In an application, 100+, 100W 24V halogen lamps need to be individually controlled and run simultaneously. They stay on for a few seconds at a time and need to have very constant irradiation.

Furthermore, It needs to be as cheap as possible, but have fine control of the lamp with at least in the 0.5% granularity of the maximum power output.

This system is enclosed and is for lab use.

The first attempt was to rectify and filter main (230VAC) and use a DC/DC buck converter to go from 300VDC rectified to 24VDC being controlled by a dsPIC33.

However, I miscalculated the inductor required and the main issue is, with the correct inductor, the Duty cycle of about 8% which at a frequency of 20kHz only leaves very large granularity for the speed of the dsPIC to generate the PWM.

Another option is to have a large (or several) AC/DC supply to go from the main to 25V or so and then a buck on each lamp, but that would increase the cost significantly, about 2-3k$ for the supplies and large currents will have to be handled.

Is there a cheap, more appropriate topology for this need, or is there a way to increase the nominal buck duty cycle without wasting power and having reasonably sized coil ?


Perhaps a possibility would be to have directly 230VAC and regulate the lamp directly with a Triac although the ripple would be too big at 50 (100Hz) so there would need some sort of circuit behind as well.


Someone came up with the idea of using lead-acid batteries which would lead to the use of a much smaller power supply and drive the lamp directly in PWM without buck, not sure about EMI and cold start.

  • \$\begingroup\$ Why would the ripple be too great? Since a triac works on both cycles of the power line, the ripple frequency would be 100 Hz, which is well above flicker frequency and almost certainly above the filament thermal time constant. \$\endgroup\$ Commented Apr 2, 2019 at 5:56
  • \$\begingroup\$ Please go into much more detail about what you are doing. You seem to be saying that you plan to run 230 volt halogen bulbs at 24 Vrms, which to a first approximation says that you're looking for 1 watt out of your 100 watt bulbs. Or are you saying that your bulbs are 24 volt units? Do you have a choice of bulb voltage? Do you need to control the irradiance of each bulb individually, or are you looking at the overall output? \$\endgroup\$ Commented Apr 2, 2019 at 6:00
  • \$\begingroup\$ It's 24V bulbs running at 100W that should be adjustable from 10-100W. I've edited the question @WhatRoughBeast \$\endgroup\$
    – Damien
    Commented Apr 2, 2019 at 8:58
  • \$\begingroup\$ If you expect each bulb to run at a similar load point simultaneously, running them in a series string would be much easier. You can add a adjustable/programmable current shunt to each bulb for 5-10% adjustment. \$\endgroup\$
    – sstobbe
    Commented Apr 3, 2019 at 16:20
  • \$\begingroup\$ Yes but they need to be individually controlled. would need some sort of current bypass system. \$\endgroup\$
    – Damien
    Commented Apr 8, 2019 at 6:04

4 Answers 4


How many lamps are you running simultaneously? I ask because dropping the mains to somewhere in the 24-48V region makes your control problem easier, and if for example only say 20 of those lamps are on at any one time then you are into the place where cheap surplus 'telecom rectifiers' are a thing.

A 2kW 48V telecom rectifier is not expensive on the surplus market, and 48V would increase your duty cycle in a PWM arrangement significantly, possibly even allowing simple PWM (Watch the difference between RMS and average here) without much in the way of inductors.

TH lamps have significant thermal time constants so flicker is negligible by the time you hit even a 1kHz switching rate.

