Wavy line interference, faint flickering on Super Famicom

[Xan]

Hello everyone,

when checking one of my SFC units the other day (SHVC-CPU-01 revision), I noticed some wavy line patterns in the RGB feed moving downward on the screen. This interference only seems to appear after 20 minutes from turning the unit on and happens on all power adapters I've tried, which are two different Roland 9V linear adapters and the 10V PAL Megadrive one. The same issue is showcased here: https://gbatemp.net/threads/snes-screen ... ce.469795/

Now I also checked a couple of other SFC consoles that do not exhibit this effect, and I've noticed that every single one has a different 7805 variant in it. The one with the wavy line issue has a Panasonic AN7805, while others that do not have the issue have Samsung-made MC7805C and KA7805, and a Hitachi 17805, respectively. From the datasheets I gathered that apparently, the AN7805 is specced for 1A max, while at least the Samsung-made units can output more than 1A with adequate heatsinking (couldn't find a datasheet offhand for the 17805).

On that note I should mention that the issue seems to only occur when using the Super Everdrive V2 cart (w/ DSP installed), I was not able to see the effect with a real Yoshi's Island cart. Does anyone know how much more power that flashcart draws compared to real cartridges, and could an upgrade to the newer X6 cart solve this issue? I've seen a mention of the FXPAK Pro having a capacitor that can supposedly improve video output quality in some fashion.

Additionally, with the Everdrive all consoles show a very faint flicker that's generally just visible in the lower IRE ranges, putting up the 10 IRE pattern in the 240p suite really highlights the effect. The flicker seems to be a tad worse on my SNS-CPU-RGB-02 unit. I know that people suggest putting in a L78S05CV with 2A current rating, which fixed the wave interference problem for the person in the aforementioned thread, but I'm curious to what extent the video issues might be caused by the old Super Everdrive v2 cart. There's a bit of faint noise visible on said pattern, too, plus the crosshatching on my SNS-CPU-RGB-02, but I'm disregarding those as known.

[xeos]

That looks like 60hz noise from failing power filter caps. I'm only familiar with the problem on the NES but I'm guessing it's the same issue. Here's my writeup on the NES and a sample photo:

https://alantechreview.blogspot.com/202 ... video.html

odd that it only happens after it's warmed up but it's not shocking that the caps might behave inconsistent near end of life.

[Xan]

Well I'd actually noticed something in that area but got sidetracked after reading that mention of the regulator swap fixing the issue. Here's how the caps in that unit look:

https://i.imgur.com/yzFtk8H.jpeg
https://i.imgur.com/Cab2F0I.jpeg

Both read >20 Ohms of ESR, C64 still has most of its capacitance while C63 reads about 4uF, which checks out from the picture, C63 clearly leaked most of its guts onto the PCB, fortunately it went where there's not much in the way of traces. I do have a hot air tool, guess that's the safest way to get those caps off, and using aluminum foil to shield the rest of the PCB? I'd imagine it'd be a nightmare trying to heat that compromised old solder with an iron and twisting them off seems risky to me. At least my other systems don't seem to have leaky caps in that area.

[xeos]

Disclaimer: I don't know anything (about the super famicon). But:

(1) you can't check properly test capacitors if they are still wired into the PCB so you readings are probably meaningless. That said most people don't even have a ESR meter so maybe you know a trick I don't!

(2) the caps that typically fail are not those surface mount shown in your photo but the larger thru-hole with the plus sign stamped on the top (see my photo). I've never seen a small surface mount cap like that wear out, and you'll note it doesn't have the stress release cross on top that indicates it's meant to carry a large load (and fail without exploding). That said, the pcb does look yucky so maybe the did fail after all. I still wouldn't go for them first though. bonus is that replacing the thru-hole type is super easy compared to surface mount so you could just try it. Or at least check it with your meter.

[Xan]

I did those measurements with an Atlas ESR60 which does explicitly claim to support in-circuit measurements, that being said it's true those aren't always reliable, it seems to depend on the capacitor's placement incide the circuit. In my experience it's much more likely to show something like 0 ohms than >20 ohms when a measurement has obviously failed. In this case I was able to get in-spec capacitance on C64 every time while C63 would sometimes show "in-circuit/leaky" and one time I'd get those 4uF, so combined with the situation in the photo I do believe at least those two caps are toast.

