Factual updates and other interesting changes are reported here. Adjustments and tweaks are not listed.
Updates for 2020
Sunday, 12th July
Having recently split the remaining Fujitsu reed information into separate FES-5, FES-8, FES-9 and FES-4 pages (with FES-1/2/3 having been split out previously), I have now expanded the details of FES-8 according to what little of the Fujitsu journal article I am able to read and understand. For example, it seems that Fujitsu offered both レベル出力 (level output) and パルス出力 (pulse output) Hall effect options. The are no photographs and measurements of any Fujitsu reed switches yet, as it may be some years before I lay my hands on any of these types. What is disappointing is that no-one with an illuminated FES-8 or FES-9 switch has ever checked what part number is written on them. It may be a long time yet before anyone encounters the Hall effect types, however.
On the subject of Hall effect, the encoding and output page recently gained a section on Micro Switch’s two-of-N code direct encoding, using binary encoders to convert the two-of-N codes directly to ASCII codes without the need for a microcontroller.
Wednesday, 24th June
Having now received my order from Incotel (as mentioned in the previous update) I have now been able to examine the M76-0899 switches. These have six terminals, and I determined that two of them were functionless, until I opened one up and realised that the first switch I tested was defective. They are in fact double pole double throw (DPDT) alternate action, with each side having the opposite operation (the output terminals are NO+NC on one side, and NC+NO on the other). The mystery of M75–M78 is finally revealed: where needed, these switches have side-by-side pairs of movable contacts, with the stationary contact also split into two. The triple and quadruple pole switches work this way, as do the double-pole alternate action switches which reserve one side for the latching wheel.
In addition to being round mount, these have silver prisms for higher electrical capacity, the first time this has been observed in M7. M76-0899 is a very interesting model, that is most likely a customer-specific part.
Monday, 15th June
UncleFan found a store in Lebanon—Incotel Electronics—selling a couple of Cherry types: M73-0800 and M91A-11NN. On further inspection I also turned up M76-0899 and some RAFI and Siemens models. The nature of M76-0899 can’t be confirmed until it arrives, but based on the data for M76-0890 (a catalogue part in 1982), it should be a 1A+1B alternate action switch.
M91 however was an interesting find, because until now I believed that M91 did not exist. In fact, the part number chart for Cherry M9 has a gap where M91 could be listed. The way that the lines are drawn suggests that the M91 entry was not removed, and some of the boxes of M91 switches at Incotel were made in 1986, after the chart was last revised. This would suggest that M91 was considered, and nobody ever filled that box in after it was made available.
Incotel seem to have a good supply of M91A-11NN (with clear stems) in 800-piece boxes dating to 1985 and 1986, as well as an unspecified black stem version that may be M91A-41NN (as per the switches in the Ultratech Miniprint).
Monday, 18th May
The Hall effect keyboards is now updated to reflect a couple of discoveries about how Hall effect keyboards are designed, as well as additional details not previously covered.
As someone who does not understand much about electronics, and someone who does not have a large stockpile of keyboards to examine, these discoveries are interesting in that they make sense, but they are also presently impossible for me to prove. It’s a strange experience documenting a subject that you don’t understand, but hopefully the result is that more people will have a much better idea about Hall effect keyboards than they did before, a subject that remains very poorly covered.
Similarly, the encoding and output page has seen numerous updates recently.
Sunday, 3rd May
There is a particular keycap family that has for years resisted identification. This family was used on most BBC Microcomputer keyboards (those made by Wong’s and PED, but not those from the SMK-made keyboards). An almost identical type can also be found on at least one SAGEM keyboard, seemingly from a TX-20. I concluded some years ago that the BBC Micro keycaps always seemed to be Comptec, but Signature Plastics never recognised them. Nonetheless, they have distinctly Comptec characteristics, that do not fit with any other manufacturer’s product range. This includes the internal step that is present in each of SA, DSA and DCS.
