Modifying LED lighting/CC-driver

[jitter]

I'm presented with a (hypothetical) challenge that needs to be solved on a limited budget.

I need to halve the light output of an existing LED module, preferably using the existing constant current source LED driver.
I could do that by shorting out half the LEDs, but then the current source would need to drop below its lower voltage threshold.

The driver we're talking about does 300 mA CC in the voltage range of 24-40 VDC. The LED-driver is powered from 230 VAC mains.

Right now, the LED module is driven at 36 V to achieve the 300 mA CC.
If I halve the number of LED's that need to be driven, it would need to drop to 18 V to achieve the 300 mA CC, right?

But that's below its 24 V lower limit, so what would happen?
Note: I don't have an actual driver to try on my DC load, at this point it's a hypothetical exercise.

Thanks for any insight.

[HackedFridgeMagnet]

Semi opaque diffuser? A few bits of tracing paper maybe.

[jitter]

Assuming its isolated and a flyback what may happen is the Vcc winding of the controller may not get high enough to keep the controller on. If it is isolated look for how they are sensing the current and adjust the current sense resistor. If it's primary side regulated adjust the current sense resistor in the fet source leg.

I did have the opportunity to have a look at the driver's insides but, unfortunately, it's potted.
It was indeed of an isolated design, the label carries the double square symbol, and one of the components sticking out from the potting clearly was a transformer with lots of yellow isolation tape, which, if it were just an inductor, I doubt it would have.

For a bridgelux V8 for half the light output the current must be reduced to 150mA from the typical drive current of 350mA, the typical voltage drop goes from 15.5V to about 14.5V.

I also had the opportunity to drive the LED-module with my lab supply set to a current limit of 300 mA. And you know what... I utterly forgot that when I turned the current limit down to 150 mA, the voltage only dropped by a couple of Volts. But in this situation, all LEDs stayed connected, and I only manipulated the current.

Surely by halving the number of LEDs while maintaining the same current, there's no other option than for the voltage to be halved too?

edit

Oh yea, you may get some color shift at the reduced drive current. The manufactuer only gurantees CRI and wavelevgth at specific drive currents. I've personaly never noticed any diference. My experince is limited to Cree and Bridgelux LEDS though.

I'm aware of that, but the project is of a utilitarian nature, a slight shift in CCT or deterioration of CRI are acceptable.

[jitter]

Semi opaque diffuser? A few bits of tracing paper maybe.

Cheap indeed, but the aim is also to lower power consumption.

As is the unit would draw 12.5 W total. The current units to be replaced draw 8.5 W total. We're talking about 10 units and roughy 5000 hours operation per year. An increase of 200 kWh/year while having the same amount of light is not particularly desirable. But thanks for the suggestion anyway.

[jitter]

Series/parallel.

6 strings with each 12 LEDs in series are parallelled, giving a total of 72 LEDs.

[jitter]

I will consider this option, it would also negate the shift in CCT but possibly at the cost of an uneven light pattern in the opaque cover.

Thanks for thinking along with me.

[Ian.M]

You are only going to save about 25 Euro / year if you halve the consumption of all 10 units.   That's only 7.5 Euro/unit over three years, so whatever you do has to be *CHEAP*.   

First check if the LED drivers are dimmable.  If so, and there aren't too many separate circuits,  a TRIAC dimmer with the control pot locked at 50% would be the cheapest/easiest energy efficient option.

If not, as others have suggested, short out four LEDs in each string for 2/3 the light output and a similar reduction in power consumption @ total Vf of 24V.  That should get the consumption down to about the same as the old units +/- a bit depending on the driver efficiency at low loads.  Live with the extra light output unless anyone bitches enough to make it worth fitting additional diffusers.

[Marco]

Find the shunt and double it?

[jitter]

You are only going to save about 25 Euro / year if you halve the consumption of all 10 units.   That's only 7.5 Euro/unit over three years, so whatever you do has to be *CHEAP*.   

Electricity costs about € 0.21 per kWh over here. An increase from 8.5 W to 12.5 W would mean in increase in cost of about € 42 per year or € 12.60 per unit per three years. But the units will need to be in service for a lot longer than that, adding up costs.
The lowering of the output is indeed going to be done while weighing the benefits against the costs, and it was already clear to me that the budget for that would need to be pretty limited.

