Cherry’s RealKey tech promises uber-responsive, ghost-free keyboards

Future Cherry MX keyboards may be more responsive than ever—and they may have better, more reliable anti-ghosting, too. That's all thanks to Cherry's new RealKey technology, which the company announced at CES this morning.

As Cherry explained, all keyboards have something called a "debounce delay" that dictates how quickly keystrokes are registered. When a key is pressed, the contacts inside the switch come together, and the key "bounces." The keyboard's controller needs to wait a certain amount of time for the "bounce" to end in order to avoid accidentally registering the keystroke more than once.

Today's keyboards use digital controllers, and Cherry says their debounce delay is 20 ms on average. The company's RealKey technology uses an analog controller that's able to shorten that delay to just 1 ms. The graphs above show how the two systems differ. (Use the buttons underneath to switch between them.)

RealKey's analog mojo has another advantage. Because every key and key combo produces a discrete analog signal, ghosting "simply does not happen," Cherry says. In other words, no matter how many keys are pressed at the same time, the controller won't fail to register some of them, nor will it register incorrect keystrokes. Some current mechanical keyboards already implement anti-ghosting countermeasures, but as I understand it, RealKey should make those obsolete.

Interestingly, Cherry doesn't plan to license RealKey to other keyboard makers just yet. The company will instead premiere the tech on a keyboard of its own design, the MX Board 6.0. Priced at $219 and scheduled for a spring release, this offering is aimed squarely at professionals. It features a 108/109-key layout, an aluminum housing, an "anti-fingerprint" coating, a palm rest (which will latch on magnetically, if I understand correctly), and red backlighting.

Comments closed
    • destroy.all.monsters
    • 5 years ago

    Not a big fan of red (sorry Sammy Hagar) but a big fan of Cherry keys. I’d get this at 60 dollars cheaper. Reminds me of SteelSeries keyboards design wise along with the wrist rest.

    • Krogoth
    • 5 years ago

    Why?

    This guy belongs in the same room as Bigfoot NICs and Ultra-high DPI “gaming” mice.

    Overpriced marketing gimmicks to people who do not understand input lag and fall victim to the placebo effect.

    Protip: The biggest bottleneck of input lag is still and will always be the human musculo-skeletal system and nervous system.

      • ptsant
      • 5 years ago

      Actually things are not so simple.

      First of all, debounce is NOT about being able to press the key 50 times per second but rather about the fact that switches never give a perfect 0-1 signal. Debounce is the time given to the circuit to let the signal stabilise as a single keypress. Without debounce, a single keypress appears jaggy and can be falsely interpreted as multiple keypresses. More details here: [url<]http://www.labbookpages.co.uk/electronics/debounce.html[/url<] Now, if the debounce time is reduced by 19 or 9ms (let's take the lower estimate), you are reducing the input lag by 9ms because the keypresses are acknowledged sooner. Is that significant? Well, an average pilot scores something like 200ms visual reaction time. Online tests give results that range from 180 to maybe 320ms with a median of 250ms (http://www.humanbenchmark.com/tests/reactiontime/statistics).

      A few ms of monitor input lag, a few ms of keyboard input lag (+polling rate), the FPS rate and most importantly the network may add up to say 30 or 40ms. This is much less than 250ms so indeed, the biggest bottleneck IS the human. However, the key point is that a 10ms benefit should not be compared with the median reaction time, but with the VARIANCE between players.

      To understand why, imagine that if ALL players had a 1000ms reaction time but ONE player managed to obtain a 995ms reaction time, he would have a consistent, permanent advantage. Similarly, a 0.1 sec is a ridiculously short amount of time compared with 10 sec, but it makes all the difference between being a world champion at a 100m sprint and a third place. For everything that follows a normal distribution (human reaction times do seem to follow a normal distribution), deviation from the mean is always measured by the standard deviation while the mean itself is not important.

      Anyway, you won’t become pro with a keyboard change, but if you manage to add up 30-40ms between all the latencies, you do get a real difference.

        • Krogoth
        • 5 years ago

        Debouncing is only problem with speedtyping. Eliminating it is only benefits a small minority of typist/coders that normally operate at such speeds. For the vast majority, it is a non-issue.

        Input lag is mostly a problem with human body not the computer unless you have faulty or just slow equipment (old LCD mointors and keyboard/mouse pre-2000 stuff). That’s why drug abuse is a big problem in eSports (stimulants) along with using bots (computer) to by-pass the problem altogether.