Recently I have dug into this again. As Acorn has closed down, all of their records are gone. Chris Turner, Acorn’s chief engineer at the time, reports that the BBC Micro keyboards arrived from the manufacturers fully-assembled, and it seems that Acorn did not choose the keycap manufacturer. I had already encountered an excerpt from a DRAM Electronics Ltd catalogue—listing BBC Micro spares—where one of the keycap manufacturers was given as “Comtec”, which I took to be Comptec spelt incorrectly somewhere (perhaps even by Acorn themselves). DSA family is structurally very similar, with the same thin walls and the same low internal step, and as a DIN-compliant family it likely goes back to the early-to-mid 1980s. This means that the reduced-plastic design seen with modern Signature Plastic keycaps could be as old as the 1980s. The BBC Micro keycaps tend to have writing on the inside where Comptec placed it, although unlike DSA and some old DCS, it seems to simply give the mould cavity numbers or some other internal code. (Comptec had more than one factory and more than one mould design per family, so these characteristics vary.)
For the mystery keycaps, we can rule out SA, as they are clearly not SA family: SA uses much thicker plastic and a much larger internal step. They are also not any of Comptec’s DIN-compliant families (DSS, DSA and DCS) due to their height. The Comptec-like BBC Micro keycaps are all stepped profile, and I remembered that there was one more family that has long since disappeared: SS. Since SS was sculptured, a row from this could be used to make a stepped profile. Thus, these mystery keycaps are likely to be Comptec SS family.
A keycap profile demonstration props topic at Deskthority from 2015 contains an unattributed chart of DSS, SA and SS. The middle row profile from SS is precisely the same shape as the BBC Micro keycaps. That row is 0.410″ tall. Signature Plastics did not state precisely where on the keycap that height is measured, but assuming it’s the centre over the base, the BBC Micro keycaps are 0.419″ tall at that point, which is very close. This chart seems to verify that the Wong’s and PED keycaps on the BBC Micro were Comptec SS family.
The SAGEM keycaps are flat (“all rows”) profile and the same height as shown for SA on the same chart, which suggests that SS also had a flat profile option, similar to SA but made using the revised reduced-plastic moulding.
Having suggested this theory to both Signature Plastics and a former Comptec employee with whom I was corresponding, neither have responded, but at this point it seems likely.
(While gathering more data on these keycaps I also talked to Professor Steve Furber about the implementation of the BBC Micro computer keyboard, which I have since documented in more detail on the encoding and output page.)
Sunday, 19th April
A new page is up on Hall effect keyboards. It is comparatively brief, but it provides what could be considered a few interesting details.
There is also now a page that collects together the details on TESLA Hall_ICs. They offered two types: full-height (used with switches based on Micro Switch SW) and low-profile (used with switches based on Micro Switch SD). Each size came in a level type and a pulse type, for a total of four models. Unfortunately it seems that TESLA declined to write anything about them, so the behaviour has to be determined from the specifications and circuit diagrams.
While looking for photos of the SD clone switches with TESLA switches, something else interesting turned up. Buried deep in the Redmaus Megasale topic is a Zentec Zephyr DD/ID220 keyboard with switches that appear to be a plate-mount version of Stackpole’s grid switches. While Stackpole interlocking was discrete but still PCB mount, these seem to be discrete plate-mount units. This is a new discovery that went unnoticed last year.
Thursday, 16th April
In August last year I bought some Micro Switch 1PB871 switches. I included a 1PB2001 in the order purely for the sake of illustrating the Micro Switch PB page. 1PB2001 is a combination of an SM basic switch (“microswitch”) and a 2ED “untimed pulse contact buffer circuit”. The latter is a special circuit that provides debouncing, so that the switch unit can drive logic circuits directly. I never paid it that much attention until I started looking into sensing and encoding in more detail. The 1966 Bulletin ED contains the circuit diagram and, after some pondering, I can see what they did. (Those who are more electronics-inclined would find it obvious.) Micro Switch took advantage of the presence of both normally-open (NO) and normally closed (NC) terminals on a basic switch, to drive a flip-flop. Operating the switch activates the flip-flop, which remains active until the switch is released, and the normally-closed current clears it. This way, the bounce period is ignored as the flip-flop remains active throughout. This principle can be extended to any switch series with NO-NC as a supported configuration, including Cherry keyboard switches from the 70s.
Additionally, I found a brief PDF on datasheetarchive.com that finally confirms that Fujitsu FES-8 and FES-9 effectively differ only in profile. Aside from a brief overview of the family characteristics and some hard-to-see photographs, that is all the detail provided.
