Ghosts αSNES

[YveltalGriffin]

αSNES (alphaSNES): A modern, zero-compromise, single-PCB Super Nintendo Entertainment System portable.

Why?
After getting I2S audio from the SNES in 2020, I thought it would be a shame to build a SNESp with digital audio, yet settle for composite video, a hand-wired cartridge slot, and all the other old-school aspects of SNES portablizing. Not to mention existing commercial SNES handhelds like the SupaBoy and FC16Go have terrible build quality, crunchy audio, noisy composite video, and are based on 2-Chip clone ICs with poor compatibility. (see nauseating images below)


αSNES Design Goals
Here are the features I'm planning on implementing in the portable:

• Reimplemented 1Chip Super Nintendo
  • 1Chip because it offers the highest quality native analog video of any SNES revision (at the expense of some minor visual glitches in games I don't care about.)
  • 2Chip does have the potential for lossless video quality via the debug digital video bus, but the additional complexity of the 2Chip and CPLD/FPGA work needed to make it work with an LCD isn't worth the hassle vs the solution I've come up with for the 1Chip.
• Internal SD2SNESflashcart with cartridge multiplexing
  • An internal flashcart will allow me to carry the entire SNES library with me on trips, without lugging dozens of carts around.
  • With cartridge bus multiplexing, I'll be able to swap between the physical cartridge slot and internal SD2SNES at will, letting me play my physical Super Metroid or Super Mario RPG carts without the hassle of dumping their save files and moving them to the flashcart.
• Integrated GBS8200 with GBS Controlfirmware
  • The GBS8200 scaler board (based on Trueview 5725) can linedouble 15kHz RGBs sources and output 480p VGA. With the custom GBS Control firmware by ramapcsx2, the video quality approaches that of high end system like the OSSC but for much cheaper. Scans of the board here.
• 5" 640x480 LCD driven with linedoubled VGA from GBS8200
• (Possibly) integrated LCD driver (may not fit on the PCB lol)
• I2S audio mixed with analog cartridge audio,
  • My idea here is to use the LM49350 to mix the SNES's I2S with the cartridge's (and internal flashcart's) analog audio for full compatibility with Super Game Boy and MSU-1 romhacks. We'll see if I can manage it.
• Controls: probably an SN30 Pro d-pad or similar, with regular SNES face buttons. Maybe Switch Pro Controller triggers if I can CAD up the case to fit them.
• Case: Recessed cartridge slot, so the cart is flush on the back of the handheld. Need to make some mockups of this to get a feel for it... Perhaps an all-black color scheme? We'll see.
• Miscellanous: Dual 18650s or lipos, AV out

Current Status
I did some basic preliminary PCB work a long time ago, but only this month have I actually started working on soldering up and testing the prototype PCBs. I have zero MCAD work done, but I'm hoping the structure of a worklog and the summer contest will help motivate me!!!

Next post: Assembly and testing of the initial prototype PCB design seen above

 

[Wesk]

You know I'm gonna have to ban you if you keep showing us up like this.

 

[YveltalGriffin]

A while back I stripped and scanned a SNES Jr (SNS-101) to help verify the existing 1Chip SNES schematics floating around on the internet. Never a good policy to blindly trust someone else's work without cross-checking it yourself! You can find the scans here if you're interested. I've also just uploaded a .pdn that has the WRAM and cart slot pinouts to the scans folder.

Using these scans to cross-reference arzi84's public PAL 1Chip schematic, I put together a SNES Jr 1Chip schematic in Eagle. The 1Chip SNES is just complex enough to make this process fun and rewarding, while not grueling. Unfortunately my schematic is hideously messy. I didn't really expect anyone else to ever see it, but that's no excuse... At least I know where everything is and can edit it easily.

Rather than jumping straight to a final form-factor PCB, I decided to make a minimal proof-of-concept prototype board in the spirit of Marshall's 3x3, to verify I had captured the SNES core correctly. I added some modern support circuitry too, mostly known-good design blocks I've used on previous boards. I omitted top and bottom ground planes for aesthetic purposes. Idk if it's justifiable but hey, I'm not getting this thing FCC certified.

