H Bridge High Side Mosfet Driver

n9352527

Joined Oct 14, 2005
1,198
Originally posted by bigbigblue+Mar 19 2006, 02:24 PM--><div class='quotetop'>QUOTE(bigbigblue @ Mar 19 2006, 02:24 PM)</div><div class='quotemain'>I've had a look at the datasheet for the IRF9Z34 P Channel MOSFET from IRF (datasheet can be found here : http://ec.irf.com/v6/en/US/adirect/ir?cmd=...ductID=IRF9Z34).

It will pass 18amps - more than enough for my application, with an RDS(on) of 0.14 ohms (twice that of the n-channel IRF540 I am currently using).

At 8 amps, the IRF540 will dissipate 4.928 watts, whereas the IRF9Z34 will dissipate 8.96 watts (ie twice as much heat). The reason I am having a problem with the N channel MOSFETS is that they got hot due to a significant proportion of their operation being in the linear area - but I doubt whether they are dissipating as much as the P channel will when it is hard on. Therefore I am not sure I gain much from using a P Channel MOSFET (other than ease of driving them).
[post=15180]Quoted post[/post]​
[/b]


That is the only gain, ease of driving, no need to generate higher gate voltage. As you already have 32V supply handy and you don't need to generate higher gate voltage, then the advantage is not there anymore.

There are two main sources of power dissipation in switching, the conducting (saturated) power loss and the switching power loss. To minimise the conducting power loss you need to ensure the device is in saturation (enough gate voltage) and select the lowest Rds(on) you can afford. To minimise the switching power loss basically you need to ensure that the device switches as fast as possible (not counting ZVS or ZCS schemes).

If you have a scope, you could check the switching waveform and see which one is your main problem, conducting (i.e. device not in full saturation) or switching (i.e. slow switching transition).

If you could capture the switching waveform and post it here that would be very helpful in diagnosing why the MOSFET gets hot.

One easy mistake to made at high frequency switching is to mount the MOSFET too far from the gate driver circuit. Long track or mounting wire on the gate would slow it right down.

Originally posted by bigbigblue@Mar 19 2006, 02:24 PM
I wonder if I could run by you my understanding of the gate charge and how much current is required to achieve a hard switch on in a given time ?

In the IRF540 datasheet, it states :

Total gate charge = 69nC (nano-Coulombs ?)
Gate to source charge = 13nC
"Miller" charge = 37nC

(all at Id = 29A, Vds = 80V, Vgs = 10V.

So, as I understand it, I need to shove 69nC of charge into the Gate at 10V above the Source Voltage in order fot the gate to be fully open. Is this correct?
[post=15180]Quoted post[/post]​
Ignoring the capacitance variations due to different Vds and Vgs and feedback through the Miller capacitance, that amount of charge is correct. However, I much prefer to say that in order to raise the Vgs to 10V, the gate needs approx. 69nC of charge. To be fully on, the MOSFET is also affected by dynamic (transient) characteristics of the channel and the circuit (this is related to the t(on)/tf and t(off)/tr you mentioned below).

The gate charge value is very useful in calculating the gate drive loss and charging/discharging currents, not very useful in calculating the ton/toff times.

Originally posted by bigbigblue@Mar 19 2006, 02:24 PM
Ignoring the CR time constant of the gate capacitance and any losses for a moment, to charge the Gate in :

1 nS requires 69 Amps of charge current
10 nS requires 6.9 Amps of charge current
100 nS requires 690mA of charge current
1 uS requires 69mA of charge current
etc etc

Again, is this correct (I think it is OK as 1 Amp = 1 Coulomb / Second) ?
[post=15180]Quoted post[/post]​
Yes, charging the gate to 10V.

Originally posted by bigbigblue@Mar 19 2006, 02:24 PM
The input capacitance of the IRF540 is 1300pF ( = 1.3nF?). Assuming I have a 32V power supply to provide the Gate voltage and the max Source voltage is 12v, and I have a 15 v zener between Gate and Source (so I can guarantee the gate is no more than 15V above the Source).
[post=15180]Quoted post[/post]​
Is this 1.3nF the ciss? You need to differentiate the ciss with total gate charge. Have a look at the typical ciss measurement conditions/circuit and compare it to your circuit. ciss is more useful in AC analysis.

Originally posted by bigbigblue@Mar 19 2006, 02:24 PM
As I have a 32V supply, to limit the current to the Gate to 690mA (assuming I want to charge the Gate in 100nS) I would need a current limiting resistor of 46.38 ohms (=32 / 0.69).