Monday, 13th April
Having finally read through the relevant patent, I now understand how the Mechanical Enterprises Mercutronic coding switch works. (This is the full version of the simpler non-coding mercury type mentioned at Deskthority as an unidentified product.) A detailed diagram of such a switch was posted to Twitter last year by Tube Time, taken from the December 1969 issue of Electronic Engineer magazine (volume 28, no. 12) With luck, it will be possible to get the whole article scanned in by the Linda Hall Library later this year (as and when the staff are able to return to work on premise). When drawing an example diagram of two coding switches I realised that there was a major problem with the design, and that diagram cleared up a point of confusion in the patent about how Mechanical Enterprises addressed it.
Have you ever looked at late 60s and early 70s keyboards and wondered what their large arrangements of diodes are for? These diode matrices are now covered in the first part of the new encoding and output article along with information on self-encoding switches, which made a brief appearance in the late 1960s before disappearing. Only late 60s and early 70s encoding is covered at the moment, which is the important part that I have been keen to document for some time. It would be nice to also document Datanetics’ encoding system, but Meryl Miller appears not to remember how it worked, and it does not appear to have been patented (I only learnt of it from his personal notes). That approach, too, seemed to soon vanish.
To support the above article, there is now a page on Micro Switch KB encoding switches. Additional photographs are due at some point in the future.
Friday, 27th March
Although a few Raytheon keyboards are known, it appears that there is only one discovered example made by Raytheon themselves, using reed switches (the others were made by Cortron). Meryl Miller collected a sample switch at the end of the 70s or early 1980s that appears to be Raytheon — I could have had it and would have had it if I had realised it was a keyboard switch (likewise his Fujitsu FES-2 and Micro Switch 1PB870 and KB types).
A new Raytheon page collects together what little is known about Raytheon’s keyboards and switches, including a single photo of Meryl’s type and a diagram showing how it appears to work, based on his photograph and a patent for an older design.
Thursday, 12th March
At the end of the 1960s and the start of the 1970s, manufacturers were finding a surprising number of ways to design switches that could directly address logic circuitry without contact bounce. The earliest known dedicated bounce-free keyboard switch type was Micro Switch SW using Hall effect, introduced at the end of 1968. Licon put out Series 550 ferrite core (“magnetic valve”), that goes back at least as far as the end of 1970. Perhaps the strangest was Mechanical Enterprises Mercutronic, a family of designs that used mercury-containing tubes for switch contacts, available for keyboard applications from around 1970.
RAFI used Hall effect from around 1973, in the form of RC 72 series. This was not their original contactless switch: an older design is known from patents, using magnetoresistance, that has yet to be encountered. Fortunately RAFI still have a 1972 catalogue documenting these switches. RAFI magnetoresistive switches appear to have no series name, only part numbers; they are described simply as “contactless”.
These may well not be RAFI’s oldest keyboard switches, but this is the oldest catalogue that RAFI were able to find, and both this and the RC 72 catalogue defied the odds against being recovered.
Wednesday, 11th March
RAFI have now scanned in for me the complete RAFI-System 72 (RC 72) catalogue. The catalogue was printed in 1983, which explains why the Hall sensor characteristics relate to later Siemens models (models seemingly not in existence when the switch series was introduced). The catalogue includes a diagram of the internal structure of both the Hall and mechanical types, and from this it is clear that this is indeed the annular-magnet type from the patent filed in 1972, which ties in with RAFI’s unclear indication that these switches were introduced in 1973. Jacob Alexander’s example will thus be a redesigned version. Two more switch types can now be seen: RC 72 N mains power switches, and keylock switches.
All we need now is details of RAFI’s original switches, which appear to have been magnetoresistive.
Friday, 6th March
Fujitsu FES-1/FES-2/FES-3 now has a page of its own, on which the original specifications and part number format is now documented, as per the Fujitsu magazine scans previously provided by Kiyoto. Curiously these are in a totally different format to the numbers found stamped on FES-1, FES-8 and FES-9 reed switches.
Tuesday, 25th February
Research has already shown that Micro Switch SW began life with an older plunger design moulded from grey plastic. These older switches can be seen in advertisements, but they have not been otherwise encountered. All discovered SW Series switches had model numbers from 1SW11 upwards; however, model number 1SW1-R can be found on various part listing sites, albeit without any photographs or other evidence. It stood to reason that 1SW11 upwards was the circa 1970 redesign (with black plungers) and 1SW1 upwards was the original 1968 design (with grey plungers).
