by Daryl Bender - Ottawa Canada
Original site of the blue sky & cloud background
This page is about the design and construction of an unusual clock. I thought that I would finally document one of my projects and use the opportunity learn some HTML at the same time. Here is the original page that started it all. I just happened to bump into it while searching for something else and thought.."Whoaaa ... that is really cool!" I decided right then that I had to do it" :-)
Bob's Original Clock
I loved the look of numbers in thin air. Scouting further I took a particular fancy for the look of Victor Tikhonov's clock and the use of the long thin arm to really achieve that effect. I also liked the idea of seconds display so that when running the clock would be as animated as possible This meant the "noblink" program (see Victor's page below) was likely the one for me. Finally I also liked the idea of transistor LED drivers although subsequent thoughts on this are, given the PIC's drive capability and the lower maximum current on super bright LEDs, it may be redundant.
Victor Tikhonov's Clock (See also the link to his page below. More pictures of this clock can be seen near the bottom of his page)
First I had to develop an overall design. A fair search for VCR motors was unsuccessful. In the end I bought a surplus 12VDC (commutated) flange motor from a local Princess Auto store (#4210814 ~$20, see below). I also decided that if this was going to be a propeller clock then, hey, lets really go with the aircraft theme! It wasn't too long before I came up with my mechanical sketch. I always tend to do the CAD work well in advance of cutting any material as it not only identifies what is needed but usually flushes out the problems that would be encountered by proceeding blindly ahead. The mechanicals were then roughed out and some paper models of the planned board were tried for fit.
Since the motor I bought (above) was a flange motor I simply made some support columns which mounted to a steel baseplate, 6"dia x 1/2" thick. I thought this massive baseplate, perhaps a bit overkill, would act like an anvil-like anchor for any vibration (more on that later). I also made a flange cover so as to protect against the sharp sheet metal edge. This part (and the motor flange) is mounted to the top of the three 1/2" columns with some neat large head low profile 1/4"-20 Allen screws used for furniture. I decided to have a propeller spinner which would be used to hold the LED board arm and clamp the PCB at the same time. At the time the thought was to get the LEDs as high as possible to maximize the "lights in air look". This spinner turned out to be tricky to machine, run true and look nice. If I were to do it again I would have the lower flange hold the board and LED board arm, and have the spinner as cosmetic only. It only dawned on me afterwards that spinners could be obtained at any local hobby shop. The LED board arm is a piece of 5/32" drill rod with a small piece silver soldered to the end to hold the LED board with two 4-40 screws. I wanted drill rod for it's strength and stiffness. I did not want the arm to eventually look like a bent coat hanger.The board mount piece at the end of the arm is recessed in the center to allow for the LED board solder tails. It is also streamlined for the airflow.
At the same time I was making the mechanicals I also did my schematic capture and came up with the layout seen below. I tried to keep things as small and as tight as possible (even standing resistors vertically) and having my rectifying diodes straddle the secondary mounted input connector. I was only able to get a good layout with the photo diode/detector rotation sensor place off to the side, hence, it is mounted at a funny angle so as to encounter the blinder at 90° (see my mechanical sketch). Since not all of the components were on my layout database I, in some cases, used physically similar components while for others they were placed hole by hole. Electrically the basic design is very simple. First, power is received from the motor, rectified and filtered and regulated. A TO-220 regulator was used to step down the 12VDC from the motor and provide a regulated +5VDC for the PIC. I should note at this point that subsequent operation finds that the clock runs well (and much more quietly) at 7 - 8VDC and possibly a TO-220 regulator is overkill. Lord knows there is enough wind to cool it. In fact at 12VDC the clock is a bit scary and makes a fair amount of noise. That same TO-220 was laid flat under the G-forces at that speed. It's nice to know it can do it but as I stated before I prefer the low gentle whir of operation at 7 - 8VDC. Anyway, after being regulated there is some capacitance for stored reserve before the power rail is split into basically two branches. One "high power" branch for sourcing the LEDs and another low power branch for every thing else. Diodes prevent any back feeding from one branch to the other. A 0.1F memory cap is used to keep the second branch up to allow the user to set the time (after stopping the unit) with the 3 little time setting key switches. The 0.1F memory cap allows the PIC to keep time for well over a minute or two when all other power has been removed and the clock is stopped. Thus there is absolutely no rush to setting time. At the heart of it all is the mighty PIC16F84 microchip. An 18 pin 14 bit 1,000,000 instruction per second EEPROM RISC processor with some built in RAM. To buy that kind of computing power these days will set you back $6-10. The PIC is clocked with a 4MHz crystal. The PIC's output ports drive transistors which, in turn, switch the 7 LEDs ON and OFF with the exact timing to simulate a roughly 7 x 48 LED array. There is a "blinker" assembly (IR photo diode shining onto a photo transistor) on the secondary side of the board. The "shade" is mounted to the top of the motor. The pulse output of this is fed to the PIC so it can keep track of relative timing of a rotation. This basically keeps the display fixed in location no mater what the motor speed is.
