Power droop along an LED strip
I’m investigating power draw of pixels on an LED strip now that I have some better measurement tools. How much power do, say, five white pixels draw? It turns out that the answer is: it depends on WHICH five pixels, specifically, how far along the strip the pixels are from the power supply.
For testing, I used a 4m strip of 240 WS2812B’s, spaced at 60px/m. When totally dark, the strip draws 207mA. The MCU draws 20mA. The measurements below are exclusive of those values.
When powered from just one end of the strip (as is the most usual setup), here’s how five pixels’ power usage changes over the length of the strip:
Lighting pixels number 0…4 to full white draws 190mA.
Lighting pixels number 235…239 to full white draws 164mA.
That’s nearly a 15% difference in power draw from one end of the strip to the other!
However, as a next step, I powered the strip from both ends, and things changed. The power draw of the five full-white pixels was 190mA at either end. If I lit five white pixels in the middle of the strip, the draw was 188mA, so there’s still a little ‘droop’ in the middle, but very little compared to the power sag all along the strip if it’s only powered from one end.
So, my conclusion from this is that power sag along the length of an LED strip can be significant. The power delivered to the ‘far’ end of the strip is not only lower in voltage (which we already knew), but that the pixels at the far end of the strip also draw less power, fewer milliamps, and they therefore produce less light! Applying power at both ends of the strip helps tremendously with this.
I’m trying to write code to estimate and manage power usage, and you can believe me when I say that I definitely did not set out to learn how power usage varied along the length of an LED strip! So at this point I’m just sharing what I’ve observed in case it helps someone understand what may be going on on their setups.
Mark, if you work out this power budget capping auto-brightness algorithm I’m pretty sure you’re thinking about, I will mail you local (cincinnati) beer. You deserve it.
The impact of transmitting low voltage over distance is a challenge. A voltage drop of 1 volt in a 5 volt system is a 20% loss of power. Even a 1/2 volt loss is a 10% decrease in power. I will be interested in your findings.
Lets face it, the flexible PCBs used in these systems are pretty average quality. There is going to be significant resistance down the strip. Not only will this affect the current available to drive the LED chips. But track capatitance will also be rounding off the clock edging and introducing ‘ring’ to the edges too. Where this is ‘regenerated’ by the LED chip is a great help here.
Length limitations will quite quickly not only limit the power for brightness, but also prevent the data signal from arriving at all!
Ensuring the 0v reference between the MCU and the drive PSU are pretty critical too. I had a loose connection here and the the results looked like timing issues.
On very long strips I re-introduce power every meter from big fat cables (not only at the ends) now, and now considering breaking the data line at intervals too. Not only because of signal degradation, but also down to update speeds. A 1000 LEDs takes ages to push through chip-to-chip.
So when your strip ‘don’t work’, there are so many things that have to be checked. Looks like a vary long Wiki page will be needed…
Adam, I had heard of some folks using RS485 (I think) drivers to freshen up the signal when driving it long distances. Our burning man CORE project did this because all of the strips were driven from a central point. And we dropped power at both ends and every meter on the meter. (and then we took it all off before burning the Inchanted Forest - Indiana state CORE.)
You have now seen the first order effect of resistive voltage drop along the flexible PCB. With that chip, a lower voltage at a given pixel means lower current draw from that pixel (tho it may not be linear and depends on the chips used).
The other factor you need to pay attention to is how the power draw of OTHER pixels will affect this. Try measuring the on/off current difference of those last 5 pixels when the first 100 pixels are on. It will be less, because the resistive voltage drop of those pixels also affects any pixels upstream or downstream of them.
(This may be harder to measure accurately because the delta current will be a smaller proportion of total current. If so, try measuring with the last 5 pixels on/off and the penultimate 10 pixels before that on in both cases).
This is not too hard to model, at least for power from one end.
Can your multimeter measure the end to end resistance of the power conductor of the strip, and separately, of the ground conductor? (while the strip is unpowered of course)