Power consumption of keyboards

[woody]

Can't remember this discussed so far.

As a quick, dirty and not very reliable experiment, I checked the increase of power consumption from HHKB Pro2, measured by the smart battery of a laptop. Roughly 0.7 - 0.8 W reported. That's quite much.

[Muirium]

The HHKB's a notorious power hog, thanks to its USB hub. I have no trouble running capsense IBMs from my laptop all day long, however.

What's your method?

[woody]

Don't know if it's the hub or something else. Hasu might have better idea, as he did HHKB dissection long time ago.

What's your method?

Reading power consumption from the smart battery of an unplugged laptop. At least the ballpark should be right.

But I'll have to reconsider the HHKB as a travel keyboard.

[woody]

G84-4400 - about 0.1W
Filco - about 0.1W

[jacobolus]

Still not going to wipe out your battery compared to the power draw of the rest of the laptop; I’d guess your battery lasts 5% shorter with the keyboard than without. Turning the display brightness down a couple notches will have a similar effect to using a lower-power keyboard.

(Still worth criticizing a keyboard that draws too much power, but not worth losing too much sleep over.)

[woody]

Yes and no. This one particular laptop keeps below 5W at idle (and I could shave few more percents). Most of the time I just stare at the screen, and I specifically bought extra battery to have over 20 hours of being brick-free.

Having the keyboard drain roughly 1/7th of the power at idle is meh.

If I was really losing sleep over, I'd be measuring the power loss precisely.

[Muirium]

Exactly. A keyboard that's taking more than just a couple of percent of the whole system power is… at once horrifying, and ho hum. You can always pull it out when you need those extra hours, and won't be typing. But you really shouldn't need to.

Tried a Realforce or NovaTouch? From what I can tell, they're much lower power draw. The HHKB's hub is a bugger.

[woody]

Nah. HHKB was my travel keyboard because of size/features, otherwise I wouldn't touch Topre domes.
That's really not the end of the world, though.

If somebody else has done measurements - please share. Mine are too imprecise to be worth documenting in Wiki, so take them with a very big spoon of salt.

[Halvar]

I tried to measure with my USB charger doctor, but as the name says, it's made for usb chargers so the resolution (10mA = 0.05W) is not sufficient for modern usb keyboard controllers. I have some concerns about the accuracy for these small currents, too.

Nevertheless:

(EDIT: deleted, this was pure nonsense. See my post below for better values.)

[Muirium]

Knowing just enough about electronics to be able to get myself into trouble: I could measure this myself with a current meter on the USB power rail, right? Well, if I hacked up a little passthrough with appropriate USB lead and socket. Or are you supposed to measure over a resistor or something like that?

Never did get far with this stuff at school. They stopped letting me investigate things by myself when a conspicuous cloud of smoke emerged from a motor I'd illicitly upgraded… (when in doubt: more magnets.)

[Halvar]

I could measure this myself with a current meter on the USB power rail, right?

Yes. Just be sure to use the current plug on your multimeter instead of the voltage plug.

I was just too cheap and lazy to sacrifice a usb cable right now.

[Muirium]

Right. I've plenty of the buggers, so I'll give it a shot sometime.

Interested to see how much power is drawn when typing, as opposed to simply sitting there. My guess is the beamspring at least takes a fair bit when the solenoid's firing.

[vvp]

Kinesis Advantage Pro (the older model): 61.2 mA @ about 5V, i.e. about 0.3W.
The number is precise (measured with an ampere-meter and a special USB cable).
The USB HID descriptor specifies maximum power for keyboard as 64 mA and for the hub as 50 mA.
Typing increases the power consumption to about 62.1 mA.
What a power hog!

[matt3o]

maybe something like this could work http://www.amazon.com/dp/B002T8RNF6/?ta ... verfl08-20

this is more fun though

http://www.instructables.com/id/USB-Vol ... /?ALLSTEPS

[Halvar]

OK, I sacrificed a USB cable and measured what I had here with a real multimeter (currents) and the charger doctor (voltage). Unfortunately, my Filco is at work, and I didn't have a lot of other modern keyboards here, but mostly vintage stuff.

