In, out and round about.

In for a penny.

Okay, so I've got this tape deck and radio combo I have to occasionally repair, because it tends to uhm, not go right. Anybody would think I'd just go out there and get a new one, junk the old one, you know. Well, One - I don't have any money, so two, I like to repair my own things where I can. Saves ME money, saves other people's time telling me they can't fix it because they no longer have the repair manual, yada yade yada, gives me the satisfaction of repairing it, and so on. Only ... somewhere in the back of my mind is the nagging thought of "what happens if I can't repair it this time?", and then I get a little worried.

I thought "Right, I've had enough"—I feel like I can't be bothered with having it in the state it was in; namely, not working. Now "not working" to me, means not performing as it was originally built to - in this case, the tape mechanism wasn't playing... something was jammed. I wasn't sure whether it was the transport mechanism or not, though I'd almost ruled out any of the circuitry beyond the playback/record head, and settled on the problem being almost purely mechanical. Those sorts of problems, I can fix. I'm no electronics whiz, can't measure a transistor's hFE without hauling it out of circuit to do so, and besides which, mechanical problems are a bit more amenable. At least, that's what I was hoping the first time I headed inside the case.

Okay. Speakers unplugged, unconnected and hauled off. Then, it's on to the screws. First five screws unscrewed, off comes the back.. hang on, no it doesn't, there's something else holding it on. Damn, where is that fifth scre... ah, there it is, on the base, along with the other six screws I have to probably undo, just to get the sides off. Off comes that screw, the other six screws, all neatly laid out in order so I can put them back on when I've fixed the problem in seven minutes, or so. Then off comes the back, gently, while I figure out which of the three leads have to come off... in this case, I decide, all three can come off. Gently does it, that's AC Flex I'm unbending there. Right. That's the back done. Now, how does... ahh, that's right - once the back comes off, the sides will come off in one assembly. Ease that dial thingymagig off, I'll need that later for the pot on the fine tune control. The buttons will stay in place - they're nice like that. One was missing long before I got the tape deck, so I don't worry about stereo/mono switching any more, though I do toggle the switch back to stereo while I notice it.

I reach inwards to see if I can twirl the motor around, the usual fix for this particular problem - oh, I can't get to it. Now, how the heck did I do this last time, twenty two months ago? (Or was it longer? I can't remember) Ahh, that's right—there's a central screw tying the front to the circuit boards, just in behind the transformer cables from the diode bridge. I reach for that screw, have it out in a matter of seconds. I'm doing well now, I can take the front off, and finally look at the little flippy thing that turns the head around for playing the tape in the right direction. I figure it's sort of up, but not really, and reach for the wheel in the back to twirl it around again, only I still can't get to it. I'd need fingers like spiders legs with the strength of cabling to deal with that... right. Time to take the whole tape transport system out - luckily it's all on one lump of metal, secured by four easy-to-reach screws. I have them off, also in a matter of seconds, also laid out neatly so I can put them all back in the order I took them out in. You can't be too careful when you're doing this job - no sense in having three more screws left over when you've finished than when you've started. It's off with the rubber band for the digits capstan, that always comes off anyhow, so it might as well come off properly now.

I finally get my fingers on to the wheel concerned, spin it around some until I hear a satisfying click from the mechanism, don't bother to check the head's flipped around, as that's what usually happens anyhow, right? Everything looks good, so I put mechanism back, screw the four screws back into place, and even remember to put the capstan band back on. I'm hot to trot as I put the front back on, uhm, wait a minute, something's not fitting. Oh, that's right. The spike for the direction-switch mechanism popped out from underneath the screw that was holding it down so the tape mechanism would go around to begin with. This is beginning to be a little unfunny. Undo the teeny tiny little screw (I've had to find a third screwdriver to take this one off; as it's so tiny, none of the other screwdrivers have heads that little), and set it aside, with the teeny tiny screwdriver head pointing at it so I can find it again. Put the front on again, good, it all fits.

Put in the little screw, then the sides, then put the back on... oh, hang on, I've got to connect those three leads back so they don't short to anything inconvenient, like the hot side of an AC cord, for example. This puppy's not earthed, so being careful of that fact will save someone's life. That's Batt+, Batt -, and ... hang on, this white wire doesn't go anywhere. Oh, that's right, it's the aerial plug. Right. Pull it over to the obvious contact. I can stick the actual telescopic aerial back on it when I get it back into the bedroom. But for now, I'll stretch.... hm. Won't come that far.. Weird. Hang on, pull the back a bit further into the case so the wire will stretch to the contact. Right. That's come far enough now, let's plug it on and have out of.... wait a minute, it doesn't fit. Am I sure, or am I only imagining it? After another couple of minutes of fruitless poking and prodding, I give into the realisation that no, the aerial contact isn't that one. Where the heck is the aerial connector? It surely can't go onto the tuning capacitor, those are NEVER used as aerial inputs except for DC (direct conversion) sets and crystal sets and the like.