Edit:
My eye was just caught by a competing unit with similar output but at a far better price and possibly better quality too. I'm going to request they send me a unit and inspect its internals. Maybe that one has better options for modifications, and it would free up a bit of the total available budget to be spent on modification.

First check if the LED drivers are dimmable.  If so, and there aren't too many separate circuits,  a TRIAC dimmer with the control pot locked at 50% would be the cheapest/easiest energy efficient option.

They are not.

If not, as others have suggested, short out four LEDs in each string for 2/3 the light output and a similar reduction in power consumption @ total Vf of 24V.  That should get the consumption down to about the same as the old units +/- a bit depending on the driver efficiency at low loads.  Live with the extra light output unless anyone bitches enough to make it worth fitting additional diffusers.

Humans are not good at detecting differences in light output below 30%, especially when light sources are not positioned next to eachother. That's why I am now indeed considering the shorting out option of about 2/3 of the LEDs and keep things very cheap indeed. But I would prefer to underrun the LEDs and get a multiple of the expected lifetime making the fixtures a fit and forget item for at least one or two decades to come.

The challenge I'm facing is the need for big round impact resistant fixtures (30 cm diameter). But they typically come only in 1000-1200 lm consuming 12-14 W. That's a bucket of light especially when you consider the 690 lm of the current non impact resistant units already comfortably exceeds the need. Those 8.5 W are going to be a target to meet or exceed (uhm... stay under, I mean).

[jitter]

Find the shunt and double it?

Be my guest.

[Kilrah]

I utterly forgot that when I turned the current limit down to 150 mA, the voltage only dropped by a couple of Volts. But in this situation, all LEDs stayed connected, and I only manipulated the current.

Easy then, just put a 5W 240 Ohm resistor in parallel with the LED string, so that about 150mA are directed away form the LEDs  It's as cheap as it goes and your voltage will as you mention stay within range.

Of course that would mean equal power consumption for half the light output, but since the goal has been presented as "halving light output" and not "saving half the power" this could be acceptable.

[Ian.M]

Another option that *may* work is to stick a PWM circuit (e.g. 555 + MOSFET) between the driver and the LEDs.  The driver should go into a low power CV mode if the LEDs are disconnected, but it may stress it more to be toggling between that and full output >100 times a second.

[Marco]

Be my guest.

That through hole resistor mounted vertically near the output is a very likely candidate.

[Marco]

Oops, yeah swapped it around.

[jitter]

Time to bump up this thread.

I decided not to go with the fixture that this thread was about so far and go for the other unit instead. Not only cheaper, it also seems to be more sensibly desinged.

That first fixture was specified at 12 W and 1200 lumen with a power factor of 0.9. In reality, it did 12.5 W with a PF of 0.77. The light output was that of the LED-module without the losses in the opaque cover.

I now have one of the other fixtures and I must say I'm more impressed with this one. Specified a much less efficient 14 W and 1000 lm, it seems to do a similar light intesity, but with the losses in the opaque cover accounted for. Also, the real consumption measures 11.6 W at a PF of 0.95. Not bad. I wonder why they don't specify it at 12 W...

It has a similar CC driver that does 300 mA in the range of 27-40 V.

The major difference is the number of LEDs. The former unit having 72, this one having 96. The same 300 mA is now divided across 8 parallell strings of 12 LEDs each (vs 6 strings of 12).
That means that the LEDs are driven less hard. A significant difference is that this is a normal PCB, not a metal core one. The (large) copper pads are all that provides heatsinking. But it doesn't seem to run any hotter than the previous unit.

But I still need to lower the power consumption and luminous flux. The good news is that the driver is not potted. I would still like to modify this driver to a lower current, and I am now looking for the current sense resistors. I have added some photos of the driver.
Personally, I'm thinking that maybe R1..R4 might be them (they're all parallelled) as they're in series with the source of the FET (JCS 4N65B), and as suggested by AcHmed99 that would mean primary side current sensing. Agreed?

[Ian.M]

Nothing else makes sense.  There certainly isn't any secondary side current sensing, and there are no other primary side low-ohm resistors that could sense the actual chopper current.  I'd start by removing one of them to see if the output current drops by 25%.  If that works as expected you should be within half a watt or so of the old fitting.

