21 Jan
2012
21 Jan
'12
3:40 p.m.
With this discussion of interfacing PC keybaords to vintage machines,
I've been thinking about a project I've had in mind for some time now..
Firstly soem backgroud. The HP9915 is essentially an HP85 in a
half-rack-width metal case. It could be run on a bench, or there was an
HP mounting kit top put it in an instruemtn rack.
Electornically, it's similar to the HP85. There is a buit-in tape drive.
Ther is no printer. There is no built-in monitor, but there's BNC socket
o nthe back that outputs TV-rate composite video (so getting a display is
no problem). The keyboard controled IC is present, but obviously there's
no built-in keyboard. There are some user-definable buttons on the front
which are wired as the programamble function keys on an HP85.
The keybaord row and commn lines are also buffered to TTL levels and
broufht out on a DB25 socket on the rear of the machine. There was an HP
keyboard that conencted here, this is basically just a matrix of switches
wired to a DB25 plug. It's the same matrix as the HP85 keyboard, and that
one is docuemtned in the service manual (although with at least one
typo!). However, the keyboards are much harder to find than the machines.
It's a 10*8 matrix with separate lines for the shift, control and
capslock keys (these lines are simply grounded by the the appropraite key)
My first question in designing this is :
Is there ever a time when you press 2 keys simultaneously on an HP85
(other than shift or control along with another key). In onter words, is
there ever a time when 2 locations of the keyboard matrix are closed at
the same time?
The reason I ask is that I can't think of one, and if such a case doesn't
have to be considered there's a very simnple interface (a mutlipezer to
sleect a scan line drivign the enable input of a demultiplexer to select
an input line) that could be used. It'd be 2 or 3 cheap TTL ICs.
I'd not want to connect a microcontroller directly to an external
connector in any case (risk of damaging it, I notice that HP had much the
same idea about the keyboard controller IC in the 9915, all the signals
brought out on the connector are buffered). Thing is, it's a lot easier
to replace a non-programmed TTL IC than having to program a repalcement
microcontroller. So I'd either want to add buffers even if I was goign to
try to use the microcontroller direcrly (which for reasons we've already
discussed I think it 'pushign it'). The mux/demux ICs wouldn't need any
more buffering IMHO, they could be simply wired to the connector.
Of course if I do need to be able to have multiple 'keys pressed' at the
same time, then I would haev to use the crosspoint switch or dual port
RAM solutions we've already discussed.
-tony
21 Jan
21 Jan
10:57 p.m.
Tony wrote:
Is there ever a time when you press 2 keys
simultaneously on an HP85
(other than shift or control along with another key).
Yes. n-key rollover.
Image you are typing "hello". A very large proportion of the time
a typist will still be pressing 'e' by the time their finger starts
pressing 'l'. 2-key rollover allows for two such keypresses, as is
the maximum that most keyboard input drivers deal with.
Also, it's electrically not possible to do 4-key rollover with a
matrix keyboard, as as soon as three keys are pressed the four
key on the other corner of the "square" formed by the three
keys being pressed is seen as being pressed. There is nothing
software can do about this, it's an electrical function of
matrix keyboard.
For those who don't know how to suck eggs, imagine the following
matrix keyboard:
| |
---A----B----
| |
---C----D----2
| |
1
If A, B and C are pressed, if the software looks to see if D is
pressed (by looking at lines 1 and 2) the electricity can flow
through C, A and B, making it appear that D is pressed.
The reason I ask is that I can't think of one, and
if such a case
doesn't
have to be considered there's a very simnple interface (a mutlipezer
to
sleect a scan line drivign the enable input of a demultiplexer to
select
an input line) that could be used. It'd be 2 or 3 cheap TTL ICs.
A simple n-to-2^b multiplexer to select a column and a 2^n-to-n
demultiplexer to sample the rows is sufficient on the hardware
side. It's better to have the modifier keys (Shift, Ctrl, etc)
as a seperate row so they can be sampled seperately from the
other keys.
This is how the BBC does it:
http://mdfs.net/Info/Comp/BBC/Circuits/bbckbd.gif with a
4-to-16 output multiplexer to select the rows and each column
samplable with a 8-to-3 demultiplexer.
--
J.G.Harston - jgh at mdfs.net - mdfs.net/jgh
22 Jan
22 Jan
12:27 p.m.
On 22 Jan 2012 at 3:57, J.G.Harston wrote:
Also, it's electrically not possible to do 4-key
rollover with a
matrix keyboard, as as soon as three keys are pressed the four
key on the other corner of the "square" formed by the three
keys being pressed is seen as being pressed. There is nothing
software can do about this, it's an electrical function of
matrix keyboard.