  • \$\begingroup\$ All of them are on simultaneously which is more than 100. I have to use products that are in production on the market as this is a commercial product. I looked at using power supplies, it's possible but I'm looking if there is another way possible. \$\endgroup\$
    – Damien
    Commented Apr 2, 2019 at 3:04
  • \$\begingroup\$ So you want 10kW++ of power, split over 100+ channels for <$1,000? I am not sure you will pull that off. Note that even if you do a switcher off a DC bus, you probably need a PFC input stage at that power level, and may want three of 'em to allow three phase input. I am sort of wondering about small flyback transformers and varying the frequency to vary the power, no need for output caps if you keep the rate above a few kHz. Use a single modest FPGA to produce all the drive signals. What are you actually trying to achieve, as in what is the end game here, this feels like an XY Problem. \$\endgroup\$
    – Dan Mills
    Commented Apr 2, 2019 at 9:47
  • \$\begingroup\$ Yes, it's quite a challenge. I've searched into flyback using small transformers but I didn't found a cost-effective solution so far. FPGA for drive signal is quite an interesting idea, however, I have no experience with them but will look into it! What do you mean by PFC Stage, Power Factor Correction? The idea was to spread the lamps across the phases somewhat equally. For the phase shift we will have to see upon test but we might leave that for the factory to correct for inductive loads. \$\endgroup\$
    – Damien
    Commented Apr 2, 2019 at 10:55

Cheap and dirty is to use a standard 50 HZ 24V Halogen lamp transformer .Then use a Basic phase controlled light dimmer .Now replace the dimming pot with a LDR .Tie a LED to the LDR to make an optocoupler .This trick has been used for making a crude lead Acid battery charger.

  • \$\begingroup\$ The cost would be way too high. about 10$ a piece without all the other required regulation parts. \$\endgroup\$
    – Damien
    Commented Apr 1, 2019 at 11:07

"However, I miscalculated the inductor required and the main issue is, with the correct inductor, the Duty cycle of about 8% which at a frequency of 20kHz only leaves very large granularity for the speed of the dsPIC to generate the PWM."

What about modulating the duty cycle? Take a running average of your duty cycle and compare it to your target duty cycle. If the average is higher than the target decrease the duty cycle and vice versa.

The trade off is that you increase resolution but will experience more overshoot and undershoot around your target. How much that is a problem I suppose depends on how much halogen lamps behave like low-pass filters and how much light variation you can tolerate in your application.

There is another problem however and I don't know if you've considered it; The voltage you are supplying the halogen lamp with does not have a linear relation with its power consumption. A given voltage at any time does not always correspond to the same power consumption (even if warmed up). A given voltage does not necessarily equate to similar power consumption with another (warmed up) lamp.

If you want precise control over the amount of light you need some feedback at the very least.

  • \$\begingroup\$ This is a valid idea. I had thought about using a secondary clock to hash the PWM with a AND gate, but I need to further check switching losses. \$\endgroup\$
    – Damien
    Commented Apr 3, 2019 at 8:01

4.2A Halogen Bulbs * 24V =~ 100W will draw 40A cold and thus switches must dissipate a lot of heat if there are rapid changes.

Each FET then becomes more expensive than the bulb.
e.g. 45 mOhm @ 44A = 87W surge $46/pair

I suggest you use 240V 100W Halogen Bulb then use 20A Triacs from 240Vac with any feedback you need to regulate.

This also has a significant load to AC supply problems on transformer THD and breaker trip ratings.

Your bulb specs overlooked the cost of switching low voltage Halogen cold currents.

But this will be very noisy for EMI.

  • 1
    \$\begingroup\$ We can control the startup to limit the current. Unfortunately, due to lamp design, and optical properties we cannot go beyond 24V \$\endgroup\$
    – Damien
    Commented Apr 1, 2019 at 11:14
  • \$\begingroup\$ Your requirements are incomplete on cost, thermal rise, per channel and bulb datasheet \$\endgroup\$ Commented Apr 1, 2019 at 11:25
  • \$\begingroup\$ The cost should be as cheap as possible, the current solutions I have runs around 3-4k$ using auxiliary supplies for 100 bulbs. bulbs are custom manufactured, not sure what you mean by "thermal rise". What kind of information you would need ? \$\endgroup\$
    – Damien
    Commented Apr 1, 2019 at 11:27
  • \$\begingroup\$ Whoever voted -1 is this an opinion on lack of a good solution or technical content? \$\endgroup\$ Commented Apr 3, 2019 at 0:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.