SNES PSU

The place for all discussion on gaming hardware
User avatar
NewSchoolBoxer
Posts: 369
Joined: Fri Jun 21, 2019 2:53 pm
Location: Atlanta, GA

Re: SNES PSU

Post by NewSchoolBoxer »

The time has come. Good news and bad news.

-----------------------------------
USB 5V to 9V USB Adapter
pic
Spoiler
Efficiency is slightly worse than switching PSU and better than linear
Image
Spoiler
Good news is that the USB to 9V adapter does indeed power the Super Famicom and ran Mega Man X and Yoshi's Island no issues. I can confirm this $9 adapter meant to power guitar pedals is negative center polarity.

I'm not sure if it's really a charge pump or a boost converter but, either way, can still generate low MHz switching noise. Best to hold out buying one until I measure the noise with an oscilloscope. If low kHz range that 7805 can filter then just need to be careful of the USB 5V source. Lose what you gained with a cell phone charger's switching noise but a power bank should be fine.
-----------------------------------

Rechargeable 9V Battery
pic
Spoiler
No load slightly above 9V is proof of regulation and comparable to switching PSU
Image
Spoiler
Rechargeable 9V battery can work while being charged. Bad news is that it powered on the console but games wouldn't start. I tried when not charging too. This means it put out enough voltage to activate the 7805 but not enough current. Looking at the watts on Poniie meter confirmed. Should note that a red LED light reaches max brightness at 20mA. It turning on means the 7805 is outputting some power but not necessarily 300mA or more to run any game.

Next step is testing two 9V batteries in parallel to output double the current. This is necessary when different 9V brands both specified they charge at a rate of (5V) x (500mA) = 2.5W. Poniie meter confirmed 2.6W. Console uses more like 4W. Power to console needs to be less than the power charging the batteries.

Alligator to alligator clips and a $2 parallel 9V to barrel connector on the way! If I need to wire more than two 9V in parallel - that must be charged while working - then this setup becomes convoluted. No one makes an adapter that connects more than two 9V together. Got to alligator clip them.
-----------------------------------
Regular 9V (don't do this)

Just as a test, I was able to boot Mega Man X on a regular (non-chargeable) 9V battery. Battery will last 30 minutes max before dropping to 7V and cutting off the 7805.
-----------------------------------

Isopropyl Alcohol Cleaning
pic
Spoiler
Image
Spoiler
I took VajSkids Consoles' advice and cleaned carts and console with 99% isopropyl alcohol + Q-tips. All were dirty though Mega Man X and Yoshi's Island already with the best chance at booting were slightly less so. Now I can get SFC Donkey Kong Country and SFC Mega Man X3 to work. Mega Man X and Yoshi's Island boot 100% of the time now versus 30-50% before. Best 10 minutes I ever spent.

Other five games I tried still don't work. I didn't unscrew the carts to clean the whole of the connector, just what I could reach while carts are intact. Real solution may be replacing cart electrolytic capacitors that are guaranteed to dry out and fail given enough years.
-----------------------------------

PAL Notes
pic
Spoiler
Image
Spoiler
I imported a PAL SNES + Super Game Boy + Honey Bee adapter (bypasses NTSC lockout for most games) from Italy after watching eBay for a month. I have two displays that can take PAL. Waited out Street Fighter II Turbo and Super Mario World from US sellers. Fairly confident the SD2SNES flash cart will work when it's made by Europeans.

I was thinking PAL emits less EMI due to no TTL csync. Turns out I had the PAL schematic this whole time and the truth is no longer clear.

PAL SNES uses a cleverly designed but convoluted (even by Nintendo standards) 2 stage Dickson charge pump that requires AC supply for the two inputs pulses to be out of phase. The AC sine waves are really supposed to be square waves and the low 50 Hz forces high 10uF capacitors.

Standard Zener diode regulator fixes output to +12V. Charge pumps are noisy like I was saying, meaning it's no longer certain that PAL noise is less. Difference must be tested.

The thing that caught me off guard is the +12V line clearly can't work without AC input. People including me and the Bible say PAL can run off AC or DC voltage. I have no SCART but could stick a probe on the +12V multiout pin to confirm it doesn't work on DC input. Way I understand things, entire purpose of replacing TTL csync with +12V is to tell PAL television with SCART to use RGB in the cable for video versus Composite that gets used as sync instead. Some PAL televisions could adjust the input with a menu option.