Don't forget to budget for buying a couple of spare units before the suppliers change the design on you!

[jitter]

Nothing else makes sense.  There certainly isn't any secondary side current sensing, and there are no other primary side low-ohm resistors that could sense the actual chopper current.  I'd start by removing one of them to see if the output current drops by 25%.  If that works as expected you should be within half a watt or so of the old fitting.

Don't forget to budget for buying a couple of spare units before the suppliers change the design on you!

Thanks for confirming that, I will try.

As is, there are three 6R80 and one 9R10 in parallel giving a combined resistance of 1.81 Ohms.
I will remove one 6R80 resistor and measure again.

I am currently testing the driver on my electronic DC load as it left the factory, but it seems that I can only use its CV mode, all the others (CC, CP, CR) instantly send the driver into protection. Guess the control loops in both devices are fighting eachother, kinda expected that. I will also try it on the LED panel and a resistor.

However, with the DC load in CV mode, I get a useful result, and it seems that the driver will go quite a bit lower than the specified 27 V min. before becoming unstable.

There is a lot of ripple on the output leading to quite a bit of flicker, and I'm sensitive to that. The first fixture had way less of that, but it had much beefier filter capacitors on the output, totaling 660 uF whereas this one must make do with a mere 100 uF for, basically,  the same load. Hmmm...

How would I improve upon that, an electrolytic parallel to C4 (output)? Or would a suitable (400 V) electrolytic cap parallel to C1 (primary) be a better option?
Why did they skimp on the primary side ripple filtering? To save cost or get a better PF (or both)? I don't think this driver has PFC...

[jitter]

Can you get the controller part number? I can't make out what one in those pics.

Hope you can help, it's one of those Chinese make ICs that seems hard to get data on... or my Google Fu is not strong today...
The code seems to read "IAFSF". Posted it in the "What's this please?" thread too...

[Marco]

The last F looked a bit dodgy, smd iafs got a hit on the first page. MP4026, the I turned out not to be an I either. They really like to make lookup easy.

The secondary side average current is a simple inverse linear function of the mosfet source resistor to ground, the other resistors are part of a ZCD.

[jitter]

Thanks for thinking along, it's very helpful. So it does have active PFC after all...

By experimentation, I had already confirmed that R1..R4 are indeed the current sense resistors.
I found that I need to remove one 6R80 and the 9R10 to get close to the desired light output, so I can modify these units without any cost.

The result is a 6.8 W unit with still more than the current unit's light output @ 8.5 W.

I took a photo with 4 units side by side to view the comparison on screen (my eyes see less difference than the camera).

From left to right:
1) New unit, driver modified to 170 mA.
2) New unit, driver modified to 223 mA.
3) New unit, driver not modified, 300 mA.
4) For reference the current unit, not modified (note: 2700 K, the others are 3000 K).

Tech.data:
1) CC 170 mA @ 32.8 V; 5.6 W; mains consumption: 6.8 W (PF 0.89), about 82% efficient.
2) CC 223 mA @ 33.1 V; 7.4 W; mains consumption: 8.9 W (PF 0.93), about 83% effiicient.
3) CC 300 mA @ 33.5 V; 10.1 W; mains consumption: 11.6 W (PF 0.95), about 87% efficient.
4) Ref. unit: CC 350 mA @ 21.5 V; 7.5 W; mains consumption: 8.5 W (PF 0.76), about 88% efficient.

Tonight I will compare the different units and choose which mod. I will apply.

Edit: with 170 mA current flowing, the LED panel doesn't even get warm, so a multiple of the specified 35,000 hours should be possible. I don't know if the driver will do that too, though...
The usual suspects in the limiting of the driver's lifetime will most likely be the electrolytics. Though Aishi may not be the most well known brand, in lighting products they almost seem to be a given. About 90% of the CFLs and LED-lamps/-drivers I tore down (from whatever brand) used Aishi. Their failure rate also seems quite good, on par with the big boys.