Now, taking the (CR) time constant of the gate into account:

46.38 x 1.3 nS = 60nS (to achieve 66% of the charge)
or 120nS to achieve 88.4% of the gate charge (66% + (66% of 34%))

etc etc

So, I should be able to get something close to my wished 100ns turn on time. I am of course ignoring rise times and turn on delays of the MOSFET here. Or have I got this badly wrong ?
[post=15180]Quoted post[/post]​
You've said it yourself, ignoring the tf and the tr. In fact, when you drive the gate sufficiently hard, it is no longer the deciding factor in tr and tf. The datasheet usualy gives these values, such as tf = 20nS and tr = 100nS with near perfect (very large current capability) gate drive under specific Vgs, Vds, Id and load type. These would be the best timing that you would get under similar conditions. Your load type (inductive in your case) and circuit would also affect the switching time.

<!--QuoteBegin-bigbigblue
@Mar 19 2006, 02:24 PM
Now to a problem I forsee. As I have a 32v supply and a current limiting resistor of 46.38 ohms (giving a 690mA current), the power dissipation of that resistor is 0.69 x 0.69 * 46.38 = 22 Watts approx - one big resistor - and I need one for each high side MOSFET Gate.

All comments on the above are most welcome.
[post=15180]Quoted post[/post]​
[/quote]

You do not take into account the fact that the 690mA current only flows for a very very short duration (a few nS?) with relatively low duty cycle. Once the gate is charged the current ceases to flow (in fact it starts decreasing the moment the gate starts charging). The average power dissipation is, therefore, much much lower than the 22W you are suggesting (ignoring any current shunted by the zener). There is a way to calculate this, but you have to get hold of the thermal transfer characteristics of the resistor that you are going to use (quite difficult to get and a waste of precious time). The easiest way is just trial and error, calculate the shunted current and find the appropriate resistor power to sustain that power and just add 0.5W to start with. Increase the rating if it was too hot to touch during operation. Use metal film if you can.
 

Thread Starter

bigbigblue

Joined Mar 15, 2006
42
Originally posted by n9352527+Mar 20 2006, 01:15 PM--><div class='quotetop'>QUOTE(n9352527 @ Mar 20 2006, 01:15 PM)</div><div class='quotemain'>That is the only gain, ease of driving, no need to generate higher gate voltage. As you already have 32V supply handy and you don't need to generate higher gate voltage, then the advantage is not there anymore.
[/b]

Ok thanks - that makes things clearer. I think I will stick with n-channel MOSFETS.

Originally posted by n9352527@Mar 20 2006, 01:15 PM

(not counting ZVS or ZCS schemes).
Does ZVS mean "Zero Voltage Switching" and ZCS mean "Zero Current Switching"?

Originally posted by n9352527@Mar 20 2006, 01:15 PM
If you have a scope, you could check the switching waveform and see which one is your main problem, conducting (i.e. device not in full saturation) or switching (i.e. slow switching transition).
I have a scope and it is definitley due to slow switching transition.

I was attempting to design a controller which could drive stepper motors with a current requirement of between 500mA and 16 Amps, with a max motor supply voltage of 24V. I now think I was being too ambitious in my original design parameters :)

The problem comes when attempting to regulate the current to a low level (say 500mA) with high motor supply voltage - the switching rate of the controlling logic is high (it gets up to about 420KHz !)

At that frequency, the gate charge is too slow and the gate does not get saturated during the "on" pulse time - so the MOSFET is in the linear region most of the time and gets hot.

When I turn up the motor current, the repetition rate of the switching logic drops considerably. I have seen it go as low as 1KHz @ 4.2 amps with a 12v motor supply. At this point, the gate driver manages to saturate the gate and it remains saturated for a high percentage of the on time (90-95% range) and not suprisingly, the Mosfet runs cool.

Originally posted by n9352527@Mar 20 2006, 01:15 PM
One easy mistake to made at high frequency switching is to mount the MOSFET too far from the gate driver circuit. Long track or mounting wire on the gate would slow it  right down.
Is this due to stray capacitance of the trace / wire? If so, I have made this mistake - the gate driver is on a breadboard at the moment with trailing wires or around 6 inches! How close should the driver be to the gate to avoid the problem?

