Help with Mosfet Selection

[Deadpixel732]

Hey guys,
Here's my first post so here goes.

So as the title says i would like some help I choosing mosfets. I will be wiring these fets up in an H bridge design so i will need recommendations for both N-Channel and P-Channel Mosfets. My load is a Peltier element (TEC) at 5v running peak at 8A. [The H-Bridge is used to vary the voltage (via duty cycle) sent the TEC {this circuit is part of a temperature control circuit using a LTC 1923}. Because the I want to be able able to clock the mosfets to switch at 1mhz [while varying duty cycle] I need mosfets that can switch fast enough (guessing 50ns or lower switching times) and that can handle the high current (so it has to be able to dissipate lots of heat). Also I would like so be able to switch these fets using a 5v logic signal (generated by the LTC1923). My problem right now is that I cant seem to find any mosfets that can both switch fast enough (low Qg) and handle the current (low Rds on resistance).

In summary, I would like recommendations for N channel and P channel Mosfet I could potentially use and I'm also wondering if it is necessary to use a mosfet driver (The LTC1923 can drive most mosfets fine but because ill be switching high current I for some reason think it may be necessary.)

Thanks for your help.

Heres the link to the datasheet for the LTC1923: http://cds.linear.com/docs/en/datasheet/1923f.pdf
and attached is the general schematic layout i'll be following.

[Psi]

1mhz is quite fast, are you sure you need that?

It seems unlikely the device being heated could change temperature quick enough (due to its thermal mass) to need that sort of switching speed.

[iwl]

Are there TO92 fets to find switching  at 3 Volt allready preferable cheap on eBay.

I use npn at time but loose 0.4 Volt Collector Emitter and Blue Led needs 3.5 so not much margin with a Lipo 4.2 Volts.

[Deadpixel732]

1mhz is quite fast, are you sure you need that?

It seems unlikely the device being heated could change temperature quick enough (due to its thermal mass) to need that sort of switching speed.

Well the main reason I want to switch that fast is keep the size of the inductors and caps down. (want to keep board size relatively small as there will potentially be 100 duplicates of this circuit on the same board (yes i will be controlling 100 TEC elements). However if finding mosfets that can switch that fast is really impossible I guess something around 700khz would do.

[Psi]

I'm no expert on TEC elements but do you actually need to smooth the switching with an LC filter? can't you just drive them directly from the H-bridge with a pwm square wave at full rail voltage?

[ejeffrey]

TECs don't handle voltage control very well.  The IV curve is sharp and temperature dependent, like a diode, so at a constant voltage the current can fluctuate wildly.  They also don't like unfiltered PWM because the parasitic resistive heating is maximized in that case.  Inductor filters solve both problems.

How power constrained are you?  If you can tolerate a lower efficiency and some heat dissipation in the controller, just use a linear current regulator.  You can still use PWM for the feedback, but because it is only signal level, the filter components will be tiny.  You will still need some switches for polarity reversal unless you make a push-pull driver circuit, and you will have a small deadband or hysteresis around zero.

[johansen]

DO NOT use pwm on a peltier.

select the optimum size of the peltier chip based on your google fu ... (there are calculators to get you within a few percent of the optimum value)

here's the easy way of understanding it.. peltier chips are operated best at 100% duty cycle.

heat moved because of electron flow is proportional to the AVERAGE of the electrical current. (Not RMS)
heat moved from the hot side to the cold side is proportional to the difference in kelvin.
Heat generated by the electron flow is related to the Square of the RMS current, and its conducted to both sides of the peltier chip.

it is thus possible to send enough current through the chip to heat both sides of it.

generally its best to select a peltier chip that is sized for about 60% of the manufacturer's rated current flow.
(since they come in 60, 50,40,30 mm square sizes, its not hard to choose)
and you want to run them at 0% current ripple, meaning all of the current is DC.

so just filter your pwm with a buck converter and call it good.

[Psi]

Putting a H-Bridge around a Peltier is not needed and a waste of 3 FET's .. unless you want to be able to transfer heat from either side ?

For normal one way heat transfer you either power the Peltier or you don't. You don't go reversing the power to it - which is what a H-Bridge is for.

Given the laser temp control appnote that he posted i think he's building a TEC cooled laser diode module.
Some diodes require a very precise temp to get the right wavelength and efficiency.
Its quite possible he needs an H-bridge for heating and cooling to maintain a temp within the typical ambient temp range.

[Deadpixel732]

Putting a H-Bridge around a Peltier is not needed and a waste of 3 FET's .. unless you want to be able to transfer heat from either side ?