I was a bit surprised that LEDs play a significant role for power consumption in modern keyboards, as they use 30 to 50 mW. If the user was typing or not didn't make a difference on any of these boards.

The most energy-saving board of mine was the Noppoo choc mini with only 0.009 W, the most power-consuming was the Model F-122 with a whopping 0.8 W (with the original controller plus including Soarer's converter). So there's actually almost a factor of 100!

xwhatsit's beam spring controller is really frugal (I only tested without solenoid though), my 3278 board with controller uses less power (0.106 W) that the teensy of Soarer's controller alone with no keyboard attached (0.124 W).

(obviously, divide by 5V to get approximate current in mA)

Code: Select all

                                   Switch             Converter # LEDs lit  Power/mW
Soarer converter w/o keyboard                                   none             124
IBM Model M SSK                    Buckling Spring    Soarer    none             532
IBM Model M full size              Buckling Spring    Soarer    none             542
                                                                all              655
IBM Model M-122                    Buckling Spring    Soarer    none             679
IBM Model M2                       Buckling Spring    Soarer    none             563
                                                                all              708
IBM Model F-122                    Buckling Spring    Soarer    none             793
IBM 3278A2 xwhatsit rev3           Beam Spring        (USB)     none             106
Realforce 91UBY                    Topre              (USB)     none             150
Cherry G80-3000 USB                MX                 (USB)     none              20
                                                                all              161
Cherry G80-3000 PS/2 winkeyless    MX                 Soarer    none             449
                                                                all              519
Compaq/Cherry G80-11801            MX                 Soarer    none             131
                                                                all              262
Keybo/Cherry G81-3000              MY                 Soarer    none             130
                                                                all              266
Noppoo choc mini                   MX                 (USB)     none               9
Tipro TMC                          MX                 Soarer    none             371
SGI Granite                        Alps               Soarer    none             257
                                                                all              377
Monterey K108                      Monterey           Soarer    none             465
                                                                all              541

[Muirium]

As I've been saying for a while. Capsense: less thirsty than you think.

What really matters is the age of the controller. Not its fundamental method. New capsense (Xwhatsit) trumps old capsense (IBM), and old Ohmic: like Model M and Cherry. Hell, he even beats Topre!

Intriguing that the Noppoo is so good. Does it have even diodes in its matrix, I wonder? Those are a drain by themselves.

I'm a little surprised the Teensy draws as much without a keyboard. But ATmega isn't particularly power efficient, and Soarer's converter is busy work, while looking for a keyboard to hook up, evidently. LEDs, meanwhile, are the bastards that I've heard. It takes daring and stupidity to put them in a wireless board, let's put it that way…

Speaking of which, the way we're going to get good battery life out of SSKs and HHKBs is with custom controllers. And not running on Teensy 2s for a change.

[jacobolus]

I'm a little surprised the Teensy draws as much without a keyboard. But ATmega isn't particularly power efficient, and Soarer's converter is busy work, while looking for a keyboard to hook up, evidently.

I’m sure if you got a team of systems engineers working on optimizing the code, you could get some keyboard firmware running on the Teensy with much lower power draw.

[matt3o]

this actually reinforces my belief that we should switch to a 3v controller (eg: teensy3) if we wish to go wireless and more efficient BT modules (BT4) of course.

The only way to really optimize the teensy 2 is to downvolt/underclock it to 3v/8MHz

More results all without additional controllers/adapters

V60 (has backlight but turned off) - 0.05A
HHKB - 0.12A
Leopold 660C - 0.06A
Model M - 0.13A
ML4100 - 0.01A

backlight at max brightness goes at 0.24A

[woody]

Nice work, Halvar.

With LEDs in the 30-50 mW range, they must be putting cheap LEDs at about 8-10 mA current.

I like the G80-3000 consumption, that's quite reasonable. I suppose the G84-4100 should be similar (it's not even NKRO), but all mine are PS2 and I don't have PS2 anymore.

Optimizing the firmware is not that hard and ATmega can also have low consumption. Going to 3.3V or even lower will incur some loss in the regulator or impose complex DC/DC solution.

EDIT: @matt3o - the numbers seem off?

