EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: damien22 on April 02, 2019, 10:29:10 am
-
Hi There,
I have somewhat a bit of a challenge. For equipment, I need to control more than 100, 24V halogen lamps simultaneously. They are on for only a few 3-4 seconds at a time, every 15s.
Each lamp is controlled through optical feedback with a photodiode.
The thing is, I want precise control while keeping it as cheap as possible.
The current implementation is a large AC/DC power supply bringing down the voltage to 25V, then each lamp has its own buck converter controlled by a micro that takes the photodiode as feedback and adjusts the PWM signal.
Another implementation would be to bypass the power supply, just have rectifier/cap and directly feeds 300VDC to a buck converter, but this would require either very large coil or high frequency.
For the high frequency, the problem arises that the duty cycle would only be around 8% for maximum power. On the controller, it then becomes difficult to control the lamp as the definition of the PWM signal is too large.
One idea was to chop the PWM signal with another, fixed and higher frequency signal through an AND gate to generate a composite hashed PWM, but I'm not sure if that would be a good solution and if then the switching losses would be acceptable.
Another concept is to chop the main AC with a triac before to drop the voltage, rectify it and feed it to the buck, but in that situation, I'm not sure what it would be to draw 10kW of power on the main being chop only at specific point on the sine.
Any other ideas or pointer would be very welcome as I'm kind of stuck so far, the 24V supply so far is the only viable choice I see.
-
The 25V PSU and PWM circuit is the most sane solution: no buck converter, just directly PWM the lamps.
Why use halogen lamps? Use LEDs of course. It will make this much easier. Even if it costs a little more, the reduced energy consumption will make it cheaper in the long run.
There are LED driver modules available which can take a low power PWM signal and change the LED brightness, without having to faff around with buck converters or power supplies or switching high currents.
-
What's the current of each lamp? You mention 10kW ... that would mean 10 000 / 100 = 100w per lamp?
My thoughts went to using 24v lead acid batteries (or two in series) to power segments of your 100 lamps... for example 10 groups of 10 lamps each to get the power consumption more even instead of so pulsed.
for example
power first 20 lamps for 3 seconds from battery
power the other 80 for 3 seconds from power supply
now top up the batteries for those 2 groups of 10 leds for 8 x (3s + 15s) = 144s ... you could use lower current to top the batteries... so you'd have 10 lamps at 100w each = 1kW and 1000/24= 40A discharge from lead acid battery is doable... then charge for 2 minutes at 10-20 A
power lamps 21 to 40 from battery
power lamps 1-20 and 41-100 from power supply
top up batteries for those 2 groups for 144 seconds
Maybe drive more than 2 groups of 10 lamps from batteries... basically get your power supply constantly topping up batteries to get them behaving as buffers in a sense.
-
Definately DO NOT PWM incandescant lamps directly. The changing filament resistance (startup current) will reak havoc on your drive transistors, spending most of the energy trying to keep them glowing at all. You want to use triacs and chop the incoming AC to the bulbs. Just feed 24VAC down a few bus lines and chop it at each lamp with a few control lines.
-
Another option would be as many 28V AC transformers (to have a bit of headroom for switching losses, and so the drive doesn't need to go all the way to 100%) as is convenient to split up the total load current, and a TRIAC + OptoTRIAC for each lamp. Do the firing for phase angle control with a fast MCU driving a multiplexed 10x10 matrix of the Opto LEDs. The multiplexing frame rate would have to be high enough that the variation in firing angle between lamp groups is insignificant.
An alternative to the matrix would be a 13x 74HC595 daisy-chain driven by a fast SPI port, which would let all the TRIACs to be triggered in a phase angle timeslot be triggered simultaneously.
Low cost multiplexing for the photodiodes is more problematic as its desirable to average their readings over at least one mains cycle. It may be possible to put them in a matrix, in series with low leakage signal diodes with a capacitor across each photodiode for averaging, to minimise the number of photodiode amplifier channels required.
-
Definately DO NOT PWM incandescant lamps directly. The changing filament resistance (startup current) will reak havoc on your drive transistors, spending most of the energy trying to keep them glowing at all. You want to use triacs and chop the incoming AC to the bulbs. Just feed 24VAC down a few bus lines and chop it at each lamp with a few control lines.
How is phase control any better than PWM?
PWM is pretty standard for dimming a filament lamp run on DC. I've done it before without any problems. The filament resistance will always be lower on start up, just make sure the switch, be it a transistor or thyristor, can handle the surge. In any case, the power dissipation in the switch will be less, when the lamp is dimmed, as the low duty cycle will outweigh the increased switching losses.
-
If the lamps are sufficiently high power that startup current is a concern, then it's a good idea to limit that current to ease the requirements on power supply etc, and also prevent accidentally blowing something up.
