shmups.system11.org

This is super clean, not too bright, no wash out.
@vajskid (twitter) for some pictures that really don't do it justice
Uses an an analogue LT6550



One tiny discrepancy to my breadboard prototype is the pull-downs are @ input pins / negative side of AC coupling caps.

No sync BS, won't kill composite. No worrying about "is there a resistor on my cable to bring it down from TTL" etc.

I have to try it in a phat model now, when i get time.
I really should bust open the SCART cable just to 100% confirm what's in it.
I did a customers install of one of Borti's bypasses and the cable he sent didn't work at all, where as mine did. (Also an NTSC SFC JR).

Wish I could find a 1 chip.

Easy to find, but not cheap. I asked for serial numbers and got a PAL one dirt cheap, sold it for a lot less than they go for now. I don't like 50hz.. yeah there's a mod but nah.

Trivially easy: all SNES Jr are 1chip.

That's what I did, just got a Jr. Easy-peasy. But, I'd sure love to have a 1-chip phat model just for nostalgia-sake of liking the old design.

Keep your eye on the JP listings on eBay, they're all ridiculously overpriced atm, cheapest I can find is 228AUD free shipping for a phat... but I scored one for 172AUD inc. shipping going off current conversion rates... that was from 'mitsuya-shop1'
I got the mint condition JR for 200AUD WITH all cables, shipped, and he threw in a CIB copy of Donkey Kong (JP version). This was from 'japan-kyoto28'- also don't be afraid to inbox them as these guys deal consoles and actually go on the hunt for your requests (he can get me any colour fully boxed JP N64, for example).
Just keep checking the listings, make sure you watch ones you're interested in as sellers always send out offers.

https://i.imgur.com/UmYx53x.jpeg

OSSC/ scan lines / 4:3 / IPS monitor
hoping small issues found are due to a mess of jumper wires on a breadboard with components held in finger tight

Edit: Also switched 5k1 to 5k6 ...

This has being revised

Now direct coupled, now removing onboard attenuating resistors which are still in place on other boards.
Fixed ghosting / smearing.
Board is absolutely tiny.
Hopefully no mistakes.

That's cool to see another opamp in use. As much I use what I see other people using, when selling things to the public, we shouldn't use THS7374 just because the other guy did. At first THS7314 was standard until people realized it was a problem that the LPF couldn't be disabled.

Strictly comparing datasheets between LT6550 and THS7374, I see LT has no LPF, which I wouldn't say is an issue when you only want it before ADC to a scaler and popular scalers have their own LPF. They dissipate effectively the same amount of power which I consider high.

THS has better phase characteristics, meaning less harmonic distortion, though not much less. THS also has better bandwidth for a better gain-bandwidth product, although I don't see how that matters in 6 MHz small voltage case here. You generally want high input impedance and low output impedance to have a buffer that protects the amping from what happens before and after. THS has better input of 800 kohm and output of 0.7 ohm max while LT has input of 300 kohm while output is 0.7 ohm but peaks at 3 ohm for expected input. Worse but I wouldn't say is significant.

LT has more than double the slew rate, meaning how fast it can amp the voltage. This is desirable for less of a delay between input and output and other reasons and is a significant factor in cost. LT has better differential gain and phase at 0.09% and 0.09 degrees vs 0.5% and 0.5 degrees. Lower is better where 0 is ideal because each channel is amped exactly 2x and phase is altered the same. LT also has better channel separation, meaning your R, G and B lines don't project noise on each other as much when passing through a tiny opamp.

What's annoying is LT gives standard voltage and noise densities as 12 nV/sqrt(Hz) and 8 pA/sqrt(Hz) while THS gives overall SNR of 70/78 dB, depending on the weighting. Not sure how to compare but neither figure is bad.

tl;dr, THS in most all other bypass mods is 37 cents apiece in US with order of 100 while LT in this bypass mod is $3.90 with order of 100. I'd prefer LT for better channel separation and higher slew rate. Is in a higher pricing tier that is a further sign of being better for typical use case. If you need a LPF because your scaler / analog to digital converter doesn't have one then go THS.