Ah, it's up over in the corner, right next to the AM rod. Figures. Right. Let's get my fat little fingers in there—no, they're not actually fat, but they certainly feel like it in this job. Can't quite reach down into the case to get the connector onto the spike. Rats. I know this came off, so it can jolly well go back on! I need more room to work in. Ah, that's right, I just put the sides on, let's see if moving those aside does give me any more room. Reach over to the sides, pull them off in one piece, set them aside. Yup, now there's plenty of room to put everything back. Cool! Put the recalcitrant aerial connector back on, fit the Batt + and the Batt -, and get ready to put the sides back on. Grab the sides, slide them back into place, put back the six screws I pulled out from the base earlier, and eventually reach for the back to put that back into place. I'm almost done now. Reach for the five screws, put them back into place, and tighten them down. We'll test it now.

Out for a pound?

Plug into wall, find tape (Jean Michel Jarre's "Concerts Lyon/Houston") and close door. Press power. Nothing. Flick power switch—handy, that. No lights. Oh, hang on - I can't hear anything without earphones. It's time to grab those from beside the mouse hub, uncoil them and insert relevant plug into relevant hole. Ah, that's right. This tape deck needs that lever depressed so it can play tapes. Depress lever some and screw the tiny little screw back into place. Press power again... hm. No satisfying "Clunk" as I hit power. I flick the switch, only to realise I've flicked it already. I toggle the power switch—nothing. Darn. Here I was thinking this job was going to be simple. Turns out it just got a bit more complicated, and I have to dive back inside to figure out what else is wrong. I've never had this problem with this tape deck before, guess there's a first for everything.

Back into the case, undo the five screws—hang on, gotta power off and unplug. No point in frying myself—this is an unearthed case, remember. Undo five screws; can I do this job without having to remove the sides? I seem to remember that if I don't have to remove the sides, that saves me about six screws, and a bit of work. Take off the back, remove the Batt and aerial lead, set aside screws in the usual order. Reach for the bottom of the circuit board; darn, gotta take off the base anyhow. Guess the sides are coming off after all. Take the screws out, take the bottom and sides off, have a quiet look around to see what I can see on the bottom (copperised tracks, but still somewhat conducting) or the top. Nothing's obviously wrong, so I figure I must have jarred something loose. I put back the sides, the back, plug back into the wall, just to check out. Hm. Nada. Obviously not connected yet. Time to remove the back and the sides again, and have more of a nosey around inside. Hm, better check the fuse while I'm here, I guess. It looks a bit coppered up too, come to think of it. Pull it out, looks good to me, put it back in. Plug into the wall, flick switch, nope. No juice. Unplug, scratch my head and wonder what the heck happened. Pull meter down from bench, set to AC, 1000V, put neg probe on the case of the transformer, the pos probe on somewhere else hopeful. 4V. Nothing really useful. Try DC volts instead ... 4V. Well, at least it's consistently f00kd. Scratch my head some more, flick the power switch off again, look at the back, flick the voltage change switch a couple of times, figuring anything'll work. Plug back in, and flick the power switch. Hey! There's lights this time! Wahey!!! I think my job's almost done.

Now I've sorted that out, I can put the case back together, so I do that, remembering this time to put the back on first so I can connect the aerial lead, then I can put the sides on, then put the back on properly. Plug back into wall, and I decide I'd better run one final test to make sure everything's all right. Maybe it was just the power glitch that caused this. Flick switch, satisfying lights glow. Reach over for the earphones, get them plugged in—radio goes. That's at least good. Flick the switch back over to tape, press play. Uh oh, nothing happens. Oh, that's right - this wants its lever depress... hang on, it's already depressed from when I screwed in that little screw again. Oh, great. so the original problem is still not fixed.

sigh

Undo the first five screws (again), remove back (again), remove only the four screws that actually hold the sides on, because now the base can stay on as I don't need underneath it. Remove the central screw inside, as I need back into the tape mechanism again. Pop off the front, taking that little screw out too... yup. gotta point a little screwdriver at it so I don't lose it. Off with the capstan band, off with the four screws that hold the tape mechanism in place, and I ease it out to have a further look at it. I ease out the tape from the place where it's sitting, press the play head, and it doesn't move. So I press play - the play button depresses and locks satisfyingly. But the head doesn't actuate. Weird. I jiggle it a little, and it flicks, and pops back down into place. Ah, I see. Now it should actuate. Press play, nope. Doesn't actuate. Strange. Jiggle around some more, then realise I have to flip the head around manually anyhow. Do that, then press play. Clunk. The head actuates, finally. Good. Press stop, the head drops back down. I think I'm done here, so I'll wrap up. Put the mechanism back into place, not quite as easily this time, as I have to contend with the base being in place instead of neatly out of the way. Still, I manage it, and put the front back on, remembering to put the little screw back into place to lock that lever down. Put the central screw in to hold the front on, then put the sides on, put the back on, oh, hang on - have to put the aerial back on, so it's off with the sides, on with the aerial connector, then on with the sides, then on with the back. Put a tape into the mechanism, plug in the earphones, plug in the plug into the wall, flick switch, flick power on. Press play, ahh. NOW it works. I even hear sound, though it's .... hang on, it's backwards. Whoops.

