159 replies to this topic

[batari]

Given LHE's current absence and scarcity of the chips he used, maybe it's time for a new mod, and make schematics and Gerber files available.

I'm not very experienced with analog design, so some of this post may contain speculation so it might be incorrect. Since other members probably know some of this stuff better than I do, feel free to correct me or clarify anything I'm unsure of.

To start, I decided to reverse-engineer the original CD4050 circuit posted in the FAQ section of AA:
http://www.atariage....=2600#composite

It seems to work like this:

• The CD4050 buffers the LUM0-2 and SYNC signals from the TIA.
  
• The different-sized resistors are for a D/A resistor ladder to give the 8 different luminance levels.
  
• The series diode and resistor are to combine the sync signal into Luma. The diode is apparently to only affect the circuit when Sync goes low, and to protect the CD4050 from driving too much current when it goes high.
  
• Luma and Chroma outputs both have transistors, and these are configured as emitter followers, not amplifiers as previously thought. As I understand, emitter followers do not provide any gain, and they may even attenuate the signal. Their purpose is to give high impedance to the output signal.
  
• The 75 ohm resistors at the emitters of the transistors is to match the impedance of the video inputs into the television.

I could find nothing really unnecessary with the CD4050 circuit except for the large caps and the diode. There's really no reason for the 100uF cap, and even the 10uF is unnecessarily large. As for the diode, since Sync is a digital signal and must just pull the whole thing low when Sync goes low, and the TIA is open-drain, perhaps the buffer and diode here aren't necessary.

I decided to reverse-engineer the LHE circuit as well, and it appears that he just took other schematics and combined various parts of them. In particular:

• The Luma portion of this circuit is copied verbatim except for the resistors in the D/A.
  
• The Chroma portion that combines BLK and COL from the TIA is essentially the same as how the original 6-switch does it, except BLK is buffered on the LHE circuit.
  
• The resistors are roughly twice the value of the CD4050 circuit.
  
• The series 0.1uF cap and 75 ohm resistor at inputs are suggested from the FMS6400 datasheet (for AC coupling, so any DC offsets go away.)
  
• 75 Ohm resistors at output are as suggested from the FMS6400 datasheet.

Now, there's nothing wrong with taking other circuits and combining them, as that's essentially what I'm going to do. But, maybe if I understand how all this works, it may work better, and use fewer parts.

LHE didn't follow the CD4050 verbatim. He changed the size of resistors in the D/A resistor ladder at LUM0-2, which I think aren't critical, but the resistors at LUM1 needs to be twice the value of LUM0, and LUM2 is 2x LUM1. The size seems to only affect output current, not voltage. For now, I'll keep these as they are and experiment with sizes if the output current matters.

There are some design decisions that I don't understand about the the LHE mod. The CD4050 mod has two transistors and his has one. He used the emitter-follower transistor for Luma but used the same method on the six-switch schematic of combining COL and BLK for Chroma, then he added the 75 ohm resistor on both with series capacitors. One problem here is the Luma will have correct 75 ohm impedance due to the transistor but Chroma will not (due to the 600 and 1k resistors.) It's possible that his video chip doesn't care about the input impedance and only the series caps would be needed. If this is the case, and I suspect it is, there is no need for the transistor at Luma, or the various resistors around it.

Another thing I don't understand is why he buffered BLK, but still used the same method of combining the signals as the 6-switch. One issue is that TIA outputs are open-drain but the CD4050 is not. This could mean that you can't just copy the method of combining BLK and COL from the six-switch schematics, as when BLK is high, it will tend to pull up the signal with a CD4050 but the TIA would not. So either a diode should have been added here (similar to the one on the buffered Sync signal) or BLK should not have been buffered.

Also, if everything else is buffered, you might as well buffer COL as well. I read a post where someone asked why this wasn't done, and I think LHE said it was an analog signal, but it is not - it is actually a digital signal, but delayed so it will be out of phase as the phase of the signal determines the color. I read some comments on the video mod thread about the propagation delay of the buffer being an issue, so maybe combining the BLK and COL signals first then buffering them would be the best idea. This of course would rely on fairly even propagation delay at the CD4050 which probably isn't completely accurate but probably better than not buffering at all.

So...

I will start from the basic CD4050 mod at the link above, go over everything and see what can be removed (the simpler the circuit is without removing functionality, the better.) What I've done so far is eliminated the diode, the buffer on Sync, the 10uF and 100uF caps (only regular 0.1uF bypass caps should be needed for the chips), most of the 75 ohm resistors (only series resistors at output were used), combined COL and BLK before the buffer, then ran through a buffer gate. I've also tried to use the same parts wherever possible (such as same cap and resistor values for different places) as this makes for simpler construction. Total savings is 12 parts, I think.