Finally, visual proof of early 1SW1 has arrived, in the form of a “3D model” of an unidentifiable SW Series keyboard, manufactured around 1969 judging by the dates on the switches. The switches are (surprisingly, for a keyboard that old) marked with their part numbers, with all visible markings being of model 1SW1 from week 26 of 1969. This is the oldest SW Series keyboard found to date. The model label is present but the stamped inscription is either absent or missing from the model; the only visible identifying code is that of the PCB, catalogue listing SW-10033.
This is not proof that 1SW1 was the older-style switch; photos of the next-oldest known keyboard do not appear to show whether those switches were stamped or not, and thus we cannot determine yet whether this is when the part numbers changed. The part number change could have instead corresponded to a change of some other characteristic, such as a redesign of the Hall sensor. As always, more evidence is still required.
Monday, 10th February
In October 2019, Deepak Kandepet introduced me to an unknown reed switch type, depicted as switch 2 in a gallery of then-unidentified switches. This type is always marked “ПКМ 1Б” (“PKM 1B”). Some investigation found that some were also marked with a logo that somewhat resembled a smiling face. Later, I came across the Russian Virtual Computer Museum. Armed with a freshly-redrawn logo derived from its brief appearance in Chyros’s TC7063.02.A003.01 Soviet keyboard review (the logo shown in the video is depicted more precisely than on the bottoms of the switches), the museum staff were able to track down the logo to Ukrainian firm Magnit (Електромеханічний завод «Магніт») in Kaniv. Formerly part of the USSR Ministry of Radio Industry, they are now a separate business.
It is possible that they may still remember these switchces, but at present they cannot be reached as their mailbox is full!
There are still many more unidentifiable Soviet switch types, but now one of them is at least partially understood.
Sunday, 2nd February
The history of German Hall ICs is now both clearer, and less clear, after Wolfgang Zucker provided some catalogue entries for Siemens Hall ICs from 1977–1985. This in turn made it possible to locate the origin of the PDF previously used as a reference. It seems that some models were quite short-lived, being in production only during the early 1980s. This might be a clue as to why, as early as 1986, RAFI were using HFO Hall ICs. We now also have more details on SAS 201–221, albeit not the supply voltage ranges that could be tied up to RAFI’s RC 72 specifications.
Sadly, still no word from anyone at RAFI at all.
Tuesday, 28th January
In 2016, OleVoip at Deskthority posted a photograph of a lone RAFI RS 76 C switch. The Hall IC is marked only with an orange stripe (which he claimed “marks the border between the rest position and the activated position”) and the plunger is uncharacteristically natural plastic colour.
The catalogue pages for Siemens Hall ICs indicate that SAS 251 S5 is marked solely with “orange”, suggesting that this switch has a Siemens IC, and thus that the reason for the different plunger colour is that it offers dual open collector outputs instead of single open collector with enable input. There are certainly sufficient RS 76 C part numbers unaccounted for (obsolete before 2001 and removed from the catalogue) to allow for this, and we know that RC 72 offered a variety of Hall sensors.
From discussion of HFO and Siemens Hall ICs over at the Robotrontechnik-Forum, it seems certain that this is a Siemens IC, based on the plastic texture, terminal plating (silver plated instead of tinned) and housing shape (the latter remaining unexplained). However, it is always good to be as certain as possible. Wolfram Zucker in Germany has a collection of Siemens Hall ICs for sale and he has most kindly added photos of them to his analog ICs page. Most are marked in full or entirely blank (the latter indicating defective examples), but some SAS 261 chips are indeed marked with nothing but a blue stripe, and blue is the colour associated with that model.
Siemens practices remain a mystery (so far, they have declined to respond to questions) but at least it now appears certain that OleVoip’s single loose RS 76 C switch has a SAS 251 S5 Hall sensor from Siemens.
Thursday, 16th January
With the help of Holm Tiffe and the Robotrontechnik-Forum, we have made some progress understanding the Hall IC scene of the past. It seemed unlikely that VEB HFO B 461 G was around at the introduction of RAFI RS 74 C, and this was compounded by RAFI RC 72 C offering a variety of outputs (including separate level and pulse types) beyond what HFO are known to have produced.
According to the Ältester B461G? (oldest B 461 G) topic at the Robotrontechnik-Forum, B 461 G is known to go back to 1981, which corresponds with the age of TGL 38658, from June 1982. This indicates that RAFI must have been using an older product range, and this is reported to be Siemens Hall ICs. Unsurprisingly, the Siemens data matches up exactly with the RC 72 C data.