For normal one way heat transfer you either power the Peltier or you don't. You don't go reversing the power to it - which is what a H-Bridge is for.

Given the laser temp control appnote that he posted i think he's building a TEC cooled laser diode module.
Some diodes require a very precise temp to get the right wavelength and efficiency.
Its quite possible he needs an H-bridge for heating and cooling to maintain a temp within the typical ambient temp range.

Psi is right. I do need the H-Bridge to allow for both heating and cooling to maintain a very precision temperature set point. Also I'd like the ability to quickly and very accurately (1 sec time scale) change temperature. (Im building a not TEC cooled laser doide however. Im building a 10x10 Peltier floor array.{Its for a scientific research application}[If your curious I can post more on this if you'd like])

In terms of power constrains, I dont mind low efficiency as long as I'm able to properly dissipate heat from everything. That being said I do still want to minimize on the number of heat sinks (preferably only limited the the mosfets of the H-bridge) on my board. If possible id like the mount all the mosfets on the bottom of the board and have one giant heat sink cooling them all.

[Rufus]

In summary, I would like recommendations for N channel and P channel Mosfet I could potentially use and I'm also wondering if it is necessary to use a mosfet driver (The LTC1923 can drive most mosfets fine but because ill be switching high current I for some reason think it may be necessary.)

The data sheet contains all the basic information on the trade offs of MOSFET selection. There is no perfect or correct solution. Big MOSFETs give low RDSon but have more gate charge and so won't turn on and off as fast with a given gate drive. A lot of power is dissipated during the on/off transition. Switching and gate drive losses are proportional to frequency the higher the frequency the lower the efficiency, however, higher frequency means lower resistive losses (and a bit more core loss) in the inductors for a given cost and/or size.

Aiming at the top end of the controllers switching frequency capability doesn't feel right.

The biggest decision is probably if you are going to add gate drivers which would allow larger MOSFETs to be switched faster and so be able to keep the frequency higher. If you are going to use gate drivers then using all N channel MOSFETs is a possibility. An additional power supply for the high side MOSFET gate drivers would add little cost when shared between a lot of channels.

[Deadpixel732]

In summary, I would like recommendations for N channel and P channel Mosfet I could potentially use and I'm also wondering if it is necessary to use a mosfet driver (The LTC1923 can drive most mosfets fine but because ill be switching high current I for some reason think it may be necessary.)

The data sheet contains all the basic information on the trade offs of MOSFET selection. There is no perfect or correct solution. Big MOSFETs give low RDSon but have more gate charge and so won't turn on and off as fast with a given gate drive. A lot of power is dissipated during the on/off transition. Switching and gate drive losses are proportional to frequency the higher the frequency the lower the efficiency, however, higher frequency means lower resistive losses (and a bit more core loss) in the inductors for a given cost and/or size.

Aiming at the top end of the controllers switching frequency capability doesn't feel right.

The biggest decision is probably if you are going to add gate drivers which would allow larger MOSFETs to be switched faster and so be able to keep the frequency higher. If you are going to use gate drivers then using all N channel MOSFETs is a possibility. An additional power supply for the high side MOSFET gate drivers would add little cost when shared between a lot of channels.

The reason I chose to switch as 1mhz is because the evaluation board for this product [http://www.linear.com/demo/DC491A] switches at 1mhz. I wont mind adding gate drivers if I can use all N channel Mosfets (never done this before). If I use all N channel mosfets how I would I have to modify the schematic to make that work?  Also one last request if you could recommend me a gate driver I could use to drive the bridge with all N channel mosfets.

Thanks.

[Deadpixel732]

So I read up on controlling a H-Bridge with all N channel Mosfets and it looks simple enough. Im looking at the Allegro Microsystems A4957 or A4940 {Same chip just different package}
Datasheet: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CC0QFjAA&url=http%3A%2F%2Fwww.allegromicro.com%2F~%2Fmedia%2FFiles%2FDatasheets%2FA4957-Datasheet.ashx&ei=nPAUUtLqN_Kj4AO_kIGYDQ&usg=AFQjCNFBK6MGljM7BdYFPiBKUAq7NuoNJw&sig2=ksk1XlpSwz40HA7RkHgppA

I like this chip because it seems simple enough to configure, can handle all four mosfets on one chip, and seems to be able to switch fast enough.

If you guys have any other recommendations for Mosfet drivers that would be great or if there was any reason why i shouldnt potentially use the a4940.

For the N-channel Mosfets Im looking at the IRF6711S.
Datasheet: http://www.irf.com/product-info/datasheets/data/irf6711spbf.pdf

Im considering this mosfet because it can switch relatively fast, has a low rds(on), can handle my 8A load while being able to dissipate lots of heat.