After completing the schematic and doing the layout I found that board fabrication was going to be very expensive. Far more than I had originally thought. Most of the cost was in set-up and usually with a 2 panel minimum. Price was ranging from $600-$1000. I did find a prototype shop, APCircuits of Calgary Alberta, that would do it at reasonable cost but there would be no silkscreen or solder resist. This seemed like a fair trade-off. As a result I had to go back into the layout and demarcate the boards with copper lines and convert some text to copper as well. The picture below shows the primary side (top) and secondary side (bottom) of the board. The lines demarcating the board outlines can be seen on the secondary side. I just cut them out fairly close to the line and belt sanded (main board) and filed (LED board) them out to size. I must say that APC did a great job of the boards. Without the solder resist however, and its tendency to contain the solder to the pads, one does have to be ginger when soldering or the solder will spread and flow up the tracks a bit. This is really not a big issue and later when I did assemble the boards they went together quickly and easily. Most of the components were obtained locally from Active Electronics or from Digi-Key. I made custom insulating sil-pads for insulating the boards from the metal mounting flange and spinner by photocopying a drawing onto a plastic overhead sheet and cutting/folding them to shape (see my mechanical sketch).
A lot of thought was initially given to getting power from the motor up to the board. The classic problem here is that while there is power available feeding the motor it is not rotating. Alternatively there is also power available on the rotating interior of the motor but one must get it out past the bearings. Many different approaches have been taken before. After a fair amount of consternation on the different options I decided to take my own approach and bore the motor shaft axially in past the top bearing and then out at 90° being careful not to scratch the windings. The pictures below show these operations underway. The left is axial drilling the motor shaft on my lathe while the right is cross drilling the shaft on my mill. Center drills must be used to start the holes as twist drills will wander.
The following picture shows how beautifully simple the result is. Power is picked up from the commutator and simply routed out the end of the shaft. The result may be simple but this method does require a lathe which may be an issue for some.
Here both the motor/board power cable and LED board cable are being dressed. It was important to have no slack in the motor/board power cable since it is routed past the the "blinker". If there was slack it could be rubbed/cut as it passed over the "blinker" shade. To do this I simply held the wires taut next to the connector and cut them were they would end in the contact then dressed them. All cables were confined and protected with heat shrink. Here also one can note the size of the clock by the inch rule placed beneath it.
The picture below left shows the original experimental balance counterweight. Balance on such a large whirring mass can not be ignored. On initial power-up (no balance or board) despite the large base plate I can only describe the effect as "rotary jack-hammer". One must remember that at 12VDC a clock this size, while smaller in weight, has a scale comparable to an Austin Mini wheel at 70 mph! Needless to say a firm grip was required. Later, after I had received and assembled the board I made the above simple "L" bracket with protruding 4-40 screws. This allowed the easy addition of weights. I also initially made a bunch of 0.125" x .75" x 1.5" counterweight plates but found even one was too much. In the end a stack of 4-40 nuts did the job and worked well as each nut individually only had a tiny incremental effect. Later once the weight required was known the "L" bracket was replaced with a more cosmetically exciting design (shown below). It was confirmed that having the counterweight mass a little higher than the opposing LED board arm had the best effect as the board (below) and counterweight mass (above) average nicely to the plane of the opposing LED board arm. In the end there is a slight residual "hum" but when placed on a mouse pad it can run indefinitely without moving and the very slight "hum" can only be noticed if felt.
