Karl Pfleger's Armrest Mounted Split Keyboard Setup

Normal angle keyboard only
Skip down to the full set of enlargeable photos

Intro

This page describes my $150 homemade ergonomic setup. (Note: I'm not selling anything, just describing.) I got fancy keyboards after I got first aches of RSI (Summer 2000). My setup is a less mainstream than just buying a fancy keyboard like the Kinesis or Goldtouch, but is cheaper than many fancy keyboards and in many ways better.

Update (Nov 7, 2003): Thanks to Chris who just e-mailed me saying: "I thought you would be interested in knowing that I recently bought a brand new 2003 usb model of the ergoflex keyboard and it still has the key swapping problem you describe. Namely, keys on the left side dominate over the keys on the right side, even when pressed a fraction of a second after the keys on the right side." (May, 2004) The engineers from Comfort say this is puzzling since the problem was fixed by the date Chris describes. Perhaps it was a randomly bad keyboard.

Update (Oct 2002): Thanks to Darryl who reports that the new keyboard he got from Comfort recently (an Ergo Magic, essentially the same thing as the Ergo Flex I describe here) does not have the problem my earlier Ergo Flex keyboard demonstrated. This seems like good confirmation of the e-mail I got from someone at Comfort saying they think they fixed the problem. Perhaps it is worth going back to this setup now.

Update (June 2002): I wrote this web page 2 years ago to describe my homemade keyboard setup. As I describe below, I stopped using it after half a year because the fully-split keyboard I used did not function properly when typing at high speed. I have not gone back to this setup. I have been living with a regular keyboard at home and the Kinesis contoured keyboard I got for the office, which I really like. Two years ago there were only two fully-split keyboards, the one I tried (the ErgoFlex version of the Comfort keyboard, for $150) and another sold by Interfaces by Cramer, who don't appear to sell keyboards anymore. Kinesis now sells something (which looks like the same one sold by Interfaces by Cramer) as the Kinesis Evolution (listed in the TI-FAQ). It's more money ($299 without chair) than the ErgoFlex but might work better. I haven't seen one in person to try it. Anyway, that's all I know at the moment. Good luck. And if anyone has one of these keyboards, test it with the test I describe below and let me know if it passes and also let me know how you like it in general and what your setup is.

Update (Dec 2000): I've stopped using the keyboard due to the problems described below and just use the laptop keyboard on the keyboard tray at the moment. Comfort's customer support has been very spotty in their responsiveness over the last few months. I'll send an e-mail, get a response, send an e-mail, and then not get a response for weeks. Then I'll start the process over again. I seem to have fallen off of their queue 2-3 times already. I feel like every 2nd or 3rd e-mail goes into a black hole. They haven't sent any replacement numerical keypad section as they suggested some time ago that they were going to do. They said it would take some time to get a new one from their plant, but have given me no indication that this is actually happening, nor any indication of what they think the problem is or what they have (would have?) changed in the replacement part they are or said they might send. Hopefully I can get these problems resolved or find another suitable keyboard for all this sometime before my symptoms come back. I still think the idea is great. If someone would only start selling a keyboard designed for this kind of thing (with or without the mounting hardware) for under $500 and it worked well, that'd be great.

My problem

Here I'll describe my particular typing related problem, but note that this setup might be useful in helping to aleviate or prevent many different problems since it encourages much better ergonomically neutral positioning and is extremely comfortable. I had already switched from right handed to left handed mouse use a number of years ago, and then subsequently even switched to the Dvorak keyboard layout. My new problem, exclusively in the right hand, was obviously extreme ulnar deviation.

I touch-type and pretty much hit all the keys with the "appropriate" finger. I also program and use Unix command lines a lot. This means the right pinky gets overworked, being responsible for return, backspace, right-shift, and up arrow (uesd a lot for Unix command line history). It also gets used a lot for braces and brackets (for programming and regular expressions), and pipes and forward and back slashes (for Unix command lines, LaTeX commands, path separators in Unix and Windows, etc.). Plus, in the Dvorak key mapping, the pinky is responsible for the more common letters L and S as the home and above home row keys instead of P and :/;. So for example, the very common Unix command "ls -l" is, besides one space, 5 pinky keys! (Including the return.)

