2
\$\begingroup\$

I'm designing this general purpose mosfet controlled output, and I'm used to putting D1 due to possibly inductive loads.

enter image description hereHowever I was looking into some other implementations and this guy suggests putting a shunt capacitor (C24 in my schematic) to deal with some ringing when driving cables. Looking some more into it, I couldnt find any version of this snubber without a resistor, so I wonder how do I estimate a good value to deal with some possible ringing? Is there a ballpark I can work with? Is it even necessary? I've driven long cables and motors just with the diode without apparent issues. However if there is a remote possibility of needing it I'd like to leave the footprint on the pcb so I can populate if needed. Im looking into driving up to 10A at 12V~30V max.

I'm also looking into putting a totem pole gate driver, is it a good idea to still leave a pull down(10~100k) from Gate to Source? Again I can leave it unpopulated on the PCB, or I can remove it from the design if its absolutely useless/unnecessary.

\$\endgroup\$
  • \$\begingroup\$ Please pay close attention to schematic drawing schemes that you use and use to depict. Your REFDES D1 (referred in your question text above) is not very evident in your schematic. \$\endgroup\$ – Chetan Bhargava Oct 8 '15 at 5:41
  • \$\begingroup\$ Your bipolar devices driving the gate appear to be backwards (I would expect the PNP emitter to power, NPN emitter to ground). \$\endgroup\$ – Peter Smith Oct 8 '15 at 7:21
  • 1
    \$\begingroup\$ @PeterSmith The drice transistor pair polarity and connections are correct. This is a classic dual emitter follower gate driver.. When Vin (ASE1 is above Vgate the upper NPN pulls the gate up as an emitter follower and when Vin is below Vgate the lower PNP pulls ikt down as an emitter follower. There is a 2 x Vbe dead spot in the middle which is usually crossed at speed by the drive weaveform BUT which gives informal shoot through protection by not allowing boith bipolars to be on at once. \$\endgroup\$ – Russell McMahon Oct 8 '15 at 8:12
  • \$\begingroup\$ My anser has been substantially added to based on your qiuestions and 'other'. \$\endgroup\$ – Russell McMahon Oct 8 '15 at 12:10
  • \$\begingroup\$ For interest - where are you based? 100% tax sounds "rather nasty" :-(. ....> ah- Brazil -yes, others have noted the ratyher harsh taxes on imports :-(. \$\endgroup\$ – Russell McMahon Oct 8 '15 at 12:28
1
\$\begingroup\$

(1) As @autistic says - having a snubber is useful but you must use series resistor with C24 - with no series R and ideal components Icap into the FET is infinite at turn on. At a minimum, total R in cct from cap via FET and power supply must limit I peak to what FET will withstand repetitively. Also the cap current may cause it problems but the FET would often fail first.

(2) The driver shown is an excellent one as long as there is enough gate drive voltage available. The two emitter followers have unity (actually just less) voltage gain BUT narge current gaion - and current drive is what the FET gate capacitor needs for fast switching.

(3) A totem pole driver has its place if of the correct polarity and if voltage is limited - it will allow you to use a common emitter upper stage (PNP, emitter to +5V, collector to gate resistor) to carry the gate essentially all the way to the (here) +5V rail BUT the existing driver is better if more voltage is available - which you say it is. As shown you cannot get Vgate above a Vbe drop below Vin so if you are driving with 0/5V PWM adding a 12V supply for the driver helps only minimally. You can add one more transistor between Vin and driver in common emitter mode with a pullup resistor (lowish value - say 1k to 10k) to eg +12V to transform the PWM from 0/5 (or even 0/3) to ~= 12/0. This inverts the PWM signal which must be allowed for. In any case, adding a small "speedup" capacitor (usually 1 nF or less) across RB (ir R from Vin to first driven stage) will give you a useful gain in switching response. This works by forming a voltage divider with the base capacitance to "sweep charge" from the base when turning off the transistor.

