Nintendo Virtual Boy Displays: Reaching out for help, to create new larger displays

[T-Squared]

The Nintendo Virtual Boy. A flop, yet an engineering curiosity. There's an entire community out there set on keeping interest in this oddity.

One problem, though, are the displays. The original LED displays are failing, and trying to bring them back can be difficult, as the connections between its flat-flex cable and the PCB were manufactured in a substandard way, and soldering them for permanence can lead to problems of the traces being lifted.

Much of what I have learned I got from https://furrtek.org/?a=vbtvout.

Trying to do this research on my own of trying to find a replacement interface has yielded some results, but I am still in preliminary stages, and I need some help from those with knowledge in electronics engineering, mainly because of the odd, but simple way the display produces its signature 4-shade brightnesses.

How the displays work, to my understanding:

A long single row/array of LEDs (224 of them) blink in a sectional pattern, to produce a vertical segment of a projected "screen". This is reflected into your eye(s) by a mirror. The mirror vibrates at the same speed as the blinking of the LEDs, building up this "screen" in vertical slices thanks to persistence of vision.

What I have found is that the company that produced the LED array essentially made something similar to today's LED Drivers. I already know I need 224 outputs to drive the LEDs, which modern LED Drivers can provide through serial daisy-chaining.

However, with the selections I've found, I don't know how to implement the PWM shading that the system uses, and a lot of the IC intricacies go right over my head:

TI TLC5955 Datasheet
Found on my own, has 48 outputs, would only need 5 to reach the 224 mark, but is missing a latch, CHIP SELECT/ENABLE, and CLEAR signal pins.

TI TLC5957 Datasheet
Suggested by a Tech Tangents Discord member, would only need 5 to reach the 224 mark, but is missing CHIP SELECT/ENABLE, and CLEAR signal pins, and requires a MOSFET gate to drive the LEDs.

Analog Devices LT3746
Found on my own, but is expensive, would need 7 to reach 224 mark, missing a latch and CLEAR signal pins, and requires a MOSFET gate to drive the LEDs.

Lumissil IS31FL3248 Datasheet
Seems to have all that I need, but unsure about voltage needed for LEDs, as the originals run off of 5v, and the daisy-chained implementation shows 12v. (Maybe I'm misinterpreting it)

What I need in a driver is:
Shift Registers-I initially found a 128-channel IC that could do this, but I've found that modern LED drivers have this.
A Latch -Some modern LED drivers have this. I know this is used to "lock" the picture data.
PWM/Brightness Control- Again, some modern LED drivers have this. PWM, if I have the reverse-engineering from Furrtek correct, is only used to dim/fade the screen, and is provided by the VB's mainboard.
Properly-sized LEDs- 01005 LEDs are the closest I can get, but it should not be difficult, because I found the Virtual Boy's projected "screen" size too small. My idea might enhance it.
LED-Addressing- The circuitry on the VB Display takes care of this. However, to save costs, apparently, they use the picture data, basic logic gates, and transistors to drive the LEDs, not any protocol.

Questions/Comments posed to me:
Your LEDs are too big.- I'm well aware of that. If anything, this might not be a bad thing, as, again, in my opinion, the projected screen is very small. I want to try to get the projected screen as large as I can get it. I'm pretty good at modeling 3D prints. I got a new 3D printer that should be able to print a new optical pathway tube.
What about the old lenses? They won't focus properly.- Again, well aware of that. I want to try using a different focusing lens.
What about the physical size of the Virtual Boy's casing?- Again, well aware of that. I'm pushing the limits of the casing, hoping that it can fit.

 

[Eudimorphodon]

Properly-sized LEDs- 01005 LEDs are the closest I can get, but it should not be difficult, because I found the Virtual Boy's projected "screen" size too small. My idea might enhance it.

I found a datasheet saying these are .25mm across in the shorter dimension, or 4 per mm if you can pack them right next to each other? If the original array has 224 pixels in 10mm your new array is going to be almost six times larger? (Over 2 inches wide verses 2/5ths of an inch?) Obviously this will work with none of the original optics, nor will the original vibrating mirror be of any use. So what is the plan here? At this size the scan area for the 384x224 screen array is going to be around 2.5 by four inches, that’s far larger than the original case; to do binocular vision with two screens that large would probably require a system of optics similar to a really large set of binoculars to “shrink” the image to correct for the fact that a viewing distance that puts 2.5” vertically far enough from your face won’t let you put them horizontally at the right place without overlapping.

Given the tech advances since the virtual boy came out wouldn’t a more practical solution for a total screen replacement be to capture the data output, render it into a framebuffer, and display it on a small LCD screen? I would bet you could find screens with enough pixels that fit in the space of the original screen assemblies. Driving a standard LCD interface is far fewer signals than 224 LEDs.

 

[T-Squared]

As difficult as this may seem, I want to keep the spirit of the console intact. Sure, an LCD screen would do it, but to me, it's like running a modern electric engine in an old 1950s roadster, or replacing the guts of a vintage computer to make a modern sleeper setup. As far as I know, that's not what we do around here in the VCFed.

