Source for CXA1645 or CXA2075 chips?

[D-Lite]

Hey folks,

I'm new in here so be gentle.  Matt and Lawrence have shown me the door, now I'm going down the dark path to video mods on consoles.

Specifically Turbo Grafx/Duo/SuperGrafx systems.

Problem is I need an RGB-to-S-Video converter and ain't no easy source for the Sony chips.  I'm of the opinion that the CXA1645 chip from early PSX (thanks Matt) may do the trick.

Anyone know where to get these?

I'd really not spend $80 for the JROK each time I do the mod.....

[Lawrence]

Jrok's PCBs don't work with PCE/TG16 consoles.  I've talked to him about it and he's as totally baffled as I am about the results: solid murky-green screen.  We went through a lot of troubleshooting, but he was unable to help me fix the problem.

The other issue is signal strength.  The PCE outputs very weak RGB and it needs to be amplified before it's used.  I have another encoder here with a Sony chip inside and it wouldn't give good results without using the amp.  The only easy method I've seen is the amazing Micomsoft XAV-2s, which uses Sony chips (the 1135 in early models, the 2075 in later ones) and happily uses an un-amped signal from a TG16 PCB.  It also handles arcade signals and pretty much anything else with aplomb.  It's as expensive as the jrok, but has dual inputs and far more adjustment options.  You can get 'em new in Japan for about $100, or I've got a couple of used ones for sale in the store.

I'm not aware of any source of bare sony chips.  A company in Japan sells a device called the AV Demiro which uses Sony chips to (Supposedly) convert RGB to composite/Svideo, but their boxes are, while small and consumer-friendly, very useless.  They don't work with arcade PCBs tho they say they do, and they don't seem to last long.  I got about 10 minutes out of one before it gave up and output only black and white.

IMO it's better to buy these things unless you plan on building them in bulk.  They're hard to make, fiddly as hell, and when you add up your costs and time, not really worth remaking.

Gutting older systems for parts sounds attractive, but the cheap consoles (Genesis 1 or 2) use the antiquated CXA1135, and the CXA1645 is only available in consoles that cost considerably more, like PSX units or Genesis 3s.  Add your removal time to the mix and you quickly get out of the worthwhile range.

L.

[smellyguineapig]

I modded a Turbografx with a cxa1645, the results where amazing, except for one problem, vertical bars that showed up on certain colors, I tried amplifying the signal, among many other things, same problem.

Funny thing, the cxa1645 would work fine with any Jamma RGB signal I threw at it.


You can also tap what looks to be Y/C signal off the video chip on the pce, but it does not look great, sharpness is there, but the colors are wrong.

Tried modifying/amplifying the signals with no luck.

[Lawrence]

FWIW the PAL TG16 uses a CXA1135 without a video amp, but most outboard convertors seem to expect brighter levels.

[matt]

Just dropping in here.  For the record, I did a TG16 mod a couple of years ago that used a 2075-based converter, and it worked perfectly.  I sold it to D-Lite last week so he could take a look at it.

I haven't tried using a 1645 with a TG16, but I've used them on a lot of my JAMMA projects and other console hacks to good effect.

The other issue is signal strength.  The PCE outputs very weak RGB and it needs to be amplified before it's used.  I have another encoder here with a Sony chip inside and it wouldn't give good results without using the amp.

I found the opposite to be true.  Normally, you need to use resistors to drop the RGB signal levels low enough to work on the Sony chips, but the TG16's RGB is weak enough that the CXA2075 will accept it directly.  The 1uF caps on the RGB input lines are required, however.

In addition, the CXA chips make really good RGB amps for TG16 mods.  It boosts the signal to just the right level, and it looks really sharp - a lot better than any transistor-based mod I've tried, and really cheap too.  I've taken to using CXA1145 chips yanked from old Genesis decks on TG16 mods.  You don't need to build the full circuit - just what's necessary to get the RGB passthrough running.

I'm not aware of any source of bare sony chips.

If you're in Japan, you can buy CXA1645p (DIP) chips from www.akizukidenshi.com for about 1k yen each.  They also sell RGB converter kits (which use the 1645p) for 26k yen.  There's an outfit in Argentina that seems to sell the CXA1645m, but they're expensive and I don't know how reliable they are.

