The curse of Poor Power Factor

[___]

Hacked the Carlo Gavazzi into a lights circuit to check what these MR16 down-lights really use. As expected, the real power is on the low side (mixture of 5W and 7W lamps totalling 32W, plus some losses on the transformers). But look at the apparent power!

Theory: by retrofitting the MR16 to the existing wire-wound iron-core transformers, I now have a 5W lamp on a 50W transformer. Unladen transformers have notoriously bad power factor, typically less than 0.4.

With all the lights on at night, we're running maybe 200 watts, but over 600VA. I don't like that.

So some solutions. First one, rip out the MR16s and go GU10. Another solution is to place more of them in parallel on the same transformer. This would mean I can never go back to halogens, not that I would want to... but a future home-owner or tenant (if the unthinkable happens and we have to sell or lease it in future) might. Third option, replace them all with the electronic variety, some which will actually work fine in this low-wattage application, eg:

http://www.gelighting.com/LightingWeb/na/images/led-mr16-transformer-compatibility-table_tcm201-23431.pdf

http://za.rs-online.com/web/p/lighting-transformers/8117140/

Right now I'm thinking, just parallel them up and add a 5A fuse for safety, in case someone tries to use a halogen.

What say you guys?

[Jaco De Jongh]

@plonkster, I went for GU10, did not see the logic in DC-AC conversion and on the light side AC-DC again. To eliminate  the transformer losses, I Used 3 watt 220vac globes. If i sell in the future, the next owner can use what ever he prefers.

[DeepBass9]

Can you explain this for us simpletons.....

[Guest]

Same as JdJ, took out all transformers, 220v lights sockets across the house.

[___]

1 hour ago, Jaco de Jongh said:

on the light side AC-DC again

But that happens anyway, it's just the means that differ. An AC-driven LED simply has a different kind of SMPS in the back, but it does the same thing: Change it back into DC to drive the LEDS. It just does it as significantly better power factor (about 0.7 by my tests, which isn't brilliant, but a great deal better).

The background to the story is that the house already had the MR16s when we bought it. The simplest solution at the time was a straight swap to LED, which worked perfectly. Even if some power is lost in the old iron transformers (about 1 watt, believe it or not), it still does significantly better than the halogens that were in there before, and when driving them from Eskom power (where you pay for real power and not apparent power) the power factor really doesn't matter. It would not have been worth the extra investment in changing all the fittings, even if I had known about the problem back when I started :-)

Of course I had known about the problem for about a year, ever since I bought an old oscilloscope from gumtree. This is just the first time I looked at the totals per circuit.

I now want to improve the power factor. If it was a simple problem of a trailing current waveform (ie inductive load) you could throw capacitors at it, but the problem is more complex than that. I'll throw in my stock picture below. I think that loading the transformers down more will help, I'm just concerned about what happens if for whatever reason you want to go back to a halogen lamp. Don't want to burn anything. I can fix it easily by fusing the DC side. Just wanted to know what you guys think.

Here is an article about why the power factor is so poor at low load. Basically you have too much magnetic flux in a too large iron core that you're dragging around instead of pumping it out on the secondary side.

1 hour ago, DeepBass9 said:

Can you explain this for us simpletons.....

Okay, so us solar people know that P = IV, power is volts times amps. Now this is true for DC, but with AC it might not be. The first thing you have to understand about AC is that the voltage we normally use in everyday speech is the RMS voltage. It's an average, so to speak. If you think about the sinus waveform with the hills and valleys, and you take the top 30% of the peaks and dump it into the valleys, you'll get a flat line at 230V. The peak is actually at 325V.

Further, we know from Ohms law that I = V/R. Where R is constant, this means that your current waveform looks exactly the same as the voltage waveform. The most current is drawn when the waveform is at the top (325V), and when the voltage is at zero the current is also at zero. This is what we call a purely resistive load. The VA is exactly equal to the Watts, and the power factor is said to be unity or one.