Press stop, off with the power, unplug, five screws, four screws, central screw, one tiny screw, front. Remove tape, fiddle with head, flip it around. Tiny front screw back in, central screw, etc etc... only this time I don't put the back on. Just put the tape back in, press play. Yay, now I get sound; though it's a bit muffled, it is recognisable and playing in the right direction. Cool! While the music's playing, I put the sides on, aerial contact didn't come off this time, so it's the back on after the battery contacts go back into place. Plug the power in - yup, still goes, still plays the tape. Good. I'm done.

And then I realise the capstan band is still out on the floor.

Expletive deleted by request of owner

Suffice it to say, that after the job was finished (and after a suitable couple of deliberate drops onto the floor)—yes, that did fix the audio, not muffled now), I'd spent nearly three hours on this simple little tape deck. It now works, thankfully, which means I can plug my MP3 player into it and hear music through the speakers. But boy, was this an adventure I don't wish to repeat.

All at sixes and sevens

Oh no! Not more cubes!

"What? More cubes?" I hear you say. Yes, though I'm behind the times in telling you this, further advancements have been made in developing larger Rubik's Cubes than just The Professor. When I last wrote about the Professor, I was aware that working models of the 6x6x6 and 7x7x7 cubes had already been demonstrated, they were only waiting for the manufacturing process. The developer of this style of cube has said that methods to produce cubes up to 11x11x11 currently exist, they merely have to bring them to market. The general assumption is given that this method could actually be extended further, but I don't think I'll be paying for a 13x13x13 any time soon, as the price points for the 6x6x6 and 7x7x7 cubes are between US$46 and US$55.

So what? I couldn't even solve the original!

Don't worry, all is not lost - there are solution pages for these new cubes already, provided by the company, as well as other pages by www.bigcubes.com; other solution pages already exist for even the little 2x2x2 version, though if you know how to solve the corners of any cube, you already know how to solve the little cube. All is not lost. However, in solving the big 7x7x7 cube, there's one important difference between it and the 3x3x3 and Professor. The centres can move in relation to each other, meaning you have to make sure you put the centres into the correct orientation first, as shown by the eight corner cubies.

Why's the big one curved?

That's right, it's not quite a cube any more, as the picture in the Wikipedia article shows. Because of the size of the respective pieces, and indirectly the method of manufacturing, when a face is turned around 45 degrees, the possibility exists that a piece will simply fall out, as it is no longer securely held by adjacent pieces. The curve combats this effect, by overlapping the area occupied by the piece, and will probably be a "feature" of all the larger models. Even the 6x6x6 cube has been modified somewhat so that pieces don't fall out - the summary is that the corner rows are actually 2mm wider than the inner four rows, allowing things to actually work.

Mirror mirror

Of course, other twists exist on the cube, many of them playing on other aspects. This cube differs from others in that all faces have a mirror-like finish of a shiny reflective sticker on each cubie; in addition, each cubie is a unique size instead of the same size as their opposite face equivalents. This makes the cube look really weird when it's mixed up. Hey, at least you'll puzzle the heck out of everyone else when they try it out.

Of mice and batteries, and calculating.

Charge it!

No, that's not the plastic fantastic, which is probably just as well in the current financial situation. Instead, this is just a wee update to my post from April 6th, 2008 about batteries in my mouse, when I posted that Alkaline batteries seem to average in my mouse about seven months. Well, when I bought some NiMH rechargeable batteries a little while ago, I decided to give them a try, after previously having disappointing results with a couple of sets of NiCad rechargeable batteries (500 mA/h). I was charging the older ones up, and was lucky to get a week out of each pair. Anyhow, when I charged up the new batteries, I put them into the mouse on about the 14th of November. They ran for more than four weeks of use before changing them, and it's the 27th of December before I had to put another pair into the mouse. Frankly, I'm somewhat surprised, even though the batteries are only the "cheaper" 2000 mA/h capacity, and not the 2450/2700 mA/h batteries I really wanted. I also heard that someone wasn't too impressed with Eveready batteries, which were exactly the brand I'd bought. Never mind, I'll see how the batteries behave.

The batteries were a steal—at only $20 for four batteries, and with a normal 18 hour charger effectively thrown in for free, it was a better price than paying out $32.50 for one 18 hour charger and two batteries that I'd seen by Energizer; that wasn't including the $13.45 cost of two more batteries to make up the foursome. Other "cheap" batteries I'd seen were only $19.85 for four batteries, not including a charger. Again, that was for 2000 mA/h capacity, just a store brand for K-Mart. I'd hate to think what it would have cost for a fast charger—either a one hour charger or an eight minute one, apparently neither of which are any good for the batteries.

Hack it!

