GaN is faster than Silicon ,
Even more efficient than GaAs used in Military AESA radar
Any good GaN transistor for Class A amplifier?
Even more efficient than GaAs used in Military AESA radar
Any good GaN transistor for Class A amplifier?
GaNFETs are mainly designed for switching, not linear duty. However the secondary breakdown performance looks good - for instance see this paper: https://epc-co.com/epc/Portals/0/epc/documents/product-training/SafeOperatingArea.pdf
Note that there are different manufacturers and the devil is in the details, different devices may be different.
However is it worth it? Class A doesn't require speed, and efficiency and "Class A" don't belong in the same sentence do they?
Surely the whole appeal of GaN is to be able to put a few 100W through a module the size of a postage stamp and not need a fan to cool it, ie better class D performance??
Note that there are different manufacturers and the devil is in the details, different devices may be different.
However is it worth it? Class A doesn't require speed, and efficiency and "Class A" don't belong in the same sentence do they?
Surely the whole appeal of GaN is to be able to put a few 100W through a module the size of a postage stamp and not need a fan to cool it, ie better class D performance??
That, as well as the fact there are no known P-types, the temperature dependance is quite high and for the above GaNsystems: you need a custom pcb because of the footprint and they're awfully sensitive to ESD, so much so, that hand soldering really is tricky.
Then again, being "fast" does mean they're even easier to drive, which makes the front end a walk in the park.
But in general, I agree with Mark it's not a walk in the park, for hardly any benefit.
Then again, being "fast" does mean they're even easier to drive, which makes the front end a walk in the park.
But in general, I agree with Mark it's not a walk in the park, for hardly any benefit.
There are also types out there that are a mix of regular silicon and the newer GaN.
Apart from these types, the Vgs for "genuine" GaN are way lower (around 6 volts) than regular fets. Iow, they're usually not interchangeable.
One of the good things is temperature stability.
Apart from these types, the Vgs for "genuine" GaN are way lower (around 6 volts) than regular fets. Iow, they're usually not interchangeable.
One of the good things is temperature stability.
If you review the spec sheets on GaN transistors you will notice that the safe operating area curves (SOA) do not spec operation for more than a few milliseconds. There is no DC curve offered. I discovered this the hard way when I bought a tube of Transphorm GaN fets and blew them up at 20 volts and 1 amp, which of course is 20 watts.
They make great switches for Class D, but at this time I would say Class A is right out, as is Class B and AB.
They make great switches for Class D, but at this time I would say Class A is right out, as is Class B and AB.
Member
Joined 2009
Paid Member
<CRAZY_AND_OR_STUPID_IDEA>
Whiling away the eternity, looking for in-production transistors that would exhibit triode-like behaviour, I've stumbled upon an enticing picture (Exhibit A, comes from this presentation) showing the 3rd quadrant / reverse conduction characteristics of a GaN transistor. Reading up on these a little, it appears that - unlike J/MOSFET, where the PN junction / body diode spoils the fun (I still can't make heads or tails of how BJT fit into all of this) - GaN will cheerfully conduct in the reverse direction in what I can only describe as triode-like manner.
So far so good...
Now, taking heed of Mr Pass's admonition that the GaN parts like to burst into flames as soon as look at you, I went thru their data sheets looking for ones w/ explicitly specified DC SOA, and found a couple (and a half) that fit the bill:
Besides DC operation, these guys even have positive VGS(th) tempco! Unfortunately, they come in minuscule PiP [Pain-in-Posterior] SMD packages and would most probably dissipate their tiny little hearts out before any meaningful power could be extracted from them.
But if one went the ZV9 / F3 way and cascoded a few of them paralleled*, mounted on an IMS adapter board... (* That kinda excludes the Infineon parts as they are much too pricey.)
A quickly cobbled up simulation gives a 1.5W avg dissipation per one lil' bugger in a quad, and you can either run the sim yourself or check the ~1W / 8R distortion profile attached.
