New 1-Chip RGB bypass

The place for all discussion on gaming hardware
Ryeno
Posts: 89
Joined: Tue Nov 05, 2019 6:50 am

Re: New 1-Chip RGB bypass

Post by Ryeno »

NewSchoolBoxer wrote:
VajSkids Consoles wrote:mouser is my new go to.
element 14 send bit by bit and once some zener diodes turned up nearly a year after order lol.

I'll do some research on how to measure slew on my scope shortly and post results as these ICs are very clean, legitimate as far as I can see.
A year later :D they only needed to be past 90 days to prevent a chargeback. I'm thinking you need at least a 100 MHz scope to estimate a 250 V/us slew rate when you can't do much beyond 1V to 2V at 3.3V supply. At least you don't need a function generator. Just turn on or off a DC supply with a resistor. Perhaps the easiest thing to measure that is significant is the two capacitive load charts on page 7 of the datasheet. High capacitive loads make opamps unstable.

Random output swing, let me know if that doesn't go away under 75 ohm or 150 ohm load. I can pull up SPICE models and play around or ask on Reddit.

I took out the gain factor of 2 in the minimum bandwidth needed in my above comment since that's for the gain-bandwidth product. Must not be given for fixed gain opamps for surely they can handle their own gain. Calcs I took from here show 0.5% max error for opamp based on being 10x faster (120 MHz to -3 dB point?) than the gain bandwidth product and 0.005% for being 30x faster. At 700mV to 1.4V white, the 0.5% error is 7mV, which is exactly 1 IRE unit that we definitely want to be under. More than 30x faster (350 MHz?) for max 1/10 of an IRE unit starts becoming overkill imo.

Explanation of higher frequency being better is greater opamp's bandwidth, faster it realizes it overshot the gain of 2.

I searched triple video opamps on Mouser for 15 minutes and found a few more that would be totally fine to use at sub-$5/100: AD8141ACPZ-R7, ADA4856-3YCPZ-R7, ADA4859-3ACPZ-R7, ISL59830IAZ, THS7303PWR, THS7353PWR, with the THS having optional filters.
Can't you use a CLC4601?
User avatar
buttersoft
Posts: 383
Joined: Sun Jul 24, 2016 7:49 am

Re: New 1-Chip RGB bypass

Post by buttersoft »

VajSkids Consoles wrote: The LT6550 is available on Ali express for a reasonable price.
Got a link? Best i can find is a single listing for 1x at $4.71 + $6 shipping. there are two other entries on Aliexpress for like $37, though they are probably for 10x though they don't say so.
User avatar
NewSchoolBoxer
Posts: 369
Joined: Fri Jun 21, 2019 2:53 pm
Location: Atlanta, GA

Re: New 1-Chip RGB bypass

Post by NewSchoolBoxer »

Ryeno wrote: Can't you use a CLC4601?
Thanks for showing me this. It's an extremely powerful opamp at 350 MHz bandwidth and 1500V/us with low noise. Use it if you got it sitting around. The reason why you see most people use video opamps is they already have the resistors inside to fix the gain to 2. CLC4601 datasheet on page 10 recommends 2x 510 ohm resistors for gain of 2. In other words, 3x2 = 6 total for RGB. You want surface mount 0.1% tolerance to keep the gain consistent across channels but 1% through hole is fine if white on TV is correct white. Very high slew rate can lead to instability or high frequency noise but we're talking about a gain of 2 on 0.7V at 6 MHz input or less, so not concerned.

I feel like I'm hijacking VajSkids Consoles' thread so I'll lie lower after this.
VajSkids Consoles wrote:There is up to around ~200mV of random output swing so I took measurements on the input
-That is very high. SNES power supply is far from ideal. Can add 1uF and 0.1uF bypass capacitors in parallel to Vcc and see if that helps. Else measure power supply ripple.
-The 1uF capacitor you add on input is common to AC couple but not strictly needed. See if removing helps. If not, if opamp has a high value resistor to ground like THS ones do, a 0.1uF or 0.047pF cap may work better.
-The cap + 5k resistor is high enough of a time constant to cause ringing. If 1k resistor works fine then use that.
-Small output with zero input is expected due to opamp's transistors being powered on from Vcc and needing small bias current. We don't know the internal resistor values for gain but 1/2 the value for a pull-up resistor on the + input to ground can help by giving bias current another path to flow. I'd guess 250-300 ohm.