With regard to the previous sets of batteries, I at least was able to save the cost of the power for charging them by using two solar panel garden lights I'd found at the side of the road. The first one took one rechargeable AA battery, and could happily be left in the sun, to give me a charge over two days after having removed the LED. The second solar cell garden light was a bit interesting, as I had to hack it a bit to get a pair of AA batteries added beside the space for a single AAA battery. In essence, I hacked the battery cradle out of a dead TV remote, got out my soldering iron, and set to work melting plastic and rewiring the cradle, then mounting the cradle onto the back of the cell, opposite from the solar panel. In the end I did all this without having to strip the single AAA carrier out. Once that was done, I had charging for free, for three NiCad AA batteries and one AAA battery, all in the sun.

Unfortunately, given the restrictions for NiMH batteries, I can't use this same method for charging them, and instead have to use the provided charger, to make sure I don't overcharge the batteries by accident.

Hack it again!

While I was at it, I thought I'd convert a Casio FX-82TL (cheap scientific calculator used for secondary school) over to a FX-85TL/FX-300TL (effectively), though with a slight twist... instead of using the normal LR44 the FX-85 has inside, I'd retain the AA battery from the FX-82 (so I didn't have to wreck the case any more than necessary), and make it rechargeable via the solar panel instead. The wiring was rather simple, as all I needed was wire to the solar cell (which I'd already "filched" from another solar four-function calculator), and a diode to prevent the battery discharging through the solar cell again. So, first off, I soldered wire on to the contacts of the cell very carefully—copper pads on glass takes no time to heat up, even with a little soldering iron like I use. Then, I connected the solar cell in place of the battery, turned on the calculator, checked it works—yup, no trouble. Added in the diode (the right way around), checked that the calculator still works, which it does, though it now needs a bit more light due to the voltage drop across the diode. Finally, place the rechargeable battery in place, cover the solar cell so there's no light, and test the calculator one last time. Yup, all working.

So for a little bit of solder, a solar cell from a four-function calculator and a diode, I have a scientific calculator with what is effectively a new ID and a battery that'll last just about uhm, forever? I vaguely thought about the same trick on my FX-82MS too, but there isn't the same space for the solar cell, as the battery space takes up 90% of the space above the display. At least on the TL, I'd be able to cut out a slot and have space to spare, if I wished to do a tidy job, and not just a sellotape job over the top of the case.

I also wanted to try this trick with one of my other calculators, but I don't know enough people with calculators with solar cells on for sacrificing (I'd need at least three cells)... what a shame. Not only that, but two other aspects of this calculator lean me away from solarising it—first off, it uses alkaline batteries, though this isn't necessarily a bad thing. It affects the voltage I have to provide the calculator, of course. Secondly, the reason for the alkaline batteries: this calculator draws more current than the FX-82 I converted over. I'd also have to fit a special plug so I could detach the solar circuit from the calculator, which would remove its usefulness somewhat.

What I was thinking of initially would be: three little cells (or maybe even four) wired together in series to provide either six volts, or eight volts to allow for halfway usable voltage at low light conditions, connected via diode to charge the batteries. Again, being NiMH batteries, I'd have trouble making sure the batteries weren't overcharged, and I'd have to build the same sort of circuit into the charging as a normal charger has. I.E. a bit more sophisticated than two wires, one diode and a solar cell. So for the moment at least, that project will either never get off the ground (usual for me) or I'll actually have to put some cash into the project to ensure it doesn't look like an ugly hack. If I were to do justice to the job of course, I'd put the whole calculator onto a "charging" cradle like the Palm Pilots used to have, and simply solarise that instead, my only problem would be sourcing (or making) suitable cradle to take one FX-9750G+ on its end, with plug. Interesting, but probably will never be done. So, what other mean hacks have been done to the humble calculator?

Poor brother to the big king

The Beginning

Gawd, the fifth edit... mainly to stick back in all the pictures that disappeared from other places. Anyhow, on with the show. Whoops, make that the sixth edit to finally show a proper screen from the grocery program, more below. I have just added to my small collection of calculators, and thought I'd describe something about them in this post. Casio make a line of programmable graphical calculators: the FX-9750G, the CFX-9850G, the Plus versions of these, and so on. My latest acquisition is the FX-9750G+. It's just about the bottom end of their range of this sort of calculator, and doesn't have a three colour screen (orangey-red, green, and blue) like the big brother. But it'll do. Yes, there is a difference between this and the previous FX-9750G: this one has had some finance functions added, which I've already taken advantage of. This programmable graphical calculator might be the little brother of the family, but there's enough in here to keep me reasonably happy.

Anyhow, I thought I'd compare it to two other calculators that I have here - the Casio FX-82MS, and the Canon F-804P. Neither calculator is programmable, though the Canon can accept a small number of user-provided functions and store them for later recall. I'm deliberately not comparing it to the HP 50g, because there frankly isn't any real comparison that can be made. The Casio's not RPN, doesn't advertise 2,500 functions nor widespread usage among professionals. However, for what it is made for, it does seem to be a good performer.

First off, I'll introduce the FX-82MS, and here's a huge version, so you can count the pixels on the display (image originally copyrighted by unknown author, and shown on a website that no longer exists). Look at how crisp they look.

