Designing a better S-Video mod, also maybe YPbPr

[samiam]

OK, here's what I'm thinking about an op-amp design.

- Op-amps can be configured in so many ways, with so many factors weighing against each other, it boggles the mind. However, a follower configuration, in which the output is simply equal to the input but can be driven at high currents, is about as easy as you could ask for. 

- The LT6552 from that document has two important parameters: an input resistance of 300k, and a maximum input voltage of 3V. Since pin 40's luma output peaks at 5V (4.2Vp-p wave riding on 0.8V) we can't put it directly into this op-amp's input. Also, if we wanted to use a resistor voltage divider to bring the signal level down, we would have to use very large values, and if it turns out that the stated 300k input resistance varies at all with the signal's voltage level, this approach would result in distortion. 

- I think that a good approach for an op-amp would be to buffer the signal from pin 40 with an NPN transistor, use a potentiometer on the output to scale the signal level down so that the waveform is 1Vp-p (standard luma), and feed this into an op-amp in follower configuration:


(Simulator link)


To pin 40, this would look just like the on-board amp. If you calculate the parallel resistance between the on-board amp's first transistor's emitter and ground, it comes to 1390 ohms, and that's almost exactly what we have between the emitter and ground here (370 resistor and 1k pot). We could even use a 2SC945 clone for the NPN. The 5.6k resistor, is almost certainly unnecessary and is there to make me feel good.

Since this is so easy to set up, I'd like to give it a shot. The next thing for me to do is see if there are any remotely suitable op amps available locally. Your typical off-the-shelf basic op-amp is no good for signals above 1MHz.

[dshadoff]

For a list of potential parts, I went looking at Mouser, based on equal-to or better-than specs (with an easier-to-use package), and came out with this list:
www.mouser.ca/Semiconductors/Amplifier-ICs/Video-Amplifiers/_/N-4gxg3?P=1z0xwwoZ1z0z5l0Z1z0y3waZ1z0z7v3Z1z0y3j8Z1z0z7zfZ1z0yhi1Z1z0z7v5Z1z0y2bm&Rl=4gxg3ZgjdhplZ1yvm5fdZ1yvm31nSGgjdhpmZ1yvcy6vZ1yvm47mSGT&Ns=Pricing|0

Not sure if that will be easy for you to access, so I'll explain it here:
1) It's actually under the topic of "Video Amplifiers" rather than "Operational Amplifiers".
2) Gain-Bandwidth product of 75MHz or more
3) Input bias of 20uA or less
4) Package type closer to PDIP, SO-14, or even SSOP
5) sort by price.

You'll see a list of possible candidates (which you can use to look elsewhere or whatever you like), and they all have links to datasheets.
Some of them even have multiple channels.

Hope this helps with the search.

[samiam]

Just as I suspected, the local shop's best op-amp tops out at 2.5MHz.

Thanks for the link, Dave.

AD818
MAX4310
LMH6715 (dual supply)
ADA4310
AD8072
OPA350

A few to investigate later on. A dual-channel solution would be nice, so that we could put chroma through the same type of amp and not have any phase-shift. Seems wise in principle.

If there is a really nice looking op-amp that needs a dual supply, a weak negative supply could be provided with something like a TL7660 and two capacitors. As long as the output never actually swings negative, it should be OK. My understanding is that running a dual supply op amp without the negative supply, even if it's supposed to be rail-to-rail, can result in some funny-business in the output near 0V.

[samiam]

After staring at the datasheets for a while, my assessment is that it's between the AD8072 and the LT6552.

The AD8072's advantages are that it comes in a big, fat, friendly 8-DIP package and that it's dual-channel. Its disadvantage, and this seems to be true across many so-called "single supply" op-amps, is that can't deal with input close to 0V unless you give it a negative voltage supply.

This is where the LT6552 excels. With a single supply, its minimum input voltage is still a straight 0V. This is important because I'll need to feed whatever op-amp I choose a pre-scaled signal, and that signal will go very close to 0V. However, to use the LT6552, I'd have to not only get two of them, but they'd be a hassle to work with because of how small they are.