This is what I have so far. Again, feel free to comment.



Also, I'm doing a parallel design with the MAX9512 chip, which appears to have similar capabilities as the FM64xx chips but is cheaper, more available and inputs can be DC coupled so (I think) you save the two series caps at input. The only issue with the MAX9512 chip is its extremely small size so most won't be able to solder it.

[akaviolence]

Did you read his website with the history of what he did and why?

[batari]

akaviolence, on Fri Jan 28, 2011 4:47 PM, said:

Did you read his website with the history of what he did and why?

Thanks. It only explains his experiments with combining the COL and BLK lines, though. Looks like a combination of what the Jr. and six-switch did (which has potential issues, as I noted above.)

[LittleJoe]

Excellent! I'm going to assemble one of these now and give it a try.

[batari]

LittleJoe, on Fri Jan 28, 2011 5:13 PM, said:

Excellent! I'm going to assemble one of these now and give it a try.

Can't wait to hear the results! In case it isn't clear, the connections on the left, from top to bottom, should be TIA pins 2, 5, 6, 7, 8, 9, 12, and 13. On the right, from top to bottom, it's VCC, Luma, Comp, Chroma, Audio R, Audio L, GND.

Not sure if the LHE instructions say this, but in this circuit, TIA pin 2 should be isolated either by lifting the pin and soldering directly to the TIA, cutting a trace or removing a resistor or something.

If it doesn't work well, we can always tweak it until it does.

[snafles]

Someone may want to try and get in touch with Chris Wilkson (cwilkson) who made the CyberTech board. He may be willing to release that design or at the very least add very useful knowledge to the project. The CyberTech mod seems to be one of the better ones.

[akaviolence]

batari, on Fri Jan 28, 2011 5:03 PM, said:

akaviolence, on Fri Jan 28, 2011 4:47 PM, said:

Did you read his website with the history of what he did and why?

Thanks. It only explains his experiments with combining the COL and BLK lines, though. Looks like a combination of what the Jr. and six-switch did (which has potential issues, as I noted above.)

It said something about why he used the 4050E and thought that might be relevant.

[batari]

That was because he thought the basic CD4050 was too slow. However, since mine buffers COL, the gate delays might not matter so much (though a faster part is probably better nonetheless.)

[RevEng]

Excellent start batari... this is definitely a needed project!

[LittleJoe]

Are TIA pins 6 and 9 supposed to be lifted along with pin 2?

[batari]

LittleJoe, on Fri Jan 28, 2011 11:15 PM, said:

Are TIA pins 6 and 9 supposed to be lifted along with pin 2?

Yes, these pins also need to be isolated either by lifting, cutting traces or removing parts from the board.

[LittleJoe]

batari, on Fri Jan 28, 2011 11:35 PM, said:

LittleJoe, on Fri Jan 28, 2011 11:15 PM, said:

Are TIA pins 6 and 9 supposed to be lifted along with pin 2?

Yes, these pins also need to be isolated either by lifting, cutting traces or removing parts from the board.

That's what I thought. My build from the schematic is complete. Here's the results thus far. I'm still searching for any errors I may have made. The second attachment is the color bars.

Attached Thumbnails

• 
• 

[wood_jl]

LittleJoe, on Sat Jan 29, 2011 12:05 AM, said:

batari, on Fri Jan 28, 2011 11:35 PM, said:

LittleJoe, on Fri Jan 28, 2011 11:15 PM, said:

Are TIA pins 6 and 9 supposed to be lifted along with pin 2?

Yes, these pins also need to be isolated either by lifting, cutting traces or removing parts from the board.

That's what I thought. My build from the schematic is complete. Here's the results thus far. I'm still searching for any errors I may have made. The second attachment is the color bars.

You are incredible. After I read you were going to try this, but before I scrolled down and saw this message, I was expecting a wait of weeks. Merely hours later, you've built one! Wow.

[batari]

The squiggles indicate an issue with sync, which is passed through Luma. You might need to add the 1.6k and 2k resistors back in (I thought these were to bias the transistor, but they could be a resistor divider to maintain a voltage level.) If you aren't getting proper sync, you may not see color either.

[LittleJoe]

Added the 2K and 1.6k resistors back in. Getting better. Color bars and greyscale:

Attached Thumbnails

• 
• 

[batari]

The colors are off - this is likely a result of buffering COL, which may delay the color phase a bit. To be sure, mark the location of your color trim pot on the 2600 board (so you can turn it back) and adjust it and see if it can be adjusted enough for the colors to look right. If it can't, we probably can't buffer COL.