Open to other analysis. I'm not actually testing these. Mods aren't my thing.

Thanks for the comparisons
I am a hobbyist, but I have taken some basic readings

With the 130r pull-downs

Red is 680-700mV
Green: 660-680mV
Blue: 640-660mV

I am switching components on the new board to 0.1% tolerance, but I think this is more the system than anything else?
They were taken with a full white screen in 240p test suite.

I was going to use an LT6557, but overkill - they're high bandwidth HD amps which I used in a Megadrive bypass. I got a heap of them cheap locally, but they're now very expensive.
The LT6550 is available on Ali express for a reasonable price. I will make a Github repository for this soon as WIP as I can only test and make installation docs on SFC and SFC JR.
I am getting a US 1-chip SNES as a trade in towards a NES RGB system and can test on that soon. Pulled the PPU yesterday
PAL 1-chips I am not interested in and have sold mine.

My main concern with existing bypasses for this system is that they aren't attenuated from scratch - they are leaving the factory input pull-downs in place, then using more pull-downs to correct the brightness.
Because of this, they recommend different methods of brightness correction dependent on console - if you just attenuate the input from scratch, it should be universal, as the RGB lines are driven from the same proprietary IC.
I'm also not sure why they bothered making the board bigger and placing footprints for already existing components on them as optional.. you could just leave or remove these from the SNES mainboard.

All my ghosting and smearing issues disappeared when direct coupled with factory onboard attenuation resistors removed. They are also sloppy tolerances and I had up to 5ohm difference per channel with the onboard pull-downs. The JR was a little better with only 1 to 2ohm difference.

Running the amp @ 3.3v as the diagrams on the datasheet show better performance @ 3.3v.

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

Folder with a few pictures off my partners phone which is a little better than mine
Keep in mind, that's a work monitor ...and I haven't bothered calibrating it. Geo/Convergence is off. (+glare, reflections etc.)

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

I think I got the noise comparison figured out. Internet told me the nV/sqrt(Hz) is rms voltage. We know RGB voltage is 0.7V peak so /sqrt(2) for rms.
edit: read opamp tests that used 150 ohm source and load resistors so know how to compare and use same THS 5.9 MHz bandwidth



LT signal-to-noise ratio is 84.5 dB unweighted, dropping to 73.8 dB if you use 5k ohm resistor that is about 30x higher than THS datasheet used to get 70 dB unweighted. Higher SNR is better, LT SNR is a massive improvement and probably goes back to paying for a higher pricing tier opamp. THS with optional LPF is extremely impressive at 37 cents don't get me wrong, but if your project can justify a $3-5 opamp, seems like good idea.

I like your effort to find the best resistor value and 0.1% is tighter than the 1% I buy! Np, was good thinking exercise for me, as was comparing the noise. I have EE degree but transitioned to computer science career long ago. I'm a hobbyist in this space too. Difference is electrical fundamentals were hardcore beaten into me and I used green screen analog scopes.

There is such a thing as overkill - LT6557's 2200V/us slew rate and 400 MHz bandwidth just to amp 0.7Vp to 1.4Vp at sub-10 Mhz. The improved channel separation and differential gain error of 0.02% vs 0.09% (3.3V) or 0.05% (5V) - I'd be surprised if that is humanely perceptible at low voltage, low MHz. $5.03 apiece with order of 100 is decent price at least.



Again, I like the effort you're putting in this, in a space that was, I think, one person's mod making a micro improvement over another's.

That's amazing if no one doing bypass mods thought to just replace antique resistors on the video lines versus make a more complicated mod to compensate. Analysis I read is every 1000 hours of use ages resistors a "year" and the tolerance can tick up or down a little based on the years used. 1% turns into 2% and beyond. Curious if JR used tighter and therefore more expensive resistors since it was a blatant cost cut.