Whenever you introduce magnetics, inductive stuff, this breaks down. Now, when the voltage is at zero the current isn't (because the magnetic flux in the device is still trying to hold the current up going the other way), and then halfway into the next cycle it has finally finished fighting the magnetic flux of the previous cycle and the current drops to zero and changes direction, but this happens long after the voltage has done so. So your current and voltage waveforms no longer line up, peak current is no longer drawn at voltage peak, and neither is zero current drawn at zero voltage.

With the LEDs the problem is actually a little worse. The current waveform IS in phase with the voltage, but it does almost all the work by the time the voltage waveform hits 40 volts, and then tapers off, so you have more of a saw-tooth waveform for the current rather than a nice smooth sine wave. You have a 5W lamp driven at 40V for all practical purposes, which means it will draw 120mA at the peak of the current waveform.

We then calculate the VA as the peak current times the RMS voltage: 0.12 * 230 = 27VA.

The real problem: The inverter cares about the peak current. It will still only take 5W out of the battery, but the power electronics are limited by their current capabilities. If you have a 100VA inverter (for arguments sake) you can literally only run three 5W lamps from it with such a poor power factor.

Below is my own measurement of one of these lamps, and this measurement was done on the DC side and doesn't even include the poor power factor of the wire-wound transformer. Notice how the lamp does all of its work using only half the voltage cycle (the saw-tooth waveform is the current waveform).

[Jaco De Jongh]

24 minutes ago, plonkster said:

But that happens anyway, it's just the means that differ. An AC-driven LED simply has a different kind of SMPS in the back,

@plonkster, when in conversation with you I always fall in the same trap, I assume that you know what I want to say, and then say to little, I am aware of the conversion at the end, but chose this route because of the "better conversion". Will try and elaborate more in the future, but then I feel it gets to long. I had to change all the fittings also, but in my case it was called for, the old fittings were shot after years and years of excessive heat from the halogens. 

[___]

1 minute ago, Jaco de Jongh said:

chose this route because of the "better conversion".

I get it. I can get deliberately technical and all "well-actually" on people, I do apologise :-)

For me the obvious reason to go GU10 is that it is cheaper on a new install. Lamps cost about the same, plus no transformer (which adds about another R100 per lamp). My house already had MR16. Transformers are very efficient already (if you ignore the power factor issue, which should not be confused with efficiency), so no real advantage there. Plus, I really have a feeling, having deconstructed both GU10 and MR16 lamps that have failed, that the SMPS in the MR16 is less stressed, so that I really would expect the MR16+transformer to last a bit longer. On the MR16s I find that one of the series LEDs fails, but the SMPS is fine. On the GU10s I find that the LEDs survive but the SMPS fails. My experience might be just that: my own limited experience.

[Jaco De Jongh]

Just a question, I haven't tested this , but if you want to stick to transformer, maybe worth wile testing the efficiency and power factor of a 220-12Volt 5A or 10A Switch mode Power supply. Maybe use one per room supplying all the lights in there, Last price I got on a 10A MeanWell was something like R260.

Do you know of a reason that a Switch Mode cant work in a application like that?

 

Edit: A few Small wiring changes in each room.

 

[___]

26 minutes ago, Jaco de Jongh said:

Do you know of a reason that a Switch Mode cant work in a application like that?

Nope, should work just fine. Only thing, it has to go in the ceiling, so I want to stick with stuff that is unlikely to blow up in spectacular burning-down-the-house fashion. Iron-core transformers strike me as okay in that regard. The lights in the ceiling are already daisy-chained, I can literally find the first one, and just chain the rest off the secondary side without even getting into the roof. I think that's what I am going to do, just fuse it for safety. If that doesn't work, I'll use an electronic transformer to drive pairs of lamps. Or something.

[SilverNodashi]

GU10, it just maked life so much easier. No need to worry about COC and regulations about cables fixtures in the ceilings, etc. 

BUT, if you want to go absolute pure (i.e. run 100% efficient) then you should use 3.7V LED globes and 3.7V wiring throughout the house. You'll need very thick cables though. So there's a trade-ff for convenience + easily replaceable lights in the future VS running everything 100% efficient.