LM3409: Can I replace the recirculating diode with a MOSFET?

Hello David,

It could work (the LM3409 wouldn't know the difference since it's regulating peak current) however I would look at the TPS92641 if you are considering this.  There are a lot of operating conditions to check if doing this.  Shunt FET dimming operation would need a close look at on-time/off-time to prevent current runaway.

I would have to find out about negative voltage on the current sense.

Best Regards,

Thank you again Irwin,

I'll look into the TPS92641.

Regards,

David

Hello Irwin,

I have examined the TPS92641 but the current sense resistor must be grounded which prevents me from driving the cathode grounded LEDs in my application.  I know that maybe with an additional, high speed high-common mode differential amplifier, I might be able to move the CS resistor from the cathode to the LED anode, but there are several design issues that would have to be addressed and my application is very cost sensitive.

My design will include 20 independant LED drivers (each with their own controller and separate iAdj inputs).  All LEDs will be turned on/off at the same time, if PWM control is to be used.  The design will be air cooled (with a fan) and cannot have any additional heat sink.  Each driver will be capable of outputting 26W (4A at 6.5V) and must turn on/off within 20us.

I was trying to avoid the recirculating diode because of the high power dissipation (4A * 0.6V or so, about 2.2W at my lowest duty cycle of 7%).

The diode alone reduces the efficiency by about 8% and will require some very thoughtful thermal management.  I could parallel two diodes at a minimum risk of thermal runaway since they will be sharing heat, but at an increased cost and PCB real estate.  The two-diode solution would still dissipate about 1.6W total.

Do have any thoughts about this?

Thanks,

David

Hello David,

Unfortunately a non-synchronous buck running at low output voltage and low duty cycle will have lower efficiency because of the freewheel diode Vf.

I'm not sure of your input voltage range, if it is fixed at 24V I would consider looking for a low Vf 30V diode or diode pair.  Vf can be around 0.3V, as it heats Vf will drop however reverse leakage goes up so looking at all operating ranges is important.  Letting the diodes run at higher temperature is a way to improve efficiency.  Adding the synchronous MOSFET (and additional circuitry) will add cost as well.  Do you have a switching frequency range you are considering for this?  LM3409 is variable switching frequency.

Best Regards,

Irwin,

The switching frequency has not been specified, but because of the 1.5V minimum output voltage and 24V input requirements, the minimum Ton time of 210ns limits the maximum frequency to about 300 kHz.  I also realize that the variation in output voltage will result in a frequency change that is dependent on the output voltage (which can be from 1.5V to 6.5V).  With a few calculations and a fixed Toff time of 3us, the switching frequency would be between 230kHz and 310kHz, depending on the output voltage.

I will plan on using two parallel diodes and with PCB testing, maybe I could DNP one of them.

My original design was using a sync. buck voltage mode controller (as opposed to a peak current model controller) with LED current as feedback.  But I was not able to meet the turn-on/off time requirement and there was a lot of analog "glue" circuitry and control loop compensation issues that added cost and PCB real estate. 

The final PCB will drive 20 independant LEDs.  I'm not sure about interference between the drivers since there is not synchronization on their load of the input power (I had all 20 drivers interleaved in my previous design).

The only other issue is that I must be able to parallel the outputs of all drivers (i.e. 80A output into an array of paralleled LEDs).  I didn't see an issue with that concerning the LM3409, but maybe I am missing something.

Thank you,

David

Hello David,

The LM3409 is a current source, you should be able to parallel them.  The current sense is before the inductor feeding a fixed voltage so there should be no issue if they aren't synchronized though the ripple current in the LEDs may look a bit odd.

Best Regards,