The big problem is that many of these keys, especially return, backspace, and up arrow, are located very far to the right. So while a normal keyboard encourages some ulnar deviation (where the hand is bent at the wrist away from the thumb and towards the pinky) due to the fact that the keyboard isn't as wide as the distance between the elbows, moving to these extreme rightmost keys without moving the arm but just by bending at the wrist increases the ulnar deviation considerably. But when typing quickly or wresting palms, the hand often is the only thing to move. I could tell that this was the cause of my problem because as discomfort would start, I would soon catch myself using my right pinky to up arrow or backspace 3-4 times (or both in a row) and feeling a particular burn afterwards.

My solution

Many keyboards solve ulnar deviation reasonably well, including the very common fixed-split keyboards, such as the MS Natural Keyboard. However, I strongly dislike these keyboards, not because I can't type on them, but because I find the typing posture of bowing my elbows out like a chicken uncomfortable, and I don't like the fact that ulnar deviation is the only problem they solve while they exacerbate other problems. For instance, they are generally thicker, increasing the problem of high desk height. Tenting keyboards have the additional advantage that they turn the wrist closer to the more natural "gun-slinger" angle where palms face each other instead of the floor, but this creates even more height between the surface the keyboard rests on and the average height of the hands. Keyboard drawers can solve the height problem for relatively flat keyboards, but the drawers can only be lowered so much and still allow room for one's legs. The solution is to do away with having a horizontal surface to rest the keyboard on, and to separate the two halves of the keyboard.

I had seen and even briefly tried the Comfort Keyboard System, at NASA Ames Research Center in Mountain View, CA when I spent a summer working there during 1998. This company also sells the exact same keyboard without the base and tilting mechanism for almost half the price under the name ErgoFlex. I mounted the two main halves of this keyboard to the armrests of my desk chair, which is technically an old rolling, reclining kitchen chair. This was the only keyboard I knew of when I bought it that was sold with the explicit intention of separating the two main halves completely. The company even mentions that they began selling the keyboard this way after getting requests from people who wanted to mount the two halves to their chairs just as I have done. Still, there are no instructions or hardware included to help with such a mounting, and I'm not particularly into building things with tools, so I just did the simplest thing I could think of doing. I'd love to hear or see how other people have mounted this or any other keyboard to their own chairs. If you've done this or know of a web page that describes someone else who has, please drop me a note and let me know about it.

I recently found out about the keyboard made by Interfaces by Cramer, which is specifically designed to mount on many standard office chairs. It comes with mounting brackets, but it is more than 3 times the price of the ErgoFlex. They also sell the only chair plus keyboard combos on the market that I know about, but these cost $700-$1000. (Well, there was also an old Slashdot article about a nealy $6000 ultimate ergonomic workstation, but this doesn't really count.) Compared with the Interfaces by Cramer keyboard, the main two halves of the ErgoFlex are much smaller since the numeric keypad section is completely separate from both. This is good if you don't need to use that section a lot, and if you do you could probably figure out a simple way to mount it so that it sits right in front of you, projected from the desk. I saw one person comment that the Interfaces by Cramer keyboard was so big he felt cramped by it. From the picture it looks like it would make sitting down and getting up from the chair difficult unless in a fully verticle position. More on alternatives below.

Photos and adjustability

Click any of these photos for a larger version. The white keyboard got a little washed out due to the camera flash, so you can't see all the keys with perfect clarity, but you can certainly see the structure and positionoing well enough.

