers ratios with the pixels???

maxis wrote:

When you see these large irregular bars and you change the brightness/contrast on the monitor do these bars move left/right?

No, they do NOT move.

tvalenca wrote:

he is using just plain wires to connect R, G and B signals directly, and CSYNC signal goes through a LM1881

Correct.

maxis wrote:

Direct YPrPb to the digitizer of LCD panel...

Correct me if I am mistaken, but not quite that. 9928 outputs YPrPb, which is converted, internally in the Expert, into RGB, which is finally conected to the digitizer.

This conversion, inside the Expert, may be at fault...

maxis wrote:

If the large irregular bars will stay the same after altering brightness, then this is not the LCD power supply, but the result of the digitizer aliasing.

I really do not thing this is the case:
1. I don't think i have the irregular stripes you mentioned.
2. The do not vary when changing brightness.
3. The monitor does not present these bars when connected to a PC.

maxis wrote:

The termination resistor installation is simple. Cut the trace right at the VDP pin. Put the series resistor of 22-33Ohm of the 0603 size.

How can I come to know that this is indeed my problem - mainly with the non-integer ratio between pixels and stripes?

Thanks a lot everyone of you!!!!

By hit9918

Prophet (2868)

Аватар пользователя hit9918

25-09-2016, 00:23

yes the bars would have to be in sync with the chars to fit the vram acess theory.
and that on screen 0 you get a 6/8 different figure fits the vram acess theory, too. because screen 0 is the special 6 pixel mode.
but when things go 3.1 4.1 that is a mystery.
symptoms of a theory plus a mystery Smile

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.

Страница 3/5
1 | 2 | | 4 | 5