Originally posted by n9352527@Mar 20 2006, 01:15 PM
Is this 1.3nF the ciss? You need to differentiate the ciss with total gate charge. Have a look at the typical ciss measurement conditions/circuit and compare it to your circuit. ciss is more useful in AC analysis.
Yes it is the Ciss

Originally posted by n9352527@Mar 20 2006, 01:15 PM
You do not take into account the fact that the 690mA current only flows for a very very short duration (a few nS?) with relatively low duty cycle. Once the gate is charged the current ceases to flow (in fact it starts decreasing the moment the gate starts charging). The average power dissipation is, therefore, much much lower than the 22W you are suggesting (ignoring any current shunted by the zener).
The duty cycle is actually near to 70% when the switching is happenning at 420KHz, but yes you are quire right for low duty cycles I had not taken account of this.

Could you post a schematic of a transistor gate drive? I was assuming that it was simply a resistor and transistor in series between the gate supply and 0v with the junction of the transistor and resistor tapped off to the MOSFET gate, with the transistor switching on and off, however if this is the case, the resistor will be passing current 100% of the time (either through the Gate of the MOSFET or through the switching transistor). However as you mention this is not the case, I must have misunderstood how the transistor based gate switching works.

<!--QuoteBegin-n9352527
@Mar 20 2006, 01:15 PM
The easiest way is just trial and error, calculate the shunted current and find the appropriate resistor power to sustain that power and just add 0.5W to start with. Increase the rating if it was too hot to touch during operation. Use metal film if you can.
[/quote]
I will do.

Many thanks for your help, I really do appreciate the effort it takes to reply to my posts. Hopefully one day I will be able to reciprocate.
 

Thread Starter

bigbigblue

Joined Mar 15, 2006
42
Oh I forgot to mention that I think I now understand why the IRF2304 MOSFET driver does not work for me.

I believe it REQUIRES HIN to be pulsed, as the bootstrap capacitor only gets charged whilst HIN is low. In the start up condition of my circuit, HIN is high, as there is no current flowing in the motor, therefore my control circuit calls for the MOSFET to be switched on.

In an application note from IRF there is a circuit which uses a separate charge pump which runs whenever the HIN is high, this might be a solution, otherwise I will need some way of holding HIN low for a period at startup to allow time for the bootstrap capacitor to charge.
 

n9352527

Joined Oct 14, 2005
1,198
Originally posted by bigbigblue+Mar 20 2006, 04:11 PM--><div class='quotetop'>QUOTE(bigbigblue @ Mar 20 2006, 04:11 PM)</div><div class='quotemain'>Does ZVS mean "Zero Voltage Switching" and ZCS mean "Zero Current Switching"?
[post=15208]Quoted post[/post]​
[/b]


Yes. These are not relevant to your design.

Originally posted by bigbigblue@Mar 20 2006, 04:11 PM
The problem comes when attempting to regulate the current to a low level (say 500mA) with high motor supply voltage - the switching rate of the controlling logic is high (it gets up to about 420KHz !)
[post=15208]Quoted post[/post]​
Are you trying to limit the current by switching off the MOSFET when the current reaches a certain value? If so, do you wait for a predefined time lapse to switch the MOSFET on again or do you wait until the current falls below a certain value?

If that is the case, I usually call this kind of control scheme bang-bang switch mode controller. It bangs onto the upper limit, switches off, bangs onto the lower limit and switches on. Kind of poor man PWM :p Just be careful with spurious frequencies all over the place, especially if you have any sensitive circuit around.

Originally posted by bigbigblue@Mar 20 2006, 04:11 PM
At that frequency, the gate charge is too slow and the gate does not get saturated during the "on" pulse time - so the MOSFET is in the linear region most of the time and gets hot.
[post=15208]Quoted post[/post]​
Have you compared the falling/rising times at 420kHz and at much lower frequency, say at 25kHz? If they are significantly different, then you have some problem with your gate driver circuit. It would be good if you could post the tf and tr at those two frequencies here so we could get an idea where the problem is.

420kHz is a high frequency for PWM, but it is certainly not impossible to design. I've designed a few switching circuits at 500kHz before, so it is definitely achievable.

There are two problems in trying to switch at high frequency efficiently. The first is you have to be able to saturate the MOSFET repetitively. For example, if you have a bootstrapped gate drive supply, is the charge reservoir large enough for high repetition gate charging without dropping below the recommended saturated gate voltage (~10V)? Does it have time to recharge the reservoir back up?

The second is to switch fast enough, is the reservoir able to supply the current needed? Is your gate drive able to pass this amount of current efficiently?