I think I got the gain-bandwidth product figured for LT6550. I zoomed in and measured it out to be between 135 and 150 MHz, which is above the 120 that I wanted to see.
Spoiler
Image
Reason why video amps have gain of 2 is 75 ohm series resistor + 75 ohm termination in TV form a 1/2 voltage divider. Gain gets set back to 1 for correct level, versus being 1/2 of the correct level. In effect then, 0.1% error for 1 IRE at the amp output is just 1/2 IRE to the TV or scaler and it'd be 1/4 IRE error with 0.35V input for medium brightness. So I still think 120+ GBW product is totally fine but more is nice.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

that was under load, connected to the PVM but prior to hitting the set. TBH I was expecting the sets 75r pulldowns to be active at the other end of the cable and bring it back down to .7vpp, but it jumps around anywhere from 1.4v to 1.6v after its gain of 2x.

inputs are rock steady and each channel jumps between the measurements stated. (+/- 20mV)

200Mhz scope - SDS 1202X-E
very new equipment for me, lots of settings to play with.

I've jumped on the system for 15 minutes or so where ever possible and played a variety of games, it's extremely crisp and nice... so the measurements haven't really bothered me.

it would be good to get a known list of games or scenes in games where ghosting artifacts are prevalent.

the ones in the original proto are all clean now.
no missing scanline phenomenon like on the THS7374 circuits either.
Last edited by VajSkids Consoles on Fri Jun 17, 2022 1:55 am, edited 1 time in total.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

Also, you keep referring to the original post of the junked prototype schematic.
I will draw a new one, so you don't keep getting confused.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

https://imgur.com/a/pIofOnH

**I meant grab RGB points off the removed resistors, but caps work too.

Edit: I found a game with a lot of scenes that cause some ghosting - megaman x. They say after lots of testing it was found that 470nf was the minimum. I have 1uf on hand and that's not helped at all on both a JR and Phat SFC. (C11 fix)
These projects are very easy to fall down a rabbit hole with, I'm also thinking I would like to try 135r pull-downs as oppose to 130r, but every time a change like this is made, its ugly botch jobs or a new order with mouser. $60AUD minimum for free shipping. I just dropped another 75 not long ago on new components. Let's say little projects like this can add up to MANY hundreds of dollars very quickly - could make some back on pre-mods in the future but probably never break even :P

I won't lie either. I killed a $250 JR doing ugly botch job prototyping when I shorted out a resistor under the board and killed the VRAM low (it's broken into 2 banks, high and low).

Unfortunately switching out the VRAM didn't fix it, tried a few. So it's the actual proprietary IC that was damaged. Either way, another JR is on the way, and i won't make that mistake again :P
Trust me guys, run some wires across to a breadboard for all your prototyping!

Also the first 0.8mm batch of proto PCB's weren't that cheap!

I spoke to Rama and he says the ghosting is pretty much remnant of Nintendo's cheap DAC solution... Why C11 hasn't helped on 2 consoles, I don't know.
That's the only game I can find at the moment with a bit of ghosting.

It probably wouldn't bother most people, but when you're doing things like this, you get a thirst for perfection. Overall 80% happy.
(also, on most ghosting, strangely enough -the angle at which this PVM is viewed can make it appear/disappear)

edit again: there's one more component on the SNES mainboard i am interested in. Another resistor which I came across on another site which is suppose to be a single resistor brightness fix, as oppose to others (which this doesn't use). Let's keep this as WIP :) You also need to try the PCB's, to make sure it's not your mess of wires breadboard causing issues as well. It's all time and expenses.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

So the component of interest is an in series resistor to the proprietary '1-chip' IC - R3

Strange , very strange.

I can see from other documentation, this is a 1k6 resistor on the phat and an alternative brightness fix has being increasing R3's value. (as oppose to 'increasing the load' on R6:R8 by adding further pull-downs, or in my case -simply decreasing the value of these resistors).
Now no one goes probing around a board when trustworthy information/ documentation exists...

but can anyone explain how when i switch this out to a 2k resistor (I knew this would further darken the screen) the NTSC sub carrier is now hitting my set? I've got R18 removed, and additionally removed C50 as the composite pin buzzed to this component as well. using clean C sync, there is no sub carrier. Sub carrier is on the disconnected composite line. My brain is melting. My PVM shows which sub carrier is hitting the set via 3 illuminated lights (Yes, even on RGB SCART).

Yes, there's 2 sub carrier frequencies, PVM shows PAL60/NTSC443 , straight NTSC and straight PAL. NTSC illuminates, disappears when back to 1k6.