Not strictly true! If one uses a *good* keyboard with an isolation
diode at each key, n-key rollover (where n is greater than 2) is the
result. For a scanned music keyboard, this is a necessity if chords
are to be supported. But many PC keyboards implement this also. So,
unless a non-n-key-rollover keyboard is laid out carefully,
"ghosting" can get to be very annoying for a fast touch typist who
pays more attention to the "make" of a key than the "break" (try on
an electric typewriter for example).
I'm not aware of any crosspoint switch ICs that have a diode at each
junction, but perhaps they exist.
--Chuck
2:08 p.m.
Chuck Guzis wrote:
But many PC keyboards implement [n key rollover] also.
I wouldn't say "many", as it was always a small fraction. The vast
majority did not have diodes or any other technique to prevent ghosting,
and only offered two key rollover, which is generally satisfactory for
most typists. It was common for some of the "special" keys (shift,
control, alt) to not be part of the scanned matrix, so you could detect
any number of those, but not reliably detect more than two "normal" keys
at once.
The USB HID spec stupidly limits USB keyboards to reporting a maximum of
six key depressions. I have not seen *any* USB keyboards with diodes or
other antighosting. Maybe some of the high-end "gaming keyboards" do.
Eric
2:32 p.m.
On 22 Jan 2012 at 11:08, Eric Smith wrote:
Chuck Guzis wrote:
The sentence before the one quoted said "good keyboards". "Many", I
suppose is subjective--"n-key-rollover" was touted more on older
systems than on $10 keyboards offered today. Keytronics used to
advertise it heavily on their high-end keyboards. The Toshiba
luggable laptops (3000 and 5000 series) all featured n-key rollover.
Perhaps, given the large number of keys on a modern keyboard (other
than laptops), careful distribution of keys minimizes the effect of
ghosting.
Six-key rollover seems to be a good compromise; I suspect the most
anyone sees in the real world is 3 or 4 key.
Or perhaps today's crop of keyboardists never learned to type on a
typewriter.
--Chuck
But many PC keyboards implement [n key rollover]
also.
I wouldn't say "many", as it was always a small fraction. The
vast
majority did not have diodes or any other technique to prevent
ghosting, and only offered two key rollover, which is generally
satisfactory for most typists. It was common for some of the
"special" keys (shift, control, alt) to not be part of the scanned
matrix, so you could detect any number of those, but not reliably
detect more than two "normal" keys at once. 
3:07 p.m.
Six-key rollover seems to be a good compromise; I
suspect the most
anyone sees in the real world is 3 or 4 key.
For ordinary typing, ie, using a keyboard as a text input device, I
suspect you are correct.
But when using a keyboard as a multi-button input device that happens
to have lots of buttons, which (for example) many games do, it's far
less clear. It would not surprise me to see rollover requirements go
as high as nine or maybe even ten in extreme cases; certainly when
playing the piano I need N-key rollover for N above 6 - fortunately my
piano, unlike most computer text keyboards, is designed for that sort
of thing. (It's an electronic piano, not a mechanical one, but a
pretty good one as such things went when I bought it; it definitely
supports N-key rollover for N well over 6.)
/~\ The ASCII Mouse
\ / Ribbon Campaign
X Against HTML mouse at rodents-montreal.org
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B
5:57 p.m.
On 22 Jan 2012 at 15:07, Mouse wrote:
But when using a keyboard as a multi-button input
device that happens
to have lots of buttons, which (for example) many games do, it's far
less clear. It would not surprise me to see rollover requirements go
as high as nine or maybe even ten in extreme cases; certainly when
playing the piano I need N-key rollover for N above 6 - fortunately my
piano, unlike most computer text keyboards, is designed for that sort
of thing. (It's an electronic piano, not a mechanical one, but a
pretty good one as such things went when I bought it; it definitely
supports N-key rollover for N well over 6.)
I believe I mentioned music keyboards as an exception, as in the
ancient Casio WK-1200 sitting at my right elbow here, which could
probably render the Tom Turpin style of playing (i.e. sit on the
keyboard, or so it was reported) just fine.
--Chuck
4:53 p.m.
Chuck Guzis wrote:
Six-key rollover seems to be a good compromise;
While the USB HID spec allows for six keys, I've never seen a USB
keyboard that reliably did more than general (not "special" key) two key
rollover.
