New 1-Chip RGB bypass

[VajSkids Consoles]

I *really* didn't want to do this - but the video circuit ..no matter what you do, obviously isn't the issue.
SO here's some absolutely useless information.

The S-CPUN A '1-chip' has a total of 8 x 5v connections tied directly to the output of the 7805 regulator.
7 are tied internally, but connected to 5v planes externally anyway.

The 8th is pin 155 which isn't internally tied and is supposed be the analogue video power connection.

The S-CPUN A will not run @ 3.3v.

The S-CPUN A will happily buzz along with all 5v connections removed from circuit bar 155. The analogue power connection is somehow enough to keep the entire system running along perfectly, albeit, with a checkboard effect on the video output and a very faint jailbar effect. All the other 7 interconnected legs (tied inside the S-CPUN A IC) will have be live with 3.9v when the system is running with only pin 155 of this IC connected to 5v.

None of this helps with ghosting at all

[NewSchoolBoxer]

Mario Paint update

Spoiler

The dog icon is an undo while the eraser icon brings up a bunch of delete animations. Kid me remembers. The jump rope figure brings up animation mode where you can have Mario Paint scroll through 4 or more pics. I frame advanced on the speed settings. In frames to advance to next pic in loop:

1: just over 1 fps = 58 frames to advance
2: just over 2 fps = 29 frames to advance
10: 12 fps = advance every 5 frames
11: 20 fps = advance every 3 frames
12: 30 fps = advance every 2 frames
13: 60 fps = advance every frame

So this is passable way to check for ghosting or other interference on video in motion versus static image. Has some Super Mario World icons to place as well. I think the checkerboard palette options have testing potential. Crude audio testing in music mode if you don't have a function generator. Saving took about a minute for 1 picture. Good time for fast forward.

You what I'm tired of? People who are experts at modding or experts at making 1 small change to existing mod, doing things they don't understand or lack of doing remotest due diligence or choosing the cheapest option possible, unlike you.

Today I can search on Mouser for video amps and sort price low to high. The cheapest that can be bought at individual amounts is THS7374 with the mandatory LPF that was used in everyone's 1CHIP bypass at the start. Second and and fourth are from brand I've never heard of and rest non-bulk under $2 are THS.

THS7314 that RetroRGB Bob and Voultar said to replace THS7374 with is on first page of cheapest results. Has mandatory LPF filter for 720p/1080i that doesn't attenuate SNES video but does shift the phase down and adds unnecessary noise and time delay compared to sync. I estimate the gain-bandwidth product from gain-frequency graph in datasheet to be 40-50 MHz, which is insufficient for any analog video without adding distortion.


C11
SNES is probably most studied console in existence. Got searchable PDF of whole PCB here that shows NTSC C11 to be 82pF. NTSC and PAL scans and pics floating online show C11 has no printed value on it. Hardware revisions, let's wait and see.

My time has come. I can tell by capacitors in series (HPF), capacitor in parallel (LPF) and inductor in parallel (HPF) that this is a 4th order (passive) bandpass filter with 1 input and 2 outputs.

I simulated the C11 section of PCB the 3.58 MHz NTSC color subcarrier travels through to the video encoder. 82pf for C11. Guess what? It's designed to bandpass the color subcarrier of 3.58 MHz input, which makes a whole lot of sense to clean up the pure sin wave before it gets modulated with color information. Why two waves at 3.58 MHz but different phases? From the PDF, I can make a strong wager that Vc input is color subcarrier itself and Vb is the colorburst that is frequency and phase matched to the subcarrier that gets added later in the same scanline signal. Video processor in television or monitor uses the colorburst at start of scanline to stay synchronized with color information. If console drifts in subcarrier phase or frequency, video processor will detect it in the colorburst and adjust in time.

Actually then, PAL bandpass filter version must differ in RLC values because its subcarrier is 4.43 MHz.

Swapping 82pF for ~1220x greater 1uF majorly shifts the bandpass down to the kHz range. Can expect this, larger L and C values shift filter down, smaller shift up in frequency.
What you all are doing by blind guessing is compensating the bandpass filter for the 3 aged caps that have reduced rated capacitance by replacing only 1. Perhaps the inductor shows lower rated value too. Just replace the whole circuit with new RLCs.