I grabbed C sync off a removed components footprint as per other guides. I'm baffled. I will keep findings posted here for shits and gigs. Even if i'm talking to myself :P
User avatar
Lopenator
Posts: 214
Joined: Sat Nov 14, 2020 10:40 pm

Re: New 1-Chip RGB bypass

Post by Lopenator »

If it means anything to you I read it although I don't know what it means.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

without any nerd talk, this is a source of interference.
the reason we use pure C sync is to get rid of sub carrier cross talk/ interference. It's likely leaking badly. I haven't being spending time on this project .. just walked past the 'lab' and switched out the resistor.
leaky sub Carrier is the cause of most of the jailbars/interference on megadrive /genesis.

this needs to be investigated :P
User avatar
Lopenator
Posts: 214
Joined: Sat Nov 14, 2020 10:40 pm

Re: New 1-Chip RGB bypass

Post by Lopenator »

Listen, I got faith you'll get to the bottom of it no doubt.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

well it's just a theory

no tools or probing

1: that's not the right place to grab C sync

or

2: if increasing the value of R3 increases cross talk so heavily the sub-carrier is picked up by the monitor clear as day on the C sync line, then reducing the value of R3 could decrease the crosstalk to a decent enough level it eliminates or at least reduces sub carrier crosstalk.

edit: 3: alternate method of pre amp signal attenuation , what's cleaner? in series or pull down? I.e. most people know you can drop TTL C sync to an appropriate level with a series resistor, but you could also use a pull-down resistor.

maybe we could 'rebalance' the signal. Reduce R3 which would increase brightness then reduce R6:R8 to pull the signal down further to the correct brightness after the 1-chips output to the RGB amps input.

only thing is, no spec sheets exist for these Nintendo IC's. We risk overdriving the 1-chip and burning it out.

once again, all a theory. I'm not 100% on what's happening..

I won't have time for a while to revisit this properly.
and it's way past my bed time and I just got badly growled at by the missus. I'm in trouble :P
Last edited by VajSkids Consoles on Sat Jun 18, 2022 1:53 am, edited 1 time in total.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

also not sure how this would affect S video / composite etc..
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

So I was using, at least, what i think was TTL C sync when i posted lastnight.
I added a 390r resistor (only because it's what i found first in the draw and sat between 330-470) on pin 20 in the SCART head.

Mainly fixed - either way, sub carrier noise is there, and shouldn't be.

https://drive.google.com/drive/folders/ ... sp=sharing

This is on the OSSC (yes, i forgot i was using TTL and ran it into the scaler the other day for photos, whoops). It also caused jail bars on blue/blacks which are now gone.
So that's how minor the issue is. You can't see it on a CRT.

I will try sync on Luma before I go messing with pulldowns, pullups, and different values of resistor around R3. This is rabbit hole stuff as mentioned.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

Ok projects over. Im getting my ass absolutely kicked... and i have quite a few customers mods to do.

So I'm going to release this board, provided there are no mistakes on it :P

With one additional note and fix.


These just look that little bit dim at 0.7vpp (reading off red channel, as mentioned it seems to taper down a little to green and again to blue)

They look perfect at 0.8vpp, and I don't think this is going to harm a scaler or PVM/ Consumer CRT. It just looks 'right' at 0.8vpp
This is easily achieved by adding a 10k pulldown to one of the sides of R3 - it's gives it a little kick in brightness (~+100mV) that just looks right to the eyes. (This is only for phats, so on a JR i'd advise installing a 1k6 as per a phat, then deciding whether you want the 10k optional brightness fix or not)

As for the C11 fix, as mentioned using 1uf as it's on hand - this has killed the top few scanlines on some games. I don't think this helped ghosting at all, my ghosting seemed to be a sync issue. This means I don't think TTL is suitable off these boards. The attenuating resistor must also act as a sort of filter for the sub carrier noise.

So that's it.
Slight ghosting might be something you have to deal with. I certainly don't want the top scan-lines missing either.

Install the board or a DYI mod, if you aren't quite happy with the brightness - add the 10k resistor @ R3. (The side between R32)
OTHERWISE, the value of R3 with the addional 10k becomes 1.37k, you can get these on mouser @ 0.1% tolerance. So you could switch it out with that, on a phat or JR.