I've seen some IC wholesalers who list the Sony chips, but of course they'll only ship to businesses and you'd have to buy several hundred dollars worth of merchandise per order.

A company in Japan sells a device called the AV Demiro which uses Sony chips to (Supposedly) convert RGB to composite/Svideo, but their boxes are, while small and consumer-friendly, very useless.  They don't work with arcade PCBs tho they say they do, and they don't seem to last long.  I got about 10 minutes out of one before it gave up and output only black and white.

That should be easy to fix - it sounds like the only thing that's happened there is that the oscillator's stopped working.  If it really does use a CXA chip, you can monkey around with the input resistors to get JAMMA stuff running.

Gutting older systems for parts sounds attractive, but the cheap consoles (Genesis 1 or 2) use the antiquated CXA1135, and the CXA1645 is only available in consoles that cost considerably more, like PSX units or Genesis 3s.  Add your removal time to the mix and you quickly get out of the worthwhile range.

The first time I tried making an RGB converter for my supergun projects, I used a CXA1645m chip that I yanked from a dead PSX.  It was a total pain in the butt to work with, but building your own circuit from irritating parts like that is a good way to learn how the whole thing works.  I wouldn't do it again, though.  If you want to make your own circuit, use the DIP version instead.

This page shows what's probably the most useful and cost-effective method - dead Saturns are cheap:
http://www.kitanaka.net/~tiff/arcade/video_conv.html

From a cost/benefit perspective, using Gen3s is a bad idea - you'll earn more for your work by modding them for S-Video and selling them than you would by destroying them and using the parts elsewhere.  I've seen a number of Gen2s that do use the CXA1645 - later ones only, though.  There's also a compatible chip made by Fujitsu that was used in the Gen2, but it's also pretty rare.  Maybe 1 in 10 Gen2s has a chip with S-Video output.

P.S. - The chip the Gen1 uses is "CXA1145" (I think I spotted this mistake on your Gen3 page also).

[D-Lite]

So let me get this straight.
I can pirate the CXA1145 from the early Genesis models and it'll be good enough for the TG16 mod?  I'm getting an RGB Duo-R from Lawrence (looking forward to it) as well.  Should be easy to get S-Video from that, right L?

Also, what is this:
http://akizukidenshi.com/catalog/items2.ph...1&page=#K-00018

I can't read kanji for crap.

Someone want to order me some of those CXA1645p and CXA1645 kits?

[matt]

The CXA1145 mod I was referring to is only good for RGB.  You don't want to try getting S-Video from it.

That link is indeed the RGB converter kit I was talking about.  It comes with all the necessary parts, including the chip.

[Lawrence]

the CXA chips make really good RGB amps for TG16 mods.

This is a fascinating idea.

dead Saturns are cheap

Funnily enough I was debating a similar mod, but I was going to cut out the section of the encoder and house it on a separate PCB.  Way too much effort to rebuilt the clock circuit tho, and sadly the chip and port are offset, so the size of the PCB needed kept growing and growing...

The chip the Gen1 uses is "CXA1145"

Ooops, damn.  I've been making this mistake for a long time.  ^_^

the oscillator's stopped working.

I'm making some assumptions, they used several ICs and like jrok's board all the part numbers were scratched off.  It's pin compatible with both the CXA1645 and '2075, so I'm making a guess it's one of the two.  The other ICs are likely just invertors or something, smoothing out the sync, but they're keeping me from properly sorting out the oscillator problem since I dunno what they're doing exactly.  If it worked, $40 for an encoder is a damned good deal.

Thanks for the post Matt.
 

[matt]

This is a fascinating idea.

I've been planning on drawing up a diagram of it for a while, but have never gotten around to it.  It's easy to figure out, though - just look at the 1145 data sheet and cut out everything that's obviously not needed.

Funnily enough I was debating a similar mod, but I was going to cut out the section of the encoder and house it on a separate PCB.  Way too much effort to rebuilt the clock circuit tho, and sadly the chip and port are offset, so the size of the PCB needed kept growing and growing...