(4) RG1 should be low but non zero - this LIMITS max gate current at turn on and slows FET waveform just a tad and reduces EMI AND reduces ringing. Usually around 10 Ohms - maybe lower.

(5) A very very very useful addition in all except PURE resistive load cases (ie none at 20A) is a zener at gate-source reverse biased with Vzener slightly above Vdrive max (here 5V). Mount zenera as close as possible to FET gs with shortest leads, tracks. This catches positive gate transients coupled from drain via Millar capacitance that the FETs days may be long on the face of the land. I have seen such expend FET lifetime from minutes to hours out to indefinite.

(6) Belts and braces: If you have ringing issues you can add a 'small as you like' reverse biased schottky diode from gate to source mounted even closer to g & s than the zener if possible. This clamps any negative going ringing transitions very severely. A waveform which has an unclamped positive half cycle but a negative part clamped to about 0.3V rapidly loses its enthusiasm and as amplitude is then well under Vgson will have minimal effect.

(7) The IRLR2905 is marginal for the drive voltage you are using. The emitter follower takes the better part of 1 volt drop and you need a good 4V Vgs to drive the FET at 20A and I'd try for even more. See datasheet here figs 1,2 & 3. I'd suggest a part that is very happy at 4V Cgs and adequate at 3V.

Note that, as often but not always is the case, the graphs (which are typical and not worst case) are specified at low on periods (here 60 uS) and often low or very low duty cycles and Rdson is usually higher in practice than may be expected. As a rule of thumb double the Rdson in the header or use the max value in the data sheet.

Increasing the gate drive somewhat would work well. Note that Vgs_max is 16V (data table page 1) so 12V is a bit high for comfort as a gate clamp zener needs to be somewhat higher again that 12V, leaving little margin.
Note that the above comments about the FET being marginal at 5V gate drive supply is for "proper" design where worst case specs MUST be used. The datasheet shows Vgsth as 1V-3V and in many cases it will work "well enough". If you are prepeared to 'select on test' and /or observe finished result it will probably usually work OK at 20A. You MAY get a batch of FETS with Vgsth near the max value which may get hotter than others - if that happens you have no comeback as it's within spec.

FET selection can be done using parametric tanbles from varioous suppliers. I find Digikey excellent for this. Even if you seldom buy from them (and they are good people to deal with when it suits you to do so) their component databse can be used this way. Search for MOSFETS then subset down using parameter selection. Then look at datasheets of those that may suit. In particular look at the graphs such as those I cite elesenwere in this answer that show FET performance in the areas of interest. eg in this case with Id at say 20A, Vds at say 1 V (as that is where most graphs set it) and see how much gate voltage yopu need to fit the selected point for Vgs at 3V to 4V. Digikey also give order of pricing and more - a very useful resource.


Note: @PeterSmith - The drive transistor pair polarity and connections are correct. This is a classic dual emitter follower gate driver.. When Vin (ASE1 is above Vgate the upper NPN pulls the gate up as an emitter follower and when Vin is below Vgate the lower PNP pulls ikt down as an emitter follower. There is a 2 x Vbe dead spot in the middle which is usually crossed at speed by the drive weaveform BUT which gives informal shoot through protection by not allowing both bipolars to be on at once.