PAL consoles came with 9-9.4V AC PSUs. Worded as such on the PSU because 230V yields 9V and 240V yields 9.4V. Issue of supplying AC instead DC power, besides PAL RGB situation or quality of SNES built-in rectifier, is the AC supply is not regulated. 220V makes 8.6V.
-----------------------------------

Measuring Current
Spoiler
Current needs to be measured in series since it divides in parallel. Voltage works the other way around. I thought out what to do and ordered female barrel to screw connector adapter + screw connector to male barrel adapter to plug into SFC. In between the screw connectors I can wire multimeter leads to measure current leaving the power supply and prove exactly how much is needed without having to lift the fuse or regulator pin or solder anything.
User avatar
Lopenator
Posts: 214
Joined: Sat Nov 14, 2020 10:40 pm

Re: SNES PSU

Post by Lopenator »

Maybe the Triad is best if the OEM has such ripple.
User avatar
NewSchoolBoxer
Posts: 369
Joined: Fri Jun 21, 2019 2:53 pm
Location: Atlanta, GA

Re: SNES PSU

Post by NewSchoolBoxer »

PSUs

Triad has acceptably low ripple in the kHz range from Sirotaca's measurements. 120 Hz noise is high but reasonable for a switching mode power supply. It's the only proven option of regulated supplies and therefore the best option for now.

I discovered a medical grade PSU that can adjust for either polarity. I made up a company name to ask customer service how the PSU does that. Turns out it's bundled with custom 5.5x2.1, 5.5x2.5 and 3.5x1.5 adapters that can be rotated to switch polarity. I like for $16.92.

The Chinese regulated linear PSU is discounted a little to $32.29 but add on a $4 polarity reversal adapter and I don't think anyone is justifying without oscilloscope proof.

I'll have all the current measurement supplies soon. I overlooked USB power meters. Turns out they can handle 30V/3-5A. I ordered a $15 AT35 model with better accuracy than my multimeter. They only measure positive DC but voltage, current, charge (mAh), load impedance and accumulated power on one screen is super tight.
-----------------------------------

9V Batteries Part II

So 2x 9V batteries in parallel can run SNES! Parallel wiring is red (+) to (+) and black (-) to (-) and since my 9V connector is negative center, I don't have to reverse polarity. If multimeter shows 9V instead of 18V then you did it right. That is nice with alligator clips that you can swap polarity by connecting (+) to (-) but 20" leads (50.8cm) was a bad choice. Adafruit sells a 4.5" (11.5cm) set.

parallel
Spoiler
Image
I think the accessible way is $1.95 parallel connector + 50 cent screw terminal to 5.5x2.1mm power barrel. Swap polarities. Since Vetco charges $13.95 shipping and eBay rates push bulk buy of 5 or 10, maybe there can be a small market to resell a full kit with the batteries. I bought 3 of the Vetco adapters, I'll give one away.

Brand matters.

Edit: The Enegon pair kept charging 1.4W while red at full charge out of the box. Recharging after use, they did flash the green light at full charge and drop power consumption to 0W.
Spoiler
Image
I left Yoshi's Island on for an hour and the Poniie meter's output ticked up from 4.27W (508mA at 8.4V) to 4.74W (564mA). Not necessarily a problem but not what I want to see. As batteries lose charge, their output voltage ticks down. Since the 7805 works the same off 8.0V as it does 8.4V, battery efficiency ticks up. I could see there being an equilibrium point. The 2 in 1 outlet charging cable is a plus.

The Beston pair also turn red while charging and power consumption drops to 0W when at green fully charged. Lights are brighter. I had them running Yoshi's Island for an hour and meter ticked back and forth between 1.78W and 2.20W (262mA at 8.4V). Unplugged from powering the console, they consumed 3.92-3.95W while charging, meaning they were definitely losing charge while running the SNES.

Seems Beston charging is gimped while powering SNES so cannot sustainably power. Batteries must be outputting sufficient 400-450mA at 8.4V to keep game running but that minus the 262mA is what they're losing. I found a lithium-ion USB power bank that warns its self-charging and powering efficiencies drop 80% while doing both at same time. Can only sustain 0.5A output while charging and powering out of a max current of 2.1A.

Current reading will reveal the truth. Even if the Beston can't sustainably power a console, lasting 4 hours would come close enough. Console power usage doesn't drop with multiout unplugged. I can do endurance testing.

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

Re: SNES PSU

Post by NewSchoolBoxer »

Pro of USB Meter
Spoiler
The AT35 USB meter came and it's pretty cool. Between the Poniie on the outlet and the AT35 on the USB port, I can measure PSU efficiency if it supports USB. Just (AT34 power) / (Poniie power).

