Author Topic: Gate Driver Question  (Read 776 times)

0 Members and 1 Guest are viewing this topic.

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Gate Driver Question
« on: April 14, 2021, 03:32:58 am »
I am trying to design an RF amplifier using the UCC27525 gate driver (pdf - https://www.ti.com/lit/ds/symlink/ucc27524.pdf).  This is a dual driver that has one inverting and one non-inverting gate.  I'm having trouble and have smoked a few of my transistors, and I got the suggestion to look at the output waveform of the driver without the transistors.  Could someone sanity check me that I *should* be seeing output on both outputs if both are enabled (floating in this case)?  And that output should be 180 degrees out of phase if I am driving both inputs with the same signal?

I think I broke my drivers in testing, and one of them gives me no output (presumably blown) but the other is only giving me output on "A" (the inverting output).  I only had the two on hand, and I'm assuming that second one is blown as well, but just wanted to verify.

Also, I am assuming this is ok to try as I didn't see anything in the datasheet.  (I've seen drivers that specify that they are not happy if you try to run them without driving anything.)

Thanks!
 

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Re: Gate Driver Question
« Reply #1 on: April 14, 2021, 04:51:39 am »
Attached the schematic below.  As for the layout, I'm building this on a protoboard so don't have one per se.
 

Offline perieanuo

  • Frequent Contributor
  • **
  • Posts: 558
  • Country: fr
Re: Gate Driver Question
« Reply #2 on: April 14, 2021, 06:56:39 am »
RF amplifier? analog RF amplifier or fast pulses?
I think the approach is not ok from the start, UCC2752x  states

"The UCC2752x family of devices are dual-channel, high-speed, low-side gate-driver devices capable of effectively driving MOSFET and IGBT power switches"

SWITCHES
 commutation :) not generic rf
 

Online james_s

  • Super Contributor
  • ***
  • Posts: 14481
  • Country: us
Re: Gate Driver Question
« Reply #3 on: April 14, 2021, 07:05:03 am »
An RF amplifier on a protoboard? Yeah, good luck with that. Even a switchmode regulator running at a few kHz is a challenge with that approach.

Deadbug/Manhattan it on a piece of blank copperclad, it probably won't win any awards for style but it's hard to beat the performance with anything less than a proper multilayer PCB.
 

Offline CaptDon

  • Frequent Contributor
  • **
  • Posts: 361
  • Country: is
Re: Gate Driver Question
« Reply #4 on: April 14, 2021, 02:08:08 pm »
Next question, RF Amplifier?? What frequency?? I doubt you can drive above 250khz with your circuit. That is barely R.F.
Senior Master Captain 140 passenger boat.
Private Pilot S.E.L. / CDL Class A / Motorcoach
Locomotive Electronics Engineer.
 

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Re: Gate Driver Question
« Reply #5 on: April 14, 2021, 03:29:05 pm »
Hi all,

So, this is for a class D amplifier (for use with FT8, which is constant amplitude).  So, not a linear amplifier.  Going for 7MHz (40M band).  In *simulation* it is happy, and the rise/fall/delay times of the driver/transistors look like they should work.

I usually do manhattan/deadbug.  I needed to use some SMT adapters here, and am more or less doing point to point between components, so thought I could get away with it, but will go back to manhattan (or break down and finally design a pcb...)

Blueskull - when you say "Your driving FETs do not seem to be clamped" - clamped how?  Like a zener or something to prevent Vds getting over a certain voltage?

But back to my original question - if I have just the driver chip with nothing attached to the output, I have the pins enabled, and I am feeding the same square wave into both inputs, what would you expect to see on the output pins.

Thanks for the help!
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 1979
  • Country: us
Re: Gate Driver Question
« Reply #6 on: April 14, 2021, 04:30:33 pm »
OK, at 7 MHz, given a 50% duty cycle digital input to the driver, I suspect it will burn up eventually, with no load.  With the load of a power transistor gate on each output, I would expect it to burn up in a second or two.  Figure out the energy lost in the driver's output transistor every time it charges the
power transistor gate, and then discharges it.  Then, multiply that by 7 MILLION!  That's the energy loss in HALF of the driver per second.
Double it for driving both transistors.

Then, add in the Miller effect, which requires additional gate charge, and it gets worse.

These drivers are designed for up to a few 100 KHz or so switching supplies.  Upping the frequency by a factor of SEVENTY will increase heating by the same factor.