 

[Eudimorphodon]

As difficult as this may seem, I want to keep the spirit of the console intact. Sure, an LCD screen would do it, but to me, it's like running a modern electric engine in an old 1950s roadster, or replacing the guts of a vintage computer to make a modern sleeper setup. As far as I know, that's not what we do around here in the VCFed.

I’m not going to dispute that it would be neat and oldskool and whatever to build a new screen using the same tech as the old one, but the indisputable fact here is that you’re not going to be able to make one that fits in the original case without access to factory tooling.

Have you tried building any mechanical prototypes of what this is actually going to require? Again, if you make an LED strip six times longer than the original you’re going to also need a mirror with 36 times the area of the ones the device came with. Are you confident you’re going to be able to make that much heavier mirror oscillate at the same rate and amplitude? The original screen is a diabolically complex system that not only has to display 384 lines of pixels for every oscillation, it has to adjust how long it holds each line of pixels lit to compensate for the acceleration and deceleration of the mirror at different points in its cycle. All of this is going to have to to be re-engineered to run at a different scale, with different mass parts made from different materials.

I suppose an alternative would be to use lenses to focus down your much larger lighting elements so they appear at a size comparable to the original arrays relative to the mirrors, but now you’re looking at having to cut a couple periscope tunnels out the sides of the enclosure. Or maybe you could use fiber optics to tunnel down larger LEDs to the tiny dots you need? That’ll be fun to build by hand…

Anyway, if you have some prototypes/mechanical sketches that explain this idea in more detail I would love to see what your proposed solution is.

 

[Torch]

The VB is the only system to come from Nintendo that I like, and the only Nintendo product in my game room. Red Alarm alone was a mind blowing experience and a game that still impresses me 30 years later. Galactic Pinball is another favorite.
Stores had giant pyramids made out of VBs, practically giving them away. I remember paying $19.99 for mine and almost all the games were $2.99. Jack Bros. and Waterworld never went on clearance, and I really regret not paying $30 for them.
Happy that mine has never had any issues and that I kept all the original boxes.
The VB has a uniqueness like the Vectrex, and considering what they sell for these days, I think someone can do well selling a repair kit.

 

[T-Squared]

At the moment, I'm working on a basic prototype, just to see if an LED array driven by modern LED drivers would be able to produce some kind of image. No graphics shading (as of yet), only the basic pixel data. No plastic-frame support, just holding the display PCB, connected to the mainboard, in front of the current lens/mirror assembly. That way, I can know if things are possible.

 

[Istarian]

I don't know if it can update its output fast enough, but you could use a 16-bit I/O expander like the Microchip MCP23017 or MCP23S17.

https://www.microchip.com/en-us/product/mcp23017

You would probably still need an LED driver circuit though.

Or maybe you could use one or more of these?

Adafruit 24-Channel 12-bit PWM LED Driver - SPI Interface - TLC5947

Adafruit 24-Channel 12-bit PWM LED Driver - SPI Interface

For all of you out there who want to control 24 channels of PWM, we salute you! We also would like you to check out this breakout board for the TLC5947 PWM driver chip. This chip can control ...

www.adafruit.com

-----

It's not too hard to source 5mm or 3mm LEDS, whether through hole parts or SMD. Building a compact array might be a difficult trick to pull off.

Using LED bar graphs might be an option if the resulting "pixels" being rectangular is tolerable.

E.g. a 1" module w/10 LEDs provides, in principle, 1 LED/2.54mm.

Carefully arranging 3mm LEDs in an alternating left/right pattern might also work if you come up with a frame so each just looks like a slightly wider element.

Even so you'd end up with a ~16.5" long LED array.

 

[daver2]

You can individually control 144 LEDs with the IS31FL3731. I have not read the data sheet yet though, so I am just tossing this out there...

The smallest LED I have found so far is a 0402 SMD at 1mm × 0.5mm.

EDIT. This is claimed to be the smallest SMD LED:
NanoPoint-0201 Series LEDs
SunLED Company, LLC
NanoPoint-0201 Series LEDs.

Dave

 

[rmay635703]

Maybe I’m dumb but weren’t these led strips off the shelf parts used inside the at the time revolutionary 300dpi led laser printers? (Remember those things? More compact and less expensive than the traditional machines)

If they were used in a Ricoh or some other machine by the millions you could maybe source scrap parts to play around with.
I believe certain scanners used led bars as well

 

[Istarian]

@rmay635703 That might be the case, but did such printers actually get much use in the 1980s and 1990s?

LED printer - Wikipedia

en.wikipedia.org

The above page only mentions OKI products in the context of pre-2000.

HackIt: Why Aren’t We Hacking On The LED Printer?

Strings of LEDs are a staple of the type of project we see here at Hackaday, with addressable devices such as the WS2812 in particular having changed beyond recognition what is possible on a reason…

hackaday.com

Here's a hackaday post commenting on the fact that nobody really seems to be doing any with the presumably plentiful LED printers lying around....