Normal position the way I like to type on it:

Normal angle keyboard only

My arms can rest on the armrests or if I sit up straight enough in the chair, my elbows hang just above the armrests with my hands right above the keyboard sections. The keyboard is attached to my laptop, but I've also got an external monitor hooked up for correct neck angle. The laptop touchpad sits directly in front of me at normal desk height, which is too heigh for ideal ergonomics, but I don't use the pointer extensively. The numeric keypad section of the ErgoFlex keyboard sits in the keyboard drawer under the desk. I can easily reach and type on the entire laptop keyboard including its numerical keypad keys (the arrows and pgup/pgdn being the ones I use the most) whenever I need to.

One neat aspect: reclining in the chair doesn't move my hands away from the keyboard! :-)

When I'm reading something that scrolls by hitting space (web browsers, more/less Unix command), I often recline. Another neat corollary is that swiveling the chair to talk to someone or look at something else doesn't create an awkward twist of the body with respect to the keyboard, so if you touch type, you can keep right on going.

As I've mounted it, each keyboard half has 3 continuous degrees of freedom, which allows for a lot of potential adjustment to get just the right fit. For example, each section can be easily tilted to any angle including flat like a normal keyboard or nearly vertical (palms facing each other):

Flat angle keyboard only Almost vertical angle keyboard only

Normally, I have it between these two extremes, as shown in the first picture. This adjustability allows for gradual tilting to get used to having the keyboard be non-flat, as described in the Comfort Keyboard product literaure and probably that of other adjustable tenting keyboards.

In addition, the angle of the keyboard section in its plane can be changed. In other words, the keyboard section can be rotated around its middlish keys. Lastly, the distance between the section and the end of the arm rest can be adjusted.

Unfortunately, the sections cannot be moved laterally with respect to the chair, so you can't move the keyboard sections closer together after you sit down and pull them apart again before standing up, like some kind of Star Trek chair. (Where did I see this? In Star Trek III on the captain's chair of the Excelsior I think. Okay, CP SET EXTREME GEEK MODE OFF. Hmm. Pehaps that's impossible on a page like this.) Also, the keyboard sections do not communicate wirelessly, so you have to attach them all together with cord. It uses spiraled cord like the cord between a phone handset and the base of the phone so it sort of stretches for some flexibility of movement, but even this means the chair is now tethered to the desk. You can simply unclip any of the parts and later plug them back in with no adverse affect or need to reboot, so it's not too bad. And one half of the keyboard can be attached to the other half under the chair and out of the way, so only one side of the chair has to have cord running to the desk, and it can be either side.

Assembly

The two main keyboard sections are each velcroed to a small flat rectangular piece of wood smaller than they are. Thus, the keyboard sections can be easily removed from the arm rests and replaced at will. The flat wood board is screwed into a cylindrical wooden rod, through the center of the flat board. This allows the keyboard section together with the flat board to rotate around the screw. The wooden rod is attached to the arm of the chair, which on my chair has an L-shaped opening (corner) under the armrest where the armrest itself meets its support. The wooden rod is held in place by an Ace bandage. I was going to use duct tape at first, but the non-adhesive elastic bandage allows the rod to be snug and secure, but sill allows it to rotate and to slide along its axis. So all the adjustments can be made but just grabbing and moving the keyboard section to where you want it. Sometimes holding the rod helps, but the point is you don't have to turn any knobs, flip any switches, or do any disassembly in order to adjust anything, and everything stays in place after the adjustment.

Cost

The cost is another nice bonus. Kinesis keyboards range from just under $200 to over $300, and the similar Maltrons are over $400. The Comfort Keyboard System is $300. The Interfaces by Cramer keyboard by itself is $500, or with a chair is $700-$1000. My chair, which I've been using as a desk chair for about 11 years now and which I find much more comfortable than most ergonomic office chairs I've used at schools and offices, cost $150. But I'm sure many chairs can be jury rigged in creative ways. Some methods might even allow for easier lateral adjustabiility of the keyboard sections. The ErgoFlex keyboard that is the basis for my setup cost $150, and can be had for $123.75 (before tax and/or shipping as appropriate) from Safe Computing in the SF Bay area. That's roughly $135 picked up locally in SF Bay area, $140 shipped to outside CA, or $150 shipped to within CA. Other resellers from outside CA obviously might have comparable prices without charging CA sales tax. Here is the approximate cost breakdown for my whole setup:

item approximate cost
keyboard $135
two Ace bandages $8
long wooden rod, cut in two and cut to fit $5
wooden board, cut into two small rectangles $5
self adhesive velcro strip roll, cut to fit $5
screws $1
total roughly $160