Wednesday, 15th January
At any one time, I have a considerable amount of both businesses and individuals from which I am still waiting for a response to a question or enquiry, going back days, weeks, months, or in some cases, years. Every so often, I remember or encounter some of them and give them another nudge.
The other day, I nudged RAFI again, in the hope that they might recognise the high-profile Hall effect switches found in Jacob Alexander’s Express-2 keyboard from the mid-80s. This time, I did get a reply: they are RC 72 C and RC 72 L. RC 72 is a series of high-profile Hall effect and mechanical switches. These series numbers seem to correspond with the year of introduction: RS 74 ca. 1975, and RS 76 in 1976. Tentatively these seem to be from 1973, making them slightly older than the reduced-size switches for which RAFI is better known. RC 72 is reported by RAFI to have been discontinued in 1993. Curiously, the part numbers slot in directly behind those of RS 74, which come directly before those of RS 76.
Some basic details are up; more details will come as time permits, including a complete list of part numbers. RAFI provided a single catalogue page; if they can provide further pages from it (including the extra details on RC 72), I will post those collectively, otherwise I will upload the single page available, which is just a parts list with no specifications.
Saturday, 4th January
José Soltren (XMIT) of XMIT Keyboards and now Metadot (Das Keyboard) has produced a set of photos of Das Keyboard Gamma Zulu switches (now proven to be part of Omron B3K series), revealing the part number to be B3K-T135M. The final “M” appears to indicate “Metadot” as suspected. The “35” may indicate that these have 3.5 mm travel instead of Logitech’s 3 and 3.2 mm.
This only leaves finding the part number of Creative PRES, the third customer product based on B3K.
Wednesday, 1st January2019 knowledge round-up
2019 has been another fantastic year for keyboard knowledge, so much so that it seems unlikely that 2020 will beat it. Two major successes were Micro Switch and Fujitsu. Micro Switch has seen by far the biggest gains in knowledge, thanks to charts and literature from Honeywell Sensing and Internet of Things, National Museum of American History Library and the Computer History Museum:
- “Micro Switch magnetic reed” is finally identified as the reed subset of Micro Switch KB, a series primarily advertised for its mechanical switches, with lots of data on both the keyboard and non-keyboard types now available (thus confirming that calling it “RW” would have been incorrect; RW keyboards use KB reed switches).
- There is now a wealth of information on SW Series (“First Generation Dual Magnet Honeywell Hall Effect”), including full disassembly of the newly-discovered double action types. The keystem colour coding appears to be largely explained.
- Likewise, we now know far more about SD Series (“Honeywell Hall Effect”), including how the tactile type works (using a sprung pin design, not a leaf spring).
- SN Series is now understood to be a companion series to SW, offering different assembly options to SW.
- The symbol codes on the KB reed switches appear to indicate the manufacture date of the internal reed capsule, separate from the manufacture date of the complete switch.
- FES-1, FES-2 and FES-3 comprise the original Fujitsu reed switch family, found listed for sale by UncleFan in model N860-1131-T010 (we guessed that these were called FES-1 from the keyboard model number, but Fujitsu keyboard numbering is far from straightforward!)
- FES-5 introduced tactile feedback to the reed switches; these were introduced in the early-to-mid 1970s and have yet to be seen in the wild. They were a whole new design that seemed to use magnetic separation.
- FES-8 is the familiar Fujitsu reed switch design, in its original full-height form; these have optional tactility and may instead be Hall effect for when you need to drive logic circuitry and cannot tolerate contact bounce.
- FES-9 is a reduced-height version of FES-8, believed to also offer tactile feedback; details on FES-9 are scarce.
- FES-4 is the “cross reed” low-profile type, possibly introduced for DIN compliance.
- FES-360 is the fully discrete leaf spring switch, seemingly adapted from FES-300 sheet key and FES-301 semi-discrete leaf spring keyboards. (The idea that FES-360 is “first generation” leaf spring seems unlikely.)
- FES-370 is the rare “lever action” type, conceivably a DIN-compliant replacement to FES-360, but hard to tell when you can only read a few words of Japanese.
Numerous Mechanical Enterprises types have been discovered (corresponding to unmatched patents), although most are yet to be observed:
- LFW is a mechanical type, with “gold V-bar” contacts; these may use a similar shell to T-5 and M-5.
