RAFI full-travel key switches
RAFI in Germany were so kind as to supply me with samples of each of their keyboard switch types. I have been curious about their switches for some time, primarily because they have latching offerings. I finally gained the chance to investigate their designs, having first attempted to obtain RAFI switches around April 2015. In addition to the current RS 76 M/MX “mechanical” and RS 76 C Hall-effect switches, I received two of an unidentified part from the extinct RS 74 M series (not covered here).
The following is an overview of the switch series.
I don’t know if the sample switches are from 2016 or whether they are spare parts set aside from an earlier production run; according to the current catalogue, the RS 76 C sliders are all colourless, while the switches I received have yellow sliders just like they have done in years past.
RS 76 C Hall effect series
Hall effect keyboard switches offer very long lifetime, but as each switch contains a microchip, they are expensive. As a result, they were only ever manufactured by a small number of companies, and Hall effect fell out of favour a long time ago. Most other non-contact designs were either capacitive (generally foam and foil) or magnetic valve.
The most notable Hall effect keyboard switch manufacturer was Micro Switch (part of Honeywell). The other well-known manufacturer of Hall-effect keyboard switches is RAFI, and now that Micro Switch Hall effect is discontinued, RAFI seemed to be left as the sole remaining manufacturer. However, while writing this article, I found that this situation had already changed!
(For that matter, RAFI are one of the longest standing keyboard switch manufacturers on the planet, alongside fellow German (but originally American, otherwise it would be Kirsche) manufacturer Cherry. RAFI RS 74 was introduced in 1975, and RS 76 was introduced in 1976, while Cherry’s nameless “classic” line—that concluded in the guise of Serie M7—was introduced around 1970. Some details about Cherry prior to that date exist, but no RAFI switches are known from before 1975; a high profile Hall effect switch exists, but strangely enough the only example studied to date appears to be from 1986, well after RS 76 was introduced. The force curve looks just like what I would expect to see for the 70 cN RS 76 C.)
Linear non-contact switches are known for their smoothness, and RS 76 C is no exception to this. Standard RS 76 C ticks two important boxes: it is extremely smooth, and it is an intermediate weight, in between Cherry MX Red and Black. Depending on preload, the actuation is going to be somewhere between 50 and 55 cN. The specifications are as follows (curious typography as per the catalogue):
|Switching travel off||0.8 ... 2.4 mm|
|Operating force max.||0.70 N (standard momentary switch)|
|Operating travel||4 mm|
|Switching travel on||1.5 - 2.8 mm|
Terminology varies between manufacturers. I am more accustomed to Cherry’s definition of operating force, which denotes the force exerted at the point that the switch contacts close; since actuation typically occurs at mid travel, this will be half of the force present at full travel. I notice that a 2010 Cherry brochure uses “initial force” instead of preload, and “actuation force” instead of operating force. (Preload is the intial force, at 0 mm travel, created by placing the spring into a space shorter than its natural length, which avoids loose, rattly keys.)
RAFI use operating force to denote what I call “terminal force”: the force present when the key is fully depressed. Expecting standard RS 76 C to be fairly heavy (as is so often the case with linear switches) I was pleasantly surprised to find how soft it is. Clearly the 0.7 N figure is the terminal force. (I confess that I do not understand the figures given above at all. If operating travel is 4 mm, then surely the “off” and “on” regions of the travel should sum to 4 mm? For RS 76 M, the catalogue notes only that switching travel is “1.5 - 2.8 mm”.)
To put this into perspective, here are the Cherry figures; these are approximate figures determined from their graphs, and the MX Red figures come from a Google image search as Cherry seem not to have published a graph for it:
|Cherry MX Black||40 cN||80 cN|
|Cherry MX Red||32 cN||60 cN|
|Cherry MX Blue||35 cN||60 cN|
|Cherry MX Brown||40 cN||60 cN|
In terms of terminal force, you can see that standard RS 76 C is in the middle between Cherry MX Red and Black. Although there is no official preload figure from RAFI, one would reasonably assume that it would be in the 30–40 cN region, as per Jacob Alexander’s force plot of the mystery tall switches (for which RAFI have yet to comment upon).
In addition to a smooth feel, the non-illuminated versions have a dampening ring that absorbs the impact of bottoming out. The illuminated variants omit this, because the large LED holes in the base of the switch leave no room for it. The LED holes incidentally accept full-size LEDs instead of the reduced size LEDs used with Cherry MX switches.
Perhaps for the purposes of ensuring that switches are properly secured and protected against rotational forces, an extra leg is provided at the opposite end of the switch to the microchip:
The whole RS 76 C series appears to use the same sensing arrangement: a specific model of Hall effect sensor chip from an unknown manufacturer (and identical now to the one used in Jacob’s 1986 tall-switch RAFI keyboard). Pulling the chip from an RS 76 C latching switch reveals it to be the same switch as used in the momentary switches.
RS 76 M however is a whole lot more complicated!