[matt3o]

Optimizing the firmware is not that hard and ATmega can also have low consumption. Going to 3.3V or even lower will incur some loss in the regulator or impose complex DC/DC solution.

it's actually pretty simple https://www.pjrc.com/teensy/3volt.html

[matt3o]

EDIT: @matt3o - the numbers seem off?

why?

[woody]

Optimizing the firmware is not that hard and ATmega can also have low consumption. Going to 3.3V or even lower will incur some loss in the regulator or impose complex DC/DC solution.

it's actually pretty simple https://www.pjrc.com/teensy/3volt.html

That's the trap - this is a linear regulator. Whatever current flows at the output, flows at the input, too. So just the regulator itself dissipates the current multiplied by the voltage difference (about 2V here). Yes, the MCU will dissipate less, but you still have plenty of energy lost for no good reason. Hence why you would want DC/DC regulator - they also have efficiency less than 100%, although can get quite close. The problem is they are way more complex than linear regulators and have their own bunch of pitfalls.

[woody]

EDIT: @matt3o - the numbers seem off?

why?

If you calculate power by P = V x I, for example HHKB results in 6 W.
Also PS2 is rated at max 275 mA.

[matt3o]

That's the trap - this is a linear regulator. Whatever current flows at the output, flows at the input, too. So just the regulator itself dissipates the current multiplied by the voltage difference (about 2V here). Yes, the MCU will dissipate less, but you still have plenty of energy lost for no good reason. Hence why you would want DC/DC regulator - they also have efficiency less than 100%, although can get quite close. The problem is they are way more complex than linear regulators and have their own bunch of pitfalls.

you are missing an important step: underclock! https://www.pjrc.com/teensy/low_power.html

teensy 2 undervoled and underclocked draws less than half the power than a stock one

If you calculate power by P = V x I, for example HHKB results in 6 W.
Also PS2 is rated at max 275 mA.

all keyboards are USB. the cherry ML draws 100mA at 5V (which is exactly how much cherry says)

[woody]

No, not missing it. I said already that "MCU will dissipate less".
The point is that the linear regulator will still dissipate comparable energy to the rest of the circuit (ratio roughly about 2/3). About 66% efficiency for 5V -> 3.3V conversion.

Not connected to the Teensy specifically.

[matt3o]

I trust you I know nothing about electronics (no sarcasm, just humbleness)

[woody]

If you calculate power by P = V x I, for example HHKB results in 6 W.
Also PS2 is rated at max 275 mA.

all keyboards are USB. the cherry ML draws 100mA at 5V (which is exactly how much cherry says)

HHKB should draw less than 200 mA, if I am not mistaken.
Also, USB devices cannot draw more than 500 mA. (excluding chargers / USB3.x)

How are you measuring this?

[Findecanor]

Intriguing that the Noppoo is so good. Does it have even diodes in its matrix, I wonder? Those are a drain by themselves.

First, choose a microcontroller that is a good fit to what you need. Then, use power-saving modes properly and shut off the parts of the chip that you don't use - the ATmega in the Teensy allows you do shut off lots of stuff and have several modes for saving power.
Then use timer-interrupts and sleep mode instead of busy-waiting or scanning constantly like I've seen some open source firmware do. Then if you lower the polling rate and use low-speed USB so that you don't stay awake too often, you could save even more.

Give me some time to modify a USB cable so that I can measure my Phantom that is running a Teensy with my firmware...
Edit: Dang, I don't have any spare USB cable...

[matt3o]

If you calculate power by P = V x I, for example HHKB results in 6 W.
Also PS2 is rated at max 275 mA.

all keyboards are USB. the cherry ML draws 100mA at 5V (which is exactly how much cherry says)

HHKB should draw less than 200 mA, if I am not mistaken.
Also, USB devices cannot draw more than 500 mA. (excluding chargers / USB3.x)

How are you measuring this?

sorry, I simply put too many zeros The HHKB should be in the 600mW range (not 6W! )

[woody]

matt3o, they look good now. Maybe include the dimension (A) after the numbers for future reference?

HHKB turns into the 0.6 W (600 mW) range, which matches my rough ballpark.