There are lots of anecdotes about recent PC keyboards (USB and non-USB)
supporting better than two key rollover, but most of those, on
investigation, don't hold up. Often it's just a report that some
specific three or four key combinations work OK, which proves nothing
about the general case. It's not N key rollover unless it works for
*any* N keys.
There are definitely some high-end, very expensive PS/2 interface
keyboards that have N key rollover, but not any for USB due to the
stupidity of the USB HID spec.
4:57 p.m.
Chuck Guzis wrote:
Or perhaps today's crop of keyboardists never
learned to type on a
typewriter.
He he. I learned to type aged 8 on a hugely-heavy "International"
typewriter :)
--
J.G.Harston - jgh at mdfs.net - mdfs.net/jgh
7:41 p.m.
New subject: Typewriters (was: Thinking of keyboard interfaces)
On Jan 22, 2012, at 2:32 PM, Chuck Guzis wrote:
Six-key rollover seems to be a good compromise; I
suspect the most
anyone sees in the real world is 3 or 4 key.
Or perhaps today's crop of keyboardists never learned to type on a
typewriter.
I certainly never did. I'm looking into getting a typewriter somewhat soon, both for
teaching myself to type a little better (more accurately, and separating the keystrokes
better) and also because typewritten recipe cards look so much better than printed ones or
my own child-like scrawl. Smith-Corona's Silent and Silent Super seem to be general
favorites among the portables; anyone have other recommendations? I seem to be able to
find properly working instances of the above for less than $100 on eBay/Etsy.
- Dave
9:16 p.m.
New subject: Typewriters (was: Thinking of keyboard interfaces)
On 22 Jan 2012 at 19:41, David Riley wrote:
I certainly never did. I'm looking into getting a
typewriter somewhat
soon, both for teaching myself to type a little better (more
accurately, and separating the keystrokes better) and also because
typewritten recipe cards look so much better than printed ones or my
own child-like scrawl. Smith-Corona's Silent and Silent Super seem to
be general favorites among the portables; anyone have other
recommendations? I seem to be able to find properly working instances
of the above for less than $100 on eBay/Etsy.
I've had two Smith-Corona portables in my life--one, a manual and the
other, an electric (they were also sold under the Sears name). The
electric was okay, but I really got spoiled later by the IBM Model B
Executive. A really wonderful machine--and if you get the prop-
spacing type, the quality of the output is remarkable. Selectrics
aren't too bad either.
One thing that an electric forces you into is a rhythm as your speed
increases. You can only push the mechanism so far until it starts to
fight you until you feel its natural rhythm.
It's odd, but when I started using a computer terminal, my speed went
down. Perhaps it was due to my looking at the letters on the screen
as they appeared (a good typist types "blind"); I don't know. When
you use a keypunch, the "chunk" is your source of feedback, so it's a
good thing they're noisy--and you type blind. If you use one of the
nonprinting punches, such as an 024, you don't even know what you
typed until you run the deck through an interpreter or do an 80-80
listing.
If you want a somewhat different experience (and this relates to
"chording" and "n-key-rollover", try a stenotype machine.
--Chuck
11:52 p.m.
New subject: Typewriters (was: Thinking of keyboard interfaces)
-----Original Message-----
From: fraveydank at gmail.com
Sent: Sun, 22 Jan 2012 19:41:27 -0500
To: cctalk at classiccmp.org
Subject: Typewriters (was: Thinking of keyboard interfaces)
On Jan 22, 2012, at 2:32 PM, Chuck Guzis wrote:
I'm a fan of the IBM Wheelwriter, myself. They're great machines, and they
even show up at thrift stores! They do proportional type like an Exec,
without the weight, and you can change typefaces like a Selectric.
The olympia electronic machines are nice also, but not at all common
in this country these days (I did spot one in a thrift last year, though).
____________________________________________________________
Send your photos by email in seconds...
TRY FREE IM TOOLPACK at http://www.imtoolpack.com/default.aspx?rc=if3
Works in all emails, instant messengers, blogs, forums and social networks.
Six-key rollover seems to be a good compromise; I
suspect the most
anyone sees in the real world is 3 or 4 key.
Or perhaps today's crop of keyboardists never learned to type on a
typewriter.
I certainly never did. I'm looking into getting a typewriter somewhat
soon, both for teaching myself to type a little better (more accurately,
and separating the keystrokes better) and also because typewritten recipe
cards look so much better than printed ones or my own child-like scrawl.
Smith-Corona's Silent and Silent Super seem to be general favorites among
the portables; anyone have other recommendations? I seem to be able to
find properly working instances of the above for less than $100 on
eBay/Etsy.