To sum up, NO WAY is this going to work.  Note that the continuous rating is 0.3A, and the puil-up transistor has an on resistance of 5.5 Ohms when hot.  The pull-down transistor is ten times better.  So, the pull-up is the killer.  You don't give the type or gate charge of your power transistors.  But, given that, page 24 of the data sheet tells how to calculate power dissipation.

They do give a sample calculation for 60 nC of gate charge at 12 V and 300 KHz, they get .4 W.  But, you want to run at 23 X that frequency.
And, your transistors might even have more gate charge.

Jon
 

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Re: Gate Driver Question
« Reply #7 on: April 14, 2021, 05:10:10 pm »
To sum up, NO WAY is this going to work.

Hm.  That is unfortunate.
Do you know of any gate drivers that would?  Or suggestions around what discrete circuit might be able to do this?

Thanks for heading me off at the pass :)
 

Offline DL2XY

  • Regular Contributor
  • *
  • Posts: 62
  • Country: de
Re: Gate Driver Question
« Reply #8 on: April 14, 2021, 06:12:05 pm »
To sum up, NO WAY is this going to work.

Hm.  That is unfortunate.
Do you know of any gate drivers that would?  Or suggestions around what discrete circuit might be able to do this?

Thanks for heading me off at the pass :)

Have a look at IXD630.
 

Online james_s

  • Super Contributor
  • ***
  • Posts: 14481
  • Country: us
Re: Gate Driver Question
« Reply #9 on: April 14, 2021, 06:27:20 pm »
I usually do manhattan/deadbug.  I needed to use some SMT adapters here, and am more or less doing point to point between components, so thought I could get away with it, but will go back to manhattan (or break down and finally design a pcb...)

Forget adapters, you can work with SMT parts directly. I put down a small piece of kapton tape under the pins I don't want grounded and then solder the ground pins of parts like SOIC ICs to the copperclad. Then I use wire-wrap wire to connect the other pins.
 

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Re: Gate Driver Question
« Reply #10 on: April 14, 2021, 07:55:02 pm »
Have a look at IXD630.

Whew.  Any that are less than $8 a pop?  I haven't dived into the datasheet, but feels big for the transistors I am driving.

But thanks for the pointer!
 

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Re: Gate Driver Question
« Reply #11 on: April 14, 2021, 07:56:07 pm »
Forget adapters, you can work with SMT parts directly. I put down a small piece of kapton tape under the pins I don't want grounded and then solder the ground pins of parts like SOIC ICs to the copperclad. Then I use wire-wrap wire to connect the other pins.

You seem to have better eyes/steadier hands than I do :)
But will give it a shot at some point.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 16973
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Gate Driver Question
« Reply #12 on: April 14, 2021, 09:24:15 pm »
Class D ceases to be very meaningful up at those frequencies.  Much of the time is spent slewing the gate and drain voltages, so that little time can be said to be spent purely in a saturated or cutoff condition.

Gate driver ICs have been seen, from time to time; IXYS RF used to have some beefy parts, with wide flat leads capable of actually driving that much power.  They were even more expensive, and poorly stocked if at all.  I haven't checked since they were acquired; I assume they're long since discontinued.

The standard RF approach is to just blast it with something rough, probably not very sinusoidal nor square, using another amplifier, which itself may be tuned or baseband, single-ended or push-pull (and at that, push-pull or totem-pole).  A coupling network may be necessary, say to transform a 30V output swing to the ~5V gate swing needed.

Several watts of gate drive is not unreasonable to push around power switching transistors (would have to check the datasheet to see what the indicated devices would require), and with the generally lower efficiency of RF amps (compared to low frequency switching) and the relatively low power dissipation of switching devices (compared to total rated switching area), the gain may not be all that amazing.

These are the advantages of RF transistors: while they are more expensive and have small switching area (making them perhaps less efficient or powerful for class D/E applications), the drive power, and AC considerations like reduced parasitics (wide low-Z terminals, low capacitances, and especially low feedback capacitance), are optimized for exactly this application, of course.

You might even end up saving money against super speed gate drivers, or coupling transformers!

But 7MHz is still low enough that a discrete solution will do.  Hams have been using ancient types like IRF540 for decades, at least in linears.  Basically, count on having multiple stages between signal source (which is what, a logic-level clock or something?) and final.


