Basic Electronics 201 - Diodes, Ghosting, and Myth-busting

[saint]

I have been thinking about diodes and keyboard encoders.  Why?  Nothing else to do on a Saturday night I suppose.  

At any rate, the point about keyboard hacks and diodes is moot for the most part due to improved programming in modern keyboards.

However, for the sake of argument, say you have an old keyboard that suffers from ghosting.

Why would diodes make one bit of difference?  Ghosting is not an electrical issue, it's a programming (encoder firmware) issue, right?

In a small matrix of:

..   X1   X2
Y1  A      B
Y2  C      D

If A, B, and C are pressed then D is generated by the keyboard encoder because it sees pins X1, X2, Y1, and Y2 all come high at the same time. Right?

The basic theory was putting diodes inline with the wiring would prevent "backfeed" of current from completing the circuit for D, correct?

But D isn't being generated because its circuit is being completed physically under the button. You could physically cut the traces at the point where they branch off to the "D" and it wouldn't make a difference.  The encoder still sees all 4 terminals high and generates the keystrokes.

The only way to prevent a ghost key would be a "diode" in the programming, no?  I.E. Keyboard blocking that was added to encoders.

A diode would make sense if for instance all the keys had lights on them, and we wanted to prevent D from lighting up when A, B, and C were pressed. But this isn't a physical issue, it's an encoder programming issue.

Right?  

We have several electrical types and encoder developers here now,  please weigh in and lets bury this one or explain why I'm wrong?

Thanks!

--- saint

[1UP]

Don't forget the other problem--key blocking.  I'm assuming that here the opposite is happening -- so many keys are pressed with common matrixes at once, that the encoder doesn't even know that certain keys are being pressed.

IMO the real problem here is not technical, it's cheapness, which doesn't even make sense.  There are so many good and even cheap alternatives to keyboard hacking, I don't know why anyone would do it these days.  Hell, the basic keywiz is cheaper than the cheapest keyboards out there!  Two cheap gamepads are easier to hack, and just as cheap too.  Unless you've got an extra keyboard laying around just screaming "hack me", and a seriously sadistic side, I'd say you're being penny foolish and hour, uh, foolisher...

[Grasshopper]

Actually ghosting is an electrical issue, although it's not immediately obvious. It took me a while to get my head round this.

Let us suppose that current has to flow from Y2 to X2 in your diagram to register key D as being pressed:

* The current starts at Y2

* then flows through switch C (in the normal direction)

* then flows through switch A (backwards)

* then flows through switch B (in the normal direction)

* then the current finally reaches X2

Note that the current has to flow backwards through switch A and this is why diodes prevent ghosting.

It doesn't matter how you define the 'normal' direction. You could swap everything round i.e. say that the current flows the correct way through switch A and backwards through switches B & C. You then just have to connect the diodes in the opposite direction.

However as 1UP says the real problem these days is blocking and unfortunately that cannot be solved by diodes.

[RandyT]

At any rate, the point about keyboard hacks and diodes is moot for the most part due to improved programming in modern keyboards.

It's really only "improved" in the sense that under normal keyboard usage parameters (i.e. you are actually using it for typing) problems of the past, namely phantom keystrokes, are no longer.

But if you are interested in hacking one, they moved a big step backwards.

However, for the sake of argument, say you have an old keyboard that suffers from ghosting.

Why would diodes make one bit of difference?  Ghosting is not an electrical issue, it's a programming (encoder firmware) issue, right?

Nope .

Blocking is a programming issue.  And it was brought about as a solution to the "phantom keystroke" problem.  Some keyboard firmware writers even went an additional step and limited the total number of keystrokes that would register at any given time.  This was probably done to save money, as less keystrokes means less (expensive) RAM is needed.

In a small matrix of:

..   X1   X2
Y1  A      B
Y2  C      D

If A, B, and C are pressed then D is generated by the keyboard encoder because it sees pins X1, X2, Y1, and Y2 all come high at the same time. Right?

Sort of.  Matrices work a little differently than your standard, direct input Keyboard emulator hardware.  Basically, a row gets brought high and then the status of the column is checked.  If  high, then no key is pressed. If low, then a key is pressed.  This is just an example, as there a number of combinations that do the same thing.

The problem comes when any three positions, that would form a right triangle on the matrix if a line was conected between them, are brought low through switch closures.  This will electrically feed the signal back through the lines, making the microcontroller think that the point on the matrix where the aforementioned right-triangle becomes a rectangle (or square) was also activated.  In a keyboard hacking scenario, this is a good thing, because you can circumvent this problem with diodes.  There may still be a limit on the number you can use though, due to voltage drops caused by the diodes.  This would probably vary on a case-by-case basis.