Here's what it looks like on a phat SFC:
https://drive.google.com/file/d/1Dfxg5K ... sp=sharing (optional brightness kick with 10k resistor)

*facepalm* lost me original C11 cap, stuck with missing scanlines :P

That's it - check my github from time to time if you want to do send gerbers for your own board order- or make your own board in kicad or similar- or purchase one when i finally have them up on sh*tbay
User avatar
Lopenator
Posts: 214
Joined: Sat Nov 14, 2020 10:40 pm

Re: New 1-Chip RGB bypass

Post by Lopenator »

Hell yeah dude. I had faith.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

Thanks, but other than a new amp/driver
and alternate methods of attenuation
not much achieved :P

I wouldn't recommend the 10k fix either, unless you have some ultra dark old set.. that shoots even further over standards than the original encoders attenuation.
It's fine on my PVM with the brightness right down. It's just an option.

Either way, I've wasted far too long on this project.
...and far too long messing with C11. 1uF = missing scan-lines
4.xxuF = missing scanlines
10iish uF = top scanlines become dim, but not missing.

20 to 100uF = top scanlines reappear

I've thrown in a 0.1uf so i have my scanlines back

Ghosting - Always there - Super Mario World - it's terrible, you start focusing on it. I'm not sure how C11 helped ANY system... but doesn't help on SFC phat or JR (not these ones anyway)

IMO - One chips suck. Just a big budget cut. Amalgamate everything into one shittily engineered piece of shit IC.
Softening the image would have being deliberate on 2 chip models.

At least - if you want to play super turrican so bad on an NTSC 1-chip just use the PAL ROM... it works fine, doesn't glitch out.
Peace out !
User avatar
Lopenator
Posts: 214
Joined: Sat Nov 14, 2020 10:40 pm

Re: New 1-Chip RGB bypass

Post by Lopenator »

I don't notice ghosting on my 1chip
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

you have to look for it. pretty hard to notice usually .

except on that one level of Super Mario World. I've added a photo in the folder in one of the posts. imagine that, but there's up to 4 or 5 items that do that same thing in some scenes. I was straining my eyes to see the ghosting most of the time.. but theres a photo there showing it.

When you're trying to make a console upgrade, you are super sensitive to everything is all.

it really rained on my parade lol

I'll still install the PCB when I get it ..on a JR, Phat SFC and US 1-chip SNES, but I went rabbit hole and lost a day breadboarding. There's a lot on my plate so it's really frustrating.

I had one side of C11 across the breadboard with a 1k trimpot. 5v one side, gnd the other and one side of C11 in the middle. It acted like a little gain/ focus knob and was almost as clean as a capacitor when dialled in correctly. It was cool, but ghosting remained.
User avatar
NewSchoolBoxer
Posts: 369
Joined: Fri Jun 21, 2019 2:53 pm
Location: Atlanta, GA

Re: New 1-Chip RGB bypass

Post by NewSchoolBoxer »

VajSkids Consoles wrote:that was under load, connected to the PVM but prior to hitting the set. TBH I was expecting the sets 75r pulldowns to be active at the other end of the cable and bring it back down to .7vpp, but it jumps around anywhere from 1.4v to 1.6v after its gain of 2x.

inputs are rock steady and each channel jumps between the measurements stated. (+/- 20mV)

200Mhz scope - SDS 1202X-E
very new equipment for me, lots of settings to play with.

I've jumped on the system for 15 minutes or so where ever possible and played a variety of games, it's extremely crisp and nice... so the measurements haven't really bothered me.

it would be good to get a known list of games or scenes in games where ghosting artifacts are prevalent.

the ones in the original proto are all clean now.
no missing scanline phenomenon like on the THS7374 circuits either.
Ha I love you pointing out missing scanlines from THS7374. That's convenient you got a 200 MHz scope. Could just estimate slew rate and the gain dropping off from 2 at high frequency input for bandwidth and conclusively prove your opamps are good. Do unit step function input to measure ringing.

The 200mV worth of oscillations, can test a 5 MHz or 6 MHz square wave input to 75 or 150 ohm pure resistive load to see if gain is constant like I expect. If so then oscillations are from antique SNES signal generator with low-ish rise and fall times, noise in general and/or significant capacitance on the rest of the signal path. The latter I kind of doubt since you have 75 ohm resistor on the output that helps deal with load capacitance. Then could test SNES output into pure resistive load for oscillations or distortion to narrow down the biggest contributor. Can add pF or < 1 uF capacitor in parallel to 75 ohm load to test that aspect.

Known games with ghosting, I've probably seen the same posts you did with Mega Man X healthbar and Street Fighter II Turbo or Super with missing scanlines. Capcom ftw. But ghosting in general is from high contrast switches with bright and dark colors. The test pattern for SAG correction is a white box with black borders and the white turning grey is essentially ghosting.