Looking at the gray bars, it looks like the 18k and 36k resistors might be swapped, and the 9.1k might be connected wrong.

[doyman]

You guys rule, I wish I understood what you were saying in more depth so I could help out. I will have to sit on the sidelines and cheer.

I for one want to say thank you and I really appreciate the work you are doing!


If you need anything let me know I'd love to help out.

Edited by doyman, Sat Jan 29, 2011 8:48 AM.

[jsmith73]

doyman, on Sat Jan 29, 2011 8:45 AM, said:

You guys rule, I wish I understood what you were saying in more depth so I could help out. I will have to sit on the sidelines and cheer.

I for one want to say thank you and I really appreciate the work you are doing!


If you need anything let me know I'd love to help out.

SECONDED!

Ya'll rock outloud

[LittleJoe]

batari, on Sat Jan 29, 2011 2:12 AM, said:

The colors are off - this is likely a result of buffering COL, which may delay the color phase a bit. To be sure, mark the location of your color trim pot on the 2600 board (so you can turn it back) and adjust it and see if it can be adjusted enough for the colors to look right. If it can't, we probably can't buffer COL.

Looking at the gray bars, it looks like the 18k and 36k resistors might be swapped, and the 9.1k might be connected wrong.

I was not able to correct the colors using the trim pot. A second TV drops the colors all together. Everything is black and white. I've double checked the assembly and it all looks correct. I've taken COL off of the CD4050, and it didn't make any difference on the second TV. I'll test on the original TV later tonight. Then I'll add the transistor back in later to check my work. At that point I should be very close to a stock LE mod. From a working LE mod we can peel away parts and make changes.

[batari]

LittleJoe, on Sat Jan 29, 2011 6:35 PM, said:

batari, on Sat Jan 29, 2011 2:12 AM, said:

The colors are off - this is likely a result of buffering COL, which may delay the color phase a bit. To be sure, mark the location of your color trim pot on the 2600 board (so you can turn it back) and adjust it and see if it can be adjusted enough for the colors to look right. If it can't, we probably can't buffer COL.

Looking at the gray bars, it looks like the 18k and 36k resistors might be swapped, and the 9.1k might be connected wrong.

I was not able to correct the colors using the trim pot. A second TV drops the colors all together. Everything is black and white. I've double checked the assembly and it all looks correct. I've taken COL off of the CD4050, and it didn't make any difference on the second TV. I'll test on the original TV later tonight. Then I'll add the transistor back in later to check my work. At that point I should be very close to a stock LE mod. From a working LE mod we can peel away parts and make changes.

So did the colors change, but not enough to correct them, or they didn't change at all?

I would build one, but I used up all of my CD4050 parts. I've placed an order for more so I can build one as well to validate changes.

[LittleJoe]

Quote

So did the colors change, but not enough to correct them, or they didn't change at all?

They changed, but they are too vivid.

[batari]

I was just wondering if you could get the correct tint (such as red where red is supposed to be, instead of what's there now - it should look more like this Stella screenshot.)

If the colors are too vivid, it's a Luma problem, which might be correctable with smaller resistors for the 1.6k/2k (for example, both 1k).

[Longhorn Engineer]

batari, on Fri Jan 28, 2011 4:00 PM, said:

LHE didn't follow the CD4050 verbatim. He changed the size of resistors in the D/A resistor ladder at LUM0-2, which I think aren't critical, but the resistors at LUM1 needs to be twice the value of LUM0, and LUM2 is 2x LUM1. The size seems to only affect output current, not voltage. For now, I'll keep these as they are and experiment with sizes if the output current matters.

The original values in that mod are to low and close together. The top 3 luma levels blend together so the contrast of your image goes down

Quote

There are some design decisions that I don't understand about the the LHE mod. The CD4050 mod has two transistors and his has one. He used the emitter-follower transistor for Luma but used the same method on the six-switch schematic of combining COL and BLK for Chroma, then he added the 75 ohm resistor on both with series capacitors. One problem here is the Luma will have correct 75 ohm impedance due to the transistor but Chroma will not (due to the 600 and 1k resistors.) It's possible that his video chip doesn't care about the input impedance and only the series caps would be needed. If this is the case, and I suspect it is, there is no need for the transistor at Luma, or the various resistors around it.

The FMS6400 requires a 75ohm input impedance and a .1uf ceramic capacitor. If you look at the schematic theres a 75 ohm in series and one pulling it to ground. This sets up the input impedance. The 2n3904 does add gain to the luma signal.

