1 of 10

Technology

Technology

Rudimentary navigation for this page:-

Fcalc - Scientific & Aeronautical Calculator

UV Lightbox

I can be contacted through the Weblog page...

I like to design and make things. Like this site. The site is written in php, the underlying database is MySQL. I originally wrote it in ASP with MSSQL, but when I looked for somewhere to host it, I found the Unix offerings more generous. Although Unix (Apache) systems run ASP it's a lot easiser to connect to MySQL from php, especially when you don't run the server.

I don't check for cross-browser compatibility, other than IE8 and Chrome.

So, as you can see, one of the hats I wear is engineer.

The site is in verified XHTML which should mean that it will be visible on a number of platforms. In order to maximise accessability features known to cause problems, such as links which open new windows, will be avoided.

I'll be posting some source code and circuit diagrams here.

Topics will range through composite materials fabrication, radio design, aeronautics, satellite link budgeting, antenna testing, register-level logic design, ballistics and others, when I can find the time to organise the material.

In the meantime, here's a drawing of a guitar. Maybe I'll get around to fabricating one one day...

For anybody who wonders, as I used to until I took the time to figure it out, the scale factor (pun intended) for the gaps between the frets (semitones in the even tempered scale) is the twelfth root of two.

The full-scale length is two thirds of forty inches???. Well, the only scale with all naturals runs from C thru G and then starts again A, B so maybe I should just be thankful for small mercies.

There used to be a lot of audio circuits on this page, they've all moved to the audio page. I have a lot of other stuff to go here.

Top

2 of 10

Technology

Scientific and aeronautical calculator

I wrote this DOS calculator about 20 years ago.

fcalc.zip

Despite it running under DOS I still find use for this calculator. It has many scientific constants built into it and a periodic table of the elements. You can select the constants and have them drop directly into the calculator's main register.

It also performs numerous conversions such as metric to Imperial, and does a variety of aeronautical calculations such as pressure and density altitude, true airspeed, track, weight-and-balance etc., along with all the regular statistical, trigonometric, logarithmic scientific functions and base conversions.

Next: A UV lightbox with timer.

Top

3 of 10

Technology

A UV lightbox with timer.

PCB fabrication. Many people use toner transfer, but I've always used photo-etch.

Here's how:-.

This time I decided to go with LEDs. The previous lightbox I had used fluorescent tubes installed in some fittings taken from an old camper van, I used to run it off a 12V lead-acid battery. The whole thing was constructed out of scavenged bits; a box I picked up in the street that had slots for a sliding lid that neatly accommodated a sheet of glass that I cut to fit. I lost it. How do you lose a UV lightbox?. Don't ask me.

I bought 100 UV LEDs from a seller in Nanjing, China for $10 US shipped. I had some Veroboard which I cut into 3-conductor-wide strips, this gave me 10 strips, each of which accommodated 9 LEDs.

I decided to run the LEDs in strings of 3, without a ballast resistor. The forward voltage is quoted as 3.4~3.8 volts. When I ran up the 90 LEDs to 600mA (30 strings @ 20mA) I got ~11 volts on the lab power supply, so I figured 2 rectifier diodes in series with 12V regulated from a 7812 would be about right to drop the voltage to an appropriate level. It's not recommended to run LEDs this way, usually there's a ballast resistor in every string or a constant current supply, but I broke from these practices first when building torches, and it turns out LEDs are more robust used like this than theory and literature would suggest.

I built a 12V power supply with a 20VA RS toroidal transformer I had lying around. The other components for the bridge and the smoothing caps were just 'lying around' too. I used a 78S12 regulator in conjunction with the aforementioned dropper diodes in order that the UV array wouldn't be overdriven.

Here's the LED array:-

Here it is lit.

4 of 10

Technology

And here's the coverage pattern:-

This is a sheet of printer paper lying on the picture frame blocked up about a foot over the array on some piles of books.

Arranging the LEDs was a problem. First they had to be seated firmly against the Veroboard to try and make the beams parallel. Then the Veroboard had to be held flat. I've got a piece of 3 mm fibreboard with 2 x 1/4 x 1/2in balsa strips adhered to it with doublesided sticky pads. Then the Veroboard is held to the balsa strips with thumbtacks, or drawing pins if that name is more familiar to you.

This has not resulted in as even illumination as I had hoped. The beams of the LEDs are quite narrow. This has the advantage that the overall beam does not spread too much, but I have had to keep the target sheet 12 inches away from the LEDs to ensure that the spots spread into one another. The beams mean that the collimation is not too bad, however, certainly better than the typical fluorescent setup. I'm looking for an LCD monitor to dismantle, they have some high-performance diffusion screens in them, which might improve things.

It looks a bit blotchy, and I've tweaked the way the LEDs point since, but its surprisingly effective and even in its results. I've made a 7.5 inch square board without problem. The exposure time is 480 seconds. The box I had previously only required about 4 minutes. It was 12 watts, this is only ~6. This was cheap, however, given that I had the Veroboard, it works, and it's low voltage.