        One of the first things that goes with age is your reaction speed and “fine” motor speeds. That’s why the majority of progamers tend to be teenagers to people in their early 20s. The population with progamers beyond their 30s drops like a rock.

        Besides, most of the so-called “super-fast” reflexes are actually players anticipating and reading their opponents actions ahead of time. This comes with spending countless hours of practicing and experience. It is no different from any other athletic endeavor.

          • dragontamer5788
          • 5 years ago

          Fighting Game players regularly hit 1-frame links (aka: timing precision of 16.6 ms)

          Its not only about “super-fast” reflexes, its also about consistency of timing between button-presses. Reducing the latency and “jitter” of your peripherals is a major element of any serious videogamer.

          Besides, someone needs to make keyboards for pro-gamers. I am not at that level of proficiency at PC games, but I regularly hit one-frame links in various fighting games when I’m practiced up. I’m sure PC progamers want the absolute best peripherals to hit a similar level of proficiency in their games.

          We fighting game players can feel the difference in 10ms lag on TV screens. I can assure you, any real PC Pro-gamer will be able to feel a difference of 10ms.

            • cphite
            • 5 years ago

            So basically, if you’re an extreme gamer you might notice a slight difference some of the time. Otherwise, there are better ways to spend $219

            • dragontamer5788
            • 5 years ago

            Actually, Fighting Game players regularly spend $200+ on custom joysticks.

            So $200 for a high-end peripheral is… about par for the course.

            [url=http://www.focusattack.com/torii-japan-theme-metal-fightstick-black/<]Pre-assembled Joysticks with high-end Sanwa parts cost about $280[/url<]. Slightly cheaper at ~$200 per if you build it yourself. So... in comparison, the keyboard is kind of cheap. [quote<]So basically, if you're an extreme gamer you might notice a slight difference some of the time. [/quote<] If you're at the level of gaming I'm talking about, you'll notice a 10ms difference [b<]all[/b<] of the time. Again, these players regularly have timing within 16.6ms. Reducing latency in your inputs (and outputs: video monitor lag is important!!) goes a long way to maximizing your ability. I can speak for the existence of one-frame-links (16.6ms) in the games that I'm good at. And also the fact that tournament players regularly hit that level of precision with no sweat, often while simultaneously setting up traps and focusing on the opponent's movement patterns. I have no idea how precise a human can get with 100% of his brain focused on the job (instead of being distracted by Yomi and whatever...), but it is somewhere far far more precise than 16.6ms.

            • Krogoth
            • 5 years ago

            Progamers naturally place the blame on the computer/peripherals rather than themselves out of hubris. Speed in fighting games isn’t as important as being as accurate at performing combos/chains and understanding how each “fighter” works.

            • dragontamer5788
            • 5 years ago

            I’ll… just leave this here.

            [url<]http://youtu.be/eQOswiAGLU4?t=3m40s[/url<] People do react at insanely fast speeds yo.

        • Pancake
        • 5 years ago

        Did you just google that reference? Have you any practical experience or are you just quoting? That example is a software implementation of the hardware described with the capacitor charging and Schottky diode. It works. It’s how I’ve done reset circuitry for computers (the hardware version). But it’s not necessarily the best idea for responsiveness.

        Here’s an example of how I would write some pseudo code for a 1ms interrupt. Assume it’s scanning a particular key.

        on_milli_interrupt:
        if counter > 0:
        counter = counter – 1
        end if
        else if my_state <> sensor_state:
        my_state = sensor_state
        counter = 20
        end if

        assuming our condition is 20ms for bounce stabilisation. That, my boy, is how a real man does it.

    • Srsly_Bro
    • 5 years ago

    OK, I’ve been using the same $13 keyboard since 2006. What am I missing?

      • Krogoth
      • 5 years ago

      Nothing really.

      Typist and programmers/coders prefer mechanical keyboards for their tactile feedback. Latency is a non-issue for this crowd. The biggest concern is actually ergonomics that minimize the risk of RSI (repetitive stress injuries).

        • Liron
        • 5 years ago

        Regarding RSI, I haven’t used mechanical keyboards but I’m curious; don’t the keypresses give more of a “hit” back to the finger than quiet keys do?
        I liked that my work mouse had a very strong and noticeable “click”, but I had to change it for a gaming mouse because the pain it started causing after some time on the finger’s middle joint got troublesome.

      • hansmuff
      • 5 years ago

      If you’re happy with the keyboard, nothing at all.

      Some people, myself included, really enjoy keyboards with more precise feedback than a $13 will give you.

