How to speed up this high side FET driver

Go To Last Post
27 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi there - I'm working on a high side FET driver for a 1MHz boost converter. The simulation of it looks... OK. Not as bad as it could be, but certainly not good.

I've attached a picture of the simulation for those without LTSpice, and I've also included the simulation files for those with LTSpice. All you have to do is extract the files in the zip to one directory.

The concept was that Q1 is acting as a low current level shifter, Q4 is putting a reasonable gain on Q1's current, then Q2/Q3 are the push pull stage driving the gate of the FET. Unfortunately - to get it to even kind of work - I had to make R5 and R6 really small, so it's completely not working how I'd like.

Any suggestions as to how to speed this up while lowering power consumption on the 12V line? The 12V supply, even though it's shown powering the load, will actually be a diode-capacitor charge pump/boost circuit. So it will be very low power.

Thanks!

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Take a look at:
diodes.com/_files/products_appnote_pdfs/zetex/an22.pdf

Baker clamps are quite useful. And it discusses how to speed up PNP's.

Have fun with this! (I first used baker clamps in the early 70's. Saved my bum!)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Nice application note, Jay! Thanks.

Nard

A GIF is worth a thousend words   They are called Rosa, Sylvia, Tricia, and Ulyana. You can find them https://www.linuxmint.com/

Dragon broken ? http://aplomb.nl/TechStuff/Dragon/Dragon.html for how-to-fix tips

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jaydhall wrote:
Take a look at:
diodes.com/_files/products_appnote_pdfs/zetex/an22.pdf

Baker clamps are quite useful. And it discusses how to speed up PNP's.

Have fun with this! (I first used baker clamps in the early 70's. Saved my bum!)


The idea of the baker clamp makes sense... But in practice I'm getting bad results.

I believe the problem is that I'm not driving the BJT with a bipolar signal - so the reverse diode never really turns on giving me slow turn off. So I'm not sure that a baker clamp is the right solution for this. What do you think?

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hmm... look at figure 1c again. Now, compare your schematic to it.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Add 100pf across R3 maybe? They must call em 'speedup caps' for some reason?

Imagecraft compiler user

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jaydhall wrote:
Hmm... look at figure 1c again. Now, compare your schematic to it.

Oops - I was referring to the circuit shown here: http://home.mira.net/~gnb/audio/...

I tried the 1c circuit and that didn't seem to help significantly.

However, adding a 1nF cap across R2 seems to help a good deal.

bobgardner wrote:
Add 100pf across R3 maybe? They must call em 'speedup caps' for some reason?

That helps a good deal for the turn on. However, it's the turn off that is terribly slow.

The delay is in the output of Q4. When I want Q4 off, it takes 200ns for its output to fall. The current through it seems to stay fairly steady. By reducing R5 to 500 I can speed it up to about 150ns, but that's about it. Plus, I really need to make R6 bigger, because I'm dropping a ton of power across it, which is not going to work.

So - how can I speed up Q4 while lowering power consumption?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Is the simulation important or a practical
application ? In a practical application I would use
a good gate-driver IC.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

How about less total resistance for the gate divider net? Try the same ratio for the divider, but half total resistance. That will give more current to the base of Q4, and hopefully faster switching. It won't use LESS power though, for that I would go with MOSFETs, not BJT's...

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Either use one diode, a shottkey, or four diodes. Two in series in the base. Your reference is incorrect. IMO.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jaydhall wrote:
Either use one diode, a shottkey, or four diodes. Two in series in the base. Your reference is incorrect. IMO.

You mean three diodes right? I don't see any circuits using four. I tried doing the three diode baker clamp on Q1, but the capacitor across R2 beats it. However, I didn't see a way of doing the capacitor trick on Q2-Q4, so I tried doing it on Q4 and it gives interesting results. I had to play around with the base to emitter resistor - but the best value I found still gave me about the same fall time for the FET and worse rise time. Maybe the diodes I chose are not ideal for this? I'm using these: http://www.fairchildsemi.com/ds/...