I was concerned that stencils made on an inkjet printer might not work. Transparency material for inkjet has a coating on one side to absorb the ink. This has tiny dots (imperfections) irregularly spaced presumably as a result of the coating drying. I thought these might show up on the PCB. I also didn't know if the contrast ratio between the inked areas and the clear would be sufficient to produce a good image in the photoresist. As it turns out, this is not a problem. Although the contrast could be better, it is still possible to produce adequate results down to 10 mil traces. This can be seen on the doublesided board shown later.

I made the board for the PSU using the new LED array driven from a lab PSU and did the timing with a stopwatch. It's a conventional thru-hole job, single-sided.

5 of 10

Technology

I like surface mount components. The more surface mount you have, the less holes you have to drill. There are no problems with requiring thru-hole plating, although obviously, in the case of a double-sided board, vias are necessary. I generally use a larger head on the vias than I would for a commercially produced board. This is not ideal, but the real-estate taken up is generally not critical in DIY projects.

This is the board for the timer.

Double-sided boards obviously require registration between the sides. This can be accomplished by peeling the protective plastic covering from one side and sticking one stencil to the unexposed board and drilling a few holes. You will have to remove some swarf from between the stencil and board before exposure. You can use reference marks included for the purpose, or simply drill at the board corners, or through a few vias. Then the protective film is peeled from the other side and the second stencil carefully matched to the holes. An alternative method matches the stencils against each other, taping them to prevent relative movement, and inserting a piece of scrap board with an edge coincident with one of the board edges on the stencil and then taping this in place. The unexposed board can then be inserted between the stencil sheets and butted up against the scrap piece and some tape added to obviate movement between exposures.

Here is the artwork as a 200dpi bitmap

6 of 10

Technology

Vias

Vias are all placed at the same time. Take about a metre of plated copper wire and grasp the ends in long-nose pliers, taking a couple of turns round each of the plier jaws. Hold the pliers firmly and pull them apart until the wire can be felt to give by a centimetre or so. This will put a 'set' on it, i.e. leave it straight and slightly work hardened.

Now cut sufficient 'needles' from the straightened wire to complete the board. Cut the needles with side-cutters held at a 45 degree or better angle so that a sharp pointed end is produced.

7 of 10

Technology

Place the board on a piece of sponge such as frequently found in kitchens or bathrooms. Place the needles to fill all the via holes, pushing them down securely into the sponge for about half their length. Once they are all placed, you can go round and solder them all. Detach the board from the sponge and trim the excess wire from the top surface. Now you can invert the board and solder the other side. Even if the board is subsequently reflowed in a toaster oven, the vias are unlikey to fall out due to surface tension, particularly if the wire is not a sloppy fit in the holes.

Here's the timer for the lightbox. It employs a PIC 16F887 which was selected, amongst other things, for its internal oscillator.

8 of 10

Technology

The board provides a simple timer function with a relay to control the UV light. The circuit is shown here.

9 of 10

Technology

When switched on the timer displays 000. There are 3 control button inputs, 2 for setting the period.. The first causes the display to count upwards. The count proceeds at second intervals when the button is held down until the count reaches 5, then the delay period is reduced and the count increases at a faster rate, this means that setting relatively long delays can be accomplished in a relatively short time.

The second button functions in a similar fashion to the first, but the count is down.

The third button turns on the UV light and starts the counter counting down at 1 second intervals. When the count reaches zero the UV is turned off and the display reverts to the number originally set. Subsequent exposures can be initiated by pressing the start button again.

A number of the components used are mounted unconventionally. The voltage regulator, electrolytic caps and diodes and relay are all conventional thru-hole types applied to one side of the board and 'glued' there with solder. In some cases I have used a custom foil pattern, and bent the component legs to mate. The 7-segment displays are common-anode types, each segment is ballasted with a 470R resistor. They are mounted in a 30-pin DIL socket which is surface-mounted to the PCB, this provides some stand-off and means that the components can be replaced or re-used if so desired. Since 30-pin sockets are not generally available, one can be made up from 2 smaller or cut down from a larger one (just cut off 2 pins leaving the plastic intact). Past experience with the MPU and displays (the LED forward voltage is >3V) have shown that the chip will easily support the dissipations encountered with 470R ballasts.

Where possible I like to use solder paste and reflow in a toaster over. A good source for solder paste is DealExtreme: Cool Gadgets at the Right Price - Site-Wide Free Shipping - DX, although there is a considerable delay in postage. Obviously solder paste can be obtained from sources closer at hand, but many suppliers like to ship fairly large quantities in an insulated package, to keep the temperature of the contents low. This can result in high prices and the quantities involved are greater than an amateur could reasonably use before it deteriorated. Solder paste is commonly used in commercial applications. If the paste fails to reflow suitably thousands of dollars worth of components can be affected in a production run. This can result in litigation in the worst cases. While this is understandable, the insistence on shipping large quantities in controlled circumstances is inconvenient to say the least, and the availability of small quantities from sites such as dealextreme is a way round this.