The OPA2350 is dual-channel and has a minimum input close to 0V, but I can't get it via the most convenient online supplier in Japan and it also comes in a small package. LMH6715 is also unavailable. MAX4310 is more of a multiplexer, and ADA4310 doesn't explain enough in its datasheet, so those are out.

LT6552 would make more sense if we were going to have boards made up to distribute, but I sure like the idea of being able to test this whole op-amp idea on a breadboard with an AD8072. (EDIT: Or I could quit being a baby and get a SOIC to DIP adapter board.)

Neither being conveniently available, however, I think I'll go back to messing with transistors. In the end, these may still be the most DIY-friendly solution.

[sarge]

So out of curiosity, will running the older S-Video mod potentially cause damage to my system? I've got one that was modded by keithcourage that I bought several years ago, so I assume that the mod there is the simpler one. (He also helped me out a ton when I was having issues with the laser sticking, so thanks very much for that!)

[keithcourage]

There have never been any cases of anything being damaged from the simple s-video method. In fact, I don't think I've ever come across a system with a bad Hu6260 graphics chip ever unless someone destroyed it trying to do an RGB mod. This conversation is more about trying to improve the video quality if possible and if things can be made slightly safer in the process then might as well. I wouldn't worry about your system. 

[samiam]

Jan 6, 2020 20:27:52 GMT sarge said:

So out of curiosity, will running the older S-Video mod potentially cause damage to my system? I've got one that was modded by keithcourage that I bought several years ago, so I assume that the mod there is the simpler one. (He also helped me out a ton when I was having issues with the laser sticking, so thanks very much for that!)

Damage is one thing I'm quite confident is not a risk with the older mod. The amounts of excess current flowing even in the worst case are so profoundly small, the extra heat generated probably isn't enough to warm the chip even a tenth of a degree. 

As keithcourage said, the reasons for coming up with a new mod are all about quality. The old mod is simple and gets decent enough results, but a different design might get us something even better, if only by a small amount.

[samiam]

YPbPr might not be as easy as we first hoped.

I hesitate to put in a direct link, but if you google "monochrome and colour television pdf" you should be able to download a book by that title (sans pdf). For the adventurous among you, have a look at chapter 26, where the chroma signal is discussed in depth.

Three things I've noticed so far that would be relevant to a PCE YPbPr mod:

1. R-Y and B-Y are both "weighted" or reduced in amplitude before modulation, each by a different amount.

2. In NTSC decoding, luma needs to be delayed about 500 nanoseconds while chroma is processed.

3. In chroma, R-Y and B-Y are separated by a 90 degree phase shift.

I assume that the HuC6260 accommodates these parameters, and therefore we would need to "un-weight" R-Y and B-Y and make phase-shifting circuitry for Y and whichever chroma element comes first. In other words, the number of necessary components just went up. Ultimately, if the goal is to get RGB output with the PCE's YUV palette, it might be easiest to do like what dshadoff was saying and put RGB through a digital conversion phase. That, or just stick S-Video into an old decoder IC like the CXA1621S, which is still available through certain sellers.


As an aside, quadrature amplitude modulation is some pretty far out stuff.


EDIT: The more I look at this, the more it seems like even with R-Y and B-Y separated from the start, a YPbPr mod is just too ambitious for someone like me. 

Someone probably ought to make a thread about furrtek's efforts in uncapping and scanning the PCE's chips. Here's the HuC6260, with a shot of the YPbPr DAC in there to boot. This guy has a lot on his plate right now, but he might just get around to analyzing this so that we can get precise YUV values without having to demodulate and measure the chip's analogue output.

[sarge]

Jan 7, 2020 2:09:46 GMT keithcourage said:

There have never been any cases of anything being damaged from the simple s-video method. In fact, I don't think I've ever come across a system with a bad Hu6260 graphics chip ever unless someone destroyed it trying to do an RGB mod. This conversation is more about trying to improve the video quality if possible and if things can be made slightly safer in the process then might as well. I wouldn't worry about your system. 