The differential gain and phase errors I was using before, I thought they came from error in opamp driving + and - potential difference to 0. Actually, it's a measure of composite video's luma brightening the colors at the color subcarrier correctly. Even the RGB opamp specs measure this way. The 3 major video opamp creators each have scientific papers describing the gain and phase and how to measure. Very relevant is one of them says 1% gain and 1° phase errors are visible and that's with blurry composite video to presumably consumer CRT. I'd say RGB would tolerate less gain and more phase error. SNES though with 2^15 colors, that's 2^5 = 32 levels each for R, G and B. If the console is actually accurate in splitting up 700mV in 32 increments then you have a very lenient 21.9mV of tolerance, or about 3 IRE worth.

Note that the papers say the 1%/1° error to be visible is for the entire video chain, including the console, the television receiver and its own video processing.

Design Comments - click pics for full size
Spoiler
Voltage regulator to drop 5V to 3.3V makes. I assume regulator datasheet shows low ESR ceramic caps are fine to use over tantalum, else you get oscillation. You said 3.3V tested better than 5V, which again surprises me since you have lower slew rate and bandwidth at 3.3V but I was saying less overshoot may compensate.

Fortunately Analog makes LTSpice and LT6550 so I was able to simulate the circuit for one channel. Input is 0.8V square wave at 5 MHz 50% duty cycle. Can download my asc file here. The 3pF capacitor is for modeling the load's capacitance. Does nothing at 3pF but majorly distorts at 300pF. Also, an input capacitor screws up the output. Some ringing on the square wave output is 100% expected and not necessarily a sign of bad device or setup.

What I'm not modeling because I don't know it is the source impedance before the 130 ohm resistor. Must be relatively high to form a voltage divider to drop 0.8V to 0.7V against LT6550 input impedance of 300 kohm. Without the high source impedance, the resistor barely matters.

One check is if you get no gain, the ground reference isn't connected to the opamp. Using 5V instead of 3.3V gives slightly less ringing on the output.
Image

What you're not doing is using a parallel 75 ohm resistor before connecting to an active device - the opamp. Is a mistake (even though most mods don't do it) because the second path to ground cuts down on transients and wave reflections. If you scroll down 1/3 the way on this post, you'll get Tim W's aka "viletim" equivalent explanation of how it reduces interference to RGB.

Can use a Shockley diode with the right voltage drop since you don't need negative voltages. Non-diode way for 0.1V drop needs 3 total resistors to get the same impedance as [source impedance + 130 ohms]. I used the series 75 for sake of not needing three values. This at x3 for R+G+B ramps up the design complication. Can try setup on just the R, G or B channel to see if sufficient improvement. Updated pic for setups:
Image
Last edited by NewSchoolBoxer on Mon Jun 20, 2022 3:16 am, edited 1 time in total.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

I've kind of lost interest in this project because of the ghosting, which is apparent in all bypasses (and factory encoder). I mean, it's finished anyway.
Ringing, as far as I know, is a downfall to using ceramic caps. I do have tantalum SMD 1uf's here, but off memory, the datasheet for the 3.3v LDO specifies that it's designed for ceramic capacitors.

On this particular system you can increase or decrease the gain on the R/G/B lines by adjusting the value of R3, and you can increase or decrease the gain by adjusting the values of R6:R8.
This is why existing brightness fixes increase the load on R6:R8 with pull-downs (It would much neater to switch them out instead) - this is used on existing bypasses.
I found changing these to 130r, as oppose to adding additional pull-downs give (at least on the red channel) a perfect 0.7vpp
Because phat uses 160r and JR uses 150r for R6:R8, the values of the additional pull-downs differs for the brightness fix - I think it's better to just remove these 3 resistors, and have 130r pull-downs on the bypass board.


Now if you go and adjust the value of R3, i.e add an additional pull-down (my measurements' show 10k pull-down adds 100mV to the R/G/B lines), you'd then have to lower the resistance of R6:R8 to pull the signal back into spec.

So that's how the balancing of the signals works on these 1-chips. Whether there's anyway to balance this out and at least 'mask' the hideous ghosting artifacts - I don't know. I jumped into this project without any prior research to existing boards other than noting they had brightness fixes on board and left R6:R8 in place - which i found completely out of balance anyway.