The basic theory was putting diodes inline with the wiring would prevent "backfeed" of current from completing the circuit for D, correct?

Correct

But D isn't being generated because its circuit is being completed physically under the button. You could physically cut the traces at the point where they branch off to the "D" and it wouldn't make a difference.  The encoder still sees all 4 terminals high and generates the keystrokes.

But the electrical path is still complete through the other switches that are physically closed.  

The only way to prevent a ghost key would be a "diode" in the programming, no?  I.E. Keyboard blocking that was added to encoders.

Unfortunately, you can't put a "diode" in the programming, because the diode would allow everything to work .  All you can do is watch for the thing that might cause a problem (activation of that 3rd point in the matrix where a "right -triangle" would be formed) and prevent it.

A diode would make sense if for instance all the keys had lights on them, and we wanted to prevent D from lighting up when A, B, and C were pressed. But this isn't a physical issue, it's an encoder programming issue.

Since the electrical path is unknown to the microcontroller, it cannot solve the problem.  The real issue here is that, using your example, when 2 of the switches are closed, it knows which two.  When a 3rd one is closed, the microcontroller can't tell which one. It only knows that a 3rd one, which isn't one of the first two, has closed.  Not very useful information, so it is taught to ignore that 3rd switch closure.

This is probably pretty confusing, but I tried to keep it as simple as possible.  If it still seems unclear, then ask away and I'll try again.

RandyT

[paigeoliver]

There are still reasons to do keyboard hacks.

Granted they are no good for the standard 5 joysticks, 37 buttons, 4 coin mechs, 14" monitor machines that a lot of people like to make, but they are very good for some applications.

Basically there are 3 situations they are good for.

1. Extremely limited controls. A 4-way, zero button panel, or a space invaders panel would be good examples.

2. One button at a time applications. Jukebox, gambling machine, quiz machine, or most Daphne games.

3. Wiring up those coin mechs, pause, tab, and escape buttons after you already maxed out your i-pac or joystick hack. Or, for the start buttons and mechs on a spinner, trackball, or steering wheel machine.

I don't know about you guys, but I must own at least 20 keyboards. So the effective cost of a keyboard hack to me would be zero.

[Tiger-Heli]

Reply #1 (would you believe my HS English teacher liked how concisely I could word things???)

Hi Saint,

My page - http://www.mameworld.net/emuadvice/keyhack2.html covers this in detail under "Little Known Facts and Common Misconceptions"  See also http://www.dribin.org/dave/keyboard/ for more of the "physics" of the problem.

Bottom line, imho, is diodes will work just fine on an older keyboard that shows true ghosting, but not one that shows blocking.  Also, as RandyT said, some keys have an independent pressed key maximum.

Quick test - Run ghostkey or keyhook or one of the other utilities.  (or even Notepad, in a pinch) Press a bunch of keys -

If you see keys displayed which you did not press - diodes may help, and you need to see how many keys you can press at the same time (but you will see more keys than this actually registering).

If after three or four or five keys are pressed (probably different numbers based on different combinations) additional keys stop registering, your keyboard probably uses blocking, diodes won't help.  It may still be hackable as shown on my page, but you need to map out the matrix and THEN test for a pressed key maximum.

BTW, FWIW, I've mapped out the matrix for three different keyboards and tested a bunch more.  I haven't found any that showed true ghosting and not blocking, nor any with a pressed key maximum.  They're out there, I have no doubts, but I haven't run across them.

Don't forget the other problem-

There are three (four if you count the pressed key maximum) problems - Ghosting, masking, and blocking - Short explanations -

Ghosting occurs with matrix encoders without diodes when a "Ghost" key is generated b/c the encoder cannot accurately read the matrix.  It can be compensated for with diodes, and usually isn't a problem - if the encoder generates a "Q" but MAME isn't using it, who cares!

Masking occurs when a ghost key is generated and one of the keys is released, but the encoder doesn't recognize the key release.  It results in a stuck key and can be corrected by using diodes to prevent the ghosting from happening.

Blocking is a software/firmware issue which prevents the keystokes from registering and producing ghostkeys.  It will render diodes ineffective.

-key blocking.  I'm assuming that here the opposite is happening -- so many keys are pressed with common matrixes at once, that the encoder doesn't even know that certain keys are being pressed.

Not exactly.  The encoder would normally not care.  Let's say you don't have blocking.  You could press 40 keys and the encoder would generate all of them.  It knows they are being depressed.  It just would also generate 15 or 20 ghost keys that were intended.

What happened is about 1989, keyboard manufacturers were aware of the problem, and decided that most typists wouldn't press 15 keys at the same time, and it was better if the keyboard didn't register 3 simultaneous keypresses that could generate a ghost key so they "blocked" the third key (and any subsequent keys) from registering.