The missing half is the potentially usable GPI65015TO from GaNPower in a much more convenient TO-220 package, but it's only a half "hit" because - while positive VGS(th) tempco is clearly shown in the data sheet - the SOA chart is absent altogether. (I asked GaNPower about that - they said they'd look into it; they already fixed an error in the DS that I reported.)
</CRAZY_AND_OR_STUPID_IDEA>
P.S. NexGen supposedly manufactures a vertical GaN JFET, NXG2EB120985, but there's only 1 reference to it here, and my enquiry for more details remains unanswered...
Whiling away the eternity, looking for in-production transistors that would exhibit triode-like behaviour, I've stumbled upon an enticing picture (Exhibit A, comes from this presentation) showing the 3rd quadrant / reverse conduction characteristics of a GaN transistor. Reading up on these a little, it appears that - unlike J/MOSFET, where the PN junction / body diode spoils the fun (I still can't make heads or tails of how BJT fit into all of this) - GaN will cheerfully conduct in the reverse direction in what I can only describe as triode-like manner.
So far so good...
Now, taking heed of Mr Pass's admonition that the GaN parts like to burst into flames as soon as look at you, I went thru their data sheets looking for ones w/ explicitly specified DC SOA, and found a couple (and a half) that fit the bill:
- GaN Systems' GS-065-004-1-L
- Infineon's IGLD60R190D1 - would you harken at those triode-y curves! 😀
- [the half comes in later]
Besides DC operation, these guys even have positive VGS(th) tempco! Unfortunately, they come in minuscule PiP [Pain-in-Posterior] SMD packages and would most probably dissipate their tiny little hearts out before any meaningful power could be extracted from them.
But if one went the ZV9 / F3 way and cascoded a few of them paralleled*, mounted on an IMS adapter board... (* That kinda excludes the Infineon parts as they are much too pricey.)
A quickly cobbled up simulation gives a 1.5W avg dissipation per one lil' bugger in a quad, and you can either run the sim yourself or check the ~1W / 8R distortion profile attached.
The missing half is the potentially usable GPI65015TO from GaNPower in a much more convenient TO-220 package, but it's only a half "hit" because - while positive VGS(th) tempco is clearly shown in the data sheet - the SOA chart is absent altogether. (I asked GaNPower about that - they said they'd look into it; they already fixed an error in the DS that I reported.)
</CRAZY_AND_OR_STUPID_IDEA>
P.S. NexGen supposedly manufactures a vertical GaN JFET, NXG2EB120985, but there's only 1 reference to it here, and my enquiry for more details remains unanswered...
Attachments
Ha, finally someone else found out;-)
I use the GanSystems. The drawback of it is that it only allows around 12 VSD and it's easier to implement as follower. I use it to drive really low impedance compression drivers (0,1 ohm etc) which it does straight from an opamp for the voltage amplfication.
The smd footprint is non-standard and they are extremely sensitive to static discharge (use zeners) etc. They do not sound like normal source followers, they are extremely transparent.
I use the GanSystems. The drawback of it is that it only allows around 12 VSD and it's easier to implement as follower. I use it to drive really low impedance compression drivers (0,1 ohm etc) which it does straight from an opamp for the voltage amplfication.
The smd footprint is non-standard and they are extremely sensitive to static discharge (use zeners) etc. They do not sound like normal source followers, they are extremely transparent.
Btw, i run them both symmetrical as well as in a circlotron configuration and at least at 12 (sym) / 5.5 (circl) volts between 1 to 2 amps. As long as you keep temps down, they are hard to break. I always use the 100 volts
https://gansystems.com/gan-transistors/gs61008t/
https://gansystems.com/gan-transistors/gs61008t/
mterbekke , what compressiondrivers u use that reach 0,1 ohm pretty much short circuit.. ?? u got a schematic of your circlotron with gan ?
It was years ago and purely prototyping, so many things tried: from just cutting the voice coil in 1 place and use alu solder at that same spot for lead outs, to using 1 winding of 3 mm wide 60 micron aluminium foil.
Those all on JBL 2450's, ranging from ribbed titanium to Radian Alu diafragmas.
