Paralelling 2 BJTs NPN-PNP totem pole

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

I try to lower the junction temp of a totem pole containing npn/pnp bjts. (see picture)

The unit zxgd3004E6 can give peak 8A and when working at a frequency of 500khz driving a mosfet the average current can reach 0.6A.

I need x3 more than it can give. I looked at some alternatives such as ZXTN19060 Peak 12A avearge 1.1A

But it seems that this combination has also high junction temperature.

The alternatives for low inductance cases such as SOT23 have not much more capability. I also want to avoid bigger packages because of higher inductance of bigger packages.

I had the idea if I can parallel 2 of these totem poles to achieve higher peak and avrage current rating.

My concern is timing. If one totem pole conducts a little bit late (because of temp difference. etc) . It causes a shoot through of the totem pole.

I think that the BJTs are not much dependant on temp but anyhow I am not sure?

Can I paralel them?

Attachment(s): 

________________________________ We dream of a world where current does not need the voltage to flow.

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

How about a 'speedup cap' across each Rg? Makes the waveform look like this |\_. Charges the gate up faster. And how about using 2 diodes to 'bias' the driver bases like an audio amp? Gives you .7+.7 of dead band at the zero crossing to allow turnoff.

Imagecraft compiler user

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

I did not show all the elements:

GND is actually not GND but -3.3V
There is also a active miller clamp just very close to the Mosfet gate

A capacitor paralel to Rg will slow down turn off. This can be a problem. Because I must discharge 2 caps.

________________________________ We dream of a world where current does not need the voltage to flow.

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

You cannot connect two BJTs in parallel without killing the stronger one. That requires some kind of an active driver or a special package that thermally couples both dies.

No RSTDISBL, no fun!

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

That means I should find bigger npn and bigger pnp bjts in one package.
Any recommendation for low inductance seperate package npn pnp bjts 60V 20A

________________________________ We dream of a world where current does not need the voltage to flow.

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

Why would you not use mosfets?

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

Mosfet totem pole has some drawback against bipolar totem pole. Shoot through is significant when their common gate is in transition. You need extra timing circutary etc. Mosfet channel resistance is more temperature dependant and they are much more expensive for serial production.

________________________________ We dream of a world where current does not need the voltage to flow.

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

Don't run a common gate then! Use a mosfet driver that has the circuitry to skew the high and low side drive.

Are you really worried about the cost of some jelly bean mosfets vs the cost of the main mosfet?
Semikron have a range of gate drivers for their large devices.

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

I have a gate driver , this stage is to amplify . It is a kind of current buffer.
Note that gate drivers have limited power dissipation.

A gate driver, the biggest one available as IC has 9A peak current but only 1-2W heat dissipation. If your frequency is up to 200khz you may be happy with it.

But if you want to drive a mosfet at 1.2Mhz your peak current is mayba also in range of 10-15A but the average power is more than 15-16W. Because your average current raises to 1-2A instead of 0.1-0.2A

The high power gate drivers from semikron uses also bipolar totem pole as current buffer.

________________________________ We dream of a world where current does not need the voltage to flow.

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

Note: when I mean shoot through I dont mean the shoot through of the main mosfet bridge. I mean the shoot through of the isolated supply voltage of the gate driver.As it is an amplification stage I dont want to use timing or logic circutary to prevent shootthrough.

The timing and logic in the gate driver is to prevent the control signals of the seperate mosfets to prevent bridge shoot through of the main power bridge.

________________________________ We dream of a world where current does not need the voltage to flow.

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

I think you missed my point. The gate driver drives the mosfet totem pole to drive your main fet.

I have an inverter. The output is a h bridge of igbts. The bridge dead time is determined by the pwm outputs of the microcontroller. The pwm outputs drive optos which drive a mosfet totem pole which drives a igbt. There is shaping circuitry to tailor the on/off drive to the igbt. The shaping circuitry ensures the igbt turns off faster than it turns on

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

Does your configuration have a totem pole with discrete mosfets or a complete package such as ACPL332J.

You need 1 totem pole for turn on resistor and 1 totem pole for turn off resistor.
Does your configuration have 1 totem pole for both.

If so how can you overcome the shoot through of totem pole mosfets. The concrete Ics opto+totem pole have 1 totem output. So Ron must be equal to Roff. (no seperate values possible)

optos can not be used for high frequency, they have high capacitance and coupling. they also have very high delay.I use Silicon on insulator transformers. But anyhow the mosfet totem pole is not suitable for high power gate drive (I>4A and Irms>1A)

________________________________ We dream of a world where current does not need the voltage to flow.

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

Incal, i am failing to understand what you are saying.

I suggest you do some research regarding optos. Life doesnt stop at a 4n35.
Not that i suggest using or not using optos - they are not a magic bullet.

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

Best one has 400ns delay, and aging with time. capacitive coupling is always there, shielding is also limited.
Why use optos I used HCPL4502 and 4503 over 10 years.
I swicthed to SOI technology adum4223 or si8234. far far better . Life does goes on.

________________________________ We dream of a world where current does not need the voltage to flow.

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

Just a old trick I don't know if you can use it.
If you put more transistors in parallel some will be faster and there for load more.
One way without resistors that heat up, is to deliberately turn one on a tad before the rest, and next time do it to the next transistor, that way the slowest transistor take a bigger load.

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

What they do in audio amps is put a .22 ohm emitter resistor in each output transistor to help them share current. How did you poopoo the speedup cap without any discussion? Leo Fender used this on the BrightSwitch on the Fender Bandmaster in 1955 and it worked like a champ.

Imagecraft compiler user

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

Quote:
Leo Fender used this on the BrightSwitch on the Fender Bandmaster in 1955 and it worked like a champ.

That's some interesting info. Not to be presumptuous, but how do you happen to know that?

hj

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

bobgardner: I thought about the paralell capacitor idea. But in general as it is a very high frequency gate drive (x20 more than an audio amp) (I assume that audio amps are in range of 48Khz) . Any capacitor on the gate path detroys the balance, speed and the most important the turn off speed.
The main problem, the hardest to solve is the turn off part. I cam optimize the turn on part with many instruments. But a fast oand ossi-free reliable turn off at frequencies above 500Khz is really hard and your suggestion to add a paralel cap to Rg makes it harder.

Lets think abaout the turned off stage . We assume that the device is turned off with your caps//rg . Now you have a paralell capacitance to the mosfet gate capacitance. Knowing that your cap is fully charged // to the charged mosfet capacitor. we will have hard time to discharge both in 50-100ns. But i must think the alternative that if I have seperate ron and roff . How will it fit if I paralel a speedup-cap for the Ron, but not the Roff.

For this case capspeedup//ron you have a capacitive divider cspeedup and cmosfet. It meens mosfet sees a voltage of cspeed/(cspeed+cmosfet) for the first moments of the applied voltage. That can also controversly decrease the gate speed. (if we talk abaout very high speed gate drives). In consequence your speedup range is only available for low-mid speed gate drives. above 500khz your speed-up cap behaves as a slow down cap. So imho it depends on the speed range at which working point you are driving

________________________________ We dream of a world where current does not need the voltage to flow.

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

Its a differentiator. When the gate drive goes up, the waveform is |\. When the gate drive goes down, the waveform is |/. Calc the TC so 5 TCs are the width of the halfcycle.

Fender amps all came with the schematic pasted inside the speaker just like TVs and radios had. So you could check the tubes at the drugstore. There are collections of guitar amp schematics, and they are ALL derivatives of the Fender topology in tone circuit and poweramp. Marshall, Hiwatt, Boogie all look the same with minor tone cap variances.

Imagecraft compiler user