Pics showing time response and AC sweep. Click for full size. Thick line is voltage magnitude and dotted line is phase change.

Spoiler

[VajSkids Consoles]

I'm no expert myself, but attenuating or at least, looking at attenuation on analogue signals has being my longest running 'career'.
I'm a bored comms technician.

R3 is like the tap strength in the HFC network.
R6:R8 are like attenuators to step the network strength down where required. If you can't get the line in spec and balanced out on both upstream and downstream, you have to palm the job off to the network guys They come and fix R3 - lol

Sub carrier interference is like Low insulation resistance on the copper network.
We used to get what we call 'battery' as well back in the public switching telephony network. That was horrible.
Basically these flimsy little lines that only came in 2 gauges, but for the most part were extremely thin gauge. (long runs used the good, thick stuff) ran about 46 to 52v and a little bit of moisture in a shitty joint and every damn line in the run is full voltage from other lines.

I just enjoy this stuff. Even if i don't get anywhere with it - and often I'll give up because prototyping expenses get out of reason.

So back to this 1-chip. I did notice running any 5v pin from the CPUN IC directly to the regulator with additional bypass capacitor helped a bit with ghosting.


This new circuit you mention sounds like a project for you
I would 'go back to school' given the chance, and level up my knowledge but I'm too old, and the bills keep coming *sacrifices a 1-chip to the tech gods*

[VajSkids Consoles]

New V2 board installed a fresh JR. Working. Just testing it.
I'm due for some major bad feedback shortly, as some customers are going to be waiting for their consoles.

I left a job last week, so this week free

If i'm not happy with this board, i probably won't release it - but I took notes along the way for the installation.
I probably won't be happy with it.

[VajSkids Consoles]

Well I'll be, tickle my balls and call me harry.
I'm happy with it.

Either
A: This consoles is just not as boo-ey (the ghost in mario)

Or

B: This super tight nice little circuit on a PCB is just the way, and it was noise from my mess of wire's prototyping.

I can't see the ghosting at all on SMB's worse scenes on the PVM.
Looks like I wasted a lot of time on that other board

I will throw a guide up on the Github along with the files.
The gradient isn't crushed, you just turn your monitor brightness and it's all there. It's just a little overbearing when it's cranked up.

So all this time, V2 was fine This is what drinking does to people.

Hang on:

C: I grabbed sync off the output of the S-RGB this time ???

[VajSkids Consoles]

V2.00 Final.

JR install

Removed to use C sync down composite line (sync over composite cable)
R38 - 75ohm series resistor for composite output
C54 - Coupling cap for composite output


Jumper top board jumper, let's call this J2.

In my case, I added the resistor in the SCART head to bring TTL level down within 75r C-sync spec, hence have bypassed this component/footprint with the bottom board jumper. Lets call this J1.

I've also just left the 470r resistor off the board (Anywhere from 330r to 470r will bring TTL to a suitable level)


Factory attenuating resistors need to be removed as they are still active in circuit, even though Nintendo didn't patch through an RGBS connection to the
AV header. Remove R6 to R8 and the coupling caps C6 to C8. This gives us extra pads to run our RGB connections to. These connections are closer to the modboard, so we don't need to use the tiny via's. You can still use the via's if you wish.

C-Sync jumper to modboard - You can use the 'tradional' via, or the via between the '3' and the '7' on the R37 text (just above it). This is closer to the modboard. I've chosen to use the C sync out on the S-RGB chip in this installation. Nintendo ran it through this IC - I don't want to chance it that there's ever addional noise on the sync by grabbing it straight from the S-CPUN-A ('1-chip') / S-RGB C sync input. You need to remove the screw in the regulator, and the 2 heatsink screws, then remove the heatsink to get to this IC.

Reference: https://www.retrorgb.com/snescsync.html

This is also your chance to add heatsink paste between regulator/heatsink if you wish.


Board installation
When assembling these boards, lay the LDO down quickly. They die easily with heat and you won't get a 3.3v output.
Insulate the rear with electrical tape or similar. This probably isn't required, but to play it safe.