This prototype board is 68x77mm, has 4 layers, and contains the following:

• NTSC 1Chip core
• 64-pin 0.5mm ZIF to connect to a cartridge slot daughterboard
• BQ25792 1S battery charger
• TPS61235P 5.1V boost converter
• LM49450 audio amp with 5V-2.5V logic-level converter
• ATtiny84A for housekeeping/I2C stuff

The prototype board also serves as a convenient development platform for the BQ25792. With the pogo pin programming connector, I can quickly upload test code to the ATtiny to configure the BQ25792 and see how it behaves in-circuit.

Oh yeah-- here's the cartridge board. Just a simple SNES cart to 64p ZIF breakout. It's actually "backwards" (FFC is on the same side as the cart label) and will have to be redone for the final build, I think...

Once the boards arrived I got to work pasting 'em up. An electropolished stencil is a must when dealing with 0.5mm pitch BGAs! (This was two weekends back, so please forgive the lack of index cards in my photos.)

The slightly off-center paste doesn't matter for chips this large. Surface tension takes care of it in the end. 6 minutes in the reflow oven, and...

\
Pretty sexy, if I do say so myself (minus the stubborn flux residue.) The BQ25792's inductor (SPM6530) is also uglier than I expected... Might swap it out for something sleeker down the road, since I don't think I'll be charging the cells at full bore 5A in the final design anyway.

Bringup was challenging, especially since I did the initial schematic work a while back and had to refresh my memory as I went. At first I didn't get any video and the S-APU (sound chip) was getting awfully hot. I opened up my scans and immediately noticed I had swapped two 5V and GND pins. Oops! Swapped out the cooked S-APU, lifted some pins, ran some wires, and the board started showing signs of life!

Looking over the design some more, I found other dumb errors. Missed a 5V connection on the CIC's reset line, had the NTSC/PAL pinstrapped wrong, and forgot the reset pullup resistor. So I bodged in fixes for these things.

Woohoo!! Paperboy 2 is booting! Weird wavy video though... Pored over my scans some more, and realized I had straight-up omitted the series DC-blocking caps on the RGB lines to the S-RGB video encoder/amp. Facepalm. Bodged the video passives in, fixed an incorrectly connected video ghosting cap, and rewired the I2S lines since I also transposed all 4 of those somehow. The PST529 (POR IC) is also deadbugged in these pics since the footprint I made for it was uncomfortably small.

Now we're talking! Perfect, stable composite video. I let Super Metroid's demo run for 20 minutes and it played beautifully. Next, I uploaded some code to the ATtiny to configure the audio amp... Digital SNES audio sounds super crisp. Volume was a little low, so I think I'll connect LSVdd to 5V on the next rev rather than to Vsys (3 - 4.2V).

Tested Yoshi's Island the next morning, and it worked too! Love to see it.

Now that I've verified the SNES core works as expected, I need to move on to reimplementing the GBS8200 video scaler and linedoubler! I also need to work on a prototype cartridge multiplexer PCB, so I can test my idea for switching the cartridge data lines between a flashcart and a real cart. And lastly, I need to experiment with the LM49350, an audio amp with both analog and I2S inputs that can mix the two inputs on the fly (replacing the S-MIX IC from the original SNES motherboard.) So, there are lots more custom boards in my future.

 

[thedrew]

shoooooot this is going to be good! I'm curious to see what you do with the GBS8200 circuitry.

 

[SparkleBear]

Bro! This is monumental! I love finally reading through the log so far… long time coming mate. The bodge spaghetti is a good look :p

You mentioned the mechanical aspects of controls (sn30 dpad etc) but what is the electrical element you’re planning? Little micro pi? Custom mcu solution?
I’m sure it’s something slick.

Great work as always, dude.

 

[xsping]

I hope you will win the first place

 

[YveltalGriffin]

Small update, mostly PCB design stuff this time!

All the errors in my αSNES mobo prototype have been fixed. This includes the various pins incorrectly connected to 5V or GND, missing video passives, and the other tweaks I mentioned in my last post, like powering the loudspeakers from 5V to get more power.