The next main issue I had to tackle was getting power into the motor. I decided that I wanted the clock to have a DC jack so it could run off of any 12VDC source. I also wanted to use a small C & K type switch which I already had that locked in the ON position and, more especially, OFF position. The toggle has to be pulled out before it can be moved. The idea was to prevent accidental turn on. The switch is incapable of reliably handling the motor current so a relay was also found. I soon came up with a design for a containment can along a quasi "exhaust system" theme. The can design I came up with was extremely tight, difficult to machine and weld together and also turned out to be real ornery to assemble. Yes that exhaust pipe with internal wall for the DC jack is machined from solid!
The motor's stock clunky cable was axed and a small 2 position terminal block was soldered on to the remaining wire. The switch, relay and jack (above left) were assembled so as to allow sequenced assembly into the can. It's so tight that even the switch hole had to be angled a bit so the switch didn't jamb on insertion before protruding out the hole. It was after overcoming all the hassles of my masochistic design that I discovered best operation was 7-8VDC. The minimum pick-up voltage on the relay was right at 7.8VDC. DOH! On the right is the result. The simplified (Pch) FET switch arrangement is shown (above right) assembled with a similar set of parts external. It's still tight to assemble. In this case the switch connects the FET's gate to GND (on) or to the 7-8 VDC at the source (off). A reverse biased diode was placed drain-GND so as to snub any kick-back when the unit is switched off. Even still with all the components assembled into the can it is still difficult to attach the two wires to the terminal block and then gently squeeze it all together to attach the two tiny mounting screws without any binding etc. In the end it works but I can not emphasize the words ornery and masochistic enough!
At the other end of the difficulty spectrum my DC supply is simple. A line cord, Hammond 166L12 transformer, a bridge, 10,000uF capacitor (freebie) and a DC jack style output cord. Currently the clock is running from the center-tap output of the transformer. Later I may swap the transformer altogether. A thinking person would have put the above switch assembly into this relatively roomy power supply box. Hindsight is, as they say, 20/20.
Having not been involved in software for many years I was a little nervous about this aspect of the project. I spent a fair amount of time trying to re-acquaint myself with the low level machine code so as to understand the program. In order to do so I translated the "noblink" program into some basic flow charts which I found much easier to follow. This translation process was an excellent method to do that. I also found that it gave me an appreciation of the processing speed of even this lowly PIC. While common sense it all seems a bit abstract until you see the clock screaming along at 12V and think that the PIC is running over 33,000 instructions every revolution! Executing an instruction for roughly every 0.001" of LED movement at the arm tip. Given that this PIC is running at 1MHz, processing 14 bits and not 32 or 64 bits or the staggering 128 bits of the new Mac G4 ("over twice as fast as the 800MHz Pentium III") it might make you appreciate what your using to read this page with a bit more!
After getting comfortable again (to a degree) with software the next decision to be made was on a programmer. Not knowing of anybody locally with the capability to program a Microchip PIC16F84 meant I would have to purchase one. I could have likely made one but I wanted to make propeller clocks and not a serial programmers. After a search of the (extensive) possibilities I settled on the Warp 13 from Newfound Electronics. It came with the download cable and only required a 12VAC adapter which I acquired locally for about $15. The complete MPLAB software for assembly and downloading to the PIC etc. is available free from the Microchip web site. Programming the PIC was very easy. One merely connects the programmer to the COM port of the PC, places the 18 pin PIC16F84 in the top of the 40 pin ZIF socket (see below), plug in the adapter and the hardware is ready. You use the MPASM program on the PC to find and assemble the program and generate a hex version of the code. Later the MPLAB program (seen below) is used to set the blast options (tiny bottom window on screen) and finally to download the hex file (large window on the lower screen) out the PC's COM port to the programmer/PIC. The actual MPLAB programming/verification etc is controlled from the window on the upper right. Everything went flawlessly first time and I had no previous experience serially downloading so I think that is a bit of a testament to the Warp 13 / MPLAB combination.