Awesome. Thanks, y'all.

I am considering busting in there and removing that resistor, though. I don't use composite at all, so I wouldn't miss it.

[samiam]

Jan 7, 2020 17:04:17 GMT sarge said:

Awesome. Thanks, y'all.

I am considering busting in there and removing that resistor, though. I don't use composite at all, so I wouldn't miss it.

If you do, please let us know what you think of the results. For all this talk, I'm still not sure that a "better" mod is actually going to produce a perceptible difference. I'm actually considering picking up another white PCE so I can try the old mod on it and do a comparison.

By the way, if you can easily remove the tiny capacitor on the chroma line, you should. It's not such a big deal once luma is out of the way, but just in principle, it can't be helping.

Another thing I would say is bad about the old mod is the way it filters chroma out of composite video. Ideally, it would be using is a band-pass filter that eliminates everything above and below 3.58MHz with a fairly narrow margin. Instead, it uses a high-pass filter that, depending on where you tap composite video, passes everything above roughly 2.2MHz. That leaves plenty of room for luma elements to get through.

High-pass filter:

(Calculator tool)

In the old mod, the 1000pf cap corresponds to the capacitor in the image, while the 75 ohm termination resistor in the TV corresponds to the resistor. 

[samiam]

I have determined that the Sziklai pair solution I mentioned on the previous page won't work. The problem is that the high resistance necessary to attenuate pin 40's output provides too little current to the base of the NPN. The simulator doesn't account for this, but transistor bases have a cutoff current, and we're going well below that during the low parts of the cycle.

I will bite the bullet and purchase some LT6552s to experiment with. Right now, I'm just debating what to buy: Two at 1416 yen, seven at 3206 yen, or twenty-five at 7581 yen, including shipping. That's 708, 458, and 303 yen apiece. I really don't need that many, but damn, I hate paying so much more for small quantities.

[sarge]

Hmm, if it doesn't give a huge difference, I may just leave it alone. The picture quality still seems pretty decent, and well above what I get from composite. I've been having fun messing with it and my other S-Video-capable consoles since I got a RetroTink 2X. I hate that modern TVs have done away with the S-Video jack!

I have been mulling over eventually picking up an SSD3, given how finicky my Duo can be with burned media, so perhaps picking up a cheapy PC Engine would work well and I could try the alternative mod there.

[keithcourage]

Have you tried out a different brand of CD-r? Some Duos are just picky about the colored Dye used in the discs regardless of having a new lens and or making lens adjustments. Meaning if the discs you have been using are light green bottom discs(usually cheaper verbatim or memorex discs), you might want to try some that are dark green/blueish instead. The darker colored brands would be taiyo yuden or Verbatim Data Life CD-Rs.

[samiam]

Jan 9, 2020 17:40:08 GMT sarge said:

Hmm, if it doesn't give a huge difference, I may just leave it alone. 

That's the thing, though: nobody knows. I mean, I hesitate to suggest that the difference could be *huge*, but it could definitely be noticeable.

But of course it's understandable if you don't want your system to be the guinea pig.

[samiam]

Perhaps I'm beating a dead horse here, but another thing about the old mod that's kind of dumb are the transistors suggested for the Y amp. Take a look at their datasheets:

www.onsemi.com/pub/Collateral/2N3903-D.PDF

img.filipeflop.com/files/download/Datasheet_Transistor_S9013.PDF

For a 1Vp-p video signal terminating in 75 ohms, your peak current flow is about 14mA. Therefore, you want something with a fairly flat current gain parameter in the range between 1mA and 14mA. The on-board amp transistors' gain rises by about 5% in that range, but the 2N3904 and the S9013 are more like 20% and 25%. What this means that either your brightest colors are being overblown or, more likely if the 220 ohm resistor was chosen based on how bright the brightest colors looked, that your darker colors are too dark.