But then again, these dodgy 1-chip IC's taper down in output volume from R to G , and again from G to B. To properly balance these out to perfection, you'd need to have a different value resistor on R6, R7 and R8.

There's also another brightness fix i found (a very dated one) involving lifting pin 155 of the S-CPUN A (one-chip IC).
I was going to try this, but I accidently lifted pin 156, if you want a laugh - heres my repair job: https://imgur.com/gallery/0lo29Xu

From VideoGame Perfection:

"Borti has a supposed fix for this by lifting pin 155 of the CPU and connecting a 20ohm resistor in series to vcc. The resistors on the RGB board would then need to be removed. This just seems a little bit extreme, and is quite risky given that the pins are tiny and access is limited."

"We use Borti’s fix in our workshop and we’ve had great results with it."

Either way, this wouldn't be any different to other brightness fixes. IMO - if you want to keep the stock encoder, the easiest and neatest way is switching out R3 to the correct value. (else R6 to R8)

Basically - Getting into nitty gritty tech details atm is kind of pointless.
I had no luck with the C11 fix for ghosting as mentioned.
I did find, you can also use C11 for attenuating the signal - I didn't experiment with this
Hopefully one day, a proper fix for all consoles is figured out ...but these consoles are very expensive to experiment on (unlike 2 chips)

And don't forget, I managed to pick up the NTSC sub carrier clear as day (my PVM NTSC light illuminated) when using C sync straight from the 'One-Chip' C sync leg, I confirmed the footprint for R9 led to the correct leg on the IC by soldering kynar onto that leg (I do quite a few PS2 modchips, so this is easy for me :P ) and confirming continuity from that pad to the IC's leg on the opposite side of the board.

THIS CAN'T be normal!
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

"Ha I love you pointing out missing scanlines from THS7374."

I think some websites get confused, this seems to be from the C11 fix. I ran a lead from each pad of this component to a breadboard and experimented with values as noted somewhere above :)
I even tried making completely new circuits around this component :P

Nothing helped ghosting, just killed scanlines, wiped top section of the screen completely or faded the top few scanlines.

edit: there is ONE thing i'd like to try, and that's a really low value resistor on R6:R8 then driving the 1-chip harder by lowering the value of R3 - just see if it has any effect on ghosting.
Why? Because when I increased the value of R3, it created noise within that IC and spat out the sub carrier on C sync. Maybe a lower value will have it more 'free-flowing' and decrease noise.

All just a theory. Trial and error.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

Had a big yard day, so hands are shaking when trying to solder - can't touch customers consoles.
So, quickly experimented for the sake of it.

1k pot from R3, R6:R8 20r (just had them SMD ones on hand) so left the 130's ungrounded on breadboard.

Once you reduce this resistor too far a value, you get power supply noise - cleaned this up with a bypass cap. (R3, that is)
Twisted the pot until I hit .7vpp

No effect on ghosting what so ever :P
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

Experiment 2,

remove R3 altogether, bypass cap here to clean up noise.
Send too hot of a signal through, then pull it back down to spec on the driver outputs.

ehhh ... ghosting isn't AS apparent.
This isn't following standards though.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

Ok I seem to have elminated ghosting by some sort of fluke. (masked it right up at least)
I will edit with photos, just want to try the scaler. (red channel is sitting at 744mV pk-pk, i should switch out some components but im a bit excited)
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

OSSC - I dropped the peak to peak for this.

Now this needs much more investigation and analysis. But that's a 100% improvement on a still image, I was pausing the ghost-iest scenes while prototyping.

On a moving image (whilst playing) there are tiny disturbances... but not like before. It's much, much better.

This is R3 wired straight through with a bypass cap (value to be determined)
20r resistors on R6:R8 , then additional 130r pulldowns (using what i have on hand) on my PVM shots, and 120R on the scaler.

This is WIP.

The trade off seems to be bleeding pixels where the ghosting would previously have started it's journey across to the right hand side of the screen.
You can see the black's bleeding across a little.

https://drive.google.com/drive/folders/ ... sp=sharing

edit: the bypass cap doesn't seem to be needed once additional set of pulldowns is in place. It seems the more paths to ground the better. perhaps a chain of resistors on each channel is the way to go. It's looking much better on OSSC as well. There was initially a lot more sync noise. (not there when using scan lines) im not using a coupling cap on sync either.
Last edited by VajSkids Consoles on Sun Jun 19, 2022 11:25 am, edited 1 time in total.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

User avatar
NewSchoolBoxer
Posts: 369
Joined: Fri Jun 21, 2019 2:53 pm
Location: Atlanta, GA

Re: New 1-Chip RGB bypass

Post by NewSchoolBoxer »

VajSkids Consoles wrote:Had a big yard day, so hands are shaking when trying to solder - can't touch customers consoles.
So, quickly experimented for the sake of it.

