VajSkids Consoles wrote:
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 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.
VajSkids Consoles wrote:
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.
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.