Umm, why did you buy them then?

Well, I bought the FX82MS as an upgrade to the FX-82TL (which I then gave away to a friend for his course work - who needs TWO FX-82 calculators?) because I loved the idea of the navigation keys on the MS, which the TL never had, and neither had any FX82 model previous to that. The extra 14 functions helped out too. Literally the only thing this calculator wouldn't do was base-N calculations, you had to buy at least the FX-100 for that. Which is where the Canon comes in. Just as an aside, I now have the FX-82TL back, as the previous owner hasn't used it since the coursework I gave it to him for.

I saw the Canon F-804P on a table in a store that was selling off all their little consumer electronics range, and changing the range of products they sold; as a result, I picked it up for only $20. The only reason I picked it up is because it appeared to do base-N calculations. It does, but more on that later. I most recently bought the FX-9750G+ because I wanted a programmable calculator; the fact that it is graphical is also a bonus. The calculator was also discounted by 25% - purely because it didn't have the outside box which I probably would have discarded anyhow. I wasn't going to complain at the price, or the discount, though I did try for more of a discount.

Do I have to have a degree to understand the Manual?

Studying the rather large 454-page manual of the FX-9750G+, I find a lot in here. We start off with a page informing us of the difference between the FX-9750G+ and the CFX-9850G+/9850GC+/9950G+ variants (the CFX has sort-of three colour screen, the FX doesn't). Then we follow with an insert describing how to put the batteries in, and what happens directly afterward. Directly following that, there's an additional 20 page Quick Start section. This covers just about everything the "I want to use it now" users probably ought to read, even if they don't read the rest of the manual. The rest of the subject matter is distributed among 22 chapters in the main bulk of the book, and is followed by Appendices and a not-very-complete index. My main complaint is that I find the depth of information provided in here tantalisingly superficial. All the examples they provide work as stated, though I'm having a little trouble understanding the financial section or repeating the same output for the same input; that could simply be Operator Error. When I say superficial, I mean superficial in a "Name-Your-Subject Unleashed" kind of way, as if they've covered everything they think the average user ought to know, but I don't find that it's detailed enough for me, especially in the programming area. I looked online for a manual describing the BASIC-like language that the calculator uses, but the only language I could find it in is French. Bleah.

In comparison, the FX-82MS manual is actually a fold-up booklet which covers the essential aspects of the calculator, but doesn't break a sweat doing it. If you know what you're looking for, you'll either find it or you won't - it's at least clear and the English is well written. Again, the examples work, but they don't spend much time on explaining them as it's reasonably obvious.

The Canon F-804P manual came from Japan, and it looks like it. Instead of being a large book, it's a small booklet that's been glued along the edge like a paperback. The language used in the English is certainly not the best used, featuring such phrases (apologies to the Canon people here, but this is meant as a criticism) as (for the ON key):
Turn on power, cleaning up / cleaning independent memory / cleaning statistic memory


• Push it down to open power source if the electric source is closed.
• Push it down to clean up the input characters and wrong identifiers.
• Push down [SHIFT] first and then this key to clean up the contents in the independent memory.
• Push down [ALPHA] first and then this key to clean up the statistic computational registers (statistic comprtional mode).

Eeeek. And yes, that spelling mistake is really in the manual. In the other half of the book we have the Japanese instructions - and they look far better written than the English. It's a pity I don't read Japanese, as they might explain some of the English instructions better.

The Interface

One of the more confusing aspects to the FX-9750G+ at least for me (being used to the FX-82MS, this threw me for a couple of seconds) is the menu that shows when the AC/On button is pressed. However, I simply hit the [EXE] button, and go straight into what looks like a knock-off of the HP-48G/GX/S/SX, except it isn't. Simple four-function calculations work the same as they always do, with ×,+,÷, and - all having their own keys. Instead of a = key (which is on another key) you use the [EXE] key, which makes sense from a programming sense, but not so much for a new-to-calculators user. At least it's in the same place as the = key on the FX-82 and many other models, at the bottom right. The two other calculators are dead simple: turn them on, start entering numbers. Aspects such as choosing Radians or fixing the number of digits of result displayed are easy if you read the manuals.

Command recall

One niggle I have about the FX-9750G+ and the Canon is the fact that the command history doesn't work quite the same as the FX-82MS; on that, you simply hit the up-arrow, and are presented with the previous command, and as you keep hitting the up arrow, you scroll through the last ten calculations issued. With the FX-9750G+ however, if you want anything but the previous command (reached by the left or right keys), first you have to hit [AC], then you can scroll through the remainder of the command history. That's described in the manual too, but isn't intuitive to me when compared to the FX-82MS. For the Canon, you can only recall the last command you typed by using the left or right keys. Deletions on the FX-9750G+ are handled differently too, more like a PC's Delete key, rather than the behaviour of the Backspace key that you find on both the FX-82 series and the Canon F804P. Basically, back up the cursor to the point you want to start deleting from, then hit the [DEL] key to delete characters.