1k pot from R3, R6:R8 20r (just had them SMD ones on hand) so left the 130's ungrounded on breadboard.

Once you reduce this resistor too far a value, you get power supply noise - cleaned this up with a bypass cap. (R3, that is)
Twisted the pot until I hit .7vpp

No effect on ghosting what so ever :P
I live in apt so no yard work or property taxes but no building equity. :/
Yeah I think you have the real solution of needing separate potentiometers on R, G and B. Can go digital with 3x X9315 or similar. Seems disingenuous to me now that every other mod comes with fixed value components when our antique consoles aged at different rates. I updated my setup pic in above comment. Realized someone could call me out for not specifying Shockley diode. Can hit 0.1V with low enough current + high enough temperature.

Consistent gradient, that's awesome and the pattern really does show every primary color possible on SNES - 32 each for R, G and B!
I'd like to see 240p pattern with light's secondary colors of yellow, magenta and cyan, with yellow being the brightest of the 6 in human vision. I suppose 32x32 shades of each.

Actually, I had a vision from god. It was flash carts are for the middle class, make ghosting pattern with $10 Mario Paint. Needs the mouse but Japanese Mario Paint is bit for the bit the same as US NTSC so can probably import for dirt. Can save and export pictures with save files too. I respect shmups forums for being hardcore against ROMs and counterfeits so not sharing the save that takes 2 min of work. (I own the cart + mouse, just easier to do in bsnes to demonstrate.)

That said, let's create the signal amplitude gain (SAG) test pattern on page 7 that shows off-white square if ghosting! Is done to check on the ~220uF or equivalent capacitance that creates HPF at low enough frequency (5 Hz 3 dB cutoff = 1/(2pi * 220u * [75 + 75]) not to hit 25/30 Hz interlaced info. In that sense, 330uF is safer considering electrolytic caps lose F over time.

Instructions: Left click on Mario to start. The paintbrush is like Windows Paint fill, do that in black and click near center. Get big brush for white since the eraser white is off-white. Make square-ish shape in center then click on right arrow to show picture frame for full-screen action! The dog icon clears the canvas. I don't get it either. The title screen and fly swap game from coffee icon seem like good contrast tests too.
Spoiler
Image
VajSkids Consoles wrote:So the component of interest is an in series resistor to the proprietary '1-chip' IC - R3

Strange , very strange.

I can see from other documentation, this is a 1k6 resistor on the phat and an alternative brightness fix has being increasing R3's value. (as oppose to 'increasing the load' on R6:R8 by adding further pull-downs, or in my case -simply decreasing the value of these resistors).
Now no one goes probing around a board when trustworthy information/ documentation exists...

but can anyone explain how when i switch this out to a 2k resistor (I knew this would further darken the screen) the NTSC sub carrier is now hitting my set? I've got R18 removed, and additionally removed C50 as the composite pin buzzed to this component as well. using clean C sync, there is no sub carrier. Sub carrier is on the disconnected composite line. My brain is melting. My PVM shows which sub carrier is hitting the set via 3 illuminated lights (Yes, even on RGB SCART).

Yes, there's 2 sub carrier frequencies, PVM shows PAL60/NTSC443 , straight NTSC and straight PAL. NTSC illuminates, disappears when back to 1k6.

I grabbed C sync off a removed components footprint as per other guides. I'm baffled. I will keep findings posted here for shits and gigs. Even if i'm talking to myself :P
NTSC has two subcarriers: Luma at 1.5 MHz that modulates around [250k to 5.7 MHz] and color subcarrier that sits 3.58 MHz above Luma with 4.85 MHz subcarrier modulating [2.55MHz to 5.45MHz]. The larger 1.3 MHz lower band of color is for I NTSC (or U PAL) chroma that represents cyan-greens and oranges that are more sensitive to human vision than the magenta-yellows in 0.6 MHz given to Q NTSC (or V PAL).

Switching from 1k6 to 2k resistor causes NTSC subcarrier to hit the set? Probably lowering the natural capacitive HPF from series cap enough to no longer filter the subcarrier. Higher value components lower cutoff frequency.