\$\endgroup\$
  • \$\begingroup\$ The IRLR2905 is indeed an unprotected MOSFET, so adding a Zener is a good idea. I honestly don't bother buying such (unprotected MOSFETs) these days. \$\endgroup\$ – Fizz Oct 8 '15 at 9:32
  • \$\begingroup\$ Do you guys have a suggestion for a similar protected device with reasonable threshold voltage? The device depicted in the schematic is one that is easily bought where I live. This PCB is intended for people who do not want/cant buy from mouser/digikey. \$\endgroup\$ – Wesley Lee Oct 8 '15 at 11:18
  • \$\begingroup\$ It might be easier to source parts to drive the emitter follower at 12V, would it be better? I'd change the zener accordingly. \$\endgroup\$ – Wesley Lee Oct 8 '15 at 11:30
  • \$\begingroup\$ i dont think I can just add 12V to the driver because it could bring Rbase to 12V, right? So if I do want to drive with 12V I need to add the transistor between the MCU and the driver. I'm not sure if I want to invert the signal though. Considering 10A is actually on the very high side of what I need, I might leave it as it is. I do often order from mouser and/or digikey, its just that it costs about 100% taxes where I'm from and the people paying for this project are not inclined to doing so right now :) \$\endgroup\$ – Wesley Lee Oct 8 '15 at 12:22
  • \$\begingroup\$ Also then I'd need to add a regulator from Vin down to 10~12V to cater the possibility of using 20~30V supplies because of the Vgs limit. \$\endgroup\$ – Wesley Lee Oct 8 '15 at 12:24
1
\$\begingroup\$

Keep your gate pull down ,Your complimentary follower is better than totem pole for what you are doing .The capacitor will do more harm than good the way you are setting things up .Place a resistor in series with the cap so you can reduce Q and hence ringing like what your friend wanted .The resistor also limits charging current at mosfet turnon .I saw somebody do this in 1992 on a job that drove a PCB car windscreen motor which was dealing with voltages and currents in your ballpark .He made the big mistake of using a slug diode for D1 and got bad waveforms which he "cured "with a 1 microfarad 63V film cap across the diode .The result that I inherited was molten caps !

\$\endgroup\$
  • \$\begingroup\$ Re "complimentary [sic] follower is better than totem pole". I thought they were the same thing. \$\endgroup\$ – Fizz Oct 8 '15 at 5:57
  • \$\begingroup\$ The traditional totem pole uses 2 NPN devices .It is an inverting driver. \$\endgroup\$ – Autistic Oct 8 '15 at 8:58
  • \$\begingroup\$ I see. It's a matter of terminology variation. For some people (see my answer for an example) push-pull = totem pole, but for others it means something different. In Wesley Lee's question they are the same though. \$\endgroup\$ – Fizz Oct 8 '15 at 9:05
  • \$\begingroup\$ I think I refered to the configuration that I used in the wrong name? Its not a tradicional totem pole right? \$\endgroup\$ – Wesley Lee Oct 8 '15 at 11:13
1
\$\begingroup\$

Regarding your 1st question, for the pretentious applications (where the load is known) there are how-tos, e.g. NXP AN11160 or Williams's book chapter 8 for determining the appropriate snubber values based on the load characteristic. I suppose you could plug-in a range of ballpark figures for your load[s] in there and see what the snubber would look like in the best and worst-case scenarios.

As for,

I'm also looking into putting a totem pole gate driver, is it a good idea to still leave a pull down(10~100k) from Gate to Source? Again I can leave it unpopulated on the PCB, or I can remove it from the design if its absolutely useless/unnecessary.

The appnotes I've looked at don't include a GS resistor when using a totem pole driver.

https://www.fairchildsemi.com/application-notes/AN/AN-6069.pdf enter image description here

http://www.radio-sensors.se/download/gate-driver2.pdf

enter image description here

The latter one has a fairly detailed discussion of the additional components needed for optimal operation (caps etc.) but still says nothing about adding a RGS. (This was a TI seminar [slup169], not some random musing of someone on the internet, but it's gone from TI's site for some reason.) Also note that the guy from which you took inspiration from in your schematic does have a supply bypass cap for the BJT driver pair, which all notes on this matter strongly suggest adding, but is missing from your schematic.

The only reason I can think of to add that RGS is if you have a use case where the 5V power rail [for the driver] can be removed (or powered down) but the 12V supply of the MOSFET kept up and you want the MOSFET to [immediately] turn off the load in this scenario.

\$\endgroup\$
  • 1
    \$\begingroup\$ I added the the capacitor now and losing 5V power with 12V still is a possible scenario and yes, I need it to be powered down so I will keep RGS. \$\endgroup\$ – Wesley Lee Oct 8 '15 at 11:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.