The display

The screen size for the FX-9750G+ is a nice 128 pixels wide and 64 pixels high; the font size for the text character set is 7x5, giving a 7-line text display and leaving some extra room down the bottom for drawing labels. The text size for graphical-mode text is a little smaller, at 7x4 pixels. The size of the screen is readable, but not exactly large. My wife has a hard time reading something that small, but most people should at least be able to read it.

The FX-82MS has a two-line display, the top line is 12 characters, each of 6x5 pixels. The numbers are clear and easy to understand, and the bottom line holds the usual ten digits plus sign and 2 digits of exponent (plus sign) display of seven-segment, with the nice addition of comma separation for larger values. As a side note, this toggles between the usual American/English system of 123,456,321.45 and the Spanish system of 123.456.321,45 - useful, though not for me. There are several additional annunciators for Shift/Alpha/Hyp and lots of other symbols, though three of them aren't used on the FX-82MS. I think they used a display module from the more powerful calculators that support complex numbers, matrix calculations and vector mathematics. I note that the new FX-82ES (where do they think up these designators, anyhow?) has gone to a fully pixel-addressable screen too, though it's still not large when compared to the FX-9750G+, but it's large enough to display "proper" equations like you'd see in a textbook, which is exactly what they're marketing this function as. It's an expansion on the S-V.P.A.M. theory.

The Canon's screen has a two-line display that is similar to the FX-82MS though the top line is made up of 5x5 blocks instead, and the quality of the letter shapes suffer as a result. The bottom line shows the same ten digits and sign, plus 2 digits of exponent, and these numbers all use a seven-segment display, though I find the "shapes" not as nice as the Casio. It also has annunciators, but they're down the bottom of the display, and there aren't as many as are present on the FX-82MS.

The keypad

For the keys, I'd have to say that Casio did their job really well for both models of calculators I have. The legends are clear, and the alternative functions that show above some keys are also clear, with good contrast and visibility. Unfortunately, I can't say the same about the Canon. Why am I not using the Canon F-804P instead of the Casio FX82MS? Frankly, though the features of the Canon are nice, they're put into a case that's frankly not very well designed. The keyboard is cluttered, the legends above the keys require the user to have good lighting conditions merely to see the legends, and the keys just don't feel right, having a short travel. They're also small and close together, which doesn't exactly help. The fascia (the bit surrounding the keys) is shiny, which makes the act of reading the legends harder, and though the white text on each key is quite readable, the orange and pale blue colours are anything but readable against the reflection of the fascia if the average user has light behind him shining on the calculator.

That doesn't register

The FX-9750 family (including CFX models) gives the user 28 registers—the 26 letters, rho (r) and theta, for polar coordinates. If you know Greek or mathematics, you'll know these symbols. In addition to this, they support lists, matrices and collections of lists called files. Storage of up to twenty formulae, six pictures, six "lists" of numbers, multiplied by six "files" of lists, twenty six matrices and six graphs is also provided, making quite a collection even for a programmable graphics calculator.

The 82MS has nine registers (A-F, X, Y, and M) that can be used to store numbers in, these registers are also supported as sources for numbers in formulae that are typed in; i.e. AX^2+BX+C. The FX-82ES only has seven, though the FX-82ES Plus has the full 9.

The Canon has 26 letter registers, an angle register and a memory register that is separate from these, and the same support for using the registers in functions.

Functionality

The FX82MS has (according to its manual) 240 functions, including support for advanced statistics, regression formulas and combinatorial functions. Quite a lot for a cheap student calculator. The Canon has 145 functions, has the ability to store multiple user-provided formulas, and also has ten constants from the engineering/science field. In addition, the Canon has the ability to do base-N calculations for the binary, octal and hexadecimal bases. No floating point support, but then even the FX-9750G+ doesn't have support for base-N floating point, though it can convert integers between bases. The Casio website states that the 9750GA Plus has 900 functions, and also states that it's the same as the 9750g+. This may seem like a low number of functions but purely because of its programmability, you can quite often synthesise what you need from what you already have. I've managed to code a rather accurate "natural log of Gamma" function from someone else's description on a web page.

One other thing that really surprised me about the HP-34C (one of the first programmable calculators I owned) was what I later found out to be called the Gamma function. At the time, I called it factorials on non-integers, but the function that Hewlett chose to implement is actually Gamma(n+1). This gives the same value for normal factorials as the factorial equation, but takes a little longer to generate. None of the other calculators I've come across will do it, but that old calc will. It seems that some other calculators in the HP stable will also do it, but not all of them. I still rather wish I had that, but it's gone, long gone. I did manage to get a cheap HP-50G far later, but that's another tale.