Nintendo NES/SNES Classic is 78-79%, iPhone white brick is 79-80% and generic travel charger is 80-81%. USB-C to USB-A adapter would cover Nintendo Switch.

The 80% magic number is what modern efficiency standards are based around. A 3% spread is enough for Mega Man X to pull between 5.91W and 6.14W from the Poniie, from most to least efficient.

The current at 5V that AT34 measures from power bricks isn't helpful otherwise since I don't know USB 9V adapter efficiency. Formula is (current into SNES) = (5V/9V) * (5V current) * (efficiency), assuming voltages are exact.
Image
Con of USB Meter
Spoiler
The massive screwjob with the USB meter is when I converted the power barrel to USB to measure the all-important ~9V current, the meter doesn't turn on. Game still runs so it passes through power and ground. Makes sense when I think about it that the meter powers itself from low voltage USB data pins versus unpredictable 5-30V. Well, barrel only passes (+) and (-) so no data.

I ordered Arduino jumper cables since those are thick enough to handle at least 500mA over a non-soldered connection. Then I can measure current with the multimeter by going barrel -> jack to screw terminal -> multimeter -> screw terminal to barrel.

I think I can breakout USB and pass the data pins around the barrel and into a new USB that AT34 hooks into since I already own USB to barrel adapters. Is convoluted though.
Image[/b]
Measurement through JST
Spoiler
The cheapest and best approach to measuring in series since it works with NA SNES and its custom barrel is to split between the JST connector on the power switch. I've almost figured out which type it is. With some care, you can pull the JST cable right out or put it back in. No locking mechanism.
Image
User avatar
Lopenator
Posts: 214
Joined: Sat Nov 14, 2020 10:40 pm

Re: SNES PSU

Post by Lopenator »

I think if we knew the data sheets on an OEM PSU recapped it would help us decide a lot what power is best to use.
User avatar
NewSchoolBoxer
Posts: 369
Joined: Fri Jun 21, 2019 2:53 pm
Location: Atlanta, GA

Re: SNES PSU

Post by NewSchoolBoxer »

I did 15 minutes of 'screwdriver research' and it seems Nintendo PSUs are locked with the "Double D" carburetor screwdriver. Console5 reselling it by masking the knowledge of where it comes from is kind of low but $8.84 including shipping to US isn't a bad price. I paid $5.96 to ship from California. Somewhat interesting how many eBay storefronts operate out of Walnut, CA with identical goods and prices to give you the illusion of choice. Now to wait out shipping.
--------------------

Current Draw
tl;dr: Max current needed from a real cart was 546mA, with non-enhancement chip games and Super Game Boy below 500mA. Flash cart games were 553-574mA, with enhancement chip games on the low end.


I know most people don't care about the setup so here are some answers. On the original release model SHVC-CPU-01 of the SFC that is the most common by far, the current drawn for the 6 non-enhancement chip games I measured was 438-479mA. Enhancement chip games from lowest to highest current were Mega Man X3 (CX4, but was it being used?), Super Mario Kart (DSP-1), Super Mario RPG (SA1), Star Fox (Super FX v1) and Yoshi's Island (Super FX v2). Current range was 452mA to 546mA and discounting X3, would be 479-546mA.

Super Game Boy 1 and 2 can also be considered enhancement chips and 2 always drew more current, most likely to power its own clock running ~5% faster and the link cable circuitry. Tennis took 455mA on SGB1 and 462mA on SGB2. Highest was Donkey Kong Land at 472mA on SGB2. So actually it seems Super Game Boy games consume current on the high end of regular SNES games. I can test 2 player link mode to a Game Boy but I don't expect a significant difference.

Current draw off SD2SNES was interesting. The enhancement chip games actually took equal or less power compared to normal games, with enhancement games at a low of 553mA off Super Mario RPG to Super Mario Kart at 574mA and normal games ranging 566-574mA. I think the explanation is the FPGA is one chip and isn't engineered to turn parts off it isn't using. The result is every game needs about the same amount of power. The full truth is probably more complicated.

Super Game Boy 2 emulation ranged from 559mA Harvest Moon to 568mA Tennis, despite simplistic Tennis from 1989 drawing the least current in real cart form. Again, all about the same. The ROM select screen took the lowest of any flash cart measurement at 551mA.