I like to go through all the pins hanging out on the bottom of the board and flush cut them. There are quite a few.

Sit the modboard over the AV header, use your thumb nails on the left and right side to push them down hard over the pins. The bottom of the board should come up a little. Flux and solder away.


Red: R/H side of R6 footprint
Green: R/H side of R7 footprint
Blue: L/H side of R6 footprint

I recommend all 0.1% tolerance components.

[NewSchoolBoxer]

Ahhh I was reading through the old C11 thread and saw a pic that showed it near R3. Seems I was referring to a different capacitor the whole time since the PDF labeled your C11 as C1! I have 3 SNES I can unscrew to see if the numbering is consistent.



I still think people are pulling values out of thin air but I like the scope measurements of tighter square waves to justify experimental findings.

Since you were talking about voltage regulars, I see a mess of comments in different places about less ghosting + vertical white bar by replacing the 1981 design Rohm 7805 1A with one that handles more current like 2A. That's not true. Look at MIC2915X/30X/50X/75X or MCP182X or another family of 7805 regulators and find no difference in specs between 500 mA-1A-1.5A-2A except for the max current output that comes at a higher retail price. Well, the lower current MCP ones have less of a time delay from the comparator, meaning they're a little faster at fixing the output. Can use a 1.5 or 2A LDO if you want but 1A is no worse and maybe better. I give rama credit on page 3 of thread for pointing this out first in 2017:

Yeah, I ordered a couple 78S05 the other day. It could just be too little oomph when it counts!
Will test that out, once they arrive.

Update:
The new regulator does nothing for the ghosting.
It does help with the Vcc noise floor but only a little. It is comparable to adding a good bypass capacitor to a stock machine.

The reality is, voltage regulators age like most everything else and perform worse over time. The 7805 on my childhood SNES straight out died. Good news is that, unlike opamps, $1 regulators are legit. The lack of diligence here is people pulling bypass capacitor values out of thin air versus reading the datasheet and measuring ripple voltage. Too large a cap and you can damage the regulator from the initial current spike since I = C*dV/dt.

[VajSkids Consoles]

I always get paranoid when switching on a recapped console myself, they've never held any charge before and it's like a heap of momentary short circuits between the voltage and ground rail...!
There must have being something loose on the breadboard when mentioning 10nf as too low a value for C11...!
will spend tonight playing this SFC, got the superNT and megaSG but still prefer OG HW.

This things looks absolutely fantastic!

Where did you pull a schematic for the SNES?

[VajSkids Consoles]

https://github.com/L10N37/Super-1-RGB-b ... p-SNES-SFC

here you go, of course ...if you want to support my efforts I can sell pre-assembled boards.. bare PCB's or PCB + components as a kit.

[NewSchoolBoxer]

I can't just keep being theoretical. I'll DM you about buying a PCB + components with an extra amp and LDO since I don't own any at 3.3V. Can test with square and sine wave function generator against THS7374 version. I want another soldering project or two that I can afford to make a mistake on.

C11 Master Clock 21.47727 MHz Square Wave Generator - see diagram from last reply
The C11 crystal oscillator circuit that gives modders headaches, I spent an hour reading online (lol) about similar circuits. I don't know it all, never messed with crystal oscillators when I have a 10 MHz arbitrary waveform generator. Crystal can be modeled with an RLC circuit resonator circuit. No manufacturer feels like disclosing these values or providing a SPICE model but some specs floating around for 10 MHz crystals and we know the expected output.

Explanation:
Setup is an NPN bipolar transistor common emitter amplifier using crystal's AC output at 2 as the input and the the transistor's sine wave output gets formed into a square wave by the two HCU04 inverters. Square wave becomes the Master Clock, XIN, of ~21.4 MHz that is of extreme importance. Notably, 21.4 MHz is 6x the NTSC colorburst frequency.