I even bought a cordless power drill for $30 just to drill the screw holes for this (though I'm sure I'll find other uses for it now that I have it), so even with the power drill the overall cost of the whole setup is less than many ergonomic keyboards.

Pros and cons of the ErgoFlex keyboard

I'm extremely happy with the mounting setup itslf so far, but I'm not completely happy with the typing behavior of the ErgoFlex keyboard. In some ways though, it is the only keyboard at the moment for this job and I commend the Comfort Keyboard company for making such a reasonably priced fully split keyboard. It seems to be the only reasonably priced adjustable split or fully split keyboard I can find that (a) is specifically intended to be used in setups like this one, and (b) has the 6/^ key on the correct half section. (The Interfaces by Cramer keyboard meets the requirements, but costs $500.) Also, the ErgoFlex is completely programmable, though the remappings and macros don't survive powerdown without getting some extra hardware. I use it to map left space to backspace which only takes a second, so isn't hard to do either every boot or whenever I'm about to type a lot. This is important because it shifts a heavy use key from the weak pinky to the strong thumb, which is also the one primarily unused finger of the 10. Evidently, the similarly priced Goldtouch adjustable split, tenting keyboard is not programmable according to my friend who has one, so you can't get one of the space bar halves to be backspace or some other key to take some work off of a pinky. And you can't solve this particular remapping problem in software if the keyboard sends the same code for both spacebar halves.

Positioning of the 6/^ key

Many split keyboards, and this includes the Goldtouch, Microsoft, and the Kinesis Maxim, position the 6/^ key on the left main keyboard section. To me this is completely wrong. The Kinesis Maxim is a particularly surprising case since they put it on the right side on all their contoured keyboards. I touch-type and that means I use the correct finger for each column of keys. 6-Y-H-N makes up one column, so even though the left index finger when positioned above the F is about the same distance from the 6 key as the right index finger when at home above the J key, the 6 key should be struck with the right index finger. When hitting isolated 6's or ^'s this really doesn't matter and touch typists could get used to either on normal keyboards, but when touch-typing numbers using the number row above the main keyboard instead of the numeric keypad, it is clear that each hand gets responsibility for 5 keys. I always type numbers with this row instead of the numeric keypad because it is closer to the main keyboard section and because I can enter numbers much more quickly with two hands than with only one. This includes numerical passwords and IDs, many of which are burned into my fingers motor memory by now. I also think this is more natural for the hands as it doesn't involve as much up and down movement for long number sequences once your hands are positioned over the row correctly. (The Interfaces by Cramer keyboard puts a 6 key on each half! Why didn't everyone just do this.) This is also important even for isolated 6/^ character typing for people used to split contoured keyboards such as the Kinesis and Maltron models, where the 6 key is correctly positioned since these keyboards eliminate the column tilt of a normal keyboard, or in fact for people used to any keyboard that eliminates this column tilt

Different key force options

Another plus of the ErgoFlex, not widely advertised by resellers yet, is that you can now get the ErgoFlex with keys having 3 different force levels. The default is 1.8-1.9oz (just over 50grams), which is fairly standard for keyboards, but you can now also get "soft touch" and "extra soft touch" versions with roughly 1.5oz and 1.2oz (about 35g) force. I have some friends who swear by low force, normal shape Keytronic keyboards and say that this strategy completely cured their finger pain.