 

[rmay635703]

@rmay635703 That might be the case, but did such printers actually get much use in the 1980s and 1990s?

LED printer - Wikipedia

en.wikipedia.org

The above page only mentions OKI products in the context of pre-2000.

LED lasers were very popular in the early 90’s.

And yeah everybody with a high end home prosumer office had an oki laser, my neighbor had 2. But they were very much a thing limited to 89-93 and were killed off As a marketing gimmick when 600dpi became the only standard

 

[Eudimorphodon]

Maybe I’m dumb but weren’t these led strips off the shelf parts used inside the at the time revolutionary 300dpi led laser printers?

I don't think you're wrong relating the LED element inside the Virtual Boy to Oki's LED printers, I think they are in fact the same tech and it's very likely that Nintendo bought their parts from the same factories that were churning out the LED array chips for printers. (The element used by Nintendo is around 600 DPI, which is on par with Oki's 90's vintage mono printers, so it could literally be the same elements, although it's certainly possible they churned out some custom-size silicon for them.)

That said, though, I think calling these "off the shelf parts" is carrying a lot of weight. Here are a couple PDFs about the history of developing these printers and talking about the challenges of building a 1200 DPI printhead that has to drive over 14,000 individual dots. The TL;DR is that I don't think the LED chips that make up these devices are the sort of thing that you can really handle outside of a manufacturing facility even if you could find someone willing to sell them to you. The history PDF above talks about the size of the LED arrays in the first revision of the tech (128 LEDs per chunk, and because those early versions had "margins" they had to stagger them around the drum), but I can't find any spec sheets or anything that says how many LEDs are in each chunk of the more modern printheads.(*)

Anyway, it's certainly an interesting *idea* to try disassembling one of these printheads and seeing if it would be possible to pirate part of the LED array, but I suspect actually getting the LED elements off without destroying them, let alone trying to reuse them, is going to be *very* nontrivial to do.

(* Edit: I don't know if this direct link will work, but this article about an "improved" 1200 DPI printhead developed around the year 2000 says every element had 384 pixels on it; the whole printhead assembly was 26 elements. Of note is the fact that the article, in comparing this printhead with a previous version of the 1200 DPI element, talks about an improvement in how the bonding wires are set up so it uses fewer wires than the older style element. Assuming the tech in the Virtual Boy and the older 600 DPI printers is like the older one trying to reuse these LED array chips is going to mean attaching bond wires to pads with a pitch measured in tens of microns. Sounds like a bit of a challenge.)

And yeah everybody with a high end home prosumer office had an oki laser, my neighbor had 2. But they were very much a thing limited to 89-93 and were killed off As a marketing gimmick when 600dpi became the only standard

They are still making printers based on this tech, and it's not just Oki, the Wikipedia page itself mentions how Fuji Xerox and Canon have their own spins on it.

 

[Istarian]

It would be a hell of a lot of work to wire them up or get them onto a PCB, but these ones are pretty darn tiny (1.5mm square) withour being invisible.

NeoPixel Addressable 1515 LEDs (1.5mm x 1.5mm) - 10 pack - SK6805-E-J

NeoPixel Addressable 1515 LEDs (1.5mm x 1.5mm) - 10 pack

Make your own smart LED arrangement with the same integrated LED that is used in our NeoPixel strip and pixels. This tiny 1515 (1.5mm x 1.5mm) RGB LED is designed to be surface-mount ...

www.adafruit.com

$3.95/each for quantities < 10

240 leds = 24 units x 10 leds/unit
24 units x $3.56 = $85.44
^ assuming qty disc. are automatic
^ might be cheaper direct from mouser/digikey if available

I'm pretty sure we're still talking an array over a foot long, though, if you line them up end to end.

If it suits your fancy, that doesn't seem too bad for a prototype. Could probably be built to go in a box to sit on a stand and have some simple optics.

._______._______.
|   _   |   _   |
|  (_)  |  (_)  |
:_______:_______:
        |
        |
        |
       /|\
      /   \
-----------------

 

[tetsujin]

It would be a hell of a lot of work to wire them up or get them onto a PCB, but these ones are pretty darn tiny (1.5mm square) withour being invisible.

NeoPixel Addressable 1515 LEDs (1.5mm x 1.5mm) - 10 pack - SK6805-E-J

NeoPixel Addressable 1515 LEDs (1.5mm x 1.5mm) - 10 pack

Make your own smart LED arrangement with the same integrated LED that is used in our NeoPixel strip and pixels. This tiny 1515 (1.5mm x 1.5mm) RGB LED is designed to be surface-mount ...

www.adafruit.com

$3.95/each for quantities < 10

I think neopixels are too slow for a display like this: to update 224 pixels 384 times in 1/50 of a second (or less?) - so that's about 4.3 million updates per second. Neopixels' data rate is fixed such that you can only update them around 2000 times a second, and their PWM rate may be too slow for an oscillating display as well. The "dotstar" (apa102) LEDs are faster, and may be fast enough for an application like this.