RS 76 M “mechanical” series
The mechanical series is divided up into three sub-families, each with a different contact system:
- Momentary, without diodes
- Momentary, with integrated diodes
The workings of the perimeter loop contact arrangement should be now be understood sufficiently well, having documented this in 2013 after picking up a Neve Necam 96 audio workstation keyboard using RS 74 M switches.
What I did not realise is that the diode version of RS 76 M is a different design inside. It is not the only switch to fully internalise the diode (another switch of this form showed up at Deskthority) but this design is unusual. Normally, the jumper or diode is kept separate (the switch will have a full four legs minimum), joined into the circuit by tracks on the PCB. RS 76 M however has only two legs, with the diode itself providing one of them. The other leg of the diode is routed inside the switch, where a small gold-plated contact is clipped onto it. The movable contact provides the other leg.
The diode is placed loose inside the switch, and then the lid of the switch holds the diode in place; this explains why the diode versions are a different shape.
As shown above, removal of the stationary contact is possible using a sharp knife. Initially I figured that it would slide off, and gave up after this proved futile. I then realised that, due to a mysterious overhang present in the shape of the part, that it should be possible to slip a knife into that gap and prise the contact off the diode leg. I have therefore succeeded in a 100% disassembly of RS 76 M (diode). I have yet to get the illuminated version apart, though — the slider just won’t come out!
Here is a non-electrical-engineer attempt at the schematic:
(Laugh all you like.)
According to the catalogues, the illuminated switches use bridge contacts instead of crosspoint contacts. I never understood the reason for this until the switches arrived. The LED positions are centre left and centre right, and the perimeter leaf used in the non-illuminated switches obstructs these positions. To resolve this, a different contact system is used that only occupies the rear of the switch.
You could be forgiven for assuming that “bridge contact” would be something dull, but in fact it’s more interesting than the design of the non-illuminated switches. Two stationary contacts are used, at the rear-left and rear-right positions. A horizontal leaf is lowered down by the slider such that it touches the contacts on either side, bridging them. Each end of the leaf is divided into two tines, and the ends are angled downwards.
If this was the entire design, then it would suffer from a lack of pretravel. As this is a RAFI switch, the design is indeed more intricate. This lateral leaf is not attached to the slider, and sits under a tiny helical spring. When the slider is depressed, the lateral leaf will reach the stationary contacts, and the tines will flex upwards slightly. At this point, it can move no further, so as the slider moves further downwards, the helical spring above compresses and allows the leaf to remain stationary without blocking the slider.
I have yet to figure out how to disassemble this switch, so instead I’ve studied it and produced the following illustrations:
Having yet to achieve successful disassembly, the dimensions—and the shapes of the parts—are only approximate (only a few of the dimensions could be measured with calipers). You will also have to forgive the springs, which I can’t draw, either in isometric or otherwise!
Because actuation increases the number of springs acting on the slider from one to two, the switch is progressive rate, with a sharp and distinct tactile event as the force gradient increases (although not tactile in the way that it is normally understood, since here the force increases sharply instead of decreasing sharply). This precise increase in the force gradient appears to indicate that the switch can be actuated without bottoming out, in rather the same manner as Cherry MX Clear, by typing gently and releasing the key as soon as the force increase is felt, giving a fairly soft, short-travel feel. It does leave me wondering whether the bridge contact design would actually offer a superior typing experience over that of the perimeter leaf design. Both the crosspoint contacts and bridge contacts are rated for ten million operations, but the bridge contact design only offers a maximum bounce time of 10 ms, which is twice the industry standard of 5 ms. (Bounce avoidance may be why such an intricate mechanism was selected; simply using soft, springy contacts that bend may have generated a huge amount of bounce.)
The latching action switches are all illuminated, which means that they have two tiny springs inside of them, for maximum pain during disassembly!
In addition to being notable for still manufacturing Hall effect keyboard switches, RAFI are also notable for still manufacturing latching action (alternate action) keyboard switches. The industry fascination with latching action faded in the 80s, leaving chiefly Apple whose keyboards continued to use them for caps lock until 1995 when the Extended Keyboard II was discontinued. Ortek’s AppleDesign Keyboard–style keyboards (the MAK-105) also used Futaba latching switches for caps lock, while Focus used Futaba latching switches in the 90s also.
I have a particular fascination for latching action switches, and I have collected a number of types including Datanetics. My favourite designs are Alps SKFL, Cherry, old SMK and now RAFI. SMK’s old design is notable for its solid turned brass follower, and for having a separate side module to contain it; the mechanism used in the second generation switches remains a mystery, and I have yet to succeed in obtaining such a switch. Cherry’s design is notable for using a wheel design that so far has proved to be unique to Cherry, and is used both by M5/6/7-era switches as well as MX.
The Alps SKFL design uses a slide block to hold the follower, which is curious. The follower pin is tiny, but disassembly and reassembly is relatively straightforward so long as you don’t lose that pin. RAFI’s design is conceptually similar, but it moves the latch track into the slide block. The follower is required to move in three dimensions: laterally with with respect to the course of the latch track, as well as perpendicular to the latch track to ride the ramps that control its direction of movement. Since the track is mounted vertically, holding the follower against these ramps requires a spring, or (in the case of Alps SKCL, a flexible arm).
