PT4115E LED Driver Efficency

[Fish4Fun]

I am using a pair of PT4115E Drivers (Data Sheet: https://people.xiph.org/~xiphmont/thinkpad/PT4115E.pdf ) to drive two different arrays of LEDs.... The circuit is identical to the "Typical Application Circuit" with the addition of a 10uF Capacitor across the LED arrays ( to reduce ripple, as shown on page 13 of datasheet ) and the rectifier diodes for AC operation are omitted as this is a battery operated device.

One array is 12 * 1W Red Leds configured as 4 parallel sets of 3 LEDs in series with a nominal Vf of 5-6V and a max current of ~2A.

The other array is 12 * 0.5W Green LEDs configured as 6 parallel sets of 2 LEDs in series with a nominal Vf of 5-6V and a max current of ~ 1A.

Aside from the two PT4115Es the circuit also has an LM1117 5V Regulator and an ATTINY2313a. The ATTINY2313a is used to provide PWM to both arrays to produce a "Fade-In" effect.

All discussion in this thread deals with power consumption after the "Fade-In" Period has ended and the PT4115E is in "Free Running Mode" ... that is current is regulated solely by the feedback resistor and NOT by PWM.

The target total power consumption for the Red LEDs is ~5W.

The target total power consumption for the Green LEDs is ~2.5W.

I have not attempted to directly measure the current in the LED arrays while powering them with the PT4115Es; however, using a bench supply with adjustable CC/CV outputs I have measured the Current of the LEDs at various Forward Voltages. I feel fairly confident measuring the forward of the LED arrays while powered by the PT4115E yields a good indication of the actual average current flowing through them ....

Finally, my problem // question ... Testing the PT4115Es with their respective arrays, the power consumption of the circuit varies considerably as the Supply Voltage increases.

The RED Array circuit consumes 3.8W ( 8.0V @ 475mA ) .... and the LEDs have a forward voltage of 6.28V ....

Stepping up the supply voltage in 1 volt increments increases the Current & consequentially the Power consumption:

Consuming 6.63W ( 13.0V @ 510mA) ..... and a forward voltage of 6.36V.

So, The input power changes as follows:

0.475A * 8.0V = 3.8W
0.510A * 13.0V = 6.63W

1 - (3.8W / 6.63W) = 42.7% increase in input power

The Power consumed by the LEDs changes as follows:

A forward voltage in the RED array of 6.28V corresponds to a 495mA forward current.... ==> 0.495 * 6.28 = 3.1086
A forward voltage in the RED array of 6.36V corresponds to a 555mA forward current.... ==> 0.555 * 6.36 = 3.5298

1 - (3.1086 / 3.5298) = 11.93% increase in power consumption

Efficiencies Compared Directly:

3.1086W / 3.8W = 81.8% Efficiency
3.5298W / 6.63W = 53.2% Efficiency

I have tested the circuit with a 10uH inductor, a 22uH inductor, a 33uH inductor and a 100uH inductor and the numbers are virtually the same.... Additionally while the Forward Voltages and Currents are different in the Green Array, the efficiency statistics are virtually the same.

In the original version of this circuit the PT4115E drivers were set @ 900mA and PWM was employed to limit the "final current" to 500mA in the Red array and 250mA in the Green array ... I don't have one of the old units to test, but I am 99.99% certain that I remember the efficiency increasing as the supply voltage increased ... I also remember the efficiency being in the 90% + range ....

Before I build a prototype with the current set higher and alter the firmware to control the "Final Value" .... can anyone think of a good reason my efficiencies would be so messed up?

Thanks!

Fish

 

[Arouse1973]

Hi Fish, do you have the datasheet for the PT4115E?
Cheers
Adam

 

[Fish4Fun]

Hey Adam!

Data Sheet: https://people.xiph.org/~xiphmont/thinkpad/PT4115E.pdf
https://people.xiph.org/~xiphmont/thinkpad/PT4115E.pdf
Thanks for looking!

Fish

 

[Arouse1973]

Hi Fish.

If you look at the data sheet it shows the efficiency going down as the supply voltage is increased. But it should still be above 75%. Have you tried it with the 10 uF capacitor removed? Also where and how are you measuring your current for the input and output values? Remember the input current will be switching current and not steady state. You need to factor this in when making the measurements.

Thanks
Adam

 

[Fish4Fun]

Hey Adam!

The input current is measured using the DRO on my CC/CV Bench supply .... Power = Current * Voltage

As mentioned above, the current output to the LED array is not measured directly, rather the Forward Voltage of the Array is measured when the PT4115e is powering it, and then I power the array directly from the CC/CV power supply with the voltage locked at the previously measured Forward Voltage ...

Example:

Powering the RED Array with the PT4115e and a supply voltage of 8.0V the input current is 475mA ... 0.475A * 8.0V = 3.8W .... and the LED Array has a forward voltage of 6.28V ....

Powering the RED Array directly from the bench supply with the voltage clamped @ 6.28V the current settles to 495mA .... 0.495A * 6.28V = 3.1086W

I am working on the **Theory** that under these circumstances the forward voltage of the LED Array is a reliable indicator of the current AND hence the Power. I can certainly wire in an ACS712-05 Current Sensor Module OR an Old-Fashion analog meter to verify the current, but I have no reason to suspect the current values calculated from Forward Voltages are wrong.