For the circlotron any circlo schematic will do, as long as you remember to reverse the polarity of the amplifier and the fet and keep the psu voltage per rail below 12 volts (max VSD) divided by 2 (since the max VSD, when driven is 12 Volts when 1 fet completely conducts, the other off/or still in classA, right?!).
With polarity is meant: your reference is the positive polarity of the psu, because when you flip the fet, the Source gets connected to the positive side of the psu. This also means you'll have to flip the polarity of the drivers (pnp becomes npn & vice versa).
In essence you can consider it building an amp with a "PNP" fet, while it is a "NPN" as well as a source follower, input referenced from the Source, but in conduction when the gate voltage is close to the Drain, if that makes any sense.
A circlotron end stage will usually have a voltage amplification of 1, so you do not lose any gain when connected this way wrt a normal circlotron.
I usually just took 2 single supplies and hooked them up, class A by a simple potentiometer and input via a coupling capacitor. Output for speaker as usual between the sources, as well as for the common reference point for the potentiometers by 2 resistors for the usual input ground reference, slider of the pot to the gate and other side of the pot to the other side (minus) of the psu.
I found it hard to get right because of many options for control, feedback if needed as well as the limited power supply.
So the other option is to use a standard bridged amp schematic, 2 single power supplies of 12 to 13 volts and a current source for the negative side.
Think old Sansui single supply amps with 2 same polarity end transistors.
Maybe it is better to email me, this can get really long.
The best way to look at the fet in that quadrant is that, if voltage is aplied to it, the higher the power supply voltage, the higher the voltage between gate and source needs to be (closer to the drain voltage).
When you drive the gate to the source (of course you can to +5 volts, flipped polarity, normally with an N fet this would be similar to -5 VGS), the lower it gets, the more it conducts.
Kind of counter intuitive.
It's been a few years, I might still have some drawings somewhere at my work shop, but better email me if this long text doesn't work well for you. I also have some pcb's for the basic circlotron (no drivers, just input, output and place for 4 fets, 2 in parallel) but will only send them if you're really interested (i can start Altium and send the electronically if needed).
Cheers
Those all on JBL 2450's, ranging from ribbed titanium to Radian Alu diafragmas.
For the circlotron any circlo schematic will do, as long as you remember to reverse the polarity of the amplifier and the fet and keep the psu voltage per rail below 12 volts (max VSD) divided by 2 (since the max VSD, when driven is 12 Volts when 1 fet completely conducts, the other off/or still in classA, right?!).
With polarity is meant: your reference is the positive polarity of the psu, because when you flip the fet, the Source gets connected to the positive side of the psu. This also means you'll have to flip the polarity of the drivers (pnp becomes npn & vice versa).
In essence you can consider it building an amp with a "PNP" fet, while it is a "NPN" as well as a source follower, input referenced from the Source, but in conduction when the gate voltage is close to the Drain, if that makes any sense.
A circlotron end stage will usually have a voltage amplification of 1, so you do not lose any gain when connected this way wrt a normal circlotron.
I usually just took 2 single supplies and hooked them up, class A by a simple potentiometer and input via a coupling capacitor. Output for speaker as usual between the sources, as well as for the common reference point for the potentiometers by 2 resistors for the usual input ground reference, slider of the pot to the gate and other side of the pot to the other side (minus) of the psu.
I found it hard to get right because of many options for control, feedback if needed as well as the limited power supply.
So the other option is to use a standard bridged amp schematic, 2 single power supplies of 12 to 13 volts and a current source for the negative side.
Think old Sansui single supply amps with 2 same polarity end transistors.
Maybe it is better to email me, this can get really long.
The best way to look at the fet in that quadrant is that, if voltage is aplied to it, the higher the power supply voltage, the higher the voltage between gate and source needs to be (closer to the drain voltage).
When you drive the gate to the source (of course you can to +5 volts, flipped polarity, normally with an N fet this would be similar to -5 VGS), the lower it gets, the more it conducts.
Kind of counter intuitive.