      Then there are people who like keyboards that are really short, a good reason being RSI that gets aggravated when you have to pivot your lower arm out to hold the mouse; a short keyboard brings that in some and there aren’t any/many $13 short keyboards. By short keyboard I mean those without 10key pad and often limited or more closely spaced cursor keys/above cursor function keys.

      Some mechanical keyboards give you very audible feedback about when a key gets actuated, for instance cherry blue switches.

      Mechanical keyboards let you replace the key caps with colorful ones, different shaped ones, different print ones etc. and some love that aspect.

      It’s a mix of all those things.

    • ptsant
    • 5 years ago

    +1 for the aluminium housing. I like sturdy keyboards.

      • Krogoth
      • 5 years ago

      Get a Model M or Unicomp version if you want a sturdy keyboard. The only thing that kills it is a nasty liquid spill.

        • Hattig
        • 5 years ago

        “Nasty liquid spills” are frequently a danger for many keyboard users…

        I’d like a tenkeyless version of this.

        #24 – according to the marketing, the debounce delay is to stop a single keypress registering more than once, because the signal isn’t a clean rise from 0V, but has some back-and-forth before the contact fully establishes (signal reliably over a threshold voltage). Now I don’t believe that it takes 20ms to establish a contact for normal typists, and 10ms seems high too.

        This would appear to have a controller that stores an accumulator for each key in the keyboard. When the accumulator saturates or reaches a specific value, it registers the keypress. The accumulator increases by the amount of voltage registered on the key during the normal keyboard scan. Saturation of the accumulator occurs far faster than using a debounce delay to ensure the signal is over/under the threshold to prevent false double-keypresses and false single-keypresses. As each key has its own accumulator, ghosting is prevented. I don’t see why this couldn’t have been done a decade ago in a keyboard controller, but there you go.

        • ptsant
        • 5 years ago

        I have a Cherry keyboard that has … Cherry switches and only cost me $70. It’s quite solid but not as solid as the Model M. I am tempted by the Corsair RGB though.

    • Pancake
    • 5 years ago

    The explanation above isn’t quite correct. Or, rather, isn’t how you should be doing it.

    Key press registration should be fairly instantaneous – at the sampling rate of the firmware scanning the columns and rows. Debounce delay is the time BETWEEN multiple key press registrations. So, 20ms delay means you can’t press the key more than 50 times a second to register each key press (assuming instant scan time).

    Now, I don’t know any situation where I’d be pressing a key 50 times a second. I would have thought this was a well solved problem by now. But these half-arsed Asian engineers keep surprising me by their inability to innovate and this might actually be an improvement.

    • CheetoPet
    • 5 years ago

    Nice, but even 1ns response time won’t make me game any better.

    • willmore
    • 5 years ago

    20ms to debounce a keyswitch? What idiot wrote that firmware?

      • Gershwin
      • 5 years ago

      You’re right. I think it’s closer to 5-10 ms. But, I suppose you can’t blame Cherry marketing for using the very worst example to inflate their accomplishment.

      Oh wait, you did notice it; so yes, we can blame them.

      • just brew it!
      • 5 years ago

      Pretty much all keyboards are wired in a matrix, which is then scanned sequentially by row and column. This eliminates the need to have a separate pair of wires (and a separate detection circuit) for each key. I would not be at all surprised if it takes on the order of 10ms to complete a single scan. Furthermore, to debounce a switch in a scanned matrix, you simply check whether the state changes from one scan to the next. For a 10ms scan period, 20ms would be the worst-case debounce time, if you happen to get unlucky and depress the key right after the scan has passed that key.

        • willmore
        • 5 years ago

        Yes, I understand that most keyboards are scanned as a matrix, but that’s really, really slow in processor terms.

          • just brew it!
          • 5 years ago

          With the traditional debounce method you need to wait for the contacts to completely settle anyhow, so there’s not much point in polling a whole lot faster. It even complicates the debounce algorithm slightly, since you need to keep a count for each key of the number of polls in a row where the state doesn’t change, then trigger the key down event only when the count crosses some threshold.

            • willmore
            • 5 years ago

            We’re coming up on the 20th anniversary of my first key debounce code. 😉 Trust me, I get it. 🙂

            • just brew it!
            • 5 years ago

            Cherry specs the contact bounce time for their MX switches as 5ms. So this would be your “best case” debounce time using traditional methods, if you polled at a high rate and registered the press as soon as the contacts settled.