You say 3.3V is better than running at 5V. I know where 5V is on NTSC SNES, haven't looked for 3.3V but if one voltage source is significantly closer, that's better for forming a smaller current loop. It's abundantly clear in THS datasheet that running at 5V is better than 3.3V and is typical of ICs generally.

For LT6550 here, 3.3V has much better output voltage accuracy so seems you're right. Has significantly lower slew rate but is 250 vs 340 enough to care about? Third part is 3.3V differential gain and phase are worse 0.09%, 0.09° vs 0.05/0.05. Max channel difference then of 2 * 700mV * 0.0009 vs 0.0005, or 1.26mV vs 0.70mV. That's not much.

Thanks again for the analysis and comparison
one thing wrong though...!
the original schematic was scrapped.

All I knew was Analogue slew rates left THS Amps dead in the water and jeez.. please.. not another THS7374 based design. They're in EVERYTHING.. including my RGB to component transcoder..(greenantz.. Dev lives close by)

new schematic is simple
remove factory input attenuation resistors which are still active on other designs..

direct couple RGB lines from the removed resistors pad(or via, but don't see why you'd choose the via )

new pull down value is 130r

that's it

on the output just your 75r in series and there's optional puff rated Caps on the console mainboard wired in like bypass caps... I believe they are for high freq filtering but from what I've seen don't make a scrap of difference. At least they're used like this in audio circuits.

then we have a little tiny 3.3v amp (can't remember the footprint) and that only calls for a 1uf MLCC cap on the input and output

forgive any sloppiness. Asus zenphone 8 has tiny screen..not at my PC ATM

oh and VEE is now tied to GND

I don't understand what you guys are saying but I think it sounds like I ruined my 1 chip by installing RGB bypass.

I'm so sorry! Can you remove the mod or have you fried the main CPU-PPU chip? Perhaps you could start a thread to ask for help or post in one of the RGB bypass ones.

VajSkids Consoles, thanks for the update! Glad VEE is tied to ground now, was throwing me off. No need to apologize. Curious if a 0.1uf cap makes a difference over 1uf since I see them both used. Probably not.

I think I should be helpful and explain some things for the general audience. I guess I like writing exercises. This place is similar to Reddit in that many, many times more people view the discussions than actually have accounts. Most people aren't technical even if quite a few posters here are.

Video amp of THS7374 vs LT6550


Filtering and when you need it


Electrical noise and filtering limitations

I can remove the mod. Do you think it's harming my console? I like the brightness fix if not anything else.

I don't think anything is fried. At least I hope not. The console runs fine. I'm using voultars RGB bypass.

I just don't want to do anything that'll harm the console.

SNES 1Chip RGB bypass is a relatively benign mod. I'm not concerned you're hurting your console. THS7374 on the PCB dissipates more power at room temperature than I'd like but it's not making me worry. You want to do PS1 csync mod, I'd try to talk you out of it.

Thanks. If it is not hurting the console I'll just leave it in then.

There's nothing wrong with the existing bypass my mods. I just saw room for improvement. I'm still questioning why Borti and Voultar didn't attenuate the input signal from scratch.

and dont get me wrong, both those guys would kill me on electronic knowledge... I just think it was overlooked.

retroRGBs (and bobs a cool guy) also has a pretty weird guide for brightness fixing.

I mean they're completely ignoring the factory input pulldowns and then adding even more pulldowns, on aged components .. to create a fix.

why add more pulldowns on top of the existing pulldowns, instead of adjusting the value of the existing pulldowns?

The correct brightness fix should be replacing the old onboard ones with 130r (150 on JR and 160 on phat) .

I'd be curious to see if the factory encoder really needs to be AC coupled as well? I will test in future by replacing them and just linking the On board AC coupling Caps direct to encoder.

once again, I'm not a tech expert.
I worked with analogue signals for 5 years as a comms technician
ADSL, PSTN, FTTC, FTTN, HFC ...

but my actual theory knowledge isnt anywhere near what boxers is. I really appreciate his in depth analysis in this matter .