Originally posted by bigbigblue@Mar 20 2006, 04:11 PM
When I turn up the motor current, the repetition rate of the switching logic drops considerably. I have seen it go as low as 1KHz @ 4.2 amps with a 12v motor supply. At this point, the gate driver manages to saturate the gate and it remains saturated for a high percentage of the on time (90-95% range) and not suprisingly, the Mosfet runs cool.
[post=15208]Quoted post[/post]​
95% saturation at 1kHz frequency, assuming 50% duty cycle is about 2.5ms total rising/falling times. That is awfully slow! You definitely have a problem somewhere. The reason it stays cool at 1kHz is because it switches on and off only 1000 times a second. The dominant factor is the conducting loss. If you increase the frequency and the rising/falling time stay around that figure, the dominant factor would be the switching losses and you would have a big burning problem.

Originally posted by bigbigblue@Mar 20 2006, 04:11 PM
Is this due to stray capacitance of the trace / wire? If so, I have made this mistake - the gate driver is on a breadboard at the moment with trailing wires or around 6 inches! How close should the driver be to the gate to avoid the problem?
[post=15208]Quoted post[/post]​
And the inductance. It is surprising that what works at 100kHz doesn't really work at 400kHz. The first time I went over 200kHz I blew many transistors until I realised that putting the control circuit and the power switches on a different board and flying wires are recipes for disaster. I don't really know how close, but after that experience I always put them next to each other (~ 1cm) with enough decoupling capacitor to prevent Vcc sag and ground bounce near the gate drive output stage. It might work with 1" or 2", I've never tried it.

Do you use a breadboard? The one that have rows and rows of holes where you can plug the components into without soldering and use jumper wires? Chuck that out. It doesn't work for high frequency high current prototyping. Use veroboard, or whatever it is called there. You need to solder everything securely and as close as possible. I know it is a pain when you have to replace/change things around, but that is the only way that works.

Originally posted by bigbigblue@Mar 20 2006, 04:11 PM
Yes it is the Ciss
[post=15208]Quoted post[/post]​
Just ignore the ciss, use gate charge.

<!--QuoteBegin-bigbigblue
@Mar 20 2006, 04:11 PM
The duty cycle is actually near to 70% when the switching is happenning at 420KHz, but yes you are quire right for low duty cycles I had not taken account of this.
[post=15208]Quoted post[/post]​
[/quote]

By 70%, do you mean the MOSFET on/off ratio? Because the gate drive current on/off ratio should be much much lower than this. Anyway, you got the idea.

Does the IR2304 work at 400kHz?
 

Thread Starter

bigbigblue

Joined Mar 15, 2006
42
Originally posted by n9352527+Mar 20 2006, 06:00 PM--><div class='quotetop'>QUOTE(n9352527 @ Mar 20 2006, 06:00 PM)</div><div class='quotemain'>If that is the case, I usually call this kind of control scheme bang-bang switch mode controller.
[/b]

Yep - BANG BANG ! I had considered enforcing a minimum off time (to limit the switching frequency) but I'm not sure how to do it.

Originally posted by n9352527@Mar 20 2006, 06:00 PM
Have you compared the falling/rising times at 420kHz and at much lower frequency, say at 25kHz?
They look about the same - the time to get to saturation is (from memory) about 500ns in both cases, but when it is switching at 420KhZ, with a 50% duty that rise time is about 40% of the total on time - which is too high. When it is 20KHz it is about 2% of the time - which is OK.

Originally posted by n9352527@Mar 20 2006, 06:00 PM
There are two problems in trying to switch at high frequency efficiently. The first is you have to be able to saturate the MOSFET repetitively. For example, if you have a bootstrapped gate drive supply, is the charge reservoir large enough for high repetition gate charging without dropping below the recommended saturated gate voltage (~10V)? Does it have time to recharge the reservoir back up?

And the inductance. It is surprising that what works at 100kHz doesn't really work at 400kHz. The first time I went over 200kHz I blew many transistors until I realised that putting the control circuit and the power switches on a different board and flying wires are recipes for disaster. I don't really know how close, but after that experience I always put them next to each other (~ 1cm) with enough decoupling capacitor to prevent Vcc sag and ground bounce near the gate drive output stage. It might work with 1" or 2", I've never tried it.
OK - I'll have a go at getting it closer.

Originally posted by n9352527@Mar 20 2006, 06:00 PM
Do you use a breadboard? The one that have rows and rows of holes where you can plug the components into without soldering and use jumper wires? Chuck that out. It doesn't work for high frequency high current prototyping. Use veroboard, or whatever it is called there. You need to solder everything securely and as close as possible. I know it is a pain when you have to replace/change things around, but that is the only way that works.
Most of my circuit is on a home made PCB (using a pen plotter to draw it, then etching), however when part of the circuit doesn't work, I have been prototyping on breadboard - creating a PCB for each iteration of my testing would be prohibitively expensive and time consuming. I have some Veroboard - time to wield the soldering iron !