I'm surprised about what you said regarding which cap failed though, from all I've seen the surface-mount types are a lot more notorious for leaking than bigger thru-holes considering the Turbo Duo, Laseractive and many other systems. The SNES isn't nearly as bad off as those of course. About the lack of a vent, I think that's normal for smaller caps in general as they aren't expected to be capable to build up enough gas inside to warrant the vent, bigger surface-mount caps do have vents though. By the way, North American SNES units after the SHVC-CPU-01 don't even come with the 1000uF/25V cap soldered in so as it's optional the cap might not be stressed all that much. There's a couple of theories as to what that cap exactly does mentioned here: https://forums.nesdev.org/viewtopic.php?t=13179

All things considered it's still strange to me that a regulator swap fixed the issue for the guy I mentioned initially, but maybe the newer regulators are able to deliver cleaner power to a point which might offset bad capacitors? Or maybe a bad regulator could have a fried those caps over time. I'll probably try swapping the regulator and all of the caps on this system once I get to it.

[xeos]

For the sake of science, perhaps replace just the caps first and see if that fixes it? Yeah I guess there could be some interaction between the regulator smoothing out the faulty caps. But with the caps looking bad it certainly makes sense to hit them first....

keep us posted on what the results are.

[Syntax]

It's well known that swapping the regulator for a newer 2amp model fixes a variety of issues, main one being the bright vertical bar centre screen via composite.

Another fix for the same issue is install more caps on the power rail.

[Xan]

The vertical bar is barely an issue in the first place on SHVC-CPU-01 from my experience, if at all.

[NewSchoolBoxer]

The vertical bar is barely an issue in the first place on SHVC-CPU-01 from my experience, if at all.

I'm glad you say that. I haven't seen a vertical bar on two stock SHVC-CPU-01s. My base understanding is the hardware revisions were mostly measures to reduce cost and maybe quality control or consoles not built to last was a consequence. Too bad NES holds up far and away better.

That said, I get the exact same interference from a noisy power supply on my Super Famicom and I'm not running 100V JP power supply on 120V US. It's another form of the same interference that xeos pointed with failing power filter caps. We draw AC from the outlet and the transformer converts to DC but the DC is not pure and always has rippling. This rippling cannot be used by chips that run off DC supply and ends up generating more heat and, not filtered enough, will pass into the video lines. So a cheap / aged power supply or near end of life voltage regulator / power filtering capacitors will yield the same result.

ESR meter right, 20 ohms would be insane. Can see three different charts here: https://www.eevblog.com/forum/beginners ... ytic-caps/

If ESR is greater than the table values then cap needs to be replaced. We're talking about antique electronics so may as well proactively replace all the caps in the same area of the console at once. I don't agree with the threads that say to change the value of the capacitors from the original spec but no need to get into that.

Electrolytic capacitors can lose 10-20% of their value with age as well as obviously gain more resistance. In other words, they become less ideal over time. As to why video is fine until console is warmed up a bit, I think that's exactly the answer. Thermal stress is a thing and ripple current exacerbates that. Copy pasting from a website I haven't heard of:

The life expectancy of an aluminum capacitor is not only determined by the ambient temperature, but also by the ripple current, and the ambient temperature plus the increase in temperature due to ripple current equals the operating temperature.

Do not apply a ripple current exceeding the rated maximum ripple current allowed for the capacitors as this will result in shortened capacitor life and may result in the capacitor venting or failing catastrophically.

In many cases capacitor heating due to ripple current is more severe than the ambient temperature stress, and an acceleration rate of approximately 2 for each 5-10°C temperature increase is realized.

I like Syntax saying to install more caps on the power rail. Clear distinction from what is stock. Generally you want different value capacitors since each filters out a different segment of the bandwidth. High value filters low frequency in parallel so should start there. Better to place as close to input and output as possible of what you're filtering. Some arguments to place one tantalum cap a little behind the electrolytics.

[xeos]

Thanks for the handy table. What about the question of measuring ESR in circuit, though? That seems very prone to false readings, to me.

That said since the failure looks like bad caps, it probably doesn't pay to be too skeptical!

[xeos]

Here's a hard maybe on the validity of in-circuit ESR testing...

https://electronics.stackexchange.com/q ... s-accurate

[Xan]

That said, I get the exact same interference from a noisy power supply on my Super Famicom and I'm not running 100V JP power supply on 120V US. It's another form of the same interference that xeos pointed with failing power filter caps. We draw AC from the outlet and the transformer converts to DC but the DC is not pure and always has rippling. This rippling cannot be used by chips that run off DC supply and ends up generating more heat and, not filtered enough, will pass into the video lines. So a cheap / aged power supply or near end of life voltage regulator / power filtering capacitors will yield the same result.