Any other recommendations on potential Mosfets would also be useful.

[Neilm]

I have used some Vishay FETs at 1MHz before - it had a Vds of 100V and was switching 30A. Having a quick look at the Vishay site and found this  http://www.vishay.com/docs/68823/sir462dp.pdf which is the lower voltage version of the one I used.

Neil

[Rufus]

So I read up on controlling a H-Bridge with all N channel Mosfets and it looks simple enough. Im looking at the Allegro Microsystems A4957 or A4940
If you guys have any other recommendations for Mosfet drivers that would be great or if there was any reason why i shouldnt potentially use the a4940.

For the N-channel Mosfets Im looking at the IRF6711S.
Datasheet: http://www.irf.com/product-info/datasheets/data/irf6711spbf.pdf

That driver chip doesn't look like it drives much harder than the LTC1923 and remember the LTC1923 outputs for P channel high side MOSFETS so half the outputs probably need inverting, preferably in the driver chip to avoid additional delays. With programmable dead times of up to 6uS the part obviously isn't intended for switching at half a MHz.

The MOSFET also looks a bit large. 8A and 3mohm is 190mW.. You should be looking at more than 80% efficiency which means less than 10W between all the MOSFETs and inductors. The LTC1923 datasheet has some loss calculations although from memory it seems to skip over MOSFET switching losses which depend on how fast you can drive them and the frequency.

Someone who knows what they are doing (which isn't me) would spend a day or two looking at parts and weighing up the options. No one here can sensibly just say use this or that MOSFET.

[Deadpixel732]

Alright guys. An update.

After weighing my options for an entire day ive decided to do the following.

a)take my switching speed down to 735kHz (so allow for delays due to mosfet driver and for a higher power efficiency)
b)go for a 4 N-Channel Mosfet H Bridge
c)Use the LTC1693-2 as my Mosfet driver because it can adequately supply voltage to drive the N Channels high side and two has outs; an inverting an non inverting, and looks to use little external components (as shown in the schematic ive attached)
Datasheet: http://cds.linear.com/docs/en/datasheet/1693fa.pdf
d)Ive decided to use these N channel Mosfets (IRF6710) becauase of there fast switching times, ability handle my 8A current, a low gate voltage (4.5v) so I will be able to drive these mosfets high side with 12v, and a gate capacitance of 1190pF which the LTC1693-2 can hopefully handle.
Datasheet: http://www.irf.com/product-info/datasheets/data/irf6710s2pbf.pdf

So now im wondering if any of you could recommend me a better Mosfet driver. Preferably one that can switch faster and handle a high gate capacitance so I can choose a mosfet with even faster switching speeds and a lower rds(on). Im having trouble finding one because it doesn't seem like that many have an option for inverted outputs.

O and also if any one could do a quick sanity check to OK my Mosfet driver and mosfet selection.

Thanks.

[HackedFridgeMagnet]

What about CSD16342Q5A http://www.digikey.com.au/product-detail/en/CSD16342Q5A/296-30314-6-ND/3133883 might be worth a look.

Though I didn't read your requirements thoroughly, it seems a little better.

[Deadpixel732]

What about CSD16342Q5A http://www.digikey.com.au/product-detail/en/CSD16342Q5A/296-30314-6-ND/3133883 might be worth a look.

Though I didn't read your requirements thoroughly, it seems a little better.

The switching times of this mosfet look similar and the gate capacitance and Rds(on) are both lower. I think i may use this one instead. Thanks for the suggestion . Also now since ill be dissipating less than a watt with this mosfet do you think that its still necessary to have a heat sink mounted on it. 

[Deadpixel732]

Also another question.

Since I want the drive the N channel Bridge with 12v. How many mA would I have to supply per bridge. Would it be small enough to use a 5v to 12v charge pump like the MAX662 (up to 30mA) or the LTC1263 (up to 60mA)?
 
LTC1263
http://cds.linear.com/docs/en/datasheet/1263f.pdf

MAX662
http://datasheet.octopart.com/MAX662ACPA%2B-Maxim-datasheet-112904.pdf

[Rufus]

Since I want the drive the N channel Bridge with 12v. How many mA would I have to supply per bridge. Would it be small enough to use a 5v to 12v charge pump like the MAX662 (up to 30mA) or the LTC1263 (up to 60mA)?

The


IRF6710 is specified with 8.8nC typical total gate charge (at 4.5v Vgs) pumping that charge at your switching frequency of 750kHz makes 6.6mA and proportionally more with higher Vgs. That plus whatever the driver needs gives you a ball park figure.