I tried models from both NXP and Fairchild Semiconductor for the BAS16 - they produced very similar (and not good) results.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

UNiXWHoRe wrote:
How about less total resistance for the gate divider net? Try the same ratio for the divider, but half total resistance. That will give more current to the base of Q4, and hopefully faster switching. It won't use LESS power though, for that I would go with MOSFETs, not BJT's...

It's already using way too much power. I originally planned on those resistors being way bigger.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ossi wrote:
Is the simulation important or a practical
application ? In a practical application I would use
a good gate-driver IC.

Well, I want to actually have a working circuit. I'm sure I could buy a great driver off the shelf - but where's the fun in that?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Overall comment: would this be better done with FETs? I have no inherent need to use BJTs here.

edit: messing around with an all FET driver. Not looking good.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

We have discussed this, Michael. The 1MHz parameter is the killer. You need proper drivers at such frequencies.

To all: nleahcim wants to use this high frequency to keep the inductors small. Good reason, but it will be a balance between size of inductors and complexity of the drive. Right ?

Quote:
Overall comment: would this be better done with FETs? I have no inherent need to use BJTs here.

edit: messing around with an all FET driver. Not looking good.

My SMPS runs with a small fet BS107 where you use Q1 in the schematic High Side FET Driver1.jpg above. Works better (read: faster) than a BJT

Nard

A GIF is worth a thousend words   They are called Rosa, Sylvia, Tricia, and Ulyana. You can find them https://www.linuxmint.com/

Dragon broken ? http://aplomb.nl/TechStuff/Dragon/Dragon.html for how-to-fix tips

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Make the turnoff half exactly the opposite of the turn on half. That is, an npn with gounded emitter will turn on just as fast as the pnp with emitter to +v does. Like a symmetrical audio amp?

Imagecraft compiler user

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Your schematic is very confusing. The mosfet in the schematic is part of the high side driver, or is actually the driven mosfet ?
I notice two power supply, 30 and 12, with negative connected to the common. Then your pulse signal reference is above the 12 V. You talk about diode capacitor bust converter, Plons about using an inductance, and in the schematic there is a 0.5 ohm resistor as load, so confusing.
The timing and scope plots are real or you just play on the computer ?
George.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

angelu wrote:
Your schematic is very confusing. The mosfet in the schematic is part of the high side driver, or is actually the driven mosfet ?
I notice two power supply, 30 and 12, with negative connected to the common. Then your pulse signal reference is above the 12 V. You talk about diode capacitor bust converter, Plons about using an inductance, and in the schematic there is a 0.5 ohm resistor as load, so confusing.
The timing and scope plots are real or you just play on the computer ?
George.

The FET is the driven mosfet. I put in a dummy resistor load on it. The reason for the 30V is that this is a high side FET on top of a synchronous boost converter. So I have to do a level shift. The plots are simulations made with LTspice.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Plons wrote:
We have discussed this, Michael. The 1MHz parameter is the killer. You need proper drivers at such frequencies.

To all: nleahcim wants to use this high frequency to keep the inductors small. Good reason, but it will be a balance between size of inductors and complexity of the drive. Right ?

Quote:
Overall comment: would this be better done with FETs? I have no inherent need to use BJTs here.

edit: messing around with an all FET driver. Not looking good.

My SMPS runs with a small fet BS107 where you use Q1 in the schematic High Side FET Driver1.jpg above. Works better (read: faster) than a BJT

Nard


Nard - I made some tweaks (will post them tomorrow, they're on a different computer) that got me 50ns turn on and turn off times. All with BJTs. That's starting to be reasonable, assuming reality matches the simulation. I still think I can get 1MHz! I mean, it's not like they're not using transistors in FET driver ICs. They just have people more clever than myself designing them.

I think on the board I will put a spot for a proper FET driver in case this circuit does not work - but I would rather not use it!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
How about less total resistance for the gate divider net? Try the same ratio for the divider, but half total resistance. That will give more current to the base of Q4, and hopefully faster switching. It won't use LESS power though, for that I would go with MOSFETs, not BJT's...

Quote:
I had to play around with the base to emitter resistor - but the best value I found still gave me about the same fall time for the FET and worse rise time.