It doesn't need to take a lot of space.  The part of the Saturn or Genny board that houses the encoder circuit is very small - maybe 1.5" square.  You don't need to keep the A/V out port unless you want to for aesthetic reasons.  Aside from the oscillator, everything's there and you barely need to rewire anything.  I've used this method a number of times in my supergun projects.

If you don't feel like making your own oscillator circuit (which is very easy), you can buy ready-made ones for $2-3.  3.58 mhz is a very common frequency and should be widely available.

I'm making some assumptions, they used several ICs and like jrok's board all the part numbers were scratched off.  It's pin compatible with both the CXA1645 and '2075, so I'm making a guess it's one of the two. 

Does it have a luma trap?  If not, it's a 2075.  Otherwise, the chips work almost identically, so if you're debugging the circuit you don't need to worry much whether it's one or the other.

The other ICs are likely just invertors or something, smoothing out the sync, but they're keeping me from properly sorting out the oscillator problem since I dunno what they're doing exactly.  If it worked, $40 for an encoder is a damned good deal.

Would you be able to take a snapshot of the PCB?  I could probably point out what wsa going on if I could see it.

The sine wave input is pin #6, so whatever's connected to that is naturally going to be the oscillator.

[D-Lite]

The CXA1145 mod I was referring to is only good for RGB.  You don't want to try getting S-Video from it.

That link is indeed the RGB converter kit I was talking about.  It comes with all the necessary parts, including the chip.

So that 1645 kit is OK for S-Video, yes/no?

Lawrence, would you be able to pick some of those up?
 

[matt]

Yes.

[D-Lite]

So, I've tried doing the S-Video mod with a CXA1645m pirated from a PS early model and can't get any video signal.  Something comes up on the screen as far as a blip, but no picture or even static.

This is on a US Turbo Grafx

I have the RGB lines and +5V connected to the chip, the 220 uF caps and 75 ohm resistors on the way out of the Y and C lines.  Grounded to GND 2 or GND1 doesn't seem to make a difference.  

?   Ideas?  matt, followed your system for the setup.  Question is, is the chip fried or did I just foul something up?  The system still works fine through composite or Rf video.   Do I need to put 1 uF caps on the RGB lines in?  I would assume I'd just see a screwed up screen if that were the case.  I did have them in line originally but no pic then either

Next is to try it on a Duo-R, but the cheaper TG16 is first.  I found +5V on that pretty easy on a trace running around the board.  

Also, I think I'm about to get a large supply of CXA1645P chips......  Will make this easier

[benzaldehyde]

I especially like the sound of amping the RGB from a TG16. May I enquire if this chip is used on Gen 1 models alone, or is in Gen 1 & 2. Let the butchery begin... :ph34r:    Thanks.  

[D-Lite]

The question I have, and I can't see a reason for it, is do I need to have the composite sync connected to the chip?

[Agentspikey95]

Do the newer sony chips in the CXAxxxx series ferform the same function as the 1645, 2075, etc.?

(i found docs on the 1645p and M, they might be still in production...)

[matt]

Question is, is the chip fried or did I just foul something up?  The system still works fine through composite or Rf video.

Could be either one.  But, it sounds like you've messed something up - that's really easy to do when you're working with a tiny SMT chip that you've never used before.

Do I need to put 1 uF caps on the RGB lines in?  I would assume I'd just see a screwed up screen if that were the case.  I did have them in line originally but no pic then either

Yes, you need the caps.  You won't get anything if they're not there.

The question I have, and I can't see a reason for it, is do I need to have the composite sync connected to the chip?

Of course you do!  How do you think you're going to get a stable picture with no sync?  Use an LM1881 to grab it from the composite pin - the CXA chips don't like the TG16's sync output.

[matt]

Do the newer sony chips in the CXAxxxx series ferform the same function as the 1645, 2075, etc.?

(i found docs on the 1645p and M, they might be still in production...)

They're similar, but there are differences.  Documentation on the 2075 is out there, if you search hard enough.

[D-Lite]

The question I have, and I can't see a reason for it, is do I need to have the composite sync connected to the chip?