It's been a few years, I might still have some drawings somewhere at my work shop, but better email me if this long text doesn't work well for you. I also have some pcb's for the basic circlotron (no drivers, just input, output and place for 4 fets, 2 in parallel) but will only send them if you're really interested (i can start Altium and send the electronically if needed).
Cheers
So, could this be the replacement for vintage vfet from 70s? I am not an expert but I read some documents showing they are internally normally on and vertical structured bla bla, besides some output characteristics looks like vintage vfet. I don't care efficiencies or switching speed, but i care the warm sound characteristics which I think defined HiFi from very beginning. If they are good or better, we could diy futuristic vintage class a vfet amp!
That's exactly what I am suggesting! 😀 But - as corroborated by @mterbekke - the usable VSD and power dissipation have to be worked around. That's why I proposed the cascoded / paralleled approach...
Those are 7mΩ RDS(on), starting at $12 a pop at Mouser. I may be wrong, but my understanding is that - for linear operation - higher RDS(on) is more desirable (safer). Hence, my choice of GS-065-004-1-L, with 450mΩ RDS(on), starting at just about $3 a pop. My plan is to parallel 4 per channel - with the $12 parts, I might as well get me some proper Tokin triodes... 😉
I think both are excellent in that quadrant.
The GS-065-004-1-L have bottom cooling and at 4 degrees per Watt are 8 times worse than the 0,5 degrees per watt of the GS1008T. You also have to cool them via the pcb on the bottom side, tough to do without a lot resistance added again.
Power dissipation of the top cooled part is so high that if you break them, you either overloaded the gate, or cooled them so bad that they loose the solder paste used and come off by themselves ;-)
That`s how safe the top cooled are, I can`t speak for the other types.
Honestly, I think the 650 volts types look great in reverse voltage mode, but I had trouble cooling the bottom cooled parts I used earlier (Panasonic), knew the characteristics change a lot with temperature (Gm goes down significantly), didn`t need the speed of less capacitance and also didn`t look forward to problems of paralleling them (oscillation, height/mounting/cooling differences) and more points of failures than just 1 fet. So it`s funny I considered them safer instead of less safe;-)
In normal operation, (the question about the VFet characteristics) the 650 volts part looks way less linear (IDS vs VDS) than the top cooled part, but i`d just test them on a bench to see that. I don`t know if the simulation model allows for linear operation and highly doubtful if the reverse quadrant is even more than just a very basic approach of the real world.
Also interesting to know:
The second image you showed (the Infineon IGLD60) looks more like the earlier reverse characteristics of the GanSystems (all of them). They changed that to the more confusing ones they have now, that look a bit strange (not linear), because the gate voltage in steps from +6 to 0 and then -3 is displayed. The GanSystems (i guess all of them) are not linear around that range from 0VSD to around 2VSD, you can see that illustrated in your first picture.
In other words: for low feedback systems, the linear range is less than the total allowable VSD. This is important for your operation point, PD etc.
By the way the Infineon`s you mentioned earlier look like the earlier Panasonics like the PGA26E07BA. They quit production some time ago, it seems Infineon bought it from them. They performed well subjectively, but were hard to cool because of the bottom side cooling needed.
These types seem to need a gate current when driven, the GANSystems do not. I didn`t measure the Panasonics, but verified the GanSystems: i could find no gate current untill just over -10 volts.
ST also have 1 type, that datasheet looks a lot like one of the GanSystems.
Lots of borrowing from each other, sadly not that many really original products, and a lot of other types are internal cascodes and will not work in reverse mode.
None of this type of GaN allow a higher gate voltage than -10 volts, and since the max VSD is around VSD + VGth ,the maximum VSD voltage is always around 10 + VGth = 12 volts.
Again, I don`t think you can break them if you keep them below e.g. 100 degrees centigrade (the pcb will break sooner than the fet) and within VGS and VSD spec. Pretty easy to do.
The GS-065-004-1-L have bottom cooling and at 4 degrees per Watt are 8 times worse than the 0,5 degrees per watt of the GS1008T. You also have to cool them via the pcb on the bottom side, tough to do without a lot resistance added again.