    • odizzido
    • 5 years ago

    This is keyboard technology that I can actually get excited for. Not $219 excited, but maybe $50 excited.

    • chuckula
    • 5 years ago

    Isn’t some of the issue with multi-key operation due to how USB handles keyboard input vs. the old PS2? Something about “n-key rollover” that USB just doesn’t do correctly compared to how PS2 did it?

      • anotherengineer
      • 5 years ago

      PS2 keyboard FTW!!!

    • Gershwin
    • 5 years ago

    I have a feeling that the biggest advantage may be negligible. The debounce time will decrease from 20 ms to roughly 10 ms–the graphic above shows 8 ms. I have a feeling that the decrease will be completely unnoticeable even by the very fastest and most discerning typists. I could be wrong–we shall see.

    The elimination of ghosting sure sounds nice. Well maybe not. I really don’t get much ghosting now even on my 10-ish y/o Unicomp. I may have at some point. I’m sure I have been doing multi-WASD diagonal running and dodging, simultaneously equipping/de-equipping while Shift-crouched in the heat of battle. But, to keep track of whether ghosting, some client hiccup, network delay or server response fault is the problem causing some misstep is pretty much impossible. Also, one’s mind and fingers tend to automatically adjust for unresponsiveness and missteps dynamically.

    I’ll probably grab one of these when they become available.

    It may be just hype, but there’s also the chance that this is going to make everything like [i<]buttah[/i<].

      • PainIs4ThaWeak1
      • 5 years ago

      Actually the graphic states: “0.8 ms”

        • Gershwin
        • 5 years ago

        Ahh yes, that it does. Thanks!

        By the way, do a search for “cherry realkey” on youtube to see their presentation.

      • just brew it!
      • 5 years ago

      They are claiming 0.8ms, not 8ms. So the improvement is substantial. Whether it’s actually noticeable in real games is another question however. I suspect it will make a difference to people with fast reflexes in “twitch” games, since 20ms is a bit longer than the time between video frames whereas 0.8ms is short enough that it just doesn’t matter any more.

      As far as the ghosting/rollover issue is concerned, this is already a solved problem. However, it sounds like maybe Cherry is doing away with a traditional switch matrix entirely, which would make it inherently ghost-proof (and n-key rollover capable), instead of requiring the addition of a blocking diode at every switch. Unfortunately, if they are really using a separate detection circuit for each switch, prices on these will probably remain sky-high even after the novelty wears off.

    • chuckula
    • 5 years ago

    Time for some inside the second keyboard benchmarks TR!

    • ermo
    • 5 years ago

    Analog: 1
    Digital: 0

    It’d be nice if eventually it was offered in a physical layout that positions the hands like Microsoft’s ergonomic line of keyboards do.

      • derFunkenstein
      • 5 years ago

      The old school Natural Elite was DA BOMB! I used one for a very long time, only switching to a standard layout when I got a mechanical keyboard about 18 months ago, because those ergo layouts aren’t common in mechanical boards yet.

        • Flying Fox
        • 5 years ago

        The Elite sucked with the unconventional arrow key layout. Almost every other Natural keyboards were fine.

          • derFunkenstein
          • 5 years ago

          Maybe I just got used to it, but I had one forever and loved it.

      • mnemonick
      • 5 years ago

      I’m gonna buy me one and name it [i<]Gypsy Danger[/i<].

        • Tumbleweed
        • 5 years ago

        ‘Gipsy’

          • Srsly_Bro
          • 5 years ago

          ‘Gypsum’

      • Flying Fox
      • 5 years ago

      [quote<]Analog: 1 Digital: 0[/quote<] And since almost forever we have been bombarded with digital > analog. And now being analog is a good thing. How times have changed.

        • The Dark One
        • 5 years ago

        The analog face buttons on the DualShock 2 were the best.

        • just brew it!
        • 5 years ago

        Our storage is going quasi-analog too. TLC flash stores 8 different voltage levels per cell, so the individual cells aren’t strictly digital (in the sense of 1s and 0s) any more. Hard drives have been analog-ish at the lowest level for a while too; high storage densities smear adjacent bits together on the platter, requiring the use of sophisticated algorithms like [url=http://en.wikipedia.org/wiki/Partial_Response_Maximum_Likelihood<]PRML[/url<] and [url=http://en.wikipedia.org/wiki/Noise-Predictive_Maximum-Likelihood_(NPML)_Detection<]NPML[/url<] to recover digital data from what is essentially a very sloppy analog signal.

Pin It on Pinterest

Share This