Choosing a MOSFET for the LTC3824

[void_error]

Having ditched the LM2596 in favor of the more versatile LTC3824 for my bench PSU's pre-regulator I'm a bit unsure about the characteristics of the external P-channel MOSFET I should use so yet again I'm asking for some advice.

The maximum input voltage will be 50V and the maximum output current will be 8.5A at an output voltage of 5-45V, however the current version will run at 36V max input and 5-25V/3.3A output. I've made some spreadsheets that will calculate the inductor but I'm stuck at MOSFET selection for the 50V input version and I need a bit of help in choosing the right switch - preferably a TO-247 device. I guess it has to have low input capacitance so there's less drive current and low R_DS[ON] for low losses as well as high enough breakdown voltage. The switching frequency will be set between 200-300kHz.

[prasimix]

TO-247 seriously narrow possibility but if you are fine with TO-220/TO-220FP (Full pack) take a look at Vishay IRFI family. Two of them are even available on TME.eu: IRFI9630 and IRFI9634.

[void_error]

I was thinking of IRFP9140 with 1.3nF gate capacitance. What I don't know is how much power it'll cause the LTC3824 to dissipate.

[prasimix]

I was thinking about other thing recently. If you'd like to use PFET SMPS controller such as LTC3824 (maybe you should look at LM5085 also which could stand Vin of 75V) because it can provide 100% duty cycle how are you going to select let say "pre-regulator" and "bypass" mode? You cannot do that with same tracking logic or you can? In the first case you need to inform pre-regulator to deliver voltage which is few volts larger then Vout. In the second one pre-regulator has to deliver Vin which results with 100% duty cycle operation regardless of Vout value.

[void_error]

You cannot do that with same tracking logic or you can? In the first case you need to inform pre-regulator to deliver voltage which is few volts larger then Vout. In the second one pre-regulator has to deliver Vin which results with 100% duty cycle operation regardless of Vout value.

In fact that's quite easy. For normal tracking operation the PNP transistor in the feedback loop will be on and follow the linear reg's output voltage (approximately, but good enough). If the transistor is forced off the feedback will be only through the resistor between its collector and emitter and with a value chosen so that for the highest input voltage the voltage at the feedback pin will be below the reference voltage which will result in a 100% duty cycle. That's one way to do it.

The other way is to simply short the feedback pin to ground but then the feedback resistors need to be of high enough value so there won't be excessive current flowing through them when the linear reg's output voltage is zero.

[prasimix]

 

[prasimix]

In this document (Providing Continuous Gate Drive Using a Charge Pump) is described a simple way how to get 100% duty cycle even with N-channel high side mosfet. Another possibly useable way is described in Evolving High-Side MOSFET Driver Meets Unique Requirements With Standard Components. When time is come I'll try to test it in combination with either LM5088 or LM5118.

[void_error]

Looks like someone from TI didn't have their head screwed in properly when they designed the circuit in Figure 6 ... 3.3V is a bit outside the spec for the 555 - oops... although it might work.

Other than that relatively small detail it looks like a good idea, gonna look into it more.

[prasimix]

Yes, it starts with 4.5V (at least what is advertised on their pages). Actually that is recommended minimum. Maybe it could works predictable even with 3.3V. I'll contact TI and try to get their opinion about this detail.

[Christopher]

Pd = Qt * Vdd^2 * F

will tell you (roughly) the average power dissipation of the driver. A higher Qt will mean slower charging and discharging of the gate, obviously

Lower Qt comes from having the die as close to the pins as possible (same as low Rds). So that's why those tiny packages have seemingly great Rdson ratings.


(or is it just V without the squared? My notes are at work)

[ali80]

actually I wont care that much about gate capacitance since the part you are using has good gate drive capability (2.5amps)
and also switching the mosfets too fast can cause other problems such as high frequency noise which will even pass your linear regulator.
I guess a 75-100v mosfet with an rdson in range of 10-100mOhm would be ok depending on your heatsink parameters
also choosing to-220 package for this kind of power and voltage is completely fine

[prasimix]

I'd like to add two interesting documents which discuss charge pump (TI - Discrete Charge Pump Design, slva398a) and floating charge pump (IR - HV Floating MOS-Gate Driver ICs, an-978). The IR document gives an example with gate driver IR2125 where bias voltage from 555 based floating charge pump is connected to the VB pin for bootstrap capacitor. If this pin is equal to the BOOT pin that can be found on many smps controllers than with this simple add-on a 100% duty cycle could be obtained regardless of Vin and of course with n-ch mosfet as a high side switch.

One other possibility comes to my mind: if there is a MCU in the system or i.e. aux SMPS for bias supply then it can provide a "clock" signal (possibly using opto-coupler) for the floating charge pump.

[prasimix]

I took some time and breadboarded the following circuit:

Switching frequency is intentionally ~200KHz since I'd like when time is comes to test it using aux SMPS via opto-coupler as a "clock" as mentioned in previous post.
Two scenarios was tested: with R2=3.3K when Vboot=15.3V and R2=4.7K when Vboot=12.3V (TP2 to TP3). NE555 duty cycle is 55 and 60% respectively. Please note the difference in mentioned readings with DVM and MSO (cyan=Vboot, magenta=Vin) in attachment.
Vin is 56VDC. So, now only is missing connecting Vboot to the BOOT pin of SMPS controller and push it to reach 100% duty cycle.