The problem with Q4 is it saturate. As more current you push to its base, to speed up the turning ON, the more time it needs to turn off. So ,here it comes the diodes trick. Use two or three diodes, and your capacitor for R2, and you will see good results.
Another problem, you should use faster transistors for Q1 and Q4, and with less amplification (A type, not C).
Having to deal with level translation, when your mosfet is off, the current through R6, R5 will be much smaller than when the mosfet is ON, and this will affect the turning ON time, and maybe will add some instability. Usually level translation is done with a differential circuitry.
The circuit around Q2 and Q3 can be improved also.
But don't spend too much time simulating, go ahead with the real circuit.
George.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

angelu wrote:
Quote:
How about less total resistance for the gate divider net? Try the same ratio for the divider, but half total resistance. That will give more current to the base of Q4, and hopefully faster switching. It won't use LESS power though, for that I would go with MOSFETs, not BJT's...

Quote:
I had to play around with the base to emitter resistor - but the best value I found still gave me about the same fall time for the FET and worse rise time.

The problem with Q4 is it saturate. As more current you push to its base, to speed up the turning ON, the more time it needs to turn off. So ,here it comes the diodes trick. Use two or three diodes, and your capacitor for R2, and you will see good results.
Another problem, you should use faster transistors for Q1 and Q4, and with less amplification (A type, not C).
Having to deal with level translation, when your mosfet is off, the current through R6, R5 will be much smaller than when the mosfet is ON, and this will affect the turning ON time, and maybe will add some instability. Usually level translation is done with a differential circuitry.
The circuit around Q2 and Q3 can be improved also.
But don't spend too much time simulating, go ahead with the real circuit.
George.

George - how do I know if a transistor is fast? Are you saying that lower hfe transistors are faster than higher hfe transistors? I specifically chose all of these transisters for their high hfes!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

See this http://media.digikey.com/pdf/Dat...
You need switching transistors, not with very high amplification.
Transistors with C suffix, are used mainly in the linear region, amplification, and provide less noise.
In the data sheet for a switching transistor, look for parameters such delay time, rise time, storage time, fall time to be small.
George.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

In the case of a MOSFET, also look at gate charge (smaller == faster switching, but usually also lower source-drain current...)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

angelu wrote:
See this http://media.digikey.com/pdf/Dat...
You need switching transistors, not with very high amplification.
Transistors with C suffix, are used mainly in the linear region, amplification, and provide less noise.
In the data sheet for a switching transistor, look for parameters such delay time, rise time, storage time, fall time to be small.
George.

Sorry for the slow response - things got hectic over here.

I'm noticing that most BJTs don't specify delay/rise/storage/fall times. Are there other specs I should be looking at?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Ft is the indicator used for bipolar transistors. It is the frequency where Hfe drops to 1. Also called gain-bandwidth product. As you go down in frequency from that point, the gain at frequency, F, and the frequency F have a product of 1. This, in turn, means that the 3db corner frequency of the frequency response is Ft/beta (where beta is the current gain at DC).

Jim

 

Until Black Lives Matter, we do not have "All Lives Matter"!

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ka7ehk wrote:
Ft is the indicator used for bipolar transistors. It is the frequency where Hfe drops to 1. Also called gain-bandwidth product. As you go down in frequency from that point, the gain at frequency, F, and the frequency F have a product of 1. This, in turn, means that the 3db corner frequency of the frequency response is Ft/beta (where beta is the current gain at DC).

Jim


Jim - isn't there a lot more to take into account though? I mean, I think that refers more to small signals, right? I'm more interested in switching speed. Also, I think that assumes you have a very stiff source driving the base of the BJT, which I certainly don't have!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

My thoughts to speed up the original circuit are as follows:
1: Replace Q1, BJT, with a 2N7000 etc mosfet.
2: Place a schottky diode between the base and collector of Q4 to prevent
it fron saturating. Cathode to base.
3: Reduce the value of R5 to 1K or less.
4: Replace R3 with a short, unless you are trying to delay turn on time.
5: Its good practice to add a resistor, 4.7R to 22R or so in series with the gate of the output mosfet.

BTW, Why is there a -30V supply?

Ron.