Typos, swapped letters, and n-key rollover

 Unfortunately, the ErgoFlex does not seem to be a good keyboard for me in terms of basic typing mechanics and keying logic. Something about it causes me to make many more typos than I do with a normal keyboard. I am hoping my fingers will adjust to its slightly different behavior in time, and I have improved slightly, but I believe that it isn't just a matter of adjusting to a new keyboard but a matter of something fundamentally wrong with the keyboard for fast touch typing. The worst symptom of the problem is that when I type very quickly, many letters get swapped (teh instead of the, for intance), and I get a lot of double characters. Of course, everyone makes more mistakes when typing quickly, but I'm talking about numbers of mistakes vastly exceeding those I get with all other keyboards for the same speeds. This happens even when I lay the keyboard flat on the desk with the halves correctly fit together, so it isn't an artifact of the chair mounting.

One concrete characteristic of the keyboard I can point to that I think, but can't be sure, is related is that the keyboard does not behave like other keyboards when multiple keys are held down simultaneously, which is called (or is related to something which is called) n-key rollover. For example, this keyboards handles the 4-action sequence: depress key1, depress key2, release key1, release key2, differently than any other keyboard I've ever seen, and also asymmetrically (swap key1 and key2 and you get a different result). Specifically, with enough pause between each of the 4 actions to allow key repeat to kick in you get the following behavior with the a and m key:

keyboard key1 key2 result
|-press key1   |-press key2   |-release key1
Comfort/ErgoFlex a m aaaaaaaaaaaaaaamaaaaaaaaaaaaaammmmmmmmmmmmmmm
Comfort/ErgoFlex m a mmmmmmmmmmmmmmmaaaaaaaaaaaaaaamaaaaaaaaaaaaaa
every other keyboard a m aaaaaaaaaaaaaaammmmmmmmmmmmmmmmmmmmmmmmmmmmmm
every other keyboard m a mmmmmmmmmmmmmmmaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa

Any pair of keys consisting of one key from the left half of the keyboard and one from the right will behave exactly as above with the left key playing the part of the a and the right key the m. I have confirmed on the phone with tech support that the keyboards at the company do the same thing as mine, so it isn't just a defective unit. The guy I spoke with (on 8/22/2000 or thereabouts) was very nice and said he hadn't known about this behavior before and would get back to me after talking to the engineers. I haven't heard anything since then. I'm not sure that this is related to the excessive typos I'm getting, but it seems reasonable that when you type in rapid bursts you will be more likely to overlap a few depressed keys, and that if every other keyboard deals with your pattern of overlaps in a consistent fashion a reasonable typing technique can result that would be screwed up if a keyboard delt with the overlaps differently.

Also, I could swear that there are pairs of keys which if pressed in rapid succession come out backwards, with the 2nd key depressed showing up first as long as they are pressed close enough together but not simultaneously. This is difficult to be sure about without some kind of controlled scientific experiment though. It is clear that if you press any two keys simultaneously, which one shows up first is always consistent. That is, for every pair of keys, one is dominant. And specifically, every key on the left side of the keyboard is dominant in this sense over every key on the right side. It appears that if you shrink the amount of time between keypresses when hitting a right side key followed by a left side key to a small enough time interval, you'll get the left side key first, even if the time interval was non-zero.

I've seen one other person on the net comment that she too found that she made more typos, specifically swaps, with this keyboard. After some e-mail correspondence we determined that all of her common swaps, like mine, are examples of attempting to hit a right side key followed by a left side key. Moreover, on average one is able to hit the left side key most quickly after the right side key if the previous letter was also a right side key, since this gives the left hand the most time to set up and be ready to hit its key. In all of the examples we have noticed, the left key that gets swapped is preceded by two right side keys (qwerty: ing -> ign, jus -> jsu, mos -> mso, [space]pr -> [space]rp; Dvorak: the -> teh (which in qwerty is kjd -> kdj). She initially thought the low force "soft touch" version of the keyboard might solve this problem, but it did not.