Note that your circuit is probably okay below resonance, where it will behave as a ZVS class E amplifier (but mind the body diode recovery).  Above resonance, it will probably blow up, because the leading phase angle of the (capacitive) load will leave some voltage on the drain, by the time of the next gate rising edge.  So turn-on sees a momentary short circuit, dumping current from one transistor, through the resonant capacitor, into the opposing body diode.

As far as "mind[ing] body diode recovery", you need t_rr << t_cycle.  This should be easy enough in low voltage types, but won't pass for, say, >200V MOSFETs probably.  (Body diode recovery performance tracks inversely with rated voltage.)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 1979
  • Country: us
Re: Gate Driver Question
« Reply #13 on: April 14, 2021, 10:47:13 pm »

Have a look at IXD630.
Ah, yes, the output resistance on the high side looks MUCH better on this one, about 10 X lower.  That will definitely help.
They have charts of supply current vs freqency and gate capacitance, and they go up to 1 MHz.  Extrapolating to 7 MHz doesn't look so
great, however.  Having only one gate driver per package helps, too.  The propagation delay is pretty bad.

Your circuit can't be driven with a single gate drive, you need to provide dead-time delays to be sure one transistor has shut off before
the other is turned on.

Jon
 

Offline JohnG

  • Frequent Contributor
  • **
  • Posts: 354
  • Country: us
Re: Gate Driver Question
« Reply #14 on: April 14, 2021, 10:54:29 pm »
If you can live with a 5V gate drive, you can use two LMG1025s. I have experience with the LMG1020, which is the BGA version of the same thing. I've run the LMG1020 over 100 MHz with a similar load to your FET of choice, though it was running pretty hot. It should have no difficulty handling 7 MHz.

As mentioned by others, layout is key. While 7 MHz is not that high, the above gate drivers can have transition times well below 1 ns. It's these transitions that will make or break the design if not understood. The effective frequency content for these edges is well over 300 MHz. A 1 cm wire can have an impedance on the order of j20 ohms, which will not be negligible. Sounds like you are somewhat familiar with RF, but I have found many RF engineers really think mostly in the frequency domain, and if you want to avoid overshoot, the time domains is a little more informative.

I didn't take a good look at the rest of your circuit very thoroughly, but if you were willing to spin a PCB, you could use a couple EPC2037 FETs. Similar RDSon, and you could probably parallel two or three Tinylogic UHS buffers (7NZ series dual or triple buffer/inverters) to drive them. No reverse recovery, either. But, it would be in the running for "understatement of the year" to say that they would be difficult to prototype dead-bug style  :o.

Cheers,
John
« Last Edit: April 14, 2021, 10:56:34 pm by JohnG »
"Those who learn the lessons of history are doomed to know when they are repeating the mistakes of the past." Putt's Law of History
 

Offline langwadt

  • Super Contributor
  • ***
  • Posts: 2398
  • Country: dk
Re: Gate Driver Question
« Reply #15 on: April 14, 2021, 11:40:27 pm »
how much power? can't be much with those tiny transistors

how about an ADSL driver chip?
 

Offline JohnG

  • Frequent Contributor
  • **
  • Posts: 354
  • Country: us
Re: Gate Driver Question
« Reply #16 on: April 15, 2021, 12:33:10 pm »
To sum up, NO WAY is this going to work.  Note that the continuous rating is 0.3A, and the puil-up transistor has an on resistance of 5.5 Ohms when hot.  The pull-down transistor is ten times better.  So, the pull-up is the killer.  You don't give the type or gate charge of your power transistors.  But, given that, page 24 of the data sheet tells how to calculate power dissipation.

They do give a sample calculation for 60 nC of gate charge at 12 V and 300 KHz, they get .4 W.  But, you want to run at 23 X that frequency.
And, your transistors might even have more gate charge.

Or, they might have less. If you are using IRLML2060s as FETs (that's what I could see from the schematic), I bothered to look up the gate charge: 0.67 nC, or 90x smaller than the 60 nC used in the example. Also, gate driver resistance controls transition time, not loss. The latter is a function of the gate charge and the inherent output capacitance of the gate driver with no load, plus some internal gate driver losses. Some gate drivers also have a little bit of shoot-through internally, which creates frequency-dependent losses as well.

0.67 nC is very small, and the limiting factor may be the gate driver's own internal losses. If you dig into the UCC2752x  datasheet, they claim to have negligible shoot-through. Also, the 5 ohms is after the transition, but they use a bootstrapped N-channel pull-up for higher speed than the resistance spec would indicate. The 13 ns prop delay is very good as well.

