Ed Scherer

Jump to post #27 in this thread if you want to see what I finally came up with and what's now installed and working. If anyone is interested in obtaining and installing a similar dimmer, just reply to this thread and I'll consider options for making this dimmer available.

You might also be interested in the writeup (still not quite complete as of 12-Jan-2012) Converting Porsche 928 Interior Lighting to LEDs. This is a comprehensive guide to upgrading all interior lighting from incandescent bulbs to LEDs. The instrument cluster with all LEDs is a stunning improvement. Pod switch backlighting and center console component backlighting isn't bad, either.





Last night, I got prototype version #2 of PWM-based dimming working (prototype #1 was excessively klugy: a wave of shame washed over me, and I may not have ever said anything about it here; I can't remember). This is my solution for restoring proper function of the instrument dimming control after you've upgraded your instrument backlighting from using incandescent bulbs to LED bulbs.

Version #2 uses a commercial high-side PWM dimmer module front-ended with some additional circuitry to allow it to work with the 928's original dimmer rheostat (non-trivial since it's a very low resistance—about 0 Ω to 6 Ω) and also to smooth out the spikes that emanate from the typically dirty/noisy rheostats.

I'm not going to move this into my car until I complete version #3, which will finally use a single-supply, rail-to-rail instrumentation amp in the front end circuitry rather than the cheap, common, double-supply op amp that I'm using now. Waiting on parts for that; I expect to have this done in about a week.

There might even be a version #4, with the PWM done by an SMT-packaged chip (German company, too!) I just discovered that would really reduce complexity, but I'm not sure I want to go to the extra work to assemble stuff at that small physical scale; I'll do some prototyping using an SOIC-to-DIP package adapter and then figure out whether or not it would be worth it to proceed further.

I suppose I might be influenced by feedback. How many of you out there are interested in solving the dimming-instrument-backlighting-after-upgrading-to-LEDs problem? Does size matter? Seriously, I'm talking about the size of the module that you'd install somewhere close to the existing dimmer rheostat. How much does cost matter? Ease of assembly? Are you willing to assemble something yourself, or do you want something that's more of a "cut the wires to the rheostat and insert this magic module" solution?

Version #2 and version #3 will probably be about $80 in parts (about $60 of that is the already-available PWM module I got from a UK-based company) and could be assembled using crude techniques for the front-end circuitry (small perfboard, soldering, wires, etc.). I'm not sure about version #4; the electronic components would be a lot cheaper (probably under $10), but it would likely require a proper PC board, which I'm guessing would hike the cost back up.




(Double-click the video to go to the full YouTube version, where you can view it in HD).

What you're seeing in this video is an original 928 dimmer rheostat (that's the control that I'm adjusting) controlling the brightness of 3 LEDs; the LEDs are a mix of different types that have different voltage/intensity response and are a representative sample of the various LEDs I'm recommending for the interior lighting upgrades. The PWM circuitry, though, makes them respond to dimmer adjustments in a linear fashion.



(In case you're curious... you'll notice that the above video wasn't shot in my usual garage work area. It has been so damn hot in the garage that I moved this project down to the table in our bar room. It has some advantages... )

borland

Nice work Ed.

I'm still waiting on Hans & Paul to come up with an RGB kit that does the PWM dimmer function with a microcontroller.

I've seen the inexpensive LED dimmers advertised as single color. So, different color LEDs must have a different lumination/duty cycle curves. Is this not a problem with the LEDs your using?

Did you consider modifing one of the inexpensive LED dimmers? Here are circuit diagrams for low and high side PWM dimmers. Notice that the only component that's different, is the MOSFET (load) transistor...


and the high side..


Course, the interface with the factory dimmer switch would need to also be reworked. How did you manage that with your PWM?

Ed Scherer

It doesn't seem to be a problem; that's why I'm trying it with the kinds of LEDs that I've used (and am recommending) for the related backlights.


