Parameter "n" specifying number of diodes in series

Hallo Pavel,

Pavel47 said:

'n' next to the LED that determines (I think) the number of diodes in series.

your interpretation is correct. n gives you several diodes in series, m would give you several in parallel.

Pavel47 said:

How to specify that "n"... after all, when you select an LED from the library, there is no "n" associated with the LED.

Place the diode on the schematic. And then change the description of the diode by adding n=... to the diode name. You have to edit the text itself, i.e. you need a right-click not on the diode-symbol but on the name/description of the diode at the side of the symbol.

To your second question: well at least it should be modeled correctly. If you observed something different in your simulation it would be good if you could attach your asc-file, so that one can analyse the issue you observered.

best regards

Achim

Hello Achim

Thanks. Not sure that I properly understood you.

Here is the LED that I picked from library ...

When I click on "QTLP690C" and add n=..., the property changes to "QTLP690C n=..."

Here is shematic LT3517_example.ascwhere current through LEDs changes when I change number of LEDs in series ...

Pavel47 said:

When I click on "QTLP690C" and add n=..., the property changes to "QTLP690C n=..."

fine. That will give you n of the QTL690C in series. (as demonstrated in this simpler simulation...)

Pavel47 said:

Here is shematic LT3517_example.ascwhere current through LEDs changes when I change number of LEDs in series ...

Thank you for the file. Your LT3517 is switching with 1 MHz. And on the PWM-pin you try to control it with 1 µs pulses. It think this is the reason for your observation: with this extremely fast PWM, your switching regulator will not reach the intended steady state in the PWM-on phase, it has not enough time to regulate the current to its intended value.

The datasheet says: "For best current accuracy, the minimum PWM low or high time should be at least four switching cycles". I personally would recommend to go for a even much lower PWM frequency unless you have a good reason for 500kHz you've choosen. With e.g. a 10kHz PWM the intended current is reached also for 3 diodes in series.

best regards

Achim

Achim said:

That will give you n of the QTL690C in series. (as demonstrated in this simpler simulation...)

Well ... when I right-click on N, the dedicated window opens, prompting for new value of N. With your approach it seems not be the case:

Achim said:

With e.g. a 10kHz PWM the intended current is reached also for 3 diodes in series.

With N=5 I changed value of Rt (4k instaed of 16.9k) that defines switching frequency. The value of "ON" current surprisingly decreased with 4k: 173 mA. Wth Rt = 16.9k Ion = 320 mA.

Achim said:

unless you have a good reason for 500kHz you've choosen.

Yes, the reason is that the PWM isn't for brightless control, but for communication. PWM contains data tranmitted by LEDs.

Pavel47 said:

Well ... when I right-click on N, the dedicated window opens, prompting for new value of N. With your approach it seems not be the case:

with my proposal both values are shown in one line. In your simulation the two values are in two lines. Both versions work equally well. If you are eager to have it solved just like in the example file, the only idea I have is to edit the asc-file with a text editor.

Pavel47 said:

With N=5 I changed value of Rt (4k instaed of 16.9k) that defines switching frequency. The value of "ON" current surprisingly decreased with 4k: 173 mA. Wth Rt = 16.9k Ion = 320 mA.

Having the switching frequency a factor 4 above the minimum pulselength is a necessary prerequisite for good current regulation, but not a sufficient one. You have e.g. L1=15µH in the current path. With 5 LEDs in series the available voltage drop for L1 ist 10V. That means that the current through L1 can rise not faster than 10V/15µH=0,66A/µs. That makes it hard to achieve stable 1A in 1µs.

Maybe you can achieve your intended current profile by playing around with the values of L1, C1, C2, R2. But it could be, that you optimize their values for a given bit sequence at PWM, and it does work bad for another bit sequence. The LT3517 has special features to give a good transient recovery time, but I'm nevertheless not sure if this is a good driver for data transmission over several LEDs with some code unkown to us. For that task my first simple idea would be to use a much simpler (and faster) bipolar-transistor based current sink. And maybe use an adjustable DCDC-regulator for voltage preregulation to keep the power efficency high.

Pavel47 said:

Well ... when I right-click on N, the dedicated window opens, prompting for new value of N. With your approach it seems not be the case:

with my proposal both values are shown in one line. In your simulation the two values are in two lines. Both versions work equally well. If you are eager to have it solved just like in the example file, the only idea I have is to edit the asc-file with a text editor.

Pavel47 said:

With N=5 I changed value of Rt (4k instaed of 16.9k) that defines switching frequency. The value of "ON" current surprisingly decreased with 4k: 173 mA. Wth Rt = 16.9k Ion = 320 mA.

Having the switching frequency a factor 4 above the minimum pulselength is a necessary prerequisite for good current regulation, but not a sufficient one. You have e.g. L1=15µH in the current path. With 5 LEDs in series the available voltage drop for L1 ist 10V. That means that the current through L1 can rise not faster than 10V/15µH=0,66A/µs. That makes it hard to achieve stable 1A in 1µs.

Maybe you can achieve your intended current profile by playing around with the values of L1, C1, C2, R2. But it could be, that you optimize their values for a given bit sequence at PWM, and it does work bad for another bit sequence. The LT3517 has special features to give a good transient recovery time, but I'm nevertheless not sure if this is a good driver for data transmission over several LEDs with some code unkown to us. For that task my first simple idea would be to use a much simpler (and faster) bipolar-transistor based current sink. And maybe use an adjustable DCDC-regulator for voltage preregulation to keep the power efficency high.

Thanks Achim,

According to your experience, can you suggest some other LED driver model that is more suited for communication applications. Sure, I keep in mind BJT-based current sink solution and probably I will use it once it becomes evident, that more integrated solution (based on LED drivers) isn't possible.