Originally posted by n9352527@Mar 20 2006, 06:00 PM
Just ignore the ciss, use gate charge.
By 70%, do you mean the MOSFET on/off ratio? Because the gate drive current on/off ratio should be much much lower than this. Anyway, you got the idea.
Could you post a schematic of a transistor gate drive? I was assuming that it was simply a resistor and transistor in series between the gate supply and 0v with the junction of the transistor and resistor tapped off to the MOSFET gate, with the transistor switching on and off, however if this is the case, the resistor will be passing current 100% of the time (either through the Gate of the MOSFET or through the switching transistor). However as you mention this is not the case, I must have misunderstood how the transistor based gate switching works.


<!--QuoteBegin-n9352527
@Mar 20 2006, 06:00 PM
Does the IR2304 work at 400kHz?
[/quote]
I don't think so :-(
 

Thread Starter

bigbigblue

Joined Mar 15, 2006
42
Originally posted by bigbigblue@Mar 20 2006, 06:54 PM
Could you post a schematic of a transistor gate drive?
I think I may be able to answer my own question ! Is the transistor gate drive based on a totem-pole mechanism?
 

n9352527

Joined Oct 14, 2005
1,198
Originally posted by bigbigblue@Mar 20 2006, 10:47 PM
I think I may be able to answer my own question ! Is the transistor gate drive based on a totem-pole mechanism?
[post=15232]Quoted post[/post]​
Yes, for high current gate driver totem-pole is a must. I only use pull-up resistor and sinking transistor for low current drive.
 

Thread Starter

bigbigblue

Joined Mar 15, 2006
42
Originally posted by n9352527@Mar 21 2006, 10:18 AM
Yes, for high current gate driver totem-pole is a must. I only use pull-up resistor and sinking transistor for low current drive.
[post=15249]Quoted post[/post]​
Ok thanks - time for me to go "carve" a totem pole :) I'll let you know how I get on.

Presumably the transistors in the totem pole need to be fairly beefy to witstand the initial charge and discharge currents - i.e the supply voltage / gate resistor, where the gate resistor is from the gate to the junction of the 2 transistors in the totem pole. Is this correct?
 

Thread Starter

bigbigblue

Joined Mar 15, 2006
42
I spent some time last night thinking about whether the microstepping bipolar drive I was designing was practical given the problems of the high side drive.

I have decided it isn't.

Therefore I am going to re-design as a full step only, unipolar drive. Luckily all the motors I have are hybrid types and can therefore be used bipolar or unipolar. I see the advantages and disadvantages of the unipolar drive as follows :

Disadvantages
1. Lower Torque compared to a parallel connected bipolar stepper

Advantages
1. Simplified bridge requiring 2x MOSFETS rather than the 4x MOSFETS of a bipolar H-Bridge
2. No high side MOSFET driving required
3. Due to 2 above there is no requirement for beefy transistors to drive the high side MOSFETS, therefore again the component count (and therefore cost) is lower
4. Simplified driving logic
5. No worries about shoot through currents
6. No separate higher voltage) gate drive supply required.
7. Due to 6 above, the motor supply voltage is not limited by the high side MOSFET driver supply voltage (just by the breajkdown voltage of the driving Mosfets)
 

n9352527

Joined Oct 14, 2005
1,198
Something like 2n2222 should be suitable for your purpose. I've got away with 2n3904 with very short pulsed 500mA current before, but it is not recommended.

I could not say which one is better for your application. Generally I prefer simpler unipolar motor, but in some applications it is cheaper to go bipolar rather than upgrading the motor to a bigger one. I guess it depends on your requirements.
 

Thread Starter

bigbigblue

Joined Mar 15, 2006
42
Originally posted by n9352527@Mar 22 2006, 12:45 PM
I could not say which one is better for your application. Generally I prefer simpler unipolar motor, but in some applications it is cheaper to go bipolar rather than upgrading the motor to a bigger one. I guess it depends on your requirements.
[post=15286]Quoted post[/post]​
I think Unipolar will be OK, as will full step only. I have stepper motors ranging from 180 oz in to 300 oz in to drive my machine - which can be seen at www.petertalbot.co.uk/cncroutermain.htm

Thanks once again to yourself and papbravo for your help.
 
i'm designing a motor controller for a 48 V dc motor . the motor is to be controlled using PWM and IRF2907ZP mosfets. We used IR 2110 IC to drive our mosfets. but, it burnt 2110 burnt. can anyone please tell me why this is burninig.
and if possible, please send a schematic diagram of a motor controller unit used for a similar application.

Thank you.
 
Top