The MCU itself is a TQFP44. Soldering these by hand only requires the correct technique, which bears repeating here. Carefully apply solder to a corner pad of the PCB foil pattern. Slide the IC into place and secure the one pin. Adjust the positioning until you are completely happy with it. Now solder a single pin at the opposite corner. Starting with one of the unsecured edges, solder thickly and indiscriminately all the pins along that edge, pulling the solder bead from one end to another, making sure all pins on that edge are wetted. Now solder the remaining pins in the same manner. Remove the excess solder with a solder-sucker, working quickly to use as little heat as possible. If a string or sphere between 2 pins is difficult to remove, try re-applying some solder in that area and try again with the sucker.

In order to build this you need a PIC programmer. I have a PicKit2. It's a USB device. You can get a clone off ebay, but the saving over the genuine article isn't much.

A couple of years ago I looked around for a BCD routine written under Mpasm, the free PIC assembler. I couldn't find anything. I wanted to display the contents of the 10-bit PIC A/D register on 7-segment displays, effectively to build a digital meter. The routine here does this and can be used to build a meter with some analog circuitry added to process e.g. AC current to voltage or WHY.

The board is programmed with a Pickit2 USB programmer via an ISP header which is simply a 6-pin 0.1 pitch SMT pin-header, connected to the appropriate pins on the PIC. The Pickit2 is available for a comparatively low price from regular suppliers, and even cheaper as a clone from ebay. It is very simple to knock up a board with a socket and header which will permit the programming of a wide variety of conventional thru-hole PICs using the Pickit2. Although it is possible to build programmers running off the PC's serial or parallel ports, the cost and effort mean that this is hardly worthwhile, and the Pickit2 is extremely versatile. There is now a Pickit3 available with (I believe) better debug facilities, but I have always found the PK2 to be adequate. Anyone who likes to be able to add a few digital features to a design should have one.

10 of 10

Technology

Since the display runs 000-999, a ten-bit buffer is required to contain the count. Arranging for this to overflow correctly is comparatively straightforward. The 3 input switches select amongst up- and down-counting routines while the program continuously cycles in one big loop. The loop processes the count in the buffer into 3 secondary Binary Corrected Decimal (hundreds, tens and units) buffers which are in turn processed into 7-segment display buffers. The process is comparatively straightforward, if tedious.

Each bit in the count buffer is examined and, if set, a corresponding number is added to the BCD registers, i.e. if BUF_HI, bit 1 is set, 5 is added to the hundreds count, 1 to the tens count and 2 to the units count, and so on, down to BUF_LO, bit 0, which is worth 1. When all the bits have been examined the BCD units buffer is reprocessed to move any accumulated tens to the BCD tens buffer by repeated subtraction of tens from the units buffer until a borrow occurs, when the units count is restored by adding ten. Similarly the BCD tens buffer is stripped of any accumulated hundreds. The size of the numbers involved is small enough that there is no risk of unintended overflows.

A simple delay routine at the end of the loop pads the time taken to execute to ~1 second. A second, shorter delay is switched in during the up and down setting routines after the buttons have been held down for a count of 5. This means that setting up a high count e.g. 600 does not become too tedious. The RC clock in the MPU is not guaranteed to keep accurate time, but the seconds delay can be trimmed to fair accuracy by timing the count over a minute or so and adding or subtracting a few cycles from the delay routine. On my board the accuracy is within 1 second per minute, which is plain lucky with such an uncomplicated delay routine.

The seconds delay means that, depending on where the routine is in the loop, there may be a brief wait for the system to respond to a button press, but I haven't bothered to complicate the routine by eliminating that.

The relay can easily be used with a minor modification to the circuit to switch mains voltage if mains-driven fluorescents are used. In that case only a 5V supply would be required.

The accompanying circuit diagram shows the proper correspondence between the PIC port bits and the common-anode display pins. The displays used are 0.5in character height which are readily available at a reasonable price on ebay.

Here's the BOM.

Resistors

24 R2-R25 470R
1 R27 1k
2 R28,R29,R30 10k

Capacitors

3 C2-C4 100n
2 C9,C10 2.2uF

Integrated Circuits

1 U2 PIC16F887A-PT
3 U3-U5 7SEGBLUE
1 U7 78S05

Transistors

1 Q1 BC108

Diodes

3 D5-D7 DIODE 1N4001
90 D8-D97 LED UV

Miscellaneous

1 TR1 TRANS-12V-20VA
1 J1 PIC_ICSP_HDR
1 J2 CONN-H4
1 J5 TBLOCK-I2
1 RL1 G2R-14-DC12
1 Large Veroboard

11 of 11

Technology

Finally, the code...

Top