- Dave
23 Jan
23 Jan
12:06 a.m.
New subject: Typewriters (was: Thinking of keyboard interfaces)
On 22 Jan 2012 at 20:52, N0body H0me wrote:
I'm a fan of the IBM Wheelwriter, myself.
They're great machines, and
they even show up at thrift stores! They do proportional type like an
Exec, without the weight, and you can change typefaces like a
Selectric.
The olympia electronic machines are nice also, but not at all common
in this country these days (I did spot one in a thrift last year,
though).
Can't stand the daisy-wheel ones. You hit a key and some variable
time later, something happens. It's a complete sensory disconnect.
Somewhere, I've got a Brother daisywheel typewriter like that also.
There's no sensory immediacy to it.
Maybe I'm too old since I learned typing on a full-size Underwood
manual with no labeling on the keys. "Okay class, hands home, A, S,
D, F, J, K, L, sem, A, S, D, F, J, K, L, sem..."
--Chuck
22 Jan
22 Jan
5:44 p.m.
On Sun, Jan 22, 2012 at 11:08 AM, Eric Smith <eric at brouhaha.com> wrote:
The USB HID spec stupidly limits USB keyboards to reporting a maximum of six
key depressions. I have not seen *any* USB keyboards with diodes or other
antighosting. ?Maybe some of the high-end "gaming keyboards" do.
Are you referring specifically to the operation of a USB HID keyboard
in boot protocol mode? While it is true that a USB HID keyboard input
report while in Boot Protocol mode contains one byte for modifier keys
and six bytes for keycodes, that restriction does not apply when the
keyboard is not in boot protocol mode when the OS is running and
talking to the keyboard instead of the BIOS. Guess I'm missing the
supidity to which you are referring to here.
-Glen
6:17 p.m.
Glen Slick wrote:
Are you referring specifically to the operation of a
USB HID keyboard
in boot protocol mode? While it is true that a USB HID keyboard input
report while in Boot Protocol mode contains one byte for modifier keys
and six bytes for keycodes, that restriction does not apply when the
keyboard is not in boot protocol mode
Even in normal operation (not Boot Protocol),
HID still imposes some
significant limitations. Low-speed devices cannot have a report size
larger than eight bytes, which potentially means that larger numbers of
input changes will require the reporting for some to be delayed to a
later USB frame. For normal keyboard usage this wouldn't be a problem,
but it does mean that true N key rollover will require delays, or for
the keyboard to operate at full speed or faster, rather than low speed.
I have not found any USB keyboards that operate at full speed or faster,
though perhaps some "gaming keyboards" may. The only USB keyboards that
I've examined which have a higher-speed interface are those with
internal USB hubs, and in this case the internal keyboard-to-hub
interface was low-speed.
However, even for the gaming keyboard, the manufacturers' tech support
people have stated that the keyboards to not have N key rollover, and
have blamed this on the six key limit of boot protocol. What this
really means is that their implementation can's support more keys, and
that they don't want to put in diodes or other antighosting, nor support
for full-speed or faster. Most of this is undoubtedly for cost reasons,
though better firmware would presumably only increase NRE and not
production cost.
7:38 p.m.
On Jan 22, 2012, at 2:08 PM, Eric Smith wrote:
Chuck Guzis wrote:
Das Keyboard, interestingly enough, claims to support n-key rollover when run in PS/2 mode
and 6-key rollover in USB mode. Presumably it has the proper anti-ghosting hardware,
though I do not own one and have never looked inside one.
Fun question: I have an Apple Extended Keyboard II running through an iMate (ADB to USB
converter) to my machine. It's great, except for the word (or syllable, as it's
frequently found in words) "out". If I type it exceptionally fast, I get a
ghost on the semicolon (e.g. "ou;" or "ou;t", sometimes it sees the T
and sometimes not).
I'm assuming this is a ghosting issue in the keyboard itself, but I don't recall
it from when I used it natively on an ADB machine. Of course, it's been years since
I've used it natively on an ADB machine (I use crappier keyboards on them these days),
so it's possible I'm just a faster typist now. Is anyone else familiar with this
problem? Is it just the iMate? I would be surprised.
- Dave
But many PC keyboards implement [n key rollover]
also.
The USB HID spec stupidly limits USB keyboards to reporting a maximum of six
key depressions. I have not seen *any* USB keyboards with diodes or other antighosting.
Maybe some of the high-end "gaming keyboards" do.
3:05 p.m.
I'm not aware of any crosspoint switch ICs that
have a diode at each
junction, but perhaps they exist.
I've not seen oen either.