I also fixed the MCLK circuit which sends the SNES's system clock to the cartridge. I don't know which games actually use this signal, but the circuit for it wasn't replicated properly on the first revision. Maybe some SA1 games like Super Mario RPG, which I didn't test on the proto.

I noticed some mild video banding/jailbars when I was testing the mobo redesign. I think it's mostly due to the TPS61235P boost regulator (switching regulators are noisier than linears.) The GBS8200 also struggles with noisy input signals, so the SNES core power needs to be as clean as possible to avoid noisy video. "Garbage in, garbage out" and all that jazz.

Luckily, the TPS61235P outputs 5.1V nominal. That gives me some headroom to regulate down a bit further. So I added a MAX38903C low-noise 5V 1A LDO after the boost converter. Putting an LDO after a switching regulator is a pretty common method of reducing noise on the rail, so this should help somewhat. I also put a feedthrough capacitor after the boost converter to further reduce noise.

Polygons definitely came out nice on this section.

I've also been reverse engineering the GBS8200 and reimplementing it into a test PCB just like I did with the SNES. I'm using my GBS8200 scans as a reference for reverse engineering. Only put one evening of work into the layout, so it's still early stages.

So far the new PCB is around 50x75mm and includes an ESP-06 ESP8266 module, USB-C to program it, a nicer 3.3V buck reg than the stock PCB, and a Si5351 clock generator IC. Hopefully in the future I'll be able to release the design for other folks to enjoy. But for now, progress is slow, and I'm not happy yet with the overall board layout. The parts don't lend themselves to a pretty rectangular layout like I'd hoped... I may need to abandon the ESP-06 and integrate the ESP-8266 directly if I can't figure something out. I hate big empty spaces on my boards.

Hopefully this sates your curiosity for the time being, @thedrew !

ESP-06 module. Small in the hand, but big in Eagle...

This is an old photo, but here's what a 1Chip SNES looks like after being linedoubled through the GBS8200 and tweaked to fit the 5" 640x480 LCD. The result is virtually pixel-perfect. This is honestly one of the driving forces behind why I'm on this modern SNES portable journey!

You mentioned the mechanical aspects of controls (sn30 dpad etc) but what is the electrical element you’re planning? Little micro pi? Custom mcu solution?
I’m sure it’s something slick.

Thanks man! I'm probably going to stick with the Mitsumi IC used in later SNES controllers. I have some very crusty junk ones to cannibalize. SNES controllers are shift register based (just like NES) so there's no need for an MCU or anything fancy.

I've been enjoying my SN30 Pro Xbox version a lot lately. I'm thinking of stealing the D-Pad from one of these, and then using both black convex buttons from an SN30 Pro+ and black concave buttons from a Famicom controller (seen below) to complete the all-black look for the final portable.

I hope you will win the first place

Thanks xsping. My goal is to make a portable I'm proud of, so if I can do that and finish by the deadline I'll be satisfied!

Well, that's all I've got for now. I'm gonna try to finish up the mini GBS8200 soon and order it along with the "no bodges required" αSNES mobo proto.

 

[thedrew]

Hopefully this sates your curiosity for the time being, @thedrew !

Curiosity has been satisfied. Anytime I read a worklog of yours, my mind is blown haha.

 

[d3z0rian]

This is an old photo, but here's what a 1Chip SNES looks like after being linedoubled through the GBS8200 and tweaked to fit the 5" 640x480 LCD. The result is virtually pixel-perfect. This is honestly one of the driving forces behind why I'm on this modern SNES portable journey!

Planning to use ZJ050NA-08C panel?

 

[YveltalGriffin]

Well, I struggled with the original TV5725 and SDRAM layout for a week and a half... I even had a dream about the final PCB being a compact square, and was pretty frustrated I couldn't make the board match the form factor in my mind's eye. And I was hesitant to alter the SDRAM layout, since it runs at 100MHz and changing it might introduce issues.