I honestly need to look up guides everytime I do something new with my oscilloscope

i also attenuate with my eyes and then use test gear AFTER
so to hit .7vpp on the head was done with my eyes

and again, retroRGB, voultar and borti are people I look up to

also, things haven't being going to well for me lately and I was actually a 6 pack of bourbon in when I was doing both the prototype and this final revision
yep. I get drunk and tinker with electronics

any customers who see this
NO
I never touch a customer's console unless completely sober

STRAYA!!!!!

You say it like its a bad thing. :p

I'm always interested in improvements. I have a standard ths7314 rgb mod on my snes jr., along with dejitter mod. I find the brightness is just a tinge too weak. With ossc, I can never get it dialed in quite right, and compared to other consoles with all things equal on crt, it is a bit duller. Would this LT6550 solution potentially resolve that?



Sent from my SM-G955U using Tapatalk

Ill throw my OSSC on and get some photos, but this is a lower end IPS monitor down in my 'lab' so the edges of the screen sort of bleed in with light which is nothing to do with the bypass...

I'm sure most people go 4:3 with nice scanline emulation, but this is just an OSSC switched on with no adjustments. No profile.
Hard to get good photos as usual, lights/reflections etc. and not the best camera on my phone.

https://imgur.com/a/p6ezmJU

I'm glad you asked this question. THS7314 should never have been used compared to other opamps available at the time. Yes, LT6550 will be better, as would THS7353. Probably any video amp in the $2.50-5 range at bulk pricing with no filter or filter that can be disabled is fine to use. The three reasons you shouldn't use THS7314 pulled straight from the datasheet are:

1) Insufficient slew rate
All video amps double the input voltage. The slew rate is how fast the opamp can increase said voltage. It can't pump 0.7V to 1.4V instantly. Maxim Integrated description explains how NTSC and PAL need at least 52V/us but 80V/us is better (and RGB will need still more). I think 150V/us in THS7374 is fine but more is better.
What do we see for THS7314 at 5V supply? At peak 1.4V output for RGB, it's < 30V/us. That low, the signal is definitely getting distorted.


2) Insufficient bandwidth
The higher the frequencies of the video signal, the higher you need the bandwidth of the opamp to be. I don't even see where THS7315 bandwidth is stated (bad sign) but you want at least 10 x max frequency, so 10 x 6 MHz = 60 MHz. Too low and the opamp acts as a low pass filter on the video frequencies you want to pass:
6.02 dB = gain of 2.0 from 20 * log10(2). See the gain get reduced from 2 MHz to 6 MHz? The 5.5 dB equates to a voltage gain of 10^(5.5/20) = 1.88. So your amping gets reduced over the video bandwidth and that fades the colors.

3) Filter can't be turned off
To its credit, THS7314 is lower on noise and crosstalk compared to THS7374. However, its filter is useless going to an (analog) CRT.
What's the price you pay? The non-linear phase distortion aka non-constant group delay above 1 MHz means the frequencies get compressed or stretched a little and no longer all arrive at the output at the same time. This results in the RGB square or sine wave or NTSC or PAL waveform getting distorted. All analog video filters do this but it's a necessity evil before converting to digital to combat aliasing.

And I get using THS71314 and THS7374 for costing <$1 if you're that broke but you have vastly superior options at $2.50-5. All the THS amps have relatively high differential gain and phase but the phase error doesn't matter as much in RGB vs NTSC+PAL that phase modulate the hue.

One last note, be careful of getting the real thing and not a cheaper version re-labeled on AliExpress. I'd verify the slew rate with oscilloscope since that's the main thing you're paying for. Else be sure the writing on the chip looks correct. I buy from Mouser and DigiKey and maybe from Newark and Jameco in the future. They are authorized distributors which means you get the real thing.

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.

There is up to around ~200mV of random output swing so I took measurements on the input

A year later 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.