Of course you do!  How do you think you're going to get a stable picture with no sync?  Use an LM1881 to grab it from the composite pin - the CXA chips don't like the TG16's sync output.

matt, knew I could count on you to respond.

Looking at the prototype I got from you using the CXA2075M, the CSYNC isn't input into the chip.  Ain't there.  Does it need to go through the LM1881 for the S-Video output?  The CXA1645 chips take both Composite Sync and also filtered Sync inputs, neither of which are input into the chip on your setup.  I figured I needed this, but it wasn't there on the original I'm going off of.  Thanks for the input (no pun intended)

I'll connect it all up tonight and try again with the LM1881 inline

[matt]

Looking at the prototype I got from you using the CXA2075M, the CSYNC isn't input into the chip.  Ain't there.  Does it need to go through the LM1881 for the S-Video output?

Yes, it's there - look a little harder.  If you're having trouble seeing it on the board, just follow the wire from the LM181 circuit and you'll find it.

The CXA1645 chips take both Composite Sync and also filtered Sync inputs, neither of which are input into the chip on your setup.

Not sure what you mean by this - there's only one sync input pin and it's the same for all versions of the chip.

[D-Lite]

Success!  Well, sort of.

The problems I was having were several fold, including a missed ground and another +5V on the chip.  Slapped the LM1881 in there and got it all hooked up and got a picture!  Very sharp image.

But Black & White.

Here's the strange part.  I had originally tapped the composite signal to an RCA terminal and for the LM1881 I simply tapped onto that post.  Signal is B&W as mentioned.  BUT, if for some reason I plug an RCA cable into the jack that I tapped into, the screen turns to color!  EVEN MORE strange:  Doesn't matter if that plug is connected to something else or not (meaning, grounded, at least).

?

also, the image is a bit dark in either B&W or color.  Thoughts?  I'm using a CXA1645M pirated from a PSX.  Right now I have a mess of wires, some of which might be too long and causing issues.   I do also see some faint vertical lines as well.  

!!! So close  !!!!!

And one thing I should mention is that either 1 uF or 0.1 uF caps work the same on the RGB lines into the chip.

I might also mention I'm getting the exact same results on a TG16 (RGB from the expansion bus) and a Duo-R (RGB from the Hu6260 chip)

[matt]

Black & White

Did you add the oscillator?  It's necessary in order to get proper colour.

[D-Lite]

Black & White

Did you add the oscillator?  It's necessary in order to get proper colour.

Nope.  Did not.  Not sure I knew I needed it.  Was it on the board on your chip or did you add it?  

[D-Lite]

I see on a different diagram a 3.58 MHz oscillator with a 47 pF cap and a 2.2 kiloohm resistor going to the SCIN input (pin 6).  Seems it was on the original PCB you had matt.

Damn it.  Gotta build all these things in.
This sound correct to you?

[matt]

I see on a different diagram a 3.58 MHz oscillator with a 47 pF cap and a 2.2 kiloohm resistor going to the SCIN input (pin 6).  Seems it was on the original PCB you had matt.

Damn it.  Gotta build all these things in.
This sound correct to you?

Yes, that's it - 3.58mhz.  You may be able to get by without the cap and resistor, but it's probably a good idea to leave them in.

*everything* on the circuit diagram is required.  There's a reason why Sony says you should include it.

[benzaldehyde]

Hi,

     Not meaning to be too much of a bother, but Matt, could you clarify the amplification scheme for say... a TG16? I'm looking at a CXA1145 datasheet, and I note a few caps and some 75ohm resistors are placed in the RGB lines. I was wondering how you use the IC? All you need is +5V and the RGB lines, correct? Thanks.

[matt]

You do need to add a few more components, and the sync needs to be connected also.  There was a thread by another fellow a couple months back that did a good job of explaining it; you can surely find it if you search a little.

[benzaldehyde]

Ah, thanks Matt! I managed to find the post entitled (cleverly enough) CXA1145 as an RGB Amplifier. Who would've thought looking could solve anything? (Self-depricating sarcasm...   )

[D-Lite]

I see on a different diagram a 3.58 MHz oscillator with a 47 pF cap and a 2.2 kiloohm resistor going to the SCIN input (pin 6).  Seems it was on the original PCB you had matt.