-VCC is +5V supply.
-Q1 BJT provides stability and amps and caps crystal's output voltage up to max of VCC before losses, collector current minimization and voltage dividing.
-R1 works with R3 to set the DC bias on the base of the transistor and R1 is a high value to work as voltage divider that prevents loading feedback into the crystal.
-R2 works alongside C5 at the emitter to provide circuit stability and set the gain with R3.
-R3 is to current limit the VCC supply and set the gain with R2.
-C2 and adjustable TC1 are there to match the crystal's load capacitance for a phase shift of greater than or equal to 180° and a voltage gain > 1 to start oscillation.
-C3 not sure of, I think it's to function as standard bypass capacitor on the output to reduce noise.
-C4 is standard coupling capacitor to block DC biasing from AC output.
-C5 is a bypass capacitor for transistor's magnified emitter current due to its β that prevents a large AC voltage drop across R2, that in turn, preserves the output voltage.
-C1 aka C11, so very hated, I believe works as a bypass capacitor for crystal's AC output to reduce noise from frequencies above the Master Clock's. Using a much larger value shifts the bypass region lower in frequency, making it filter noise below the Master Clock's. You could instead measure the rippled output from the crystal, dump into FFT on scope, and come with capacitor value(s) to work as better AC noise filtering. Imo that you're changing it means the crystal's Q has declined from age and wear and tear.

tl;dr
Really, before messing with C1 bypass filter, you all should have replaced the antique crystal X1 first. 21.47727 MHz remain cheap and plentiful today due to use in ham radio and anything derived from NTSC.
Datasheets for crystals made today show a variation of +/- 5ppm per year, meaning ONE YEAR's deviation of +/- 0.000005 * 21.47727 MHz = 107.38635 Hz. Evidently, deviation ppm decreases over time.
Also replace the transistor*. I saw a few guidelines for a transistor based on the oscillating frequency but don't want to be quoted on which when I'm not experienced in crystals.
Better/faster inverters made today that can also be looked into.

Oh yeah, the voltage gain from transistor is -R3/(R2+Re). Re is determined from transistor's datasheet and is small, likely under 10 ohms. At 100 kHz and below, capacitor impedance is small enough to short R2 for gain of = -1200/(210+10) ~ -5.5.

At high frequency, namely the Master Clock's, C5 gets shorted by R2 and gain becomes a very high -1200/Re for gain of magnitude over 120. Inversion of voltage isn't a concern for the clock. It's just a 180° phase shift in AC.

*New transistor may have different β that yields different Re and circuit R2.

[NewSchoolBoxer]

I always get paranoid when switching on a recapped console myself, they've never held any charge before and it's like a heap of momentary short circuits between the voltage and ground rail...!
There must have being something loose on the breadboard when mentioning 10nf as too low a value for C11...!
will spend tonight playing this SFC, got the superNT and megaSG but still prefer OG HW.

This things looks absolutely fantastic!

Where did you pull a schematic for the SNES?

I haven't recapped yet but my time will come. Best case, it lets my SFC work with flash cart. I like the idea to replace electrolytic with ceramic where I don't think timings were so exact to take into account electrolytic ESR.

I originally found schematics on Console5 SNES page. 1992 copyright suggests they were officially made for Nintendo. One newer diagram was made by someone for PAL and the other came from NTSC wiki.superfamicom.org that I used. Seems like the latter site has amazing technical information but pretty minimal on context or greater understanding. Then I found the German language scene's PAL stripped PCBs from the aptly named (in English) Strip Club.

Fixed the high frequency gain above. I haven't done transistor calcs in over 10 years. Junior level EE student should be able to handle it. But yeah, I think main thing is to replace the $1 crystal.

[VajSkids Consoles]

Postage to the states is a bit stupidly priced sometimes.
I am waiting for a parcel ATM sent via Pitney Bowes as it was literally 1/4 the price of what I was quoted. Sent the shipper the labels.
You are welcome to order the board with the files off Github and the LDO's are like dropping 50cents on the ground and not realising. Not expensive.
As for the Ali express LT6550, they look genuine ..seller is not really an english speaker. Just says they are new (using translator)
I checked on the analogue forums too, and from the pictures of fakes verse real, these look real.

[VajSkids Consoles]

I don't see a letter envelope costing much if you really want a kit

[NewSchoolBoxer]

Let's see, shipping fee from major components vendor is around $7 USD and that doesn't include PCB. I'll take it all in envelope with extra LT6550 that ships for less.
I can believe LT6550 are good. I'll get that and other opamps as free samples for sake of comparison when I form an LLC and pay $10/month for business address. Build crap to test analog cables on which is my real interest.


