Wouter van Ooijen (wouter@voti.nl)
This little project was triggered by a request from Ruud Setteur on the newsgroup nl.hobby.electronica (dutch) for a circuit that generates a voltage that rises slowly from 0 to 10 Volt and (when a switch is flipped) slowly back to 0. The request did not pose much restriction on the wave form, but a rise / fall time of 30 minutes should be possible. If it were not for this quite long rise / fall time the standard one-opamp integrator would be a perfect solution.
Being a PIC user my reply was of course: for a PIC this is a one-banana problem. That is of course correct, but only hardware wise. Some software is also required, so my contribution was not very helpfull for the requestor, who could build simple circuits from a description, but certainly not program a PIC.
After some discussion about counters and D/A converters on one side and charging capacitors on the other side I offered to design the PIC solution. Stef van Diest offered to try the discrete-logic approach (oscillator, up/down counter, D/A converter), as shown here. There was some discussion about the feasibility of a simple charge-a-capacitor circuit but no volunteer for that approach. The main problem is that for large (electrolytic) condensators the current must be so low that the leakage of the capacitor poses a problem, and for smaller condensators the current must be very small.
While thinking about the PIC program I also considered using a 3.3F gold-cap capacitor, but as these capacitors are very critical with respect to the maximum voltage (2.1 Volt) the circuit would not be very simple (4 or 5 opamps), so I dropped that idea.
By the time I had sort of finished my design Guus Mattaar posted a very simple analog circuit, cleverly (mis) using the offset compensation of a CA3140 FET opamp to create a very small charge current. I have my reservations about the stability and reproducability of this circuit, but it is certainly simpler, cheaper, easier to understand and probably closer to what Ruud wanted than my PIC solution. The (dutch) description Guus' solution is reproduced below.
Na de discussie te hebben gevolgd kon ik het niet nalaten ook in het probleem te duiken en kwam tot de volgende oplossing: Een opamp met mosfetingangen CA 3140 op 12 V. Verbind uitgang (6) met - ingang (2) en via 1M.Ohm met + ingang (3). Leg 3 met goede condensator 0,47 uF.aan -. Zet tussen nulcorrectie (1) en (5) 2 potmeters 10 k.Ohm (25 slags) parallel. Door de lopers beurtelings aan - te schakelen loopt de uitgang lineair van 0 - 11,5 en 11,5 naar 0, uiteraard na zorgvuldige afregeling van de potmeters. De condensator wordt op en ontladen met een konstante stroom, bepaald door 1 M.Ohm en de ingestelde offsetspanningen. Tijden tot 50 minuten waren ruim haalbaar.
The PIC fader uses a 16f628 because that was at the moment the cheapest hobby PIC, and it happens to have to two features that can be put to good use in this application. The first is the ER oscillator option, where an external resistor determines the clock frequency, while the resistor carries only DC (not AC as in the external RC clock option found in other PICs). The other feature is the PWM (Pulse Wide Modulation) hardware, used here as a slow 8-bit D/A.
Unfortunately Microchip does not mention in their datasheets how the clock frequency of in ER clock mode depends on the external resistor! But asked on the piclist where such information could be found and got an answer very quickly:
> The oscillator configuration section in the datasheet states
> that the electrical specifications section documents the relation
> between the resistor and the frequency, but I find no such info.
> Does anyone know where I can find it?
Right here ! Where else ? ;-))
(original post 2/12/01)
These are practical measurements for 16F628-04
R Clkout * 4
2k7 9.40 MHz
3k3 9.35
4k7 9.23
10k 8.44
22k 5.65
39k 3.50
100k 1.52
220k 0.75 MHz
470k 0.393
1M 0.192
1M5 0.124
2M2 85,600 Hz
5M6 36,400
12M 16,400
So if you ever have questions about PICs, you know where to go!
The minimum circuit is very simple. A 2M2 log potmeter and 10k series resistor determine the clock frequency and hence the rise / fall time, which can be set over a 1:100 interval that includes the range 1 .. 60 minutes. A 100k / 0.1uF RC combination provides a crude low-pass filter for the PWM output. The minimum frequency of the PWM signal is 336 Hz, so this simple filter with -3dB at 1 Hz is adequate. The output is 0 .. 5 Volt and does not tolerate much load (my multimeter loaded the circuit too much, I had to lower the resistor to 10k), so to meet the original specification (0 .. 10 Volt) and provide a lower impedance an opamp with 2x aplification must be added. (An inexpensive 1/4x LM324 will do, as its input common mode includes ground. Note that a load resistor on the output of the opamp might be needed to get an output close to 0 Volt.) The MCLR (reset) pin is tied to Vcc via a 33k resistor and the LVP input is likewise tied to Gnd. I use my Wisp HVP programmer for development, but because I did not need the pin for other purposes I left the LVP pin unused and enabled LVP in the PICs configuration, so the circuit can be build and programmed using LVP. With pin RB0 floating (pulled high by the internal weak pull-up) the PWM duty-cycle rises, with RB0 pulled low it falls.