Accuracy

Both the Casio calculators are accurate to ten decimal places on display; the FX-9750G+ is accurate to fifteen digits internally, in comparison to the FX82MS and Canons' 12 digits internally. However, I've found to my surprise, that the Canon is the most accurate calculator I currently own, bar none. I suspect that unlike most calculators I've used, which use a certain number of digits of precision, the Canon says it uses 12 digits of precision, but actually uses floating point at its core, and rounds down to 12/10 digits of precision when displaying the final result of a calculation. Take this as an example, which shows what I mean (use the symbol for PI, not the letters P and I): (PI-3.141592653)*10000000000 That's 1E10 for you engineer types. On the FX-82MS, I get a result of 5.9, on the FX-9750G Plus, I get 5.898, showing a greater degree of precision of 15 digits as opposed to 12. However, the Canon comes up with an answer of 5.897932385 so where'd those extra digits of precision come from? We've got rid of ten of them when we took away the original 3.141592653 (ten digits worth) but we're still left with ten digits, at least for simple single-stage calculations involving the constant PI. Extending the concept does eventually run into a barrier, at least for Pi, as I end up with: 3.14159265358979323846_198694562nnn... as compared with the real value of 3.14159 26535 89793 23846 26433 83279 50288 41971 69399 37510 (five digit spacing, fifty decimal digits.)

Other calculations appear to indicate that the Canon does indeed work with floating point internally, although I've found that I can't simply enter in as many digits of precision as I can, just to ramp up. The calculator will only take the first ten digits of input and discards the rest, though numbers generated internally retain their precision until display. So for a crappy-looking calculator, it's impressed me for precision. Another common test that can be done with any calculator that's scientific is called the "forensics" test. Here's the formula, and here are the results for my four little machines and the Palm SciCalc 2 program, with the extra internal digits added. Make sure you set each calculator to Degrees mode first, or else the results will look really naff.

(arcsin (arccos (arctan (tan (cos (sin (9) ))))))
Occasionally arcsin will show up as sin-1

	Displayed result	Internal result
FX-82MS	8.999998637	8.99999863704
HP 50G	8.99999864267	8.99999864267
FX-9750G+	9.000000007	9.00000000733343
FX-9750GII	8.999999998	8.99999999759468
Canon F-804P	9.000000000	9.0000000000000000000...
Palm SciCalc 2	8.9999999999999999	8.99999999999999999999999998482

It was this result, and the description in the Forensics table of the only calculators to pass the test, that showed what was happening. Calculators such as the HP-30S, Kinpo SG1, SR19S1 and SG2 were described as having the same floating-point chipset, the Sunplus SPLB30A.

To achieve the extra digits shown in the "Internal Result" column, I simply took off the Displayed Result from the Internal Result, multiplying by 1E10 to get the scale to fit, which is why you see far more digits displayed for the Canon. The same also appears to be true for this calculation: (sin 60 - 0.866025403) * 1E10

	Displayed result	Internal result
FX-82MS	7.85	7.85
FX-9750G+	7.84439	7.84439
HP 50G	7.84435	7.84435
Canon F-804P	7.844386467	7.8443864671232000709...
Palm SciCalc 2	7.8443864676372317(see next column)	7.84438646763723170752931984524

For the Canon (assuming I'm correct in my assumption) I have found I can simply keep on peeling off ten-digit slices of the result, and I've gone out to thirty digits of result so far.

In total contrast to the form factor and shoddy case of the Canon, we have the FX-9750G Plus, which comes in a fetching white and green plastic case. Say what you like about the colours, but at least the legends above the keys can be read, as they contrast with the white nicely—I suppose they could have used blue instead of gold, but I'm not complaining. I understand that other models come in slightly different colours and form factors. It's still heavier than the FX-82MS though, small wonder, as it takes four AAA cells in addition to the back up 3V lithium battery.

The Program

Previously to this, I was using the FX-82MS to motor around the supermarket, adding up what we'd bought, in about seven different categories. I came across a nice way of doing this: [ALPHA] A + ([ALPHA] X x 1.175) [SHIFT] [STO] A which looks like this, once it's in the top line of the 82MS: A+(Xx1.175)->A

Basically, that takes whatever's in register A, adds (X multiplied by 1.175) to it, and sticks the result back into register A. A variant on this removes the multiplication, and simply acts as an addition, storing the result back into the register:
[ALPHA] A + 2.25 [SHIFT][STO] A which, again, looks like this: A+2.25->A

I simply used the six letter registers for different categories of goods, used the X and Y registers for price-per-kilo costs, and combined with the previous program, substituting register letter as I went. Once I'd finished I could get a grand total simply by adding together the entries in the six registers, thus: A+B+C+D+E+F= Try doing that with the original FX-82 from years gone by. However, I didn't buy the FX-9750G+ just for the ability to do the same as the $20 FX-82MS. Instead I chose to do something else and actually write a program. Just as a note, Casio put out a number of different versions (upgrades) of the FX-82 since 1981 or so, including the FX-82A, FX-82B, FX-82C, 82D (which used fractions), 82L/LB, 82SX, 82MS and the FX-82TL. The Solar version of the 82 is the same as the FX-82D, or at least similar enough (i.e. it can also do calculations using fractions.)