The "pressed key maximum" as RandyT said, was an attempt to limit the number of simultaneous keypresses allowed, again probably as a cost savings to allow less RAM requirements.

[Tiger-Heli]

Reply continued . . .

IMO the real problem here is not technical, it's cheapness, which doesn't even make sense.  There are so many good and even cheap alternatives to keyboard hacking, I don't know why anyone would do it these days.  Hell, the basic keywiz is cheaper than the cheapest keyboards out there!

Ok, let me play devil's advocate here.  First, I'm a big fan of the keywiz!  For $26.95, plus shipping, you get a much easier package to solder to, 32 non-ghosting inputs, redefinable keys, easy to use software, Shazaaam! key support, etc., etc.

But what do you mean "cheaper than the cheapest keyboards out there"  I can find literally dozens of them on pricewatch for $4-$6, shipping included.  Broken used ones at computer fairs/swap meets go for $1-$2, or free with another purchase.  And Best Buy, OfficeMax, Circuit City, etc., routinely offer them "Free after Rebate".

You might want to rethink that statement.  (I realize you still need to buy the bulk wire, and solder, etc, and your time is worth something, but a keyboard hack will still be cheaper, in a purely monetary sense).

Two cheap gamepads are easier to hack, and just as cheap too.

Huh?  Gamepads are usually $2-$3 even on sale, and you need two of them.  Double the cost.  Even assuming you get them free, there are other concerns, namely -

Most gameport gampads don't support 11-13 inputs like the USB ones.  Even if they do, most people don't have two DB15 gameports on their computer (newer mobo's won't even have one), so you have to go USB, which means -

Can't be used under DOS or Windows NT, and . . .

Have to hack two different brands of USB gamepads, b/c if you use the same brand, then tend to swap order when the cab is re-booted.

Apart from this, it's a different interface.  MAME doesn't particularly care, but some software will ONLY work with the keyboard, making you gamepad hack useless, while some will prefer gamepads, making it better than a keyboard hack.

Unless you've got an extra keyboard laying around just screaming "hack me", and a seriously sadistic side, I'd say you're being penny foolish and hour, uh, foolisher...

Nice wording.  But yes I agree overall.  See my initial disclaimers at the start of the EmuAdvice page.  To paraphrase LilWolf:

I've known a few people who've made keyboard hacks work.  Many of them ended up buying an encoder later and were very happy they did.  I never known anyone who bought an encoder and said "You know, I really wish I had hacked a keyboard."

[Howard_Casto]

I'll agree with the peanut gallery here.  Keyboard hacks aren't viable for most projects anymore.  The ONLY time I reccomend them to newbies is if they are skilled at soldering, low on cash, and making a "classic" cab that requires less than 5 inputs.    They are also good for media player hacks and jukebox hacks but as this is build your own arcade controls I'll skip over that bit.  

I'll also agree with tiger.  Keyboard encoders are actually the most expensive option (sorry randy, even yours), but they are well worth it.  With a keyobard hack you spend a great deal of time and even when you are done it's uncertain if it'll operate properly due to ghosting, blocking, usb limitations, ect.  Joysticks are actually your second best bet, but the fact that many games just don't support joysticks and they don't have "shift" keys make it slightly less desireable.  

And it's like lilwolf said, I've also  heard many people regret making keyboard hacks but I've yet to hear a complaint about buying a keyboard encoder.  

To finish off my ramblings......  I think keyboard hacks are a pain because you can't tell what you are dealing with until after the hack is done.  This isn't a good way to start off any project imo.  I hated them when they were our only option and quite frankly I'm not at all sorry that they are almost a thing of the past.

Hey I didn't disagree with anyone in this post.  Is that illegal or something?

[1UP]

Hey I didn't disagree with anyone in this post.  Is that illegal or something?

I'll disagree with ya, since I just finished my mini-cab entry and I feel a little rowdy!  

the fact that many games just don't support joysticks and they don't have "shift" keys make it slightly less desireable.

Once again, you don't need shift keys.  This functionality is built in to Mame's key definitions.  Instead of selecting just one key for each input in the Mame menu, you can select any key combination (p1 start+ p2 start = exit, etc).  I used to do this before I had an I-Pac.

Also, my favorite pad to hack, the Dual Strike, actually does have a built-in shift key.  And most of the game pads I've owned for the last 8 years have had keystroke emulation either in hardware or software.  And with modern USB pads, the 2+ player limitation is a thing of the past.

Right then, I'm off to play my tiny Pac-Man cab.  Cheers!  

[Tiger-Heli]

Well, just a few clarifications, but so far this thread has been a good summary of the problems.

