LM3405: Using LM3405 / LM3405A to drive less than 200mA

Hi Chris, 

The recommendation on datasheet is based on corner condition results. I do not recommend to regulate a current below that. You can set 200mA and use dimming function to reduce the output to 100mA. 

And we actually has latest devices, TPS92200, fully synchronous buck LED driver, which can support 100mA. 

Could you please share your end-equipment application? 

Thank you! 

Best, 

John. 

Our application uses the chip to drive an IR LED (rated for 100mA max) for human presence detection in office-like environments, so we don’t need the temperature extremes, and we probably have loose requirements on other specs as well. Do you know how it would be likely to fail in corner testing? Do you have thoughts on how we can do our own corner testing to verify the LM3405 at 100mA, since we have fewer corners to test? 

The datasheet says “[current] must also be kept above 200 mA for stable operation”. I did see instability at very low currents with the recommended 1uF output capacitor (C2), but that was resolved by reducing it. That led to high current ripple, but we don’t care in our application.

The TPS92200 is intriguing, but our supply voltage can go down close to 3V, so the 4V Vin min would be a problem. 

The TPS92201 seems promising, but some things are not clear from the datasheet:

• I’m not finding how low the 100% duty cycle current can be set (one example uses 400mA). I don’t like the idea of the hardware being able to fry our LED if a firmware bug leaves the PWM pin stuck high.
• We need to be able to pulse the LED at around 1800 Hz. The EN pin seems to have a “soft startup” of “xx ms” (figure 8-13 looks like it’s more like 600us, but in either case that’s too slow.) The PWM pin turns 20kHz+ pulses into analog dimming, but would 1800 Hz pulses make it through?
• The 8.2 “Typical Application” seems to have some errors. For example, it talks about R1 and R2, which don’t exist.

Some of my questions could possibly be answered by the WEBENCH Power Designer, but I guess this chip is still too new to be on there?

Hi John,

Increasing C2 above about 0.1uF (1uF shown below) causes huge overshoot (and ringing) to the point that the overcurrent kicks in and switching stops briefly.

5us/div, see schematic below for probe locations.

Ch1 (yellow): 100mV/div
Ch2 (blue): 500mV/div
Ch3 (red): 5V/div

Increasing the feed-forward capacitance (C4, or C23 in mine) didn't help much, so instead of that, I added feed-forward compensation to a more effective place, from before the inductor (see below).

This worked well, though still has a little steady-state oscillation (see below 10us/div). The high-frequency oscillation on the yellow is an artifact of the compensation, not actual current through the LED.

What are your thoughts on this solution? The other option that works well in our application is to use a very small C2 (C4 in mine), like 10nF, which gives a stable output but has a bit of ripple.

Chris

Hi Chris, 

This is a good try. If you add C2 to 1uF leading phase margin decrease, this means internal zero compensation is not large enough. The external compensation just get the phase margin back, but did not change the increased bandwidth by current setting resistor, which has the risk of instability. 

Let me check with design team tomorrow to see how much margin this will be. 

Thank you! 

Best, 

John. 

Hi John,

Were you able to determine whether either of these options is preferable or recommended?:

• The feed-forward compensation above or
• Small (10nF) output capacitance (C2)

Thanks,

Chris