The minimum circuit leaves eleven IO pins unused (and one more is available when LVP is disabled), so I could not resist the temptation to add some frills. These extra features can be ignored (and the pins left unconnected). Four outputs give an indication of the current output value: minimum, larger than minimum but below half, above half but not maxmimum, and maximum. These outputs are active low and can be used to drive LEDs or (via a transistor) relais. When LEDs of the same color are used these could share a single current limiting resistor (LEDs of different colors might need different resistors to give the same perceived brightness). One more indicator LED blinks once for each step the output voltage makes. Two preset inputs allow the ouput to be forced immediately to the minimum or maximum value. One input can be used to free the current output value. One input speeds the rise and fall times by a factor 60, so the range becomes 1 .. 60 seconds instead of 1 .. 60 minutes. This can save some time calibrating the scale of the potmeter. After adding these extra's there are still 2 IO pins left, but I was out of idea's.
The software is (of course) written in Jal, my very own programming language for PICs. It first initialises the IO. Note that on a 16f628 the port A analog functions must be disabled before the port can be used for digital IO. The port B internal weak pull-ups are enabled in the option register, so switches can be connected without the need for external pull-up resistors.
Next the PWM module is initialised. The PIC can produce PWM with 10 bit resolution, but I use only 8 bit because Jal has only byte (8 bit) integers, and a higher resolution would result in a lower minimum PWM frequency.
To get a 49 s rise / fall time (1 minute minus some 10% for margins) each step must take 49 s / 256 = 190 ms. For each step the main loop must loop 120 times (factor 2 to blink the step LED, factor 60 to make a 1:60 speedup possible), so the loop time must be 1.33 ms. The Jal libraries can not cope with a varying clock frequency, hence a fixed 10 MHz frequency is specified. The expected clock frequency with the minimal (10k) resistance between ER and Gnd is 8.44 MHz. The amount of processing in the main loop is so small that it can be neglected for the timing calculation, so the delay at the start of the loop is be ( 8.44 / 10 ) * 1.33 ms = 1.12 ms.
The main loop first waits the calculated 1.12 ms. When the loop count has wrapped around the step LED is toggled, and when it is now dark the output level is adjusted one step. Next the preset inputs are checked and if one is low the output level is set as requested. Next the hardware PWM duty-cycle registers are update and the indicator LEDs are set according to the current output value. Finally the loop count is incremented and checked for wrap-around, which happens at 120 in the default mode, and at 2 when the speedup input is active (low).
I build the prototype on a solderless breadboard, which is definitely my favourite method for building prototypes. I guess that proves that I am from the computer / digital world, not from the analog world, but it always served me well. The printout is a listing of the program. There are no toggle switches: the two loose wires can be connected to the ground rail to make the relevant PIC pins low. There is no potmeter either: I just plugged in the appropriate resistors when needed. The PIC is programmed in-circuit using my Wisp proprogrammer. The second picture shows how a DIL clip is used to connect the programmer to the target chip.
The table below shows that the application uses very little of the PICs resources. There is ample room for extra features, for instance an RS-232 interface to set or read the level. Note that when significantly more extra CPU time must be spent in the main loop the interval library should be used instead of the fixed delay.
| code | data | stack | eeprom | I/O pins | |
| available | 2048 | 224 | 8 | 128 | 16 |
| used | 228 | 13 | 2 | 0 | 2 |
To modify the Jal source you will need to download it first. Any text editor can be used to edit it, notepad will do, but I prefer PFE (free!) or UltraEdit. For compilation you must download the Jal compiler. The compiler produces a .hex file that you must program into the PIC. The design and software for the programmer I use can be found on my website. The ic-prog programmer software supports a wide range of PIC (and other target!) programming hardware, and the website has references to a wide range of programming hardware designs. When you just want to program a PIC with the fader software the best way is probably a parallel port LVP programmer, but I have not found a good description on the web of such a programmer. The circuits are drawn using Eagle. Watch out when you want to make a PCB from the circuit: I just took the first resistor etc. that looked right in the circuit diagram, so you might end up with components of a surprising size.
Note that this design uses a feature that is unique to the 16F628 (the ER oscillator mode). No other PIC will work, not even a 16F628A.
 | Jal source (Jal 04-40) |
| | hex file |
| | circuit (eagle 4.01) |
| | PIC 16f628 datasheet |
| | PIC midrange reference manual |
| | Jal: my own PIC language and compiler (free) |
| | wisp: my flash PIC programmer |
| | piclist: THE forum for PIC users, heavy traffic (~ 100 posts/day) |
| | PFE (Programmers File Editor) is a free text editor, often used by programmers (including me) |
| | CadSoft's Eagle circuit / PCB editor (free version with some limits) |
| | ic-prog: software that drives lots of programmers, the site also provides lots of programmer schematics |
| | LM324 datasheet |
| | CA3140 datasheet |
| | 25-jan-2004 the previous circuits had R2 connected to +5V, this was WRONG (thanks Sietze Kooistra for discovering this problem after 2 years), 'freeze' switch added |
| | 26-jan-2002 pictures, cleanup |
| | 18-jan-2002 first version |
(c) 2000 .. 2004 Van Ooijen Technische Informatica / Wouter van Ooijen
All rights for this text are reserved. The design information however, is free for all use. This includes (as far as available on this page) the code in source and executable (.hex) format, circuit diagrams, and PCB designs. There is no warranty of any kind, except as mandatory by law.