RAFI’s approach is to use a minuscule spring inside the slider, which holds a small follower pin against the slide block. I do mean minuscule: this spring is only 2.7 mm long! From my batch of RS 74 M SPDT switches, I started out with ten springs in ten switches, and now I only have nine: one is lost forever. They are so small and light that reaching for one with tweezers allowed the latent magnetism in the tweezers to pull the the spring towards it with enough speed such that it overshot and disappeared. I did eventually find it approximately where I thought it had gone, but I don’t know that I hadn’t simply found the one I’d lost earlier in the evening.
I have already documented my first RAFI latching switch type here: RS 74 M SPDT latching
The same latching system is used by RS 74 M, RS 76 M and RS 76 C. This may explain why it is so amazingly intricate: the design has to fit into the confines of the RS 74 M dimensions. I have yet to attempt disassembly of any of the sample latching switches that RAFI sent me, as it is just so difficult to keep hold of the parts, and much time passes scanning the carpet with a magnet trying to see where the invisible spring ended up this time.
Hergestellt in Lilliput
One of RAFI’s secrets is is that their switches are assembled by Lilliputians. If you don’t believe me, here are some of the sizes of the parts inside:
- RS 74 latching system slide block: large by RAFI standards, at 2.8 × 3.9 × 1.85 mm (this is the RS 74 size: I think the RS 76 C slide block is a bit taller but I remain disinclined to dismantle a switch to check)
- RS 76 C Hall effect magnet: 2 × 3 × 1 mm
- RS 74 latching system follower pin: around 2.6 mm long, with diameter 0.5 mm at the engagement end, and 1.7 mm diameter at the end facing the spring
- RS 74 latching system pressure spring: 2.7 mm long and 0.9 mm diameter — yes you read that correctly, a spring less than a millimetre across, which is why I have lost one of them forever
- RS 76 (diode version) stationary contact: 1.7 mm wide, 2.6 mm tall and 0.9 mm high
I only have an 11-year-old ultracompact camera, so I cannot offer scanning electron microscope views of these parts (i.e. what the diminutive workers would be able to see with nothing more than a magnifying glass) but the following images will still give you some idea of just how small everything is:
No-one has yet knocked Futaba down from the top position when it comes to the sheer number of parts, with 16 parts for a mere linear switch (with more parts than anyone has ever been able to find an explanation for). The top prize for incredulity goes to Tokai, whose MM9 series is so bizarre that it appears to defy the laws of physics (the movable contact is a large gold-plated ball bearing, and a spring that interposes that and the slider that will not only raise the slider but also pull the ball up with it).
RAFI however must get the award for the smallest parts, as I have never found anything as uniquely frustrating as that 0.9 mm diameter spring.
Conclusion and thoughts
RAFI switches turned out to be far more interesting than I had expected! I am very glad that I had the opportunity to study and document them.
On the basis that no-one was doing Hall effect in the keyboard community, I had started to contemplate the prospect of a 60% Hall effect keyboard using RAFI switches. They’re not cheap, though; the MOQ is of 25 per switch type, with prices from one supplier of €3.46 each for 3.13.001.010 (70 cN) and €4.03 each for 3.13.001.110 (70 cN illuminated, to provide for backlighting). 50 such keyboards would require around 3,000 switches in total (3,050 for 61-key ANSI, and 3,100 for 62-key ISO), and for an order of 3,000 RS 76 C switches you’re looking at a reduction of around 2–3%.
Just one day later, this showed up on Deskthority: Ace Pad Chinese-made Hall effect keyboard! I was clearly correct in believing that Hall effect was a desirable avenue, but is there a place for RAFI in this picture, as a source of switches for a premium product? I don’t foresee myself ever being in possession of one of these Ace Pad keyboards (I’ve settled on my Majestouch 1 and Poker II) so for the foreseeable future, I cannot say. The Chinese-made switches sound horrible, but they take Cherry keycaps (of little interest to me, but it will definitely help with sales) and the controller already provides more eye-bleeding backlight effects than you could desire. (I do like the idea of reactive lighting, but the implementation in that controller is too harsh, without a proper fade effect.)
What RAFI does offer however is discrete switches, something that I assume will in time be an option with the Chinese product (although not as discrete modules but rather the option to purchase loose parts for assembly).
Since RS 76 C is illuminable, it would be possible to create a keyboard similar to my Poker II, but with substantially better switches. This will never happen, but I can dream! I would also love to try a whole keyboard of RAFI’s bridge contact switches, as they offer the possibility of a new typing experience.
Many thanks go to RAFI for providing all the samples. I have two spare RS 76 C 70 cN switches (standard weight, non-illuminated), so I can spare these to anyone seriously considering using them in a modern production keyboard.
No switches were killed in the production of this article. Harmed, yes, but there were no fatalities.