Yeah... I was thinking about modifying this one (from Super Bright LEDs):





It appeared to me that this just wasn't a very good option; I think it'd have to be pretty darn hacked up, and (after admittedly not spending that much time trying to reverse engineer it) it wasn't obvious how an external control signal could be injected.


I remember looking at those (and many other) PWM circuit designs. At the time, I wasn't really wanting to have to build a PC board for this; I might be changing my mind about that, though.

After finding and testing a high-side commercial voltage-controlled PWM unit (Abeltronics DIM14, nicely deisgned and built by a company in the UK), that had a lot of appeal: off-the-shelf availability and (assumed) thorough design and implementation. I was especially happy to find that there were no traces of RFI that I could detect (using my AM radio as a detector, not exactly rigorous, I'll admit ).



I've got some Infineon BTS730 chips on order. That is one nice little chip, and it's perfectly suited for (was probably designed for) automotive PWM-based dimming. I'll do some prototyping with it (assuming I can succeed at getting one soldered to an SOIC-to-DIP adapter) and might wind up using it. We'll see.

As you can imagine, the 0 Ω to 8 Ω resistance range on the factory dimmer rheostat isn't exactly great when trying to use it to control electronic dimmers. What I would up doing was putting it in series with some resistors (approximately 1 kΩ total, enough to keep the current down to a nice 12 mA or so; no need to waste a lot of current on this) between +12 V and ground to form a voltage divider such that adjusting the rheostat produces an approximately 0 to 100 mV voltage drop across the rheostat. Feed that into an op amp based difference amplifier with a gain of about 100, and you have your 0 to 10 V control signal for the PWM module. Oh yeah, and then I noticed that (even after cleaning and putting DeoxIT on) the factory dimmer rheostat is still rather noisy/spiky. I added a simple RC filter (tuned to cutoff frequency of about 5 Hz; a guess that seems to work OK), and that helps eliminate the spikes quite a bit as you're adjusting the rheostat.

For the next round of improvements, I'm replacing the crude 741 op amp (requires pos/neg power supplies, isn't rail-to-rail, requires external offset adjustment, and requires 4 resistors to set the gain in a difference amp configuration) with an INA122 instrumentation amp (single supply, rail-to-rail, internal precision offset with offset voltage < 250 µV, single resistor gain setting and then ready to roll as a difference amp). That INA122 is a very nice little general-purpose amp that I just discovered a few days ago. I got a pretty good supply of them, as I suspect I'll be using them in future projects. One thing is for sure: I sure like designing around op amps a hell of a lot more than discrete transistors.



It's fun doing this stuff again. I haven't really messed with electronic stuff all that much for quite awhile.

Ed Scherer

I forgot to mention something about your comment: I think the "single color" qualifier on those dimmers is more about being a "single channel" than anything else. There are 3-channel dimmers that provide separately-controllable outputs for the R, G, and B channels on certain kinds of LED products like light bars.

TheoJ

a PIC 16F628 with PWM and a power stage stould do the trick, where the characteristics of the different LED's are individually adjustable in the PIC program and thus one original Rheostat can be used.
If I have time I will do a tryout on the breadboard.

hans14914

Ed,

Very interesting project. With the LEDs going in these days, having a PWM solution is a very good idea. The infineon series chips are great products, but I think I concur with Theo and using a standard PIC or even a PICaxe or simple Arduino flashed Atmega328 or Ardunio Pro Mini if you wanted a pre-assembled unit would be the way to go. It would work with the stock pot, and could control a PNP transistor directly. The package size and cost would go down considerably, and you could even add a soft on/off function very easily.

Please keep us posted.

Thanks
Hans

borland

Smart LEDs...



A BlinkM MaxM (about $25) might be the easiest way to go since it already has analog inputs and a 12V power supply. Arduino development & LED/PWM libraries are free, but you'd still requires an ATMEL flash programmer ($30).

BlinkM MaxM...


more info here ... http://thingm.com/products/blinkm-maxm.html

'todbot blog' is a great source for sample BlinkM projects and makes it look simple...

http://todbot.com/blog/2011/03/22/bl...llest-arduino/

They are for RGB LEDs, but you could apply them to single color LEDs as well.

