SpritesMind.Net

And today I finally got the SMS levels checked out. Good chuck went into debugging my flashing hardware which turned out to be faulty in the end.

I got measuremests from SMS1, SMS2, MD1 and MD2.

MD1 and 2 produced same results, SMS1 and 2 I had a bit of different measurement points. The usual point at voltage divider gave huge DC offset on SMS2 and I couldn't do measurements properly because of it so I measured the VDP output instead.

Here's the measurements for MD1 and 2 :

mV, ratio, RGB
0, 0.000, 0
368, 0.387, 99
604, 0.634, 162
952, 1.000, 255

SMS1 :

mV, ratio, RGB
0, 0.000, 0
324, 0.354, 90
620, 0.677, 173
916, 1.000, 255

SMS2 :

mV, ratio, RGB
0, 0.000, 0
364, 0.350, 89
720, 0.684, 174
1052, 1.000, 255

MD1+2 : 0, 99, 162, 255
SMS1 : 0, 90, 173, 255
SMS2 : 0, 89, 174, 255
Linear : 0, 85, 170, 255

Note that the MD measuremests DO NOT match any of the MD native mode output, there is some analog trickery going on or there's separate DACs used for SMS mode output.

Here's the test ROM (it is big because I hacked one game I have been sloooowly working on) : http://www.tmeeco.eu/SMS/COLOR.SMS

TmEE co.(TM) wrote:Note that the MD measuremests DO NOT match any of the MD native mode output, there is some analog trickery going on or there's separate DACs used for SMS mode output.

Suddenly that makes bit 2 of register $00 (the one that trims bits 2-1 from each RGB component) even weirder. Given it's set in mode 4 (if I recall correctly from the Master System docs) I assumed it was just used to let SMS color go through (by overriding those two bits) but...

...actually, that'd make sense if the VDP needs to use different levels in mode 4.

SMS2 :

mV, ratio, RGB
0, 0.000, 0
364, 0.350, 89
720, 0.684, 174
1052, 1.000, 255

I was recently trying to figure how to simulate accurate RGB levels of MD1/MD2, MS1 and MS2 RGB models (that were afaik only sold in France) and was wondering about these measured voltages (that correspond to 315-5246 VDP outputs according to the initial post).

These RGB models used the same SCART adapter (called RVB adapter in France) to adapt the RGB
signals coming from the consoles for the TV, with some differences depending on the console:
- on MD1/MD2, it is connected to the outputs of the video encoder (CXA1145 chip or equivalent which is still used in RGB model)
- on MS1 and MS2, it is directly connected to the outputs of the VDP (315-5204 for MS1, 315-5246 for MS2).

I simulated the behavior of the RVB adapter (schematics are available here) with a 75 ohms resistor to ground on TV side in a SPICE tool and figured it requires around 1.4Vpp (with some DC offset) on the input to output 700mVpp on TV input.

This fits with the measured voltages on MD1/MD2 (according to datasheet, CXA1145 chip outputs 1.4Vpp for 1Vpp input so 0.952Vpp would give in theory 1.33Vpp at the output) and MS1 (0.916Vpp after voltage divider using resistors values from available schematics gives in theory 0.916x6.9/4.7=1.34Vpp at the VDP output) but the voltages measured at the SMS2 VDP outputs (1.05Vpp) seems very low to me.

I already noticed that my french MS2 had a quite dark video output on my old TV but I thought this was due to the console age and the ratio to expected voltage (75%) seems very low here. I am wondering if those measures on VDP outputs could vary depending on what is connected to the VDP (CXA1145 chip in PAL/NTSC models vs RVB adapter in RGB model) ? I found this old thread on a german forum that seems to indicate that the measured levels are different on MD1 VDP output depending on what is connected to it but I am not sure how to interpret it: https://circuit-board.de/forum/index.ph ... post506974

NB: also note in this thread how the RGB levels coming from VDP pins are linear when disconnected from the board then gives the known 'compressed' levels only when connected to voltage dividers and video encoder.

I think I made a typo and meant to write CXA output instead. Signal levels are near 1Vpp for full while there. It was so long ago so I am not totally sure anymore.

MD VDP RGB DAC output is roughly linear when unloaded indeed, but it is quite high impedance and the loading from the divider will cause nonlinearities that all models follow.

I don't know any real details about the original french hardware and what its special RGB cable does in detail.

TmEE co.(TM) wrote: ↑

Fri May 27, 2022 8:46 am

I think I made a typo and meant to write CXA output instead. Signal levels are near 1Vpp for full while there. It was so long ago so I am not totally sure anymore.

1Vpp for full white should be at the CXA input according to its datasheet with output being 1.4Vpp for full white still according to data sheet (700mVpp on TV input being obtained with 75 ohm serie resistor on the CXA output, assuming 75 ohm to ground on TV side)

Thanks anyway, the original post is indeed quite old so I understand it's hard to remember the details.