Unless there is some other gotcha, I would say there is a decent chance of success, as long as circuit parasitics don't kill you.

Cheers,
John
"Those who learn the lessons of history are doomed to know when they are repeating the mistakes of the past." Putt's Law of History
 

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Re: Gate Driver Question
« Reply #17 on: April 15, 2021, 02:34:52 pm »
Or, they might have less. If you are using IRLML2060s as FETs (that's what I could see from the schematic), I bothered to look up the gate charge: 0.67 nC, or 90x smaller than the 60 nC used in the example. Also, gate driver resistance controls transition time, not loss. The latter is a function of the gate charge and the inherent output capacitance of the gate driver with no load, plus some internal gate driver losses. Some gate drivers also have a little bit of shoot-through internally, which creates frequency-dependent losses as well.

0.67 nC is very small, and the limiting factor may be the gate driver's own internal losses. If you dig into the UCC2752x  datasheet, they claim to have negligible shoot-through. Also, the 5 ohms is after the transition, but they use a bootstrapped N-channel pull-up for higher speed than the resistance spec would indicate. The 13 ns prop delay is very good as well.

Unless there is some other gotcha, I would say there is a decent chance of success, as long as circuit parasitics don't kill you.

Yeah, I chose these parts because they seemed like they might work based on my somewhat limited understanding.  And like I say, they work in simulation at least :)

Do you have a suggestion of a next step here?  I could order more parts and try some better construction next time, I could try laying this out on PCB and ordering some boards, or I could look for different parts (or something else entirely).

My eventual goal is a 5-10W Class D at 7MHz.  I know I could go some other route (other than class D), but I'm just learning so much  :)

Thank you all for your continuing feedback!
 

Offline CaptDon

  • Frequent Contributor
  • **
  • Posts: 361
  • Country: is
Re: Gate Driver Question
« Reply #18 on: April 15, 2021, 02:48:02 pm »
If you intend to use it as an RF amplifier and put it on the air you better have some filtering or resonant tank circuits!!! If you are switching 7mhz square wave into an antenna circuit then you will have strong harmonics every 7 mhz up to about 100mhz and above. How are you getting dead time between when one FET turns on and the other turns off. All in all your idea seems horrible in so many ways. If you want to put an RF amplifier on the air it needs to amplify a clean sinewave and produce a clean sinewave with minimal distortion. Driving FETs with hard switching in class D will require a lot of filtering and harmonic suppression!!!
Senior Master Captain 140 passenger boat.
Private Pilot S.E.L. / CDL Class A / Motorcoach
Locomotive Electronics Engineer.
 

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Re: Gate Driver Question
« Reply #19 on: April 15, 2021, 03:05:15 pm »
If you intend to use it as an RF amplifier and put it on the air you better have some filtering or resonant tank circuits!!! If you are switching 7mhz square wave into an antenna circuit then you will have strong harmonics every 7 mhz up to about 100mhz and above. How are you getting dead time between when one FET turns on and the other turns off. All in all your idea seems horrible in so many ways. If you want to put an RF amplifier on the air it needs to amplify a clean sinewave and produce a clean sinewave with minimal distortion. Driving FETs with hard switching in class D will require a lot of filtering and harmonic suppression!!!

I mean, switching amplifiers are not uncommon in RF.  Of course you need output filtering. 
 

Offline CaptDon

  • Frequent Contributor
  • **
  • Posts: 361
  • Country: is
Re: Gate Driver Question
« Reply #20 on: April 15, 2021, 08:46:13 pm »
I was a warranty station for the Peavey Deca series of audio amplifiers. In that case we switched at 500khz to obtain a clean audio bandwidth 20hz to 20khz flat as a pancake. Truth is the amp could have been D.C. coupled and used as a power supply but to prevent speaker damage the low end was rolled off. The 500khz was PWM modulated at a steady carrier frequency with some R.F. filtering at the speaker output. I also worked on the Gates/Harris MW-5 5kw AM commercial broadcast transmitters. Their scheme was a bit reverse, They started at around 20kv with the modulator tube at the top. The modulator was hard driven from full off to saturation at 90khz. It was PWM modulated and had a 90khz filter between it and the class C R.F. amplifier stage which was of conventional design. So the modulator tube basically provided a varying plate voltage to the R.F. amp and that varying plate voltage was the 'audio modulation' signal. The station was 910khz. Two very different schemes, one had the modulation or PWM carrier at 25 times the highest modulation frequency, and the other scheme used PWM at 1/10 of the carrier frequency although the recovered and filtered component of the PWM was at most  1/100th of the RF carrier. All fun stuff!!! Best wishes for your project. Local hams are getting W.A.S. in just a month or two with the digital mode you speak of. I haven't messed with it yet. Way to busy fixing broken locomotives and boating season is here.
Senior Master Captain 140 passenger boat.
Private Pilot S.E.L. / CDL Class A / Motorcoach
Locomotive Electronics Engineer.
 