Power dissipation of the top cooled part is so high that if you break them, you either overloaded the gate, or cooled them so bad that they loose the solder paste used and come off by themselves ;-)
That`s how safe the top cooled are, I can`t speak for the other types.
Honestly, I think the 650 volts types look great in reverse voltage mode, but I had trouble cooling the bottom cooled parts I used earlier (Panasonic), knew the characteristics change a lot with temperature (Gm goes down significantly), didn`t need the speed of less capacitance and also didn`t look forward to problems of paralleling them (oscillation, height/mounting/cooling differences) and more points of failures than just 1 fet. So it`s funny I considered them safer instead of less safe;-)
In normal operation, (the question about the VFet characteristics) the 650 volts part looks way less linear (IDS vs VDS) than the top cooled part, but i`d just test them on a bench to see that. I don`t know if the simulation model allows for linear operation and highly doubtful if the reverse quadrant is even more than just a very basic approach of the real world.
Also interesting to know:
The second image you showed (the Infineon IGLD60) looks more like the earlier reverse characteristics of the GanSystems (all of them). They changed that to the more confusing ones they have now, that look a bit strange (not linear), because the gate voltage in steps from +6 to 0 and then -3 is displayed. The GanSystems (i guess all of them) are not linear around that range from 0VSD to around 2VSD, you can see that illustrated in your first picture.
In other words: for low feedback systems, the linear range is less than the total allowable VSD. This is important for your operation point, PD etc.
By the way the Infineon`s you mentioned earlier look like the earlier Panasonics like the PGA26E07BA. They quit production some time ago, it seems Infineon bought it from them. They performed well subjectively, but were hard to cool because of the bottom side cooling needed.
These types seem to need a gate current when driven, the GANSystems do not. I didn`t measure the Panasonics, but verified the GanSystems: i could find no gate current untill just over -10 volts.
ST also have 1 type, that datasheet looks a lot like one of the GanSystems.
Lots of borrowing from each other, sadly not that many really original products, and a lot of other types are internal cascodes and will not work in reverse mode.
None of this type of GaN allow a higher gate voltage than -10 volts, and since the max VSD is around VSD + VGth ,the maximum VSD voltage is always around 10 + VGth = 12 volts.
Again, I don`t think you can break them if you keep them below e.g. 100 degrees centigrade (the pcb will break sooner than the fet) and within VGS and VSD spec. Pretty easy to do.
Some good practical info there, @mterbekke - thanks!!
Yes, the RΘjc is much worse on the 650V parts, but I planned to have them mounted on an aluminum substrate PCB and just bolted to the heatsink, which should keep them plenty cool, I guess.
We shall see… 😉
Yes, the RΘjc is much worse on the 650V parts, but I planned to have them mounted on an aluminum substrate PCB and just bolted to the heatsink, which should keep them plenty cool, I guess.
We shall see… 😉
I hope some of you are familiar with works of Dr. Ray Ridley, power electronics specialist.
Once a year he posts something audio related on FB /groups/ridleyengineering/
I believe "GaN Class A amplifier" was posted year or two ago. Anyhow he made such a thing and it worked.
So to speak reference design does exist. Sometimes all you have to do is to ask.
Once a year he posts something audio related on FB /groups/ridleyengineering/
I believe "GaN Class A amplifier" was posted year or two ago. Anyhow he made such a thing and it worked.
So to speak reference design does exist. Sometimes all you have to do is to ask.
I have no idea who he is and I can't see the page.
A bit underwhelmed really, but I didn't have coffee yet so that might be it.
A bit underwhelmed really, but I didn't have coffee yet so that might be it.
F3 topology?
GaN system 100V parts do have high Yfs, low parasitic capacitances and negative tempco. Should also works great as follower in normal positive Vds, but then there are other options such as TI NexFET such as CSD18537NKCS with low Crss, easier to mount TO220 package and a fraction of the price albeit positive tempco.