Once again, you don't need shift keys.  This functionality is built in to Mame's key definitions.  Instead of selecting just one key for each input in the Mame menu, you can select any key combination (p1 start+ p2 start = exit, etc).  I used to do this before I had an I-Pac.

True, but only if you use MAME V36B12 or newer.  (Slower computers work better with older versions).  Also, a commercial keyboard encoder's shift function works with any program, while this only works in MAME.

And most of the game pads I've owned for the last 8 years have had keystroke emulation either in hardware or software.

None of the cheapo pads I've considered hacking have had this, except the dual strike.

 And with modern USB pads, the 2+ player limitation is a thing of the past.

The USB limitation HC mentioned was with keyboard hacks, and it IS known before you start the project.  USB keyboards are limited to six simultaneous keypresses plus modifiers (Shift, Alt, Ctrl) and thus are only suitable for very limited applications (like a Pac-Man CP, for example).

And USB pads have the limitation that they swap positions sometimes if they all use the same driver.

[eightbit]

I'll throw down here...  

My opinion is that a keyboard encoder is the way to go and Randy's inexpensive key wiz is an excellent encoder. I've never done a keyboard hack. I have many keyboards laying around. I have never felt that it would be worth the pain.

For every reason listed above to not hack a gamepad or keyboard someone can probably point out a particular brand or model that it will work on. I did a game pad hack I used a gameport version of the sidewinder for the following reasons-

You can get them for around $3.
You can daisy chain up to 4 of them in a row.
they don't swap players when you reboot.
You get 14 inputs per pad.
They use a common ground.
The circuit board is nicely labled and has test trace pads for easy soldering.

The cabinet that the hack is in gets played a lot and it works. Its only configured for mame.

[Chris]

What a lot of people making keyboard hacks don't take into account is barrier strips.  My first two keyboard hacks had the wires soldered directly to the controller board, but as the wires move when a control panel is moved or removed, these solder points get stressed and eventually break.  I'm finally getting ready to redo my keyboard hack with barrier strips.

By the time you add the cost of barrier strips, you are now approaching the cost of a KeyWiz. (Note that the KeyWiz was NOT yet available when I did my hack!)

So I am going to finish and use my hack and I would still use a hack for a simple classic cabinet with limited inputs, but if I could start over I would buy an encoder, and if I ever sell the cabinet I will definately put an encoder in, as I think keyboard hacks reduce the resale value.

--Chris

[Tiger-Heli]

By the time you add the cost of barrier strips, you are now approaching the cost of a KeyWiz. (Note that the KeyWiz was NOT yet available when I did my hack!)

Well, approaching, maybe -

The keyboard should have at most, an 18x8 matrix, so we're talking a 26 input terminal strip.  Radio Shack P/N 274-680 is a 12-pin TB for $2.49.  (Other sites are much less, but then you have shipping charges).  You might need three of these, so approx $7.50.  This is slightly more than the shipping on a KeyWiz.

The main question is "Is it worth $25-$35 additional for a solution which will allow Shift keys, double the number of inputs, any choice of inputs, re-programmability (through software), greater speed, better reliability, and no set-up/construction time on your part?"

[Chris]

The main question is "Is it worth $25-$35 additional for a solution which will allow Shift keys, double the number of inputs, any choice of inputs, re-programmability (through software), greater speed, better reliability, and no set-up/construction time on your part?"

Well, for me, setup and construction was the fun part...

That $25-$35 also has to be viewed in the context of being a mere fraction of the cost of a cabinet.  If you're already closing in or past the $1000 mark, what's an extra $25 to do it right?  Of course, that can be said about a lot of things, and those $25 chunks add up....

It comes down to both solutions having their place...

[Tiger-Heli]

Well, for me, setup and construction was the fun part...

That $25-$35 also has to be viewed in the context of being a mere fraction of the cost of a cabinet.  If you're already closing in or past the $1000 mark, what's an extra $25 to do it right?  Of course, that can be said about a lot of things, and those $25 chunks add up....

It comes down to both solutions having their place...

Agreed, for me the $25-$35 would be well worth it for the advantages gained, but some people want to spend as little as possible, and the IP was asking about keyboard hacks.

[Howard_Casto]

The main question is "Is it worth $25-$35 additional for a solution which will allow Shift keys, double the number of inputs, any choice of inputs, re-programmability (through software), greater speed, better reliability, and no set-up/construction time on your part?"

Well, for me, setup and construction was the fun part...

Anyone who considers hacking a keyboard the "fun" part is just plain sick.  You should seek professional help man.  

[tmasman]

Anyone who considers hacking a keyboard the "fun" part is just plain sick.  You should seek professional help man.  

Are you offering your "professional" services?
where's the couch? I apparently need help...

Then again, I haven't heard the voices in a while... Maybe I'm getting better...
<mumbles to self...>
uh oh... there back...