Transistor Discussion: why it's worth replacing

The superfamicom wiki site lists the crystal oscillator's amping transistor as C4699. Was very good for 1990. Might seem strange in world of voltage manipulation but transistors amp the current by factor β aka hfe and this multiplied by the resistor ratio forms the voltage gain. Rough equivalent to opamp gain-bandwidth product is the transition frequency, called ft. The huge limitation versus opamp is that β is temperature and collector current dependent but advantage is speed and gain of 100+ is no big deal.

The SNES NPN transistor made by Panasonic has β from 60-200 with 140-150 what I'd expect on the console. The ft range based on emitter current is 275-450 MHz. Collector output capacitance (less is better) is 2-6pF.

Spending fives minutes to find a modern NPN BJT that is better and still under $1 is a BFS17N.
Its β is about 100 from what I'd expect on the console and ft goes from 1-3 GHz. Collector output capacitance is 0.8-1.5pF.

Can adjust resistor ratio to compensate for lower β to get one several times faster for tighter and smoother square wave with less parasitic capacitance that slows the transistor down and creates a LPF that we don't want.
Transistors are cheap and plentiful. Your cell phone CPU has a few billion. Not even considering C4699 wear and tear from years of use as an active component.

[KPackratt2k]

I got lucky and found a 1 chip in the wild.

Damn, I'm jealous. I wish I had a 1-chip SNES.

[Lopenator]

I got lucky and found a 1 chip in the wild.

Damn, I'm jealous. I wish I had a 1-chip SNES.

I been on the hunt for literally years.

[VajSkids Consoles]

bite the bullet and grab SFC JR. I found a yellowed and cracked shell one (hence cheaper) that works and it's going to be rehoused in the clean sacrificial systems housing

[Lopenator]

That's what I was going to do if I didn't find a phat

[KPackratt2k]

That's what I was going to do if I didn't find a phat

I ended up jumping the gun and bought a 1CHIP-01 Original SNES model at a reasonable price. On one hand I kinda feel bad because it cut into the money I was saving up to buy a new computer with, but on the other hand I felt that these consoles would only get harder to find and more expensive if I waited any longer. I'll keep my 3-chip SHVC-CPU-01 revision console in the event I encounter a game that doesn't work properly on the 1-chip.

[VajSkids Consoles]

Let's see, shipping fee from major components vendor is around $7 USD and that doesn't include PCB. I'll take it all in envelope with extra LT6550 that ships for less.
I can believe LT6550 are good. I'll get that and other opamps as free samples for sake of comparison when I form an LLC and pay $10/month for business address. Build crap to test analog cables on which is my real interest.


Transistor Discussion: why it's worth replacing

The superfamicom wiki site lists the crystal oscillator's amping transistor as C4699. Was very good for 1990. Might seem strange in world of voltage manipulation but transistors amp the current by factor β aka hfe and this multiplied by the resistor ratio forms the voltage gain. Rough equivalent to opamp gain-bandwidth product is the transition frequency, called ft. The huge limitation versus opamp is that β is temperature and collector current dependent but advantage is speed and gain of 100+ is no big deal.

The SNES NPN transistor made by Panasonic has β from 60-200 with 140-150 what I'd expect on the console. The ft range based on emitter current is 275-450 MHz. Collector output capacitance (less is better) is 2-6pF.

Spending fives minutes to find a modern NPN BJT that is better and still under $1 is a BFS17N.
Its β is about 100 from what I'd expect on the console and ft goes from 1-3 GHz. Collector output capacitance is 0.8-1.5pF.

Can adjust resistor ratio to compensate for lower β to get one several times faster for tighter and smoother square wave with less parasitic capacitance that slows the transistor down and creates a LPF that we don't want.
Transistors are cheap and plentiful. Your cell phone CPU has a few billion. Not even considering C4699 wear and tear from years of use as an active component.