I should state however that I initially thought the programming was unsuccessful! It turned out that the low quality (i.e. free) 5 mcd LEDs I had placed on the LED board until my HP "SunPower Series" LEDs arrived were actually working but so dim as to be not noticed in office light. It was only by chance I noticed the faintest of LED glimmer. When I placed it under the desk the display could be seen, barely. Dim as it was it was an exciting moment! I can't emphasize enough how important bright LEDs are. When the 30°, 1750 mcd HP LEDs (HLMP-EG30-NR000 red & HLMP-CM30 green) did arrive from Newark Electronics I did a "quick-blip" test at 30mA on the green. Like a supernova I'm sure there was a light beam shining past my head onto the ceiling. I saw a spot for 5 minutes afterwards. These LEDs are very bright, and, of the two, the red is even brighter. However, when installed, one can look directly into them. I think this is because being multiplexed on and off they are only on a small percentage of the time and your eye tends to average the light intensity. I suspect the corollary of this is it is also why the persistence effect works. At any rate even using the "SunPower" Series LEDs the display tends to wash out against a daylight sky backdrop when placed in front of the window. When trying to get bright LEDs initially I also thought the narrow 30° angle would be a problem. This is why I made the clock large. All things being equal a large radius narrows the required display angle since the digit height is fixed. It turned out not to be as big an issue as I thought. There is a brighter central zone to the viewed display. This can be seen in the brighter central "02" in the lower left photo in the group of 8 photos below. The brighter central zone (not the display) also moves as you move around the clock but I find it is really not a big deal. I would not recommend narrower display LEDs though. Since I do get asked, additional displays could be added by simply adding additional "blinker" shades on top of the motor. Since this will reduce the rotational period count (see flow charts) the "period_use" value would have to be multiplied by the number of shades to maintain display width. Beyond 3 shades, however, the displays might start to run into each other and one might want to increase the "period_use" value by a lesser amount so as to narrow their width.
The last thing thing I did was have a bit of fun and finished the clocks off with a brass logo plate. To give them a finished look as it were and not so home-made. There were a lot of reject ideas/artwork before I settled on this one. Unlike the churn in artwork I immediately liked my theme slogan "Blick - Making Time Fly Since 1997" (my tribute to Bob). I decided that the artwork would have a whirring propeller display this slogan. Given the clock's lean no-frills nature and the orientation of the display I settled on the name "Aerobat". That soon led to the harmonizing idea of a full "off we go into the wild blue yonder" aerobatic take-off with streaming smoke. The flags were partially inspired by the Reno Air Races and partially by the whole concept that this is a winning idea. Finally I also developed a thematic brass warning label reminding one to "always perform clearance check before take-off!"
The brass was etched in my basement set-up which I had previously used to etch a sundial . It consists of two fluorescent UV lamps. Kodak KPR chemicals were used. The composite artwork was scanned, clean-up a bit then printed to overhead mylar for use as a positive. The brass is cleaned, covered in resist, exposed, developed, etched, painted, sanded, drilled & fitted and finally protected with lacquer.
The following pictures show the final clocks. I made a "red" one and a "green" one. Overall I'd have to rate this as an excellent project. It touched on so many different fields from mechanical design & machining through layout, build, micros & programming to artwork. Additionally along the way I got to converse with some very friendly, intelligent and creative people. At the end of it all you have a really cool clock that fires the imagination of all who encounter it. I find that, at times, it does look like a clock on the edge of a whirring propeller yet at other times it makes me think of some futuristic time-gizmo that might be found behind a seedy bar on a Star Wars set. Watching it though I must admit it truly does make time fly. After all, how many clocks do you know that consume 18W due to "windage losses"?
Links to other Propeller Clocks
May 23/2000 - Initial Upload June 2/2000 - faster loading, minor corrections, clarifications & improvements, 2 additional pictures, 2 new links, flowchart update Jan 7/2002 - my etched sundial link added, comment about original background Jan 14/2002 - add link to Mark Ursum's clock. Feb 4/2002 - link additions & updates. Apr 1/2002 - link additions. July 23, 29, 31/2002 - link additions. Sept 12/2002 - motor shaft drilling pictures added. Sept 17/2002 - add link to Otis Irace's clock. Oct 10/2002 - add link to Nebulus' clock and Henk Soubry's clock. Oct 18/2002 - add link to Alexander Telegin's Displays. Dec 12/2003 - add link to Dirk Gehrke's "Crazy" Clock. Jan 7/2004 - removal of dead links. Feb 19/2004 - link update & addition of "iBall" & Spacewriter. Mar 30/2004 - add link to Nick Dawkin's Clock. Jun 2/2005 - add link to Pierre Lando's "Bill Poster". Sept 30/2005 - update expired links Back to Back