You should probably differentiate between switching and linear power adapters, I once had severe noise issues when using one generic modern switching adapter with the SFC, but it was more like "TV static" noise pattern all over the screen instead of this wavy interference, if I recall correctly (been some years). It also occurred instantly instead of taking 20 minutes to build up.

Incidentally, here's a report of someone that poly-modded the whole board and seems to have measured lower ripple with an oscilloscope afterwards: https://forums.nesdev.org/viewtopic.php?t=20230

Here's a hard maybe on the validity of in-circuit ESR testing...

https://electronics.stackexchange.com/q ... s-accurate

C63 and C64 are evidently in parallel on this board, and have the same nominal specs, yet the meter does come up with different readings for those two caps. From this alone I'd say it can be useful in some situations. Also if you read the comments there's someone that mentioned the same ESR meter and claims it helped identify a bad cap in his subwoofer.

[NewSchoolBoxer]

You should probably differentiate between switching and linear power adapters, I once had severe noise issues when using one generic modern switching adapter with the SFC, but it was more like "TV static" noise pattern all over the screen instead of this wavy interference, if I recall correctly (been some years). It also occurred instantly instead of taking 20 minutes to build up.

Incidentally, here's a report of someone that poly-modded the whole board and seems to have measured lower ripple with an oscilloscope afterwards: https://forums.nesdev.org/viewtopic.php?t=20230

You know what, I was wrong. I had the "TV static" noise you did. Was a year ago and I dare not repeat the test. I used this cheap adjustable power supply. I actually really like it because I power GBP and PSP with it no issues. Included polarity reversal adapter for SFC and other things but NTSC SNES stupid custom large barrel. Not sure if switching or linear or switching -> LDO combination but obviously I'm aware switching generates more noise at the tradeoff of amazing efficiency.

That sounds like a smart idea to replace electrolytic with ceramic and has actual engineering change justification that large enough ceramics didn't exist in the early 90s. Not that I have the patience.

[removed xeos link to save space]

C63 and C64 are evidently in parallel on this board, and have the same nominal specs, yet the meter does come up with different readings for those two caps. From this alone I'd say it can be useful in some situations. Also if you read the comments there's someone that mentioned the same ESR meter and claims it helped identify a bad cap in his subwoofer.

That's cool the user mentioned the Atlas ESR meter I hadn't heard of. I like there being an option besides the Anatek Blue kit.

Thanks for the handy table. What about the question of measuring ESR in circuit, though? That seems very prone to false readings, to me.

That said since the failure looks like bad caps, it probably doesn't pay to be too skeptical!

Device is designed specifically to measure in circuit. I was only vaguely aware of ESR meters and no idea how to use until I saw this video by EEV Blog Dave: https://www.youtube.com/watch?v=72sJ5Infuu0
Good part starts at 10:20.

[Xan]

Finally got around to this, doing some research, it seems people these days are often suggesting that just pulling the caps off with pliers (but putting pressure downward while twsting) is a pretty safe method if it's clear the caps have leaked. Did this and indeed, both came off very cleanly with the plus leg still attached to the board.

Going to do the rest later and will probably try to just replace the caps first without replacing the regulator, which is hopefully possible without removing the heatsink.

[jd213]

You just need to take off the screw on the regulator in addition to the ones on the heat sink to get it off,don't think it will be possible to get to most of the caps otherwise. Nintendo didn't use any thermal paste on the voltage regulator, so might be a good idea to put some on if you have some. It's also said that replacing the ceramic cap on the bottom of the board with a 22uf one will help reduce the center white bar on dark screens.

[Xan]

Yeah I'll see about removing it, it's a saner idea than messing about with little space in the cartridge slot area.

I'm not quite sure about the thermal paste though, it would seem this could actually help transport more heat above the ICs in that area. My reasoning here is from reading about that heat sensitive and irreplaceable Hitachi PLL chip in early Saturns, where people had the idea to just heatsink it, but others said that this would only make matters worse because it's not the PLL that produces the heat, so a heatsink would just draw in heat from the other chips. I guess someone else could chime in about this.