Quote

Another thing I don't understand is why he buffered BLK, but still used the same method of combining the signals as the 6-switch. One issue is that TIA outputs are open-drain but the CD4050 is not. This could mean that you can't just copy the method of combining BLK and COL from the six-switch schematics, as when BLK is high, it will tend to pull up the signal with a CD4050 but the TIA would not. So either a diode should have been added here (similar to the one on the buffered Sync signal) or BLK should not have been buffered.

It is buffered to attempt to keep it more in synced with the luma signal. This signal is basically means "off screen" when the electron beam in a TV comes back across. Its not really needed but some TVs have a much better picture quality with it. Mainly older tube TVs.

Quote

Also, if everything else is buffered, you might as well buffer COL as well. I read a post where someone asked why this wasn't done, and I think LHE said it was an analog signal, but it is not - it is actually a digital signal, but delayed so it will be out of phase as the phase of the signal determines the color. I read some comments on the video mod thread about the propagation delay of the buffer being an issue, so maybe combining the BLK and COL signals first then buffering them would be the best idea. This of course would rely on fairly even propagation delay at the CD4050 which probably isn't completely accurate but probably better than not buffering at all.

You can't buffer the COL signal as IT IS analog. Buffers work by reading the value and propagating that through a gate. Running the COL signal through a buffer would basically be sampling the signal every 20ns or so and assigning it a high or low value depending on the voltage level of it.

editt// passing the COL through the buffer would probably do strange things to the phase.

[batari]

Longhorn Engineer, on Sun Jan 30, 2011 10:17 PM, said:

The FMS6400 requires a 75ohm input impedance and a .1uf ceramic capacitor. If you look at the schematic theres a 75 ohm in series and one pulling it to ground. This sets up the input impedance. The 2n3904 does add gain to the luma signal.

I am looking at the datasheet now, and there are no resistors at input, only the 0.1 uF caps, and I see nothing at all about required input impedance. At output, there is only the series resistor, and the resistor to ground appears to be after a 75 ohm coaxial cable (which I assume means an impedance matching resistor in the television itself.) It's not clear from the diagram alone, but looking in other datasheets for other video chips and other video circuits, this seems to be correct.

An emitter follower can at most provide unity voltage gain. So maybe it's not quite correct to say it provides no gain, but instead provides a gain of less than one.

Quote

Quote

Also, if everything else is buffered, you might as well buffer COL as well. I read a post where someone asked why this wasn't done, and I think LHE said it was an analog signal, but it is not - it is actually a digital signal, but delayed so it will be out of phase as the phase of the signal determines the color. I read some comments on the video mod thread about the propagation delay of the buffer being an issue, so maybe combining the BLK and COL signals first then buffering them would be the best idea. This of course would rely on fairly even propagation delay at the CD4050 which probably isn't completely accurate but probably better than not buffering at all.

You can't buffer the COL signal as IT IS analog. Buffers work by reading the value and propagating that through a gate. Running the COL signal through a buffer would basically be sampling the signal every 20ns or so and assigning it a high or low value depending on the voltage level of it.

editt// passing the COL through the buffer would probably do strange things to the phase.

I looked at the TIA schematics, and it is most certainly an open drain digital output. It is asynchronous and fairly high-speed, however, so a gate delay might mess up the phase. What I hoped to determine is if the gate was fast enough and the delay was consistent enough to work. This is my plan when parts arrive later this week.

EDIT: On another subject, I figured out your paypal address and sent you the royalties I promised.

Edited by batari, Mon Jan 31, 2011 5:34 AM.

[Longhorn Engineer]

batari, on Mon Jan 31, 2011 5:05 AM, said:

Longhorn Engineer, on Sun Jan 30, 2011 10:17 PM, said:

The FMS6400 requires a 75ohm input impedance and a .1uf ceramic capacitor. If you look at the schematic theres a 75 ohm in series and one pulling it to ground. This sets up the input impedance. The 2n3904 does add gain to the luma signal.

I am looking at the datasheet now, and there are no resistors at input, only the 0.1 uF caps, and I see nothing at all about required input impedance. At output, there is only the series resistor, and the resistor to ground appears to be after a 75 ohm coaxial cable (which I assume means an impedance matching resistor in the television itself.) It's not clear from the diagram alone, but looking in other datasheets for other video chips and other video circuits, this seems to be correct.

Read the application information on the datasheet. The Luma input needs to be driven by a low impedance source or a 75ohm terminated line. The 2n3904 causes a really high impedance on the luma line so the 75ohm resistors where necessary in this case.

If you remove the 2n3904 from the design as is it will produce a noticeably darker image on the TV. Its mainly for increasing the current load for the luma. The DAC portion pretty much sucks all the current out of the luma signal. If you got a CD4050 that could source more current you might be able to do away with it.