The History

What inspired this program was the very ability to program that I'd first come across with a HP-34C I'd come across: one with a Red LED display, and nice chunky buttons with a good solid "click" to them I've not found on any other calculator since. This calculator had the ability to store a key sequence, and also had multiple registers I could stuff numbers into and retrieve from, in addition to a prompt for number, so it had nearly everything I needed, except for reliability. The calculator was getting long in the tooth when I got it, and it was showing it - with occasional times where I couldn't see the display, the Run/Stop switch wouldn't always make the calculator change modes, and the battery hatch was falling off. In the end, the calculator was retired into the bin, though I wish now I'd kept it. I'd managed to make a really simple version of the program that simply asked me for a number, then added it to the internal register I'd set aside, displayed the total, then asked me for a number again. Simple.

When I tried to do the same thing with the FX-82MS, I was missing the programmability, which means I couldn't get the calculator to ask me for a number, then add that number to the internal totals, then display the total, then go around the loop again; I had to compromise, which is where the above scriptlet came from. The saving grace, at least for the scriptlet, is that because of the command history recall, I didn't have to keep typing in the scriptlet from the beginning, I could simply recall it with the up-arrow or left-arrow, complete with assign to register, only changing the variable amount (for example, 2.25) each time.

In comparison, not only does the 9750 support asking me for a number, it also has the ability to display arbitrary text and graphics simultaneously anywhere (mostly) on the 128x64 pixel screen. This very feature is used by the calculator itself to present functions for the top row of keys, and shows what I'd call keylabels on the bottom of the screen. These keylabels are used as a menu with up to six choices, which can be completely tailored. I decided on this means, as the program should be intuitive to use for my wife - a set of abbreviations makes more sense than simply remembering that register A is used for fruit and veges, register B for meat, register C for cleaning stuff, etc. What I've ended up with looks a bit like this:

The World?

The programmability of this calculator is quite impressive, even with only 28,000 bytes available (or so). The frustrating bit of that impressive programmability is the fact that you have to press heaps of keys to get work done, purely because the alphabetical keys aren't separate from the main keyboard, they're integrated and reached with the use of the [ALPHA] prefix key. Casio must have thought about that, because they provided this calculator with what amounts to a serial port, although it's a non-standard one. It runs at 9,600 baud, so it's not fast, and it expects different stop bits for each side of the transmission—two for transmitting, three for receiving (I think). So, I could (theoretically) type up the average program on my PC, plug the cable (not a cheap purchase, it'd cost me more than the calculator itself cost) into the calculator and the PC, and pipe the data over in less time than it'd take to laboriously hammer keys to get the program entered in manually, correct the inevitable errors, and fine-tune the results. I personally would have loved a slot where I could add memory, but for the cost, Casio won't provide that. I already have one extra cost to outlay—a 3V lithium battery used for memory backup while the main batteries are being changed. I'm simply not flush enough to afford the $139 RRP (2008) for that cable in addition (not the SB-62, which is simply two stereo 2.5mm plugs wired crossover fashion, but rather the FA-123 serial interface or its big USB brother the FA-124.) If possible, I'm going to attempt to make my own, using the rather well-known MAX232, otherwise known at Dick Smiths as a IS232 (I think); some wire, and two plugs. I hope to make the cost all up to be only about $25. I can solder the project together myself, I just need the bits. Once the cable is assembled correctly, I'll be able to put nearly anything on the calculator that I can uhm, fit, and that the calculator understands. We shall see.

What the young can teach us.

Mirrored from The Flying Brick Commentary over on Microsoft's Spaces. Hey. It's a blog.

It seems strange, but even given the length of time I've been banging around computers, I still have plenty to learn. This fact got brought home to me when my sister in law rang me with a problem her (nearly) 3 year old son had instigated. The phone conversation began with:

My delightful son has hit a couple of keys on the keyboard, and managed to turn my screen upside down. How do I get it back the right way up?

I hadn't heard of this particular little wrinkle of Windows, though it made an insane sort of sense once I thought about it for more than two seconds. Of course, her son couldn't tell her which keys he'd managed to toggle, given he's barely saying "buhss" for bottle, and "pahpah" for brother (you have to think a bit for that one). So it was up to me, the resident techmonkey, and Google Search for Microsoft problem. Thankfully, the problem came up in the first couple of hits I looked at. You see, given the key combination Ctl-Alt-Up or Ctl-Alt-Down (I'm not sure which), Windows will then go and flip the screen that way up. I vaguely wonder if Ctl-Alt-Left and Ctl-Alt-Right rotate the screen? I ought to try it, just to see if my poor old Win XP Home System with a CRT can actually do such a thing. Last time I tried it under Linux, the system locked up and I had to reboot. Yeesh.

One of the hits states

This doesn't work with all displays—only with certain drivers that have this provision for rotation.

Obviously her display supported it. To the best of my sketchy knowledge, I had thought that only certain LCD screens and laptop screens actually supported such an idea, though it turns out that it's a function of the display driver, not really of what's connected to the video output connector. Her screen is most certainly a CRT, and wouldn't have supported it unless the driver was able to do the job.

So, I learned something new today, from a three year old!