Some people also claim the CPU and PPUs in early SNES units run hot and heatsink them. Copper coolers to stick on are available but not sure what the ideal dimensions would be.

The ceramic cap I think I also read about back then and ordered it with the caps when I started this thread, so might put this in later, but SHVC-CPU-01 video quality already looks pretty decent to me on a CRT. The vertical bar might be there but I think you have to actively look for it on this revision as it's very faint. I do also have an RGB-02 and that certainly has pretty bad output though.

[Xan]

I didn't do the soldering yet, but I have cleaned the area with a q-tip and some IPA and was not able to remove those stains on the ground plane (posted pics earlier in the thread). It seems the cap fluid had seeped into the PCB and caused some corrosion there.

In the meantime I wanted to mention a bizarre issue I've been having with another SHVC-CPU-01 SFC, not the one with the leaked caps to be clear. I was playing Street Racer using my Super Everdrive v2 w/ DSP soldered (yeah, the old thing with crappy edge connector finish and databus connected through a bunch of resistors). The game worked fine at first, but when I returned to it to do the 3rd cup, it was completely broken and has been ever since. The race starts and the timer runs normally, but all the racers basically don't move and then start porting across the entire track, rinse and repeat.

I've already checksummed the ROM (which btw doesn't use saves), reseated the cart a bunch of times, and the console in question passes the burn-in ROM, and every other game appears to run fine, including other Mode 7 stuff like F-Zero and SMK. If I run Street Racer on another SFC, it runs fine.

The console with the Street Racer issue is a 2/1/3 chip revision according to Lion King (can't get whatever people are saying is in Mystic Quest to work), so it is not an early model. No idea if this is the dreaded SHVC-CPU-01 failure that people have been talking about, maybe it is an instruction that only Street Racer uses? I was thinking cartridge port as well, but the SFC that does run the game has some green oxidiation on one of its pins, and the pins on the console with the issue seem be looking cleaner.

[jd213]

I've had random glitches like that on a SFC (SNS-CPU-APU-01 from 1995) that just needed recapping, even though none of the caps were leaking. I know I sound like a "just recap it bro" person but SFC/SNES caps are pretty unreliable in my experience, maybe not as bad as a Turbo Duo or Game Gear but they're up there. I'm also paranoid about killing a CPU or PPU by using a system that hasn't yet been recapped.

[Xan]

I have considered that, it isn't actually random in this case though. That one game is acting up in the exact same way right after starting the console. And last time it also started on a black screen and needed a reset to get to the Everdrive menu, which isn't a good sign either.

The thing about recapping the whole board is, which method to use? I think the twisting off method might be good for leaked caps, but I am not so confident about doing that on caps that are in better shape. And even with a good hot air station you still need to shield the chips with kapton tape and run the risk of caps exploding in your face. With the aforementioned Saturn PLL chips, I've read posts from people that said they killed the chip while modding the console, due to soldering too close to it. So it's something that needs a bit of thought.

[jd213]

I think the twist-off method is fine for beginners, back when I did it I'd only go in one direction and wouldn't pull up until both legs were separated. It's probably safer on caps that aren't leaking, I never ripped a pad that wasn't already corroded.
Most of the caps are away from any proprietary chips, so I wouldn't worry about that. Might be easier to use through-hole caps as the replacements, console5 also recommends this and sells a version of their recap kit with all through-hole caps.

[Xan]

Been long enough, but I finally replaced the two caps on the console this thread was originally about. It wasn't too difficult thanks to using an SMD flux pen, although a better iron than what I have is certainly recommended to get "by the book" results. While doing this I have noticed that some or all of the other caps on the console test bad, I'm talking >20 ohms in the cluster near the AV out. When I went to test the console I first only got a black screen with single colored pixels, then some garbled white blocks - had to power cycle a couple times, but it works fine now. Definitely the caps being bad from sitting unused for too long. The image may be ever so slightly dimmer than normal as well.

I think it's time to accept that every SHVC-CPU-01 needs to be recapped now. Checking FFVI, I could definitely see the big vertical bar and jailbars across the screen as well on this console. Almost impossible to see on other games but the particular hue on that FFVI intro really makes it apparent for some reason. I should still have 10 and 22µF 0805 X5Rs along with some L78S05CV regulators (EOL now actually) lying around. So I could do the other caps first, then the other parts one by one to see what actually improves the FFVI screen.