I can // and will // remove the 10uF capacitor and repeat the tests; however, this capacitor was added to the original circuit ( if I remember correctly, in the original circuit the capacitor slightly improved the efficiency ) .... without it the LEDs "Flash" at low PWM Duty cycles rendering the device unsuitable for its purpose.

Again, the primary difference between the original circuit and this version is that the original circuit used feedback resistors setting the current to ~750mA in both PT4115e drivers and used PWM dimming to limit the RED array to ~500mA and the Green Array to ~250mA .... In this "new version" the feedback resistors were changed so that the Red Array current is 500mA and the Green Array is 250mA and the PWM is simply employed to achieve the "Fade In" ...

I guess at this point I need to build a second prototype like the "Original" and compare it to the "new version" ... to see if there is some reason PWM dimming improves efficiency LoL .

Thanks for Looking!

Fish

 

[Arouse1973]

Hi Fish
If you tested the original one in the same way and got different results then there must be something wrong with the current design. Building another unit is the next step I agree.
Thanks
Adam

 

[Fish4Fun]

Hey Adam!

Thanks! Gonna cut another PCB and populate it :-(

The **problem** is that I did the testing of the original circuit ~ 3 years ago and I don't have any notes on the actual test results of the efficiency ... I know I tested it, and I know I told the guy the circuit is for that the efficiency was ~92%, but I just don't remember any of the details about how I tested it. I feel certain I would have tested it @ ~13V because it is powered by a 12V lead-acid battery .... As far as "something wrong with the current design" .... I would agree except the "new design" is literally the "old design" with different current sense resistor values and a different PCB layout ... AND, of course, different firmware... in the "original" circuit the maximum PWM Dimming Duty was ~50% while in the "new" circuit the PWM is actually "disabled" at the end of the Fade-in period.

Regardless, I when I get the "New-Original PCB" populated and tested I will report back with my findings ...

Thanks!

Fish

 

[Arouse1973]

Hi Fish.
To compare results from earlier testing it would be better if the designs matched. You really do need to take into account what I mentioned earlier about the switching current. Because there is a portion of the measured input current which will be closer to zero when switching and I would be surprised if your power supply could resolve this and display it correctly.

Thanks
Adam

 

[Fish4Fun]

I KNOW everyone has been sitting around on pins and needles waiting for me to post my findings ... LoL !!

The **actual answer** is really a collection of things that add up to more than one might think ...

A primary culprit was the late bound addition of a diode to the voltage input (to prevent damage from the end user connecting the battery incorrectly) ... The SS34 diode used has a Vf of ~0.6V @ 800mA and ~0.8V @ 1.2A ...

The next culprit is again linked to diodes ... I selected SS34 diodes for the Buck Converters and I did my tests using SS34 diodes I had on hand ... these diodes maintain a Vf < 0.5V for currents < ~2A .... the diodes used in the tested units came from a different batch and while marked SS34, the Vf is (like the above) ~0.6V @ 800mA and ~0.8V @ 1.2A.

The SS34 Didoes I have on hand (and used in the original tests) are in SMC (DO-214AB) Packages while the ones supplied (and used in the recent tests) are SMA (DO214AC) Packages .... The SMC package is roughly 8mm x 6mm while the SMA package is roughly 5mm x 2.6mm. The SS14 diode is typically packaged in the SMA package while the SS34 is typically in the SMC package.

I don't know if the supplied SS34 diodes are mislabeled or simply do not perform to the same specifications as the Vishay datasheet ... regardless the point is moot because the person I designed this circuit for had the PCBs made for the SMA package and is going to use the SMA diodes marked SS34 he has on hand, and is no longer concerned about the "loss of efficiency", LoL.

Surprisingly the overall efficiency is affected considerably by how the PWM dimming is implemented in combination with the primary PWM current limit. For instance, if the current feed back resistors are sized such that Iout = Iavg then the efficiency is lower than if iOut > Iavg and PWM Dimming is employed to limit the average current.

Huh?

In the case of the RED LEDs there are a total of 12 X 1W LEDs configured as 3 in series AND 4 in parallel ... The Red LED Vf = 1.8V to 2.2V ... and Imax ~ 450mA. 450mA x 4 = 1.8A Max ....2V * 3 = 6V ... SO .... 6V * 1.8A < 12W ....

Selecting a feedback resistance of ~0.167 ohm should theoretically give a max current of 600mA yielding a max RED output of ~3W .... Which is where testing shows the perceived brightness is only slightly less than at a full 12W input. SO... in theory PWM dimming @ 100% duty cycle should yield 3.6W output.

If instead we select a feedback resistor of 0.08 Ohms this would theoretically give a max current of 1.2A yielding a max output of 7.2W ... if we then set a maximum PWM dimming duty cycle @ 50% we once again have a "average output" of 3.6W, but the actual input power is significantly reduced, and the perceived brightness is actually slightly increased!

I suspect the increased efficiency of the 50% dimming duty cycle version is achieved because there are 50% less switching losses in the primary PWM combined with a lower primary switching frequency ( because for any given inductance doubling the max current increases the on time ). The increase in perceived brightness almost certainly occurs because the "peak output" is slightly brighter for twice the current, and the PWM Dimming switching frequency is well above naked eye detection.

Other minor contributors involved the inductors. The original test circuits used 12mm x 12mm foot print inductors while the "new version" uses 6mm x 6mm inductors .... but this was a minor contributor...

I didn't crunch all the numbers, but the diodes alone contribute substantially to the increased losses ... the losses were minimized when the max PWM dimming duty cycle was set to ~50%.

Fish