Damn it.  Gotta build all these things in.
This sound correct to you?

Yes, that's it - 3.58mhz.  You may be able to get by without the cap and resistor, but it's probably a good idea to leave them in.

*everything* on the circuit diagram is required.  There's a reason why Sony says you should include it.

I'm nearly there.....

One n00bish question, what type of oscillator should I go with?  I'm getting some 14-pin TTL oscillators that match a diagram I found, but how about the regular crystals out there?  The one on the PCB you had Matt is an HC49 type I think with only 2-pins.  Hell if I can figure out how to use that type here.  I assume I actually need to build a circuit for that?

I just got a hold of a supply of CXA1645P chips!!  No more cutting up PSX's for me!

[matt]

You can use any sort of oscillator you want, as long as it's the proper frequency, is easy to install, and is cheap enough.  Don't bother with fancy multi-frequency models, since they're extra work and you'll only ever need 3.58 mhz.

If you want to save money, you can build your own oscillator from a crystal and a 7404 chip for about half the cost of a ready-made one.  This is what's being used in the TG16 I sent you.

[D-Lite]

OK, so now a couple things.

Added an oscillator (a "14-pin" one with 1= NC, 7= GND, 8 = output, 14= Vcc) to my circuit that used a CXA1645M from a PSX and now I have color!  BUT, I see a lot of green "smearing", although the image is fairly clear.  The same setup gives the same result on a TG16, PC Engine Duo-R, and a PC Engine Duo.  ??  I put the 2.2 k resistor and the 47 pF cap in with the oscillator and it seems to make no difference.

Now here's the weird/frustrating one.  Based on my partial successes with the PSX chip, I slapped all the components on a breadboard including of a CXA1645P chip (DIP) I got.  Modelled the board based on:  my previous setup, matt's prototype with the CXA2075, a JROK converter, and some web documents.  All this and all I get is a black and white horizontal rapidly rolling screen.  A second circuit I made, same set-up, doesn't even give the rolling screen.

So basically I need to know what definitely needs connecting to the CXA chip.  How important is it for me to have the Iref and Vref (pins 13 and 14 connected)?  And the YTRAP shouldn't matter at all since I'm not using this for CVideo.

So close.....  Perhaps I'll post a pic.

[matt]

So basically I need to know what definitely needs connecting to the CXA chip.

Everything in the circuit diagram on Sony's data sheet needs to be connected.  You won't get anywhere by leaving components out - Sony didn't slap extra parts onto their design just for fun.

The only things you can really get away with omitting are the Luma trap and the resistors/caps on the composite and RGB outputs (assuming you don't plan to use composite or RGB).

It's not surprising that you're having mixed results with your homemade circuits.  The RGB converter circuit is rather finnicky, and if you're not experienced you will run into problems that you know nothing about.  Not knowing anything about the basics of RGB or S-Video before going into this project really has the odds stacked against you.

This is why I suggested using the RGB converter circuit from a Saturn - most of the components will already be mounted for you (in a properly designed circuit to boot).  Not only will it be a lot less work, but you won't have to worry so much about your own mistakes.

[D-Lite]

SUCCESS!!!!!

I connected the other points that needed to be and it works!!  With the completely homebrew set-up starting from a CXA1645P, not the PSX chip.  And the brightness is even good. Compares favorably with standard composite (not HUGE, but definitely better).  The only thing I didn't do from the Sony datasheet is decouple the power with the 47 uF and 0.01 uF caps.  Also the oscillator doesn't seem to need the 47 pF cap and 2.2k resistor.  Although I think I'll add all those to see if it does make a difference, but it works great!

Thanks for all the help Matt!

Dean
 

[matt]

Congrats!  That mod is no small job for a beginner.

As for the 47uF decoupling cap, I've found that I sometimes get some rather bad "smearing" effects if it's left out, so you'll want to include it.  This is especially evident with whites and really bright colours on black.

You'll also want to test your circuit over a long period of time (at least 2 hours) to make sure it works reliably and doesn't flake out after heavy use.