Crosover frequency of LED Driver

[schnell]

We are doing a LED Driver with ZXLD1360 Hysteretic Mode LED Driver.

ZXLD1360 LED Driver datasheet:-

http://www.diodes.com/datasheets/ZXLD1360.pdf

The average LED current is being monitored with a microcontroller via a hi-side current monitor (ZXCT1081).

ZXCT1081 current monitor datasheet

http://www.diodes.com/datasheets/ZXCT1081.pdf

Thus, the microcontroller can monitor the LED current, and gradually increase it over the years, so as to maintain brightness levels.

Since we have the current monitor, and we are reading its output with an ADC port on the microcontroller, we can be assured that we are setting exactly the required current at the required time.

The microcontroller is reading the current once every 500ms and , if necessary, correcting the LED current by applying a voltage to the ZXLD1360’s “ADJ” pin.

The schematic is as follows:-

http://i46.tinypic.com/35m42zo.jpg

Considering the fact that hysteretic mode LED drivers are unconditionally stable,
-and also the fact that we are reading the LED current with the microcontroller every 500ms…
-and , if necessary , we are also applying a correction voltage to the ADJ pin of ZXLD1360 every 500ms.

….considering all this, what do you think the zero gain crossover frequency of the bode plot of the LED Driver would be?………that is, the bode plots with respect to the LED current…….

………would it be 2Hz or less?

 

[Skogsgurra]

You could, of course, do it 'the text-book way'. But is it really necessary?

There are no fast changes in LED characteristics and I wouldn't care about making this fast at all. 100 mHz is plent fast and without problems. Even 10 mHz works. Or 1 mHz.

The important thing is that you really meaqsure mean current and not switching components. That means that you shall filter well before measuring.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[schnell]

Thankyou, we are filtering the current feedback signal to the microcontroller.......just enough so that we can read the current accurately with the ADC.

The thing is, given that the LED current is only being read and adjusted every 500 milliseconds by the microcontroller, can you say that the zero gain crossover frequency must be less than 2Hz?

 

[Skogsgurra]

Yes. Definitely. Your bode plot (G) cannot cross zero at the same frequency as you sample.

But, as I said, it is more important that you filter well before measuring.If you filter "just enough so that we can read the current accurately with the ADC", then you should filter more. There is no need to create problems and you gain absolutely nothing by having a high f0 in your filter. Remember that you are compensating for a very slow change in LED efficiency and not trying to make a machine-tool servo perform.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[schnell]

Would you say that the Bode Plot *must* cross zero at a *lesser* frequency than the sampling frequency?

 

[Skogsgurra]

Why are you so concentrated on that question? And why do you absolutely need to use a Bode plot in a situation where most engineers wouldn't?


If I say that you cannot have G=0 dB at sampling frequency - then it is obvious that the crossing must be at a lower frequency. Isn't it?

Is this a test?

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[schnell]

To be honest, there is potentially a stability issue here.

For example, supposing we were to attempt to change the LED current by reading and adjusting the LED current with the microcontroller at too high a frequency, then we could end up with a feedback loop oscillation.

In other words, supposing we made too many “read LED current and adjust it” operations per second, -then ultimately, we could end up with a feedback loop oscillation.

……it is thus that we are interested in some kind of stability analysis for this ZXLD1360/Microcontroller system……and one of our experts has come up with Bode Plots for the system showing a crossover frequency of 10KHz.

…..and we feel that this simply sounds too high.

(we need to step change the LED current at certain times in the products life, as previously explained)

-Ultimately, we wish to know, for the microcontroller, what is the maximum permissible frequency for “read LED current and adjust it” operations to ensure stability.

 

[Skogsgurra]

Skip the expert. Has he/she ever tried any practical work?

This is a clear case of someone trying to force theory onto a subject that doesn't need it.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[IRstuff]

OK, START OVER. Do you even NEED to monitor the LED current and adjust at anywhere near 2Hz? From your description, the LED current degradation is something that literally takes years to transpire. Therefore, you don't need to monitor at anywhere close to 2 Hz, or even 0.2 Hz. Once an hour might be even be overkill.

And so what if it oscillates? It's not like you're going to be changing the value by more an an LSB or so, which would be imperceptible to the human eye.

Furthermore, based on the assumption that the changes are on the order of years, your monitor current will look like DC at 2 Hz, so there won't be any oscillation to begin with.

TTFN

FAQ731-376

 

[schnell]

For most of the time, even 2Hz is a bit too high a frequency, i admit that.........

....the thing is, we do need to read the LED current and trim it, since the Peak of the Analog PWM Dimming voltage shown in the diagram in the first post has a fairly wide tolerance on it.........

......the microcontroller that produces the Analog PWM dimming signal has its Supply Voltage supplied from two paralleled output pins of another microcontroller.......so this Analog PWM dimming signal is not an accurate way to produce a control voltage for the ZXLD1360 LED Driver.

....also, the ADJ pin of ZXLD1360 LED Driver is the pin to which an analog voltage is applied in order to set the LED current.....ADJ pin voltages from 0.3V to 1.25V will vary the LED current from 0% to 100%.........but the datasheet does not say how linear the current graph is in relation to the ADJ pin between 0.3V and 1.25V........so we do not know how accurate the ADJ pin will be in LED current setting.................
....therefore we need the microcontroller to measure the LED current and trim the LED current to the required value, -by incrementing/decrementing the ADJ pin voltage.

Most of the time, this loop does not have to be fast, but when the Lamp is first switched on, we may have an overcurrent which could damage the LEDs......or an undercurrent which would mean the LED's being too dim, or even flickering, -so therefore we want the LED current feedback loop to be reasonably quick.

-i admit that once the LED current has been trimmed to the demand value, the LED current loop need not then be quick, -
-but realistically, we may as well not complicate things by making the LED current loop slower as time goes on....and we are better off just having one "read LED current and adjust it" frequency, and sticking to it, since it means less microcontroller coding effort....and less calibration values which potentially have to be stored in limited EEPROM.

 

[IRstuff]

I guess I'm still failing to see why you are worrying the LED drive so much. The LED is a VISUAL indicator, and is only there for the operator as an ancillary indication, so the fact that the drive might be oscillating or might be slightly inaccurate is IRRELEVANT because the human eye response is woefully imprecise. If you are not using the current monitor circuit to control the PWM, the LED drive is IRRELEVANT.

ONLY IF you are using the current monitor to control the PWM is the bandwidth of any importance. AND with adequate filtering on the current monitor, the bandwidth is still of little consequence, as you won't have enough gain or phase lag to cause oscillation.



TTFN

FAQ731-376

 

[MacGyverS2000]

IR,

I didn't read this particular thread in detail, but I doubt the LED would be considered an ancillary indicator in this case. If this is a high-power LED driver, the LED is the item of interest. While I'm not convinced the OP is worrying about the right things (as Gunnar has suggested), he's certainly not completely out of the ballpark. Keeping lumen output within a tight range over the lifetime of an LED is a big thing in a number of applications... they eye may not see it, but inrush current is definitely something to be worried about.

On that note, maybe the OP should consider the initial turn-on phase and the normal operating phase as the two separate events they are and code loop detection appropriately...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[IRstuff]

A turn-on transient wouldn't have much to do with worrying about the bandwidth though, nor would the linear regime be an issue during turn-on.

TTFN

FAQ731-376