...but then on Thursday I saw this teardown of a similar scaler. The SDRAM is right underneath the scaler IC! This was the inspiration I needed to reroute the TV5725's SDRAM.

So after two more evenings of work, here's the end product! I decided to call it Shinobi.

The board is 47x57mm and only 2 layers. It's got the ESP-06 ESP8266 module on the top left, complete with a PCB antenna, and a CH340 + USB-C for programming/serial monitor. It also has a high current 3.3V reg with a removable series 0805 0-ohm in case you want to power the board from an external 3.3V source rather than 5V. The chonky Hershey font labels came out well too.

The power and ground network isn't ideal yet so I'm still tweaking it. But I'm very glad I was able to make the board match my original vision. Slow and steady! Next post should cover testing this board along with the updated SNES mobo.

Planning to use ZJ050NA-08C panel?

Yup! It's the perfect size for a handheld with physical SNES cartridges.

 

[YveltalGriffin]

Welp, got sidetracked with other projects for a while... My new SNES and scaler PCBs should be here in a few days; in the meantime, I've done some preliminary CAD work in Fusion 360. The case is around 190x97x33mm and has original size SNES D-pad and face buttons.

My initial attempts were hideous blocky things done the old-fashioned way with sketch/extrude, but after a while I wised up and used the Form tool to make a more organic shape. It sort of reminds me of the GBA and Game Gear... and it looks nice from every angle, which is very important to me. Curves galore!

When I started this project I intended to support both US and Japanese/PAL carts, but the US ones are so much larger and more angular that I just couldn't justify it. Most of my own carts are NTSC-J anyway so it works out.

The case is nothing more than an empty shell at this point. No port holes, mounting hardware, or screwposts to hold it together. But I did create the PCB feature, and pushed it over to Eagle. This single large PCB will have every single component except for the cartridge slot. It'll be a massive challenge to route everything I want to integrate on a single PCB like this. I'll probably need 6 layers and ICs on both sides if I want to fit the SNES, scaler, flashcart, cartridge muxing, and LCD driver all on this 187x95mm PCB.

The 1S2P 18650s fit nicely in the handle regions on either side of the cartridge slot. I'm considering having battery clips soldered into the mainboard to hold the 18650s in place without wires. There's not much vertical room in the middle region between the LCD and cart, so I'll have to keep component heights in mind when designing-- it'll be a nice challenge.

Finally here are some screencaps of what I'm thinking for the final portable's color scheme. Glossy black everything (may or may not get it in translucent resin).

My next post will hopefully cover bringup and testing of both the final prototype SNES core and the Shinobi scaler! Probably won't be able to finish everything before the deadline, but I'm not sweating it.

 

[Cholerika]

The mockups look amazing, can't wait to see the final result!

 

[screamingtrees23]

Any chance you'd be able to upload those Snes Jr. schematics you made? I blew a smd capacitor on mine and need help determining the value on a replacement.

 

[72hour]

Well, I struggled with the original TV5725 and SDRAM layout for a week and a half... I even had a dream about the final PCB being a compact square, and was pretty frustrated I couldn't make the board match the form factor in my mind's eye. And I was hesitant to alter the SDRAM layout, since it runs at 100MHz and changing it might introduce issues.

...but then on Thursday I saw this teardown of a similar scaler. The SDRAM is right underneath the scaler IC! This was the inspiration I needed to reroute the TV5725's SDRAM.
View attachment 23263View attachment 23264

So after two more evenings of work, here's the end product! I decided to call it Shinobi.
View attachment 23267
The board is 47x57mm and only 2 layers. It's got the ESP-06 ESP8266 module on the top left, complete with a PCB antenna, and a CH340 + USB-C for programming/serial monitor. It also has a high current 3.3V reg with a removable series 0805 0-ohm in case you want to power the board from an external 3.3V source rather than 5V. The chonky Hershey font labels came out well too.
View attachment 23266
The power and ground network isn't ideal yet so I'm still tweaking it. But I'm very glad I was able to make the board match my original vision. Slow and steady! Next post should cover testing this board along with the updated SNES mobo.