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Re: Gate Driver Question
« Reply #21 on: April 15, 2021, 10:09:47 pm »
Best wishes for your project. Local hams are getting W.A.S. in just a month or two with the digital mode you speak of. I haven't messed with it yet. Way to busy fixing broken locomotives and boating season is here.

Thanks!
Yeah, I like the digital modes because it lets me experiment with stuff and use very low power.  I have a little IRF510 amp that does a couple of watts and that can get me all over the country with a low dipole.

Fixing broken locomotives!  Sounds great!

I'm too busy with work/kids, so I only get to work on this stuff in tiny bits and spurts, so I often forget a bunch of things in between...
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 16973
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Gate Driver Question
« Reply #22 on: April 15, 2021, 10:22:04 pm »
Ah yes, a technique that works well with semiconductors for certain power applications as well -- sometimes a push-pull or H-bridge current-sourcing inverter is preferred, and that constant current supply is provided from a big fat choke from the input.  Which in turn has its current regulated by a buck converter.  To save on fatness, multiple inductors can be used, in series, with filter caps along the way -- you get a higher cutoff frequency with more attenuation.

Whereas for a voltage sourcing inverter, you want a low impedance (Zsupply < Zsw / 5 say, where Zsw is the switch peak voltage over peak current at nominal load), for a current sourcing inverter, you want a high impedance (Zsupply > Zsw * 5, say).  This defines the filter impedance so that it can be designed.

For the RF amp of course, the modulator bandwidth (including the filter) must exceed the audio bandwidth (which for AM is a mere ~5kHz); for general controls, it's whatever control bandwidth is needed, or is feasible.

The biggest I've seen was IGBT switched, either, uh, I forget actually which models did what, but I think there was a benchtop one around 20kW, and a floor standing one in the 300kW range (the latter being of a modular, scalable design, so that's just an example really).  Big industrial induction heating supplies.  Being big and fat, they merely opted for the big fat inductor, no higher-order filtering.

The biggest I'm aware of (not personally seen), they don't bother with PWM at all, they plonk down huge stacks of SCRs -- which must be driven with current, or quasi-resonant commutation -- these make up HVDC interlinks and the like.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline JohnG

  • Frequent Contributor
  • **
  • Posts: 354
  • Country: us
Re: Gate Driver Question
« Reply #23 on: April 15, 2021, 10:57:19 pm »
Some tips:

Build it on a ground plane.

Don't let any wires loop in the air, keep them as short as possible and flat along the ground plane as much as possible.

Gate drive leads must be short, and remember that the gate drive bypass cap is part of the FET turn-on loop. Keep the inductance low.

Consider gate drive resistors to damp the gate loop.

Take a stab at estimating parasitic inductances and add them to your sim.

Pound your head against a wall....

History alert: Here is a PWM envelope modulated transmitter from 1923 (almost 100 years ago): https://patents.google.com/patent/US1672215A/en.

Cheers,
John
"Those who learn the lessons of history are doomed to know when they are repeating the mistakes of the past." Putt's Law of History
 

Offline cadr

  • Contributor
  • Posts: 16
  • Country: us
Re: Gate Driver Question
« Reply #24 on: April 15, 2021, 11:23:24 pm »
Some tips:

Build it on a ground plane.

Don't let any wires loop in the air, keep them as short as possible and flat along the ground plane as much as possible.

Gate drive leads must be short, and remember that the gate drive bypass cap is part of the FET turn-on loop. Keep the inductance low.

Consider gate drive resistors to damp the gate loop.

Take a stab at estimating parasitic inductances and add them to your sim.

Thanks!  I will try building another on copper (mahattan/deadbug/etc)
I tried keeping everything short here, but will double down on that.
The design does have gate drive resistors.

Pound your head against a wall....

So far ahead of you there!  :)
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf