PWM dimming

I have a few questions:

1.  Is the reason for wanting to be out of the audio band due to the piezoelectric ringing of the output ceramic capacitors?  If so, there are ways around that, for example, don't use a ceramic output capacitor.

2.  The reason for "the IC and the external switch FETs need several hundred microseconds to regulate the LED current" is two fold: 

      a:  you will likely need a gate driver or some type of buffer amplifier to turn high speed FET on/off at that much current.  The simple circuit in the datasheet with a pull up resistor and pull down FET is too slow,

      b:  any converter's control loop will need to have a bandwidth of at least the PWM signal frequency so it can respond and recharge the output capacitor before the switch turns back on.  With the TPS61500 configured in a SEPIC configuration, this will be difficult and may be impossible to achieve.

3.  What is your desired dimming range?

Another option is to not use an integrated solution like the TPS61500 and use a separate converter that regulates slightly higher than 7.2V and then a discrete fast switching current sink circuit (op amp + FET).

Thanks, Jeff.

The reason for wanting to keep the PWM frequency out of the audio band is not because of ringing, but out of fear of pickup from nearby audio circuitry.

The dimming range is 1:1024.

My original plan was actually to do as I think you suggest; have one big boost converter and four current sinks. Because of different voltage drop for the different color LEDs, I'm struggling with the efficiency of this approach. Quite a bit of power is wasted for the LEDs with lower Vf.I could look at having individual converters with current sinks for each LED pair to optimize for efficiency, but that may be too expensive.

If we could shield potential audio interference, we could probably get by with a PWM dimming frequency as low as 200-300 Hz.

I have a 3V3 rail elsewhere in the system. Would it not be possible to put a MOSFET  in series with the LED for the PWM dimming and have a "weak pull up" on the FB pin to avoid that the converter runs off while the feedback loop is open? Use a simple voltage divider from the 3V3 rail to set the right FB voltage?

The concern is not the FB path opening.  The IC has an OVP circuit that clamps the output voltage when the LEDs are removed as a set by the resistors between the OVP pin and ground.  There are several problems: 1) finding a low cost FET that will turn on/off fast enough for your dimming frequency, 2) finding a low cost FET driver to driver that FET and 3) compensating the converter to respond quickly enough so that the capacitor does not droop when the FETs are re-introduced.   Please consider that if you have a PWM frequency of 25kHz and you want 1:1000 dimming then you want the FET on for only 40ns.  The FET will likely need over half that time for the LED current to settle.  The P-FET driver will need to sink several hundred milliamps of current from the gate of the FET in under 10ns.

Hmm,  good point. I hooked up a simple current sink and the FET won't fully open.

We need very precise dimming ratios, but we don't actually need to dim all the way down. Output below 8/1024, or 1/128, looks good on the box, but  is not really usable in a real world scenario for this product. It looks like we should either cap the low end of the output, lower the PWM dimming frequency, or both.

I'm considering changing to a series/parallel battery configuration. That would bring the input voltage to 2 - 3.7 V and in range of converters, like the TPS63000/63020/63030. These seem to need fewer external components, offer higher efficiency, but no OVP. Provided I can get the FB "pull-up" to work, do you think that would be a good choice?

The TPS630xx maximum output voltage is 5.5V so that family would not work unless you put your LEDs in parallel as well.  Also, finding a boost converter with integrated switch that will work efficiently providing 7.2V @ 750mA from Vin= 2V will be a challenge.  TI's TPS61175 would work as a boost if you split the power rail (2V to 3.7V) from the IC's VIN and used a separate 3.3V or 5V rail to power the IC but it is not very efficient in that configuration.  So, I recommend that you stay with the series battery configuration, use the TPS61500 as a SEPIC (or possibly as a boost if the efficiency if low enough to keep the part boosting even at high Vin) and lower your dimming ratio. 

Meant to say TPS61500 not its sister part TPS61175.  Apologies.