PDA

View Full Version : keyboard weirdness ps/2 not working as AT



carangil
October 11th, 2009, 07:18 PM
I picked up a ps/2 HP keyboard for $2 at Weirdstuff. It's not scratched up, nice and heavy, and, the best part, no Windoz key! (This keyboard is for a 486 I'm putting together.) (Also, while I was there, I picked up a new-in-box AT case. SO when I'm done, it'll be a brand-new old computer!)

I put it through a ps/2 to AT connector adapter, and it doesn't work. I get 'keyboard error.'

The keyboard tests fine on my daily desktop system!

Seems the adapter is bad?

Well, no: I have a cheap roll-up USB keyboard. (http://www.thinkgeek.com/computing/input/5a7f/) When I connect it though a USB to PS/2 adaptor, and then PS/2 to AT, it works.

Works:
(USB Keyboard) -> (passive usb to ps/2) -> (ps/2 to AT)->486

Does not work:
PS/2 Keyboard -> (ps/2 to AT) -> 486

:-(

None of my other USB keyboards work either. I know a lot of USB keyboard don't support ps/2 passive adapters anymore, I just think it's funny the cheap squishy keyboard the only one that works!

So... What's the deal here? I thought PS/2 and AT were compatible, and have a physical, passive wiring change? Any ideas?

Chuck(G)
October 11th, 2009, 07:27 PM
The timings are not identical between AT and PS/2 keyboards. Sometimes PS/2 keyboards will not work on older equipment.

I ran across this in the PS/2 to XT converter project. All my PS/2 keyboards worked fine, but I had one or two AT (Keytronics mostly) that refused to. It took some tinkering with the timings to get things right.

cgrape2
October 11th, 2009, 07:30 PM
Your new keyboard doesn't have a small switch on the underside does it?
I had a similar experience and the problem was a switch that needed to be changed.
cgrape2

carangil
October 12th, 2009, 09:27 AM
No switch. I sorta remember some keyboards having a switch. It's for AT vs XT.

I've never seen an AT/PS2 switch... seems unnecessary.

Maybe my best bet for this old machine is either trying a few more ps/2 keyboards, or find an AT one somewhere.

kishy
October 12th, 2009, 07:48 PM
No switch. I sorta remember some keyboards having a switch. It's for AT vs XT.

I've never seen an AT/PS2 switch... seems unnecessary.

Maybe my best bet for this old machine is either trying a few more ps/2 keyboards, or find an AT one somewhere.

PS/2 and AT have identical protocols and identical timings...everything is 100% identical about them except for the physical connector. It is indeed XT which has (the same protocol pinout but) different timing and that's what those switches are for.

(made a mistake there, WHY is there no strikethrough text here? grr)

The following are possible causes for the issue you're having, but how to correct them, I do not know:
-keyboard rejects request to change scancode set to what computer wants
-keyboard reports incorrect keyboard ID to controller

Chuck(G)
October 12th, 2009, 08:32 PM
PS/2 and AT have identical protocols and identical timings...everything is 100% identical about them except for the physical connector. It is indeed XT which has (the same protocol but) different timing and that's what those switches are for.

That might be true for IBM-branded keyboards, but it's definitely not true for non-IBMs. Generally, I've found that third-party keyboards (e.g. HP, Honeywell, Cherry) vary widely in their clock frequency and if they change data from the host on the rising or falling edge of the clock.

What's more many "AT-style" keyboards have nonstandard reset sequences. For example, a Multitech will send a string of "ACKs" with wrong parity, waiting for the host to send a reset command back. (Has something to do with autoswitching between XT and AT).

XT protocol is very different from AT. It's slower, unidirectional and uses different scan codes. AT sends two (or three, depending on the key) bytes on key release, where the XT only sends one. AT protocol is bidirectional (requests to retransmit, play with the LEDs, set typomatic rate, reset, etc.).

FWIW, I have a Faraday 286 board that will talk to an AT keyboard just fine, but refuses to hold a conversation with a PS/2.

carangil
October 12th, 2009, 09:04 PM
Well, I guess its time to look for an AT keyboard... or an IBM PS/2.

I hear the model M is good...

kishy
October 12th, 2009, 09:19 PM
That might be true for IBM-branded keyboards, but it's definitely not true for non-IBMs. Generally, I've found that third-party keyboards (e.g. HP, Honeywell, Cherry) vary widely in their clock frequency and if they change data from the host on the rising or falling edge of the clock.

What's more many "AT-style" keyboards have nonstandard reset sequences. For example, a Multitech will send a string of "ACKs" with wrong parity, waiting for the host to send a reset command back. (Has something to do with autoswitching between XT and AT).

XT protocol is very different from AT. It's slower, unidirectional and uses different scan codes. AT sends two (or three, depending on the key) bytes on key release, where the XT only sends one. AT protocol is bidirectional (requests to retransmit, play with the LEDs, set typomatic rate, reset, etc.).

FWIW, I have a Faraday 286 board that will talk to an AT keyboard just fine, but refuses to hold a conversation with a PS/2.

Absolutely right on AT/XT, my apologies, I meant same pinout with different protocol rather than same protocol with different clock (the costs of multitasking when posting...)

Of course when I say AT and PS/2 I'm talking about 100% compliant (that is, IBM) hardware...what noncompliant hardware does is, well, noncompliant. For what it's worth I've had none of the issues you've described but I can certainly see how they would happen in only borderline-compatible hardware.

My only true IBM hardware are PS/2 systems though, so I can only speak for using AT-type boards on PS/2, not the other way around, as far as an actual IBM system goes. As for plugging PS/2 into AT, all my non-IBM hardware gets along fine.


Well, I guess its time to look for an AT keyboard... or an IBM PS/2.

I hear the model M is good...

The Model M is godly.

Chuck(G)
October 13th, 2009, 09:18 AM
I hear the model M is good...

I'm typing on one now. I've got them scattered all around my most commonly-used systems. Yes, they're noisy, but they're heaven for a touch-typist.

...and they're darned near indestructible.