Oh, by the way, did I say my 90' S4 is still running great!

Ed Scherer

Thanks for the feedback, guys... more stuff to learn about and look into. I've been away from hardware for too long.

I'm rather tempted to start messing with some of what you're recommending (Arduino, etc.) because I haven't done microcontroller stuff before (I live in a high level software ivory tower professionally ) and need to fill that huge gap in my skill set just for hobbyist purposes.

Since I'm close to having something working acceptably along the lines I've already started on, I might just finish that off and then maybe as time permits, go back and start experimenting with higher tech approaches just for the fun of it.

hans14914

Ed,

If you use an Ardiuno, just pickup an Uno to play with, and then when you get it dialed in right, you can move the code to the Pro, which is tiny and doesnt have the extra USB connectivity, which you wont need. With the Uno, you wont need any programmers... etc, and I bet you can have what you want up and running in less than an hour, even as a newb... Those functions are very simple to use.

I have a 5v power supply schematic I can send you which I am using on my climate control which has several considerations for the automotive environment.

Thanks,
Hans

Ed Scherer

Update: first, thanks again for recommending that I look into microcontrollers. I got a bunch of AVR (Atmel) stuff to play with (mostly Arduino stuff, but also some raw chips) and will be spending some quality time with that over the next few weeks as time permits. Sometime, I probably should look at the PIC stuff, too, but I needed to start somewhere and decided initially to evaluate AVR. That stuff just arrived yesterday, so I haven't really had much of a chance to mess with it yet. The Arduino Nano is impressively miniscule, BTW.

In the meantime, I got a PCB designed that should work based on my initial (primitve, I guess ) design. Might go ahead with this, might not. Or might try to reduce the size by going with more SMT. Or might just wait until I can whip up an AVR-based solution. Need to think about it more.

hans14914

Ed,

I have a build house that can get those boards done for you in inexpensively if you need an affordable quick-turn prototype. PM me if I can be of any help

I am curious what U2 is though, looks like it has some considerable thermal requirements... must have p-channel trans in there.

Thanks
Hans

Ed Scherer

Sure, I'd be very interested in any recommendations. I was looking at ExpressPCB, but that was just based on a quick web search.

U2 is that Infineon BTS730 that I mentioned earlier. It does indeed have an on-chip high-side driver. If it weren't for the complication of wanting to use the factory 928 rheostat, this chip would provide a one chip, two-small-caps solution. Pins 4-7 and 14-17 are all Vbb and leadframe connected; need all the copper on the PCB for heat dissipation. Probably gross overkill (can drive quite a bit of current: something like 15 A) with careful thermal management. That's a side of this stuff that I never really knew how to do properly (thermal modeling, etc.); I just figure if you over design the heat sinks and try not to restrict air flow too much, you'll probably be OK. In this particular case, I noticed that some of the datasheet specs were based on the use of 6 cm2 of copper for the eight Vbb connections. I've never added up or measured the actual current requirements for instrument lighting (whatever part goes through the dimmer circuit); all bet it's under 1 A when LEDs are used. So... if I go another round on PCB design using these chips to get the size down, I'll probably reduce that Vbb connection area quite a bit.

You also mentioned earlier that you had some power supply designs for use with the microcontrollers that require 5 V. I'd be interested in those, too. I'd probably otherwise just whip up something based on the chunky old parts I have on hand (7805s, caps, etc.)

borland

I think you've lost sight of the project goal.

Wasn't the fact that there are no 'reasonably suitable off the shelf components', the reason why it was suggested to substitute a microcontroller in one of the Ardiuno compatible development boards?

With custom PCB manufacturing, wouldn't an analog circuit in this case be the simpler, quickest, lowest cost, and most reliable solution?

Ed Scherer

Actually, I'm not 100% sure of why that was recommended (Hans did specifically mention small package size and flexibility, though). Other criteria may have been cost, availability, and familiarity; I'm guessing, though.