PVM detecting NTSC from C sync is hilarious. Must come from interference. You know what I want to see? A ferrite bead to LPF sync that is basically DC as far as this sub-$1 bead is concerned. I like how Fair-Rite gives links on page to official distributors to price compare. Sony JP-21 RGB cable uses ferrite beads. Precedent is set.
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

Good information, will try that paint game :P
So far the scenes in a level of SMB are doing me well. Megaman X is pretty much fine, but SMB *was awful*
I don't know what it is exactly with R3, but any amount of resistance there to gnd causes ghosting/smearing.

So C11 fixes (albeit, with some drawbacks) seems to work on US SNES (What's the difference??) but plenty of reports of it not working on SFC and PAL units (PAL unit's i won't go near yet, why on earth are they doing 75r terminating resistors in the SCART head when that's in the set??)

Eliminating R3 has squared out the R/G/B channels - They are all equal @ the input of the amp and @ the output of the amp. I am trying to attenuate a new circuit with some stability with R3 jumpered through. I can see ~1v on it and shunting this straight to GND doesn't seem to cause any issues.

When i mention stability, the readings would only jump up and down 20mV on a full white screen (originally), now they they are all over the shop.

As mentioned it's WIP, but I can't see shunting the connection @ R3 straight to ground to cause any issues - it just drives the RGB channels on the C-PUN harder. I'm not sure if i can attenuate a stable signal like this yet. I just know as soon as resistance is added (@ R3), ghosting is back. Image is crisp, but on massive full screen scene changes there are issues. Also, amp datasheet says direct coupled is fine, but obviously I can see the DC trail when reading the RGB channels, that wouldn't be eliminated until it hits the coupling caps in the SCART head.

Also ignore any references to a bypass cap @ R3, I jump on here and ramble prior to taking readings with the scope and meter :P
VajSkids Consoles
Posts: 123
Joined: Tue Jul 07, 2020 10:22 am

Re: New 1-Chip RGB bypass

Post by VajSkids Consoles »

I also have the added possibility my C11 cap is causing these issues, as they are all the top of the screen that C11 affects.
I lost the original cap.
I don't even know it's value....? People are going 270nf or 470nf..was it 100nf? It's a pain in the ass - I have a couple 1-chip phats and 2jr's but the JR one is lost (has a 1uf in it's place), the other one I don't even want to touch until i finalise this (for a 3rd time :P)
As in, screw transplanting C11 board to board, there's enough crap floating around here already without more dismantled consoles/screws etc. floating about.

Edit: went digging through some draws and found components from other projects including abandoned ones..

Tried all these

1uf ceramic
1uf tantalum positive to C11 text.
100uf ceramic
10nf (too low a value)
22nf ceramic

The 22nf is fine, doesn't effect top few scanlines what ever, it was an 0402 and i managed to bridge it in there somehow (without a struggle either :P, by eye!)
I don't know what i was thinking when i ordered 0402 components (for a master system mainboard, which mainly worked, but got too expensive to continue prototyping). They're small enough that you only have to look away from where you dropped it out of the reel packaging momentarily and when you look back, you can't find it.

So, when i get time, I'll see what I can come up with. I can attenuate this to hit a max 0.7vpp but problem there with fullscreen scene changes and I'd like it to be more steady with the fluctuations.

Edit: So 1-chips are a major pain in the ass and seem to be a lottery win type thing to get one that is

A: not really ghosting or smearing
or
B: C11 fixes the issue

I can hit 0.7vpp spot on in so many ways, within +/- 8mV fluctuation on 240p test suite with full white screen- but ghosting on SMB is so annnoying (only on certain levels) it makes you want to smash your console up with a hammer.

You NEED a bit of resistance @ R3 to fix the issues I was having, but once there's resistance ghosting is terrible, with resistance you crush the first few bars out of the gradient.
With no R3 , then hitting 0.7vpp you get a perfect colour bar gradient, ghosting/smearing is at least masked, but then you have the scene transition issues. (big spikes on the RGB channels on blank screens)

Maybe the US SNES 1-chips are the way to go? But from what I see everywhere on the internet, they can exhibit many other issues I haven't come across - vertical bar's in the centre of the screen and so forth.

I may wait until I get my hands on a US SNES 1-chip and see how my revision 2 board goes on that.
Any sort of interference is just a no deal. I can deal with tiny hints of jail bars on a model 1 megadrive with good audio that's had an RGB bypass and sub carrier removal... but this ghosting and smearing on certain 1-chips is a deal breaker.

Apologies for the 1000 edits, I've become a bit of a pisshead in the last few weeks :P
Post Reply