- LM is a low-profile “gold V-bar” type.
- Mercutronic is a whole family of mercury-contact switches, introduced to avoid contact bounce; introduced around 1971, Mercutronic seems to be MEI’s answer to Micro Switch SW (Hall effect, 1968) and Licon Series 550 (ferrite core inductive, ca. 1970). M-5 series looks to be largely identical externally to T-5 mechanical switches, and shares the same actuator bar.
Some other discoveries made in 2019:
- The existence of SMK reed switches is now known (found by Deepak Kandepet); sadly no further details on these are recovered.
- Understanding of Alps series names and model numbers has increased significantly, including the revelation that the product codes appear to have changed twice: once from S-codes to K-codes, and then the K-codes split into SK (keyboard and Tact switches) and KF (full-size keyboards). We now know that (S)KCC green was KCC10903 before it became SKCCBK, corresponding nicely with KFL10903 (the (S)KFL equivalent), just leaving us to find its SCH part number from before that!
- Understanding of SMK series and part numbers is not as complete, but now there are lots of clues and more details by far than we had before.
- The Cherry reed switch patent is now confirmed to have gone into production, although Cherry’s reed switches seemed to have only been advertised briefly before disappearing. The models were 201-0100 (single pole) and 202-0100 (double pole). The corresponding mechanical types saw the second digit become “6”: 261-0100 (single pole) and 262-0100 (double pole). Shortly afterwards, the leading “2” changed to “M”, hence M61-0100 and M62-0100. This gives us the first hints as to why the mechanical types began at M6. Custom types (so it seems) were placed into M5, and then illuminated types were given M4. Over in Germany, their own production line took M7, and unlike the US, they placed the whole series under that designation, using the pole position to also indicate subfamily.
- Investigation into Licon, Cortron and ITW gives us some series name for their keyboards, in particular 550 for the original ferrite core keyboards and FC2500 for the DIN-compliant low-profile keyboards. Switches often have two-digit prefixes, suggesting that the switches fall under 55, 35 and 25 series just as the keyboards do. (45 was used for their CP-4550 Digitran-like leaf capacitive keyboards.)
- Digitran series are starting to make sense now.
- Maxi-Switch series are becoming clearer too.
- Controls Research Corp’s concentric spring switches are now identified as “Bi-Pac”, as noted in Electronic Design magazine in 1972.
- Clare-Pendar S820 is now known to have at least three different designs over its lifetime: one screw-fitted and two plate mount. The type from the 1986 catalogue is sold by Radwell at the time of writing.
- The orange RAFI RS 76 M diode switches are now tentatively identified as 3.13.005.002/0000, as this part number was found in a huge list of RAFI RS 74 M, 76 M and 76 C types, and corresponds with the “2” written on the switches themselves.
- Some details are now documented about RFT East German reed and Hall effect switches, although more work is in the pipeline.
- Alps SCK turned out to be a hysteresis design, seemingly a copy of Datanetics DC-50.
- The customer number on the Cherry M11-0101 drawing turns out to be the HP part number for the Datanetics DC-60 space bar switch, seemingly proving that M1 was created specifically to replicate DC-60.
- GRI KBR reed switches are now known to be Futaba MR, after it became clear that our understanding of Futaba 1st generation was flawed due to errors on the Korean sites used as reference material for series names. MR-6C is a reed type, and the standard height mechanical types that we put under that name, are just a variant of ML. The sealed type is probably MD as it seems that the two-letter prefix is a classification code, not a series name.
- There are “FD”-branded blue Alps switches!
- It now seems that Cherry had a separate US production line for MX switches, giving us three separate MX designs.
Additionally, I have now a huge list of Alps patents for anyone who can read Japanese, including several for “fat Alps”.
2019 has brought with it a couple of other changes. The site graphics are now all converted to vector, to be DPI-independent (and this includes MouseFan’s site). With the exception of the switch collection, the thumbnails are all rebuilt in higher resolution (generally 150–175% DPI, as a compromise). Although my photos range from passable to execrable in quality, the increased detail level of the thumbnails does make a real difference even on a screen at only 120 DPI (125%), with the added sharpness clearly evident.
Additionally, the presence of “.php” in URLs is no longer required. The old URLs will continue to function, but they will redirect to the cleaner URLs.