MegaDrive 1 CXA1645 mod

Can you elaborate?

 

Considering the 1145 has an internal oscillator, I think he wants you to do some microsurgery on the chip ;-) :lol:

 

It could be either.

The circuit you used is an amplifier, but with a gain of <1 (it's a current amplifier). I It's inefficient and doesn't seem well designed, but it's good enough because there's a significant DC offset coming out of the VDP. The problem is that the level is swinging too large... which is pretty weird! Typically a VDP with internal DAC will output a TV-ready signal to be buffered without any gain. To fix this level problem you'd have to attenuate BEFORE the amplifier or you mess up the impedance matching (as you're doing now with the series resistor). Impedance matching isn't especially important with SD video since the frequency is still low, but still for the best quality (correct level and no signal ringing) you want to do things correctly.

A video amp could be any non-inverting amplifier with negative feedback for linearity; in this case the common-collector amp you're using is fine and it can probably be simplified by removing the AC coupling at the input. When you do need gain though it's best to use a NPN and PNP pair of mixed topology, I'm not sure if the arrangement I'm thinking of has a name.. or use an op amp which make designing things simpler but they're obscure, expensive and most are too low bandwidth for video.

You can lower the level and still keep the correct output impedance by removing that and lowering the 10k resistor to GND a little bit, maybe try a 8.2k or 6.8k.

It's 3.58 or 4.43 MHz and generated by a square wave oscillator, square waves have exceptionally high frequency harmonics. High frequency signals can couple across close parallel wires such as PCB traces. Since the "jailbar" issue was caused by the blue amplifier being next to composite and chrominance has the highest frequency components, disabling it by way of the colorburst crystal might fix the problem without having to modify the board or build an amplifier; it'd only mean lifting a pin and the only loss would be of color through composite and S-video.

?????? No it doesn't, it has a built-in inverter and maybe feedback resistor to assist in constructing a Pierce oscillator with an external crystal. Colorburst needs <<1% tolerance so I'm not sure if it's even possible yet to bring on-chip; most chips with internal oscillators are RC circuits realized with transistors so are quite inaccurate.

 

Alright let me see if I got this right. I should remove the 470ohms after the amp. Then should I add a resistor BEFORE the amp on the RGB lines? Maybe something like 220ohm?

The 10K resistor to GND should be replaced with something slightly smaller. I don't have either 8.2K or 6.8K resistors but I have plenty 4.7K and 2.2K's I can put in series instead, so 6.9K. Would that be worth a try?

 

I think this one uses both a NPN and PNP transistor:



Will that work better? I have the parts to make it, but I'm not a 100% sure how it should be soldered to the MegaDrive :/

 

Ahh I was too sleepy when writing that, changing the 10k resistor will just adjust the amp's bias. Adding a series resistor in front of the amp might work, but it'll load down the prior stage (DAC). You can keep everything how it is and reduce the series resistor you added to the output or try another circuit. Without knowing what the VDP puts out it's hard to come up with the correct values though. I'm guessing this whole problem is that the DAC swings 1.4 V, so expects to have a series output of 75 ohms to divide the level in half to the proper 0.7 V. If this is the case I think all you need to do is connect the transistor's base to the input, collector to 5V, emitter to the 220uF cap, cap to a 75 ohm resistor which is the output.

And yup that's the two transistor arrangement but slightly tweaked, it's not needed here, but it makes a good video amp when you need a gain > 1.

 

I decided to try it out before I could read your reply and it's a great improvement 220ohm before the amp is way too small though so now I'm using a 4.7K and the image is almost perfect!

There is still a problem with the sky changing in brightness and still very little color bleeding. Although it's nowhere near as bad as before.



Here is how I got it all wired up atm:



What should I change next?

EDIT: Looking at the original picture on page 1 it's still a bit too bright. So I should add another small resistor to the input side. Wonder if that will help the bleeding/colorchanging?

EDIT2: Adding a 2.2K to the input (4.7K+2.2K) seems to give nice results. I can't see any colorchanging any more but I'll ask a friend of mine to come over so he can give his opinion

EDIT3: My friend agree with me. It looks pretty damn good now After I have assembled the MegaDrive I will do some more testing.

 

I'll give that a shot. Lifting which pin, though? I've had a look at a datasheet and it isn't obvious to me.

 

A bunch of comparison shots alternating between CXA1145 (default) and RGB amp:









The jailbars are almost completely gone!! One could argue that the new shots are a bit softer, but I doubt anyone would be bothered by that when playing. Color levels also looks pretty much similar to the old ones so I probably won't change it anymore.

I'll draw up the completed schematics later

 

Noticeably lower saturation which makes it look slightly softer (to my eyes anyway), but I'd take that over the jailbars any day of the week. Good work!

 

Thanks! I think perhaps the colors could be even better if I added yet another 1K resistor on the input side. But I don't want to bother with it right now

I'm not too sure how this will work with the 32X though... :/ I might have to move the 75ohms from the cable to the console so the signal is passed through them before entering the 32X.

I should receive a 32X first thing this week. I'll wait with the schematics until then

 

I'm glad you have something "working" but would you mind trying this?


The latest circuit you posted has too much going on, here are some problems:

-the 10R at 5V isn't doing anything helpful
-the 75R to GND isn't doing anything but wasting current
-the 10k replacement I suggested earlier won't do anything but lower the DC bias meaning if anything could have caused clipping
-the input is unnecessarily capacitively coupled which acts like a filter, but thankfully not a very good one here so it isn't noticeable (capacitively coupling the input of an amplifier should be avoided for this reason since it will only degrade.. since the signal is DC coupled out of the VDP there's no need to, and thus there's no need for the bias resistors at all)
-the new input resistor is working, but it's working by non-linear averaging 5V, 0V, and the 1.4V input and linearity is important for accuracy, though our eyes don't pick up on it too well (you're basically adjusting the screen's gamma curve a little right now heh)

 

Sure, if possible I'll try it tomorrow

 

un-related but how are getting those screen grabs?

 

32X is passthrough composite, isn't it? Is there an easy way to get RGB out of it? The lack of RGB has put me off buying one in the past.

EDIT: the XRGB has a screenshot tool, Twimfy.

 

Taken with my XRGB-3. Using the newest Japanese firmware it can send screenshots to your PC

Nope RGB Otherwise I wouldn't touch it either.

 

I'll be fucked. I wonder where I heard that? Maybe I was confusing it with a Tristar/Super 8, I know those are composite passthrough.

Time to look into a 32X.

 

It's confusing because I guess the MD divides the master clock down for the colorburst input so there is no Xtal leg to lift. You can lift pin 6 which the datasheet says is the inverter input, but they could have gotten that wrong (they have before) and it might be pin 5. Just don't cut the leg since there are no guarantees it'll fix anything. For a test I'd short pin 6 to GND and see if that clears up the problem then decide whether to lift or not. You could just keep the pin shorted probably but who knows it might burn up the output driver from the VDP.

 

Sweet, wish there was some way I could show you the picture from my MD1. It's just an RGB Scart cable straight to my 22" LCD and it looks absolutely fantastic.

Not to rub it in but maybe there's some difference between the chips or something on mine and yours.

 

I would like to see that as well. I own 3 model 1's and all suffers from the jailbars :/ You must have a very special version