One option is to use an MCU to directly implement a current-limited buck converter: many MCUs have PWM modules that can be configured with fault inputs, such that when an input is asserted the PWM cycle is cut short. If the PWM output drives a low side switch + flyback diode with an inductor in series with the load, and the fault input is fed by a comparator (possible one built into the MCU) connected to a current sense resistor, then voila, you have a current-limited low-side buck converter. The MCU can then be used to read the photodiode and implement the closed-loop optical control plus whatever else you want. You should be able to find MCUs with enough PWM modules with this capability to do at least three channels, possibly more.
-
Definately DO NOT PWM incandescant lamps directly. The changing filament resistance (startup current) will reak havoc on your drive transistors, spending most of the energy trying to keep them glowing at all. You want to use triacs and chop the incoming AC to the bulbs. Just feed 24VAC down a few bus lines and chop it at each lamp with a few control lines.
Howå is phase control any better than PWM?
PWM is pretty standard for dimming a filament lamp run on DC. I've done it before without any problems. The filament resistance will always be lower on start up, just make sure the switch, be it a transistor or thyristor, can handle the surge. In any case, the power dissipation in the switch will be less, when the lamp is dimmed, as the low duty cycle will outweigh the increased switching losses.
We're not talking about a single house lamp here, it's 10KW worth of 100W bulbs. Also, if you zero cross switch the triac properly it will make a difference compared to a hard square wave startup.
-
7 of these will give you 112 pwm outputs from a single i2c port.
https://www.adafruit.com/product/815 (https://www.adafruit.com/product/815)
-
Why use halogen lamps? Use LEDs of course. It will make this much easier. Even if it costs a little more, the reduced energy consumption will make it cheaper in the long run.
Because LED at this wavelength (that we need) and power would run somewhere around 300'000$
-
My thoughts went to using 24v lead acid batteries (or two in series) to power segments of your 100 lamps... for example 10 groups of 10 lamps each to get the power consumption more even instead of so pulsed.
That is actually a very good Idea!
-
Another option would be as many 28V AC transformers.
I checked this solution but the cost still remain higher than having a huge power supply, and I prefer not to use dirt cheap power supply as to not having a failure.
Concerning control, it is not an issue, the initial design uses a dsPIC that generates the PWM and measure the photodiode feedback, all this is controlled over a serial daisy chained bus controlled by software.
Concerning driving the bulb directly by PWM, I don't know how much EMI would that generate given that there are about 15cm cables to the lamps. Current limiting would be possible but probably would require very short timing which I don't know can be handled with standards MCU and might require some more analog control. Probably a coil would still be required somewhere or perhaps oversized MOS.
I think it's not easy to provide 26VAC, it would require quite a massive transformator.
-
Definately DO NOT PWM incandescant lamps directly. The changing filament resistance (startup current) will reak havoc on your drive transistors, spending most of the energy trying to keep them glowing at all. You want to use triacs and chop the incoming AC to the bulbs. Just feed 24VAC down a few bus lines and chop it at each lamp with a few control lines.
Howå is phase control any better than PWM?
PWM is pretty standard for dimming a filament lamp run on DC. I've done it before without any problems. The filament resistance will always be lower on start up, just make sure the switch, be it a transistor or thyristor, can handle the surge. In any case, the power dissipation in the switch will be less, when the lamp is dimmed, as the low duty cycle will outweigh the increased switching losses.
We're not talking about a single house lamp here, it's 10KW worth of 100W bulbs. Also, if you zero cross switch the triac properly it will make a difference compared to a hard square wave startup.
On switch per lamp, with the switching interleaved so not all lamps are being turned on and off simultaneously, will alleviate a lot of noise problems.
Why use halogen lamps? Use LEDs of course. It will make this much easier. Even if it costs a little more, the reduced energy consumption will make it cheaper in the long run.
Because LED at this wavelength (that we need) and power would run somewhere around 300'000$
What wavelength is that? Presumably this is used as an IR heater, rather than light.
-
Another option would be as many 28V AC transformers.
I checked this solution but the cost still remain higher than having a huge power supply, and I prefer not to use dirt cheap power supply as to not having a failure.
Did you factor in the cost of the huge copper cables and installation needed to keep voltage drop within acceptable limits? (I was once involved with an SELV traffic junction controller project, power consumption was small in comparison!).
-
Another option would be as many 28V AC transformers.
I checked this solution but the cost still remain higher than having a huge power supply, and I prefer not to use dirt cheap power supply as to not having a failure.
Did you factor in the cost of the huge copper cables and installation needed to keep voltage drop within acceptable limits? (I was once involved with an SELV traffic junction controller project, power consumption was small in comparison!).
Yes, it's actually another post on this forum, we use tin plated aluminum busbar.