Yup! It's the perfect size for a handheld with physical SNES cartridges.

the SCL network table connected incorrectly to the GND of esp06?

 

[YveltalGriffin]

PCBs arrived! Right now I'm waiting on the parts for the Shinobi scaler to get here. In the meantime I started fleshing out the audio system for the portable.

The 1-Chip SNES has a pretty straightforward audio setup: the S-APU generates I2S, which is converted to analog audio by a DAC on the mobo. The game cartridge can also generate analog audio and pass it through the cart slot, where it's mixed with the system audio. The mixed analog audio then goes to the AV port. AFAIK, only the Super Game Boy, Satellaview, and MSU-1 romhacks on an SD2SNES generate analog audio and pass it through the cart slot like this.

Instead of just replicating this setup, I want to keep the signals digital for as long as possible. That way I don't have to worry about keeping the audio signals clean while routing them all over a noisy all-in-one PCB. Also, I'm going to have one amp for each loudspeaker, to minimize how far the amp outputs have to travel before reaching the speaker.

I originally wanted to use the LM49350 or LM49352. Like I've mentioned before, these amps have dual digital inputs (with asynchronous mixing) which would let me combine the SNES's I2S with the internal SD2SNES's I2S right inside the audio amp. They also have analog inputs, so I could mix in the cartridge audio too (just to maintain full compatibility with the Super Game Boy or an external SD2SNES).

However, the LM4935X only comes in a 0.4mm BGA and when routing it I was forced to use via-in-pad plated-over... This makes the PCB cost explode. Over $200 for a tiny simple test board! So I did some more digging and found the MAX98089, which comes in a QFN package and has all the same features, with much better availability and documentation than the TI parts.

So I whipped up a 2-layer RP2040 eval board for the MAX98089. Guess I'll be ordering it along with my cartridge multiplexing test board pretty soon. There are so many functional blocks to test for this project that it gets pretty daunting... One step at a time.

the SCL network table connected incorrectly to the GND of esp06?

I'm using the newer revision of the ESP-06 which has GPIOs on three of the four outer pads. If you check out the photo of the underside of the module earlier in the thread, you can see what I mean. I've also attached the datasheet in case you're interested. SDA is connected to GPIO4 and SCL is connected to GPIO5.

You can tell the newer revision modules apart from the older ones by the 3 WiFi lines on the Ai-Thinker logo. This blog post has some more info.

I blew a smd capacitor on mine and need help determining the value on a replacement.

Which cap? Most of them are decoupling caps and the values for those aren't critical (10nF, 100nF, 1uF). If it's one connected to a crystal or analog stuff, I couldn't tell you what values those are... I've just been transplanting those from the SNES since I don't have an LCR meter to measure pF or nF.

 

[72hour]

I also made a GBS according to your design

I haven't done the test yet

 

[SparkleBear]

What came first: the scaler or the scaled?

There can be only one (gbs)!

One GBS to rule them all?

<insert your next dumb meme here>

Seriously though... @72hour , Did you really just Reverse engineer his reverse engineered design? Thats some serious Shenzhen Pirate magic.

 

[YveltalGriffin]

Small Shinobi scaler update! Here it is pasted up...

...and here's the fully assembled board! Yes, the H-Sync/V-Sync buffer is backwards. I swapped it around right after.

My SDRAM footprint was 0.508mm pitch instead of 0.500mm. I fell for one of UltraLibrarian's classic blunders! Darn imperial-metric conversions... even NASA isn't immune to them. I was still able to solder the IC on just fine though.

Turns out when programming through Windows, you can't connect the CH340C's DTR and RTS lines directly to the ESP8266. esptool.py expects you to have a two-transistor circuit in hardware to smooth over various timing glitches, and without it the ESP8266 won't be auto-reset into programming mode 9 times out of 10. Oh well. I still managed to get GBS-Control flashed. And the PCB Wifi antenna works!

Even though I used the most recent version of GBS-Control for this test, I'm actually going to use an older version of the FW that supports true 640x480 in the portable. Newer versions only support 720x480 and it doesn't look as crisp on the 5" LCD.

Battling sync issues...