These problems can be avoided by simply typing slower, and it may be that the improvement I've seen so far comes from slowing down. Clearly though, the keyboard should be able to handle typing at the same speed as other keyboards without making errors.

Shift key sticking

One other annoying little problem is that the keyboard sometimes fail to send the scancode for the release of the shift key, with the result that the shift key appears stuck every so often after using shift even though the physical shift keys are clearly not stuck. Repeatedly pressing and releasing either shift usually solves the problem quickly, but you can do a lot of damage to a text file in vi command mode before you realize what is wrong.

Verdict and other armrest mounting options

I used to think I liked the chair mounting enough to keep using the keyboard depsite these issues, but I've recently shelved the keyboard due to these problems. The Comfort folks seem like they might be working on sending me out a replacement part to fix the situation, so perhaps if they do, or if my RSI flares up again I'll go back to using it. I'd love to know whether anyone else with a Comfort Keyboard or ErgoFlex (same thing supposedly) has experienced these problems, and whether they've figured out a way to type at high speeds despite them.

It might be possible to separate some of the jointed adjustable split keyboards that are reasonably priced, such as the Goldtouch or Kinesis Maxim and use them in a chair mounted setup. [Update: The Goldtouch appears to have a very short cord going from one side to the other, too short for armrest mounting, and it would probably be hard to replace it with a longer cord.] I don't want to use either due to the 6 key problem and presumed lack of programmability of left space, but other people may not care about these things and may find these preferrable to the ErgoFlex. I haven't tried them. If anyone knows of any other candidates keyboards that would probably work, please let me know. Clearly what is needed is a wireless completely split keyboard, with good keying action. (There are of course gloves and similar hand-mounted devices that can still use standard typing motions and familiar layouts, but these are currently too expensive or not yet released as far as I know.)

Other solutions

At my Stanford office I'm simultaneously beginning to use a Kinesis contoured keyboard. My thought was that using different ergonomic setups at home and work would be good variety for my body. So far so good on this count, but it's really to early to tell. The Kinesis keyboard itself seems really nice so far. The key action is good. The contouring is not too hard to get used to. Many actions are transferred to the thumbs away from either pinky, so that's good. Plus, it's completely programmable and the remappings and macros survive being powered down. Switching back and forth between these different setups and normal keyboards at my part-time consulting job hasn't proved difficult at all. Now if only Kinesis would split their contoured keyboard into two halves that I could chair mount. That would be great!

I'm also trying out a footpedal, which is another nice way to remove keys for the right pinky to have to type. I haven't really gotten used to using it yet.

Another good tip if you suffer from similar symptoms to the ones I have is to use macros and aliases to avoid common problem sequences, such as "ls -l".

A more radical approach to controlling typing related injuries was offered up by a good friend of mine:

I've also found having a baby helps reduce the amount of time spent at a keyboard, and therefore helps with RSI, but that's a pretty expensive way to go.
:-)

Disclaimer

I'm not a doctor. I just talk about what I've learned from reading useful typing and ergonomic information in many places, such as the excellent Typing Injury FAQ, and from talking to my own doctors. I made this web page because it is more effective to show people that I talk to about my new keyboard what my setup looks like than to explain it verbally. None of the things on this web page should be taken as advice or prescription. My setup or similar setups may not work for you. I am not responsible for your results if you attempt to duplicate aspects of my own personal solutions. Reading pages like this or others on the web is not an adequate substitute for consulting a doctor if you have a real problem.

Also, note that I am not affiliated with any of the companies mentioned here or in fact any company that sells or distributes ergonomic products.

Thanks to my housemate Jason for taking the pictures with his digital camera.

Hopefully you found this page fun and interesting, and if you do set up something similar, please share the results. Good luck, and safe typing!

Other health and safety related consumer info that I've collected. Karl Pfleger
kpfleger@cs.stanford.edu
August, 2000