I'd like to more fully understand what you're getting at, though. Do you think I'm ignoring or not seeing the potential of the (for example) existing Arduino products? Or generally heading down the wrong path? Or are you wondering why I'm looking at microcontrollers at all if I want to go with a custom PCB anyway?

Could very well be. I started with almost no knowledge of microcontrollers, how much support circuitry they'd require, what existing "kits" (like the Arduino) were out there, etc. I'm just looking for a solution. Low cost is of course an advantage. Simplicity is an advantage. Appropriate packaging is a big advantage.

I've already got something that works (the DIM14 with the INA122 front end), but that's a rather crude, chunky, $80 solution that I still haven't attempted to package neatly.

I'm still waiting for the BTS730 chips (they're back-ordered through Digi-Key). Being that they're only available in a SOIC package, though, pretty much calls for a PCB.

FWIW, I think component parts are probably somewhere around $15 for a INA122/BTS730 solution (the significant parts costs being the two ICs and a nice 5 position terminal-block style connector); this doesn't include the PCB and enclosure, though. Make the PCB and stick it in a suitable enclosure, and this is a sufficient solution (for me).

To be honest, the microcontroller-based alternatives aren't looking all that obviously superior to me yet, as I think about costs and what all is involved when you address power supplies, drivers, and packaging everything into a module. However, I'm certainly not qualified to pass this judgement yet and would definitely defer to those of you who work with this stuff. Show me what you've got!

So... at this point, it looks to me like there are lots of viable solutions. The best solution is admittedly just not obvious to me because I don't have enough experience. Remember, I'm a software guy (with a little hobby and academic hardware experience in the distant past) trying to stretch a little to do some hardware stuff.

Heck, I'd be glad to just roll with anyone else's solution. Design something, build it, and sell it to me!

Regardless, I'm still interested in microcontrollers just to learn about them, if nothing else. I'm already really intrigued with what I've seen and can imagine all sorts of cool devices that would be easy to design and build.



Perhaps it would help to list the goals/requirements for the dimmer module, so everyone is on the same page. I'll offer mine:

Dimmer module must:

• interface with the factory dimmer rheostat and preserve its functional behavior (i.e., control dimming in a linear fashion from dim to full bright as the rheostat wheel is rolled)
• work for a mix of LEDs that have different intensity/voltage responses
• be able to drive all dimmable instrument lighting
• be in a single enclosure that is small enough to fit somewhere under or behind the pod
• be able to be connected/disconnected using a single connector
• be reliable
• not generate objectionable RFI (particularly in the AM/FM broadcast spectrum)

Desirable, but not strictly necessary:

• use a terminal block connector system with a terminal block header in the enclosure and a terminal block plug that can be attached to the car's wires
  (makes it very easy to install)
• be low cost
• be able to be built from readily-available parts by a typical 928er (I'm starting to doubt that this is feasible)

Here's what would make me happy (from an installer's viewpoint): I get a black box. It has one (probably 5-position) connector attached. The plug part has screw terminals. I pull that off. I cut the two wires to the car's dimmer rheostat and attach the four wires ends to four of the five terminals on the plug. I have to add a ground wire from the fifth terminal to a ground somewhere convenient. I plug the connector into the black box and then stick (or nylon-tie or whatever) the black box to something under the dash. Done.

hans14914

Ed,

Shoot me a PM with some contact info, and we will figure something out. I can certianly get you cheaper boards than Express PCB. Is that what you did the layout in? If so, send me your schematic, and I and I will capture it for you in Eagle or DesignSpark so you can get a non proprietary Gerber to shop around with.

If the infineon based circuit is tested and works, then that is a good option. However, if you have issues getting it to work, I would say we can get a stripped down version of the arduino to work. After you have the code, we can just burn it on new chips. Mot of the extras arent necessary... no USB/Serial bridge, status LEDS.... they can all be eliminated.

Stay away from a 7805 for automotive duty. Alternator load dump tends to fry them. Use an automotive grade regulator with a diode in front to prevent improper installation. A cheap 7805 substitute, which is automotive grade is the LM2940T-5.

Thanks
Hans