Oops! Sync termination resistor was on the wrong side of the coupling cap.

Bodged in a program mode button and some other stuff as well. αSNES's RGB output was strangely dark (need to debug that later) so I switched over to a stock 2Chip for testing.

Mmmm.... crispy! Protip, turn off the 'line filter' in the GBS menu. It's just bilinear filtering so disabling it helps sharpen the pixels up.

I also went ahead and fixed all the problems I discovered during bringup of the Shinobi board in Eagle:

• Various pinstrapping things on the ESP8266 (reset, chip enable, & GPIO0 not pulled high and GPIO15 not pulled low)
• Removed CH340C auto-reset and added a simple program-mode jumper
• Fixed sync input resistor
• Connected TV5725 debug output pin to the ESP-8266
• Measured remaining mystery passives with my new DT71 LCR tweezers and added their values to the BOM'

Once it arrives and I verify it's working reliably I'll post the files on GitHub. @Redherring32 noticed that thanks to the 0.1" header breakouts you can use it as a through-hole module on a carrier PCB, so maybe folks will find it useful. Happy accident.

I also made a GBS according to your design

Wow, awesome! I'm flattered. Hopefully you got the sync input circuitry right unlike me.

How are you programming the ESP8266? Is there a usb-serial IC hiding on the backside, or do you program the module separately and then solder it on?

 

[72hour]

Small Shinobi scaler update! Here it is pasted up...
View attachment 24794
...and here's the fully assembled board! Yes, the H-Sync/V-Sync buffer is backwards. I swapped it around right after.
View attachment 24795View attachment 24796
My SDRAM footprint was 0.508mm pitch instead of 0.500mm. I fell for one of UltraLibrarian's classic blunders! Darn imperial-metric conversions... even NASA isn't immune to them. I was still able to solder the IC on just fine though.
View attachment 24798
Turns out when programming through Windows, you can't connect the CH340C's DTR and RTS lines directly to the ESP8266. esptool.py expects you to have a two-transistor circuit in hardware to smooth over various timing glitches, and without it the ESP8266 won't be auto-reset into programming mode 9 times out of 10. Oh well. I still managed to get GBS-Control flashed. And the PCB Wifi antenna works!
View attachment 24797
Even though I used the most recent version of GBS-Control for this test, I'm actually going to use an older version of the FW that supports true 640x480 in the portable. Newer versions only support 720x480 and it doesn't look as crisp on the 5" LCD.

Battling sync issues...
View attachment 24799
Oops! Sync termination resistor was on the wrong side of the coupling cap.
View attachment 24800
Bodged in a program mode button and some other stuff as well. αSNES's RGB output was strangely dark (need to debug that later) so I switched over to a stock 2Chip for testing.
View attachment 24801
Mmmm.... crispy! Protip, turn off the 'line filter' in the GBS menu. It's just bilinear filtering so disabling it helps sharpen the pixels up.

I also went ahead and fixed all the problems I discovered during bringup of the Shinobi board in Eagle:

• Various pinstrapping things on the ESP8366 (reset, chip enable, & GPIO0 not pulled high and GPIO15 not pulled low)
• Removed CH340C auto-reset and added a simple program-mode jumper
• Fixed sync input resistor
• Connected TV5725 debug output pin to the ESP-06
• Measured remaining mystery passives with my new DT71 LCR tweezers and added their values to the BOM'

Once it arrives and I verify it's working reliably I'll post the files on GitHub. @Redherring32 noticed that thanks to the 0.1" header breakouts you can use it as a through-hole module on a carrier PCB, so maybe folks will find it useful. Happy accident.
View attachment 24803
View attachment 24804


Wow, awesome! I'm flattered. Hopefully you got the sync input circuitry right unlike me.

How are you programming the ESP8266? Is there a usb-serial IC hiding on the backside, or do you program the module separately and then solder it on?

yes! I use the automatic download circuit hidden in the back, reserved debugging interface! At the same time, a set of composite signal separation circuit is added. I just want to use this module alone

 

[Shankwannabe]

I'm not comprehending half of this