Somewhat amazingly, the 680mA PSU I bought from local retro game store is apparently sufficient, so long as I avoid 5V rail abuse with 100mA wireless controller or power anything off the multiout. Nothing broke 600mA. NA and Japanese Nintendo PSUs state a max of 850mA. No need to be rated higher.
--------------------

Test Setup for Current and Voltage Measurement

My JST trial and error paid off. SNES/SFC uses a 2 contact JST-EH connector between the on/off switch and the console. JST isn't a meant to be regularly connected/disconnected but you can remove it with care. Notice it has outward divots on the sides and inward on the top. I had to buy a set of 10 male cables + 10 female connectors. I'll mail 2 or 3 people a free pair in US if you DM me and agree to share a few current measurements. Be careful if you try to buy them since search results will return incompatible JSTs and everyone's pictures are small and grainy.

Defying every expectation, the power flows from the connector next to the voltage regulator through the fuse then towards the on/off switch then back towards the connector again. Despite SNES taking negative center polarity, the console reverses this by running voltage on the red wire and ground on the black.

JST manipulation isn't so necessary for SFC and PAL with 5.5x2.1mm breakout adapters but it is to bypass the custom power barrel of NA SNES. I fixed my NA model SNS-CPU-GPM-02. Can compare power metrics on this revision soon.

Now the fun part. Isn't this clean?
Spoiler
Image
Pigtail cable plugged into SFC JST-EH port next to the voltage regulator and female connector with 2 prongs connected to on/off switch cable. This lets me connect pigtail red cable through multimeter to measure current and then out to the red prong. The black pigtail cable goes to black prong with breadboard in the middle to connect 2 cables without soldering. I can avoid breadboard when 8" IC hook to IC hook cables come in the mail. Does allow more bypass capacitors to be added, however far from voltage regulator they would be.

I added the red IC hook you can see on the voltage regulator I (input) pin to measure input voltage on a second multimeter. Was between 8.40-8.49V for every game, which suggests SNES D1 diode under the heatsink is a Shockley diode as the circuit diagram indicates:
Spoiler
Image
Not sure if I read this somewhere or figured it out first but instead of using the G (ground) pin as 0V reference, can use any part of the heatsink! No more shorting the I and G pins together.

Power chain goes Poniie meter -> blue USB meter -> 5V USB to 9V barrel adapter -> SFC.
To give you one set of measurements, Street Fighter II: Turbo was 6.02W / 10.25VA from the power outlet, 4.73W (4.836V / 0.9773A) at input to 5V to 9V adapter and 3.88W (8.48V / 458mA) at voltage regulator that likely needs a min of 7V. Console receives 458mA minus ~5mA or so quiescent current of 7805 at 4.9-5.1V for about 2.27W. This gives the regulator 59% efficiency (2.27/3.88) and means it dissipates most of the lost 1.61W through the heatsink.
--------------------

SD2SNES Rant

Obviously I got SD2SNES to work again. Hint came from the official message board saying that it can corrupt its own firmware and that I just need to re-install and overwrite the files in the sd2snes folder. I was wrong to blame the makers for not working on SFC, though that a device that costs over $200 ($125 with trade-in) and has been around for years can corrupt itself doesn't look good either. My anger shifted to the flash cart blocking the power button on my stock SNS-CPU-GPM-02 console and therefore can't power on unless the console's top shell is removed. Must be a rare console revision when I could only find one similar complaint online.

Other smackdown is I think SD2SNES and its FXPAK PRO evolution steal part of the The Lion King source code, ported from SNES assembly. I wouldn't be messing with Disney IP but that's just me. Can see the addition from ancient firmware:
v0.1.4

SPC Player (contributed by necronomfive/blargg)
System Information screen now shows CPU/PPU revision (contributed by necronomfive)
Part of the reason to own The Lion King is the B-A-R-R-Y cheat code in the Options menu that, for our purposes, brings up a CPU/PPU revision statement. May also be a 1CHIP detector without unscrewing console. SHVC-CPU-01 shows up as 1/1/1 on real cart and SD2SNES System Information in the same manner:
Spoiler
Image
B-A-R-R-Y off real cart and emulated game work on SNS-CPU-GPM-02 as well. The port must have been botched, however, since going to System Information freezes everything and forces a restart. Unless you think the flash cart is reading the timing crystal wrong?

There is room to deny that ikari_01 knew where the code came from but doing due diligence on what you stick in your product comes with the territory.
I do give the flash cart credit for showing the exact frequency of the Master Clock versus having to measure it in-circuit. Important as a check of the timing crystal's drift, which can be partially rectified by adjusting the TC1 red capacitor potentiometer. Is probably related to the C11 situation.
Post Reply