[jitter]

Thanks for warning me about this. I think I'm not going to get into trouble.
Looking at the resistors in the CS/ZCD section, the CS resistors are very low value compared to the ZCD resistors. We're talking a couple of Ohms vs many kOhms. A change in the former doesn't likely upset the driver's behaviour in the protection, as the ratio is hardly changed.

I need two of the ten fixtures to give even less light, so I modified one to 146 mA. Even at that current, the panel still needs 32.6 V.

If I run it unloaded, I can see the voltage repeatedly bump into 52 V and then drop back again, presumably the OVP at work. The specified max. unloaded voltage is 55 V, and an unmodified driver indeed did the same bumping and dropping routine to about that voltage. I don't think that 3 V difference is going to give me problems.

The lowest expected temp will be about 10 C/50 F. But just to be on the safe side, I chucked the 146 mA modded driver and LED-panel into the freezer...

What is it that might happen, failure to start?

Edit:
After almost an hour in the freezer, there appear to be no problems. Driver and LED panel start without a hitch and spot on at the expected current.

BTW, in the end I decided to parallel a 330 uF/63 V 'lytic to the connection on the LED-pcb. It doesn't completely eliminate the flicker, but it's now low enough to be almost unnoticeable for sensitive people like me.

[jitter]

Thanks for the link, I will read it.

I decided to go for the easy option of putting a cap parallel to the LED-pcb connection. Total output capacitance is now 430 uF, which is a bit lower than the dual 330 uF caps of the application note, but I think it's good enough. Even good old fashioned incandescent bulbs flicker, but no one ever noticed, so it doesn't need to be 0.

One thing is clear, and that's that the driver design is minimalistic to save cost, and probably also to limit its expected lifespan to about 35,000 hours. The LEDs don't run hot at all, so they needed the driver to be the limiting factor. 
It's probably a well considered decision, because these fixtures don't go into residential or workspace lighting and some flicker is not really a problem. I chose to add the caps because I don't like the stroboscopic effects LED flicker creates.

Even freezing temps don't seem to phase the modified driver, so I'm going ahead with the rest. I installed two of the units already, and now that it's dark I can confirm that dropping to 170 mA gives the perfect reduction that's around or slightly brighter than it was. Verified with a luxmeter.

Thanks again for the help.

[jitter]

One thing is clear, and that's that the driver design is minimalistic to save cost, and probably also to limit its expected lifespan to about 35,000 hours. The LEDs don't run hot at all, so they needed the driver to be the limiting factor. 

Silicon doesn't go bad. The caps or in this case cap is likely the limiting factor on lifespan. Unusual that they have just one 100uF cap that low a value will typically have a high esr, worse with colder temperature. Over extreme temperature variation the ferrite core might crack from thermal shock to. I've never seen that happen though. All massed produced stuff is done to the bare minimum, sad.

I just rembered that when I installed the units I'm now replacing, I opened one driver. It wasn't easy as Philips decided that they wanted it ultrasonically welded shut, but I persevered... 
The bodge wire on the bottom is my doing, I snapped off a corner of the pcb... ooops.

Apart from the pcb material, this driver actually looks a bit beefier and a bit better designed than the ones I have now modified.
That's remarkable because this is a consumer fixture specced at just 15,000 hours... the others are supposed to be industrial and specced at 35,000 hours  .

The topology looks similar, but it has much better smoothing and hence no or very little flicker.
But it must be said that these are "8 W" (actually 8.5 W) 690 lm fixtures that have been superseded by a 7.5 W 620 lm model. Perhaps the driver in that is a bit more economically constructed now too.
I'm going to place these units in another part of the building that has higher ceilings, so eventually I will find out about their real lifespan. 15,000 hours doesn't seem that much, OTOH they are driving the LEDs harder... (7 sets of three parallelled LEDs in series, so 21 total).

[jitter]

Don't forget to budget for buying a couple of spare units before the suppliers change the design on you!

Well, I did buy two extra units just for that...
But while modifying the drivers, I noticed one with a more recent production date printed on the underside, same type though. Guess what, the sense resistors are different values on the newer one...

The older one (2015-10-15) has three 6R80s and one 9R10 parallelled (giving 1.81 Ohms), the newer one (2016-05-26) uses three 7R50s and one 6R80 (giving 1.83 ohms). Guess they thought "meh... close enough".