For digital signals (which keyhorad scan lines tend to be), and if you
know which are the driven lines and which are the inputs (pretty much
esenetial if you're goign to use the diodes anyway), then a digital
cicuit solution is possible. Store the simulated key states in D-type
flip-flops. The output of each D-type feeds on input of an AND gate, the
other input is driven by the appropriate scan line, the outptus of all
the AND gates sorresponding to a given input line from the keyboard are
ORed together and fed to that input. (OK, you may need to invert the scan
lines and inputs, but I mean 'logically' ANDing and ORing here).
This works with machines that allow multiple scan lines to be driven
simultaneously, like the TRS-80 M1/3/4. The downside is that htis logic
is really only easy to implement usign an FPGA or simiular. Building it
from TTL takes a lot of packages.
-tony
2:56 p.m.
Tony wrote:
No, that';s no what I meant.
If you are emulating the keyboard by a circuit that can only effectively
'press' one key at once, it won't affect this. WEhen you press the 'E'
key, the interface simulates the appopriate key contact closure. When you
press the 'L' key (with the 'E' key still down), the interface can then
simualate the 'L' key contact closure. Of course at this point it can no
longer have the 'E' contact closes from the point of view of the host
machine, but that don't matter. The host sees an 'E' keypress followed by
a 'L' keypress, which is what you want.
Is there ever a time when you press 2 keys
simultaneously on an HP85
(other than shift or control along with another key).
Yes. n-key rollover.
Image you are typing "hello". A very large proportion of the time
a typist will still be pressing 'e' by the time their finger starts
pressing 'l'. 2-key rollover allows for two such keypresses, as is
the maximum that most keyboard input drivers deal with.
Also, it's electrically not possible to do 4-key rollover with a
matrix keyboard, as as soon as three keys are pressed the four
For a simple matrix of swithces, sure. But if you add a diode in series
with each key you can have full n-key rollover (this is one reason why
such diodes are present on some keyboards (like the HP46020 IIRC). Such
diodes eliminate the 'sneak paths', if you close 3 key contacts, there is
no electrical path when the one at the 4th corner of the rectangle is
scanned (one diode is reverse-biased).
A related problem affects passive matrix LCDs. The LCD pixel/segment is
electrically a capacitor, and thus unpolarised -- in fact you energise an
LD with an AC signal to prevent electrolysis of the liquid crystal
material. For this reason, if you have the pixels in an electrical
matrix, you will get sneak paths when you turn on one particualr pixel.
This is one reason why the rivers for such displays are not simple, and
why there are mutliple drive voltages fed ot the display segments. It
also is the explanation fro a rule-of-thumb noticeed by an HPCC member --
'The cheaper a calculator, the clearer the display'. Point being that a
cheap clacualtor probably has a direct-drive 7-segment dispaly (no
multiplexing), a high-end graphing calculator is probably multiplexed
with 32 backplanes or soemthing. The former will have better contrast
than the latter.
A simple n-to-2^b multiplexer to select a column and a
2^n-to-n
demultiplexer to sample the rows is sufficient on the hardware
side. It's better to have the modifier keys (Shift, Ctrl, etc)
as a seperate row so they can be sampled seperately from the
other keys.
I am doing the reverse. I want to simulare the matrix keyboard, not scan one.
On the HP9915, as I mentioend,the shift, control, and caps-lock keys are
not part of the matrix. They are on their own. One side of each of those
is groudned, the other sides goes to a dedicateds input line (buffered
and then fed to the keybaord interface IC). All other keys are in an 10*8
matrix.
-tony
23 Jan
23 Jan
4:31 p.m.
As you demonstrated there's also the possibility of pressing two keys when
one only meant to press one:
On Sun, Jan 22, 2012 at 07:56:27PM +0000, Tony Duell wrote:
No, that';s no what I meant.
^---- ' and ; together
WEhen you press the 'E' key, the interface
simulates the appopriate key
^---- W and E together
--
Kevin
http://www.RawFedDogs.net
http://www.WacoAgilityGroup.org
Bruceville, TX
What's the definition of a legacy system? One that works!
Errare humanum est, ignoscere caninum.
4565
days inactive
4567
days old
18 comments
9 participants
participants (9)
-
ard＠p850ug1.demon.co.uk -
cclist＠sydex.com -
eric＠brouhaha.com -
fraveydank＠gmail.com -
glen.slick＠gmail.com -
jgh＠mdfs.net -
Kevin＠RawFedDogs.net -
mouse＠Rodents-Montreal.ORG -
n0body.h0me＠inbox.com