TPS92640 steady error between analog setpoint VIADJ and average LED current

Is this error present without the 1.2M resistor? Have you measured the actual voltage at the IADJ pin with a multimeter? In any case the datasheet says that for proper startup operation Roff should be less than 100xVoutxRf, or less than 1.14M in this case. Have you tried lowering Roff or Rf to see if that helps?

Dear Jensen

Q) Is this error present without the 1.2M resistor?
A) Yes it is;

Q) Have you measured the actual voltage at the IADJ pin with a multimeter?
A) Yes I measured it with a multimeter

Q) Have you tried lowering Roff or Rf to see if that helps?
A) No, I will do tomorrow

After some measure setting different VADJ and measuring ILED I see that there it appears as e gain error.

ILED = 0.6A (multimeter measure) VIADJ = 1.252 (multimeter measure)
ILED = 379mA -> VIADJ = 0.786V
IDED = 217mA -> VIADJ = 0.453V

My colleague told me that with 4 LED there are no error, but I can't confirm it today. I will double check tomorrow if this is the case. At the moment using the EVA board with 4LEDS and the following chances:
The original configuration where I only added the R11=1.2M of the demo board (Is the RFOO of the data-sheet) for a true zero current LED operation and I've cut R5 of the demo board, connecting an external signal generator to IADJ pin .

If a set the 0 gain error @ 379mA then Gm = VIADJ/ILED = 2.073 Siemens
So :

with VIADJ = 1.252 than ILED = VIADJ/2.073 = 604 mA (difficult to evaluate the error due to the 10A multimeter range)
with VIADJ = 0.786 than ILED = VIADJ/2.073 = 379 mA error = 0%
with VIADJ = 0.453 than ILED = VIADJ/2.073 = 218.5 mA error < 0.6%

Seams that there is a gain error of the (2-2.073)/2 = 3.6%
I is it possible ? What is the gain error of the TPS92640 ?

Paolo

Try the lower values to see if it helps, but can you also measure the actual CS voltage? This is a very accurate part, the CS voltage should be IADJ/10 +/- 600uV. Is it possibly error in the multimeter either due to its resolution or its averaging capability? Let me know. Worst case I can try and power up an EVM here and take a look, but it might take me some time to find one.

Now I have a second problem: I need to zero the ILED when VIADJ is 0. To do that ROFF has to be connected between Vout and CS. There are 2 equations to  calculate ROFF

Setting Vout = 24V (3Vx8LED = 24V), RF=475 Ohm (demo board) RCS=0.2Ohm (demo board) delta ILED=3mA (for example), It results:ROFF = min  [15.5K, 1.1MEG]  so ROFF has to be set to 15.5K.

If I do that 15.5K the VOUT ILED is always 0 even varying VIADJ from  0 to 1V.  If I set ROFF to 1.2MEG the linearity is good and only if I set 0V as VIADJ ILED is 0. Starting from VIADJ = 0.1V than the ILED follow is linear with VIADJ. So the conclusion is the above ROFF selection seems to be wrong and only the max (not min) rule has to apply.  

1) Are you sure of the above  formula ?

2) Connecting ROFF to VCS produces a down shift  of the  line ILED = k*VIADJ + q, changing the q value. Unfortunately VOUT is not exactly constant so theoretically “q” is not constant. Why don’t connect ROFF to a constant voltage to increase the regulation linearity ?

I am not sure about the formula in the datasheet. I was not the development systems engineer for this part so I did not write the datasheet. But there could be an error. If you want to provide me your contact information (you can do a friend request if you do not want to post it here) I can send you an analysis from the original systems engineer that has different (and hopefully correct) equations if you are interested.

As for the connecting it to a static source, I'm not sure. I believe part of the function for ROFF is to draw the small amount of minimum current you could see from the output, through RCS, and through RSNS and therefore bypassing the LED. But perhaps it would still work with a static source by pulling up CS just enough. I'm not sure, I have not tried that. If you test it out and it works fine there should be no issue doing it that way.