Please let me know how you go with this component.
If there are notable differences, I will order some for all 1-chips i come across (the NPN BJT)

[VajSkids Consoles]

Also, full documentation for installations is now on the Github. Added phat models today.
Though, I still haven't seen an SFC with the issues that plague US models seem to have. I mean according to the internet there are quite a few other issues I haven't encountered (black bars and so forth).
For this reason, I've just made a single guide for phat SFC and US model boards as they should be the same.

If anyone notes anything not quite right, let me know and I will update.
As for 1-chip 03 models, these are only missing some supporting circuitry for C sync buffering. I have being using a phat model SFC all night, with the sync just run straight to the C sync leg of the proprietary Nintendo IC's TTL C-Sync leg. I've removed all supporting on board circuitry,
Zero dropouts on the OSSC what so ever, nothing to note. It's crispy as.

So you shouldn't have to do anything for the 03 1-chip, just run the jumper. 470r resistor is all that's required for attenuation.
The guide contains info on all sync and cable options, bar Luma, which i don't see as necessary. Also bar composite as sync, in which case, you'd just use a sync on composite cable (cheapies) and leave the sync jumpers alone and not run the link wire.

Also, finally took out sandman on super punchout. That bastard!
onto the 3rd league

[Lopenator]

Awesome dude!

[NoAffinity]

Got my pre-assembled kit from VSC earlier in the week. Assembled it, and got it installed today. Coloration is definitely improved. I am able to get OSSC/flat panel and Trinitron pretty closely aligned, at least as far as my eyes can tell. Very pleased with the results. Thank you VSC!

Spoiler

[VajSkids Consoles]

Thanks for the feedback, looks great!

[copy]

C11
SNES is probably most studied console in existence. Got searchable PDF of whole PCB here that shows NTSC C11 to be 82pF. NTSC and PAL scans and pics floating online show C11 has no printed value on it. Hardware revisions, let's wait and see.

My time has come. I can tell by capacitors in series (HPF), capacitor in parallel (LPF) and inductor in parallel (HPF) that this is a 4th order (passive) bandpass filter with 1 input and 2 outputs.

I simulated the C11 section of PCB the 3.58 MHz NTSC color subcarrier travels through to the video encoder. 82pf for C11. Guess what? It's designed to bandpass the color subcarrier of 3.58 MHz input, which makes a whole lot of sense to clean up the pure sin wave before it gets modulated with color information. Why two waves at 3.58 MHz but different phases? From the PDF, I can make a strong wager that Vc input is color subcarrier itself and Vb is the colorburst that is frequency and phase matched to the subcarrier that gets added later in the same scanline signal. Video processor in television or monitor uses the colorburst at start of scanline to stay synchronized with color information. If console drifts in subcarrier phase or frequency, video processor will detect it in the colorburst and adjust in time.

Actually then, PAL bandpass filter version must differ in RLC values because its subcarrier is 4.43 MHz.

Swapping 82pF for ~1220x greater 1uF majorly shifts the bandpass down to the kHz range. Can expect this, larger L and C values shift filter down, smaller shift up in frequency.
What you all are doing by blind guessing is compensating the bandpass filter for the 3 aged caps that have reduced rated capacitance by replacing only 1. Perhaps the inductor shows lower rated value too. Just replace the whole circuit with new RLCs.

Pics showing time response and AC sweep. Click for full size. Thick line is voltage magnitude and dotted line is phase change.

Spoiler

Ahhh I was reading through the old C11 thread and saw a pic that showed it near R3. Seems I was referring to a different capacitor the whole time since the PDF labeled your C11 as C1! I have 3 SNES I can unscrew to see if the numbering is consistent.

Spoiler

I still think people are pulling values out of thin air but I like the scope measurements of tighter square waves to justify experimental findings.

I know this is late, but:

I think you got a little mixed up here, because the schematic you're looking at is only for 3-chip models (e.g. SHVC-CPU-01 etc.). All the discussion about C11 and ghosting only applies specifically to 1CHIP models. C11 in the 3-chips (subcarrier circuit) is entirely unrelated to C11 in the 1CHIPs (DAC compensation node). Obviously the 3-chip and 1CHIP designs are very different overall, and the component numbers can't be assumed to have equivalent functions/values from one design to the other.

Per Voultar, the original value of the 1CHIP C11 is 47nf, FWIW.