IGBT's as power stage?


Paid Member
2000-10-10 7:27 pm
I know you can find IGBT's at Digi-Key and you may also find the BC's there too. International Rectifiers (IRF) is the maker of the IGBT's at Digi-Key. I know I've seen the BC's at Mouser Electronics. Both of these comapanies are in the USA though.

[Edited by vdi_nenna on 01-28-2001 at 12:18 PM]
Analog devices has an app note using IGBTs as the output drivers for one of their opamps. The circuit is patented by Mark Alexander, assigned to ADI. You may be able to find the note in an older ADI databook or on their website.

The amp uses a FET input opamp, (OP42, slightly better than an LF356 or TL072) as the input buffer to a current feedback amplifier. the supplies of the op42 are generated from the emitter of a transistor. The collector of same transistor feeds into a complementary current mirror. The quiescent current of the opamp is balanced and biases the current mirrors. Any output current from the opamp is mirrored into the current mirrors as an error signal. The high impedance junction of the current mirrors is connected to a typical output buffer, albiet with IGBTs as the final driver. The original circuit was designed and patented with MOSFETs.


2001-10-13 2:25 pm
Best output device

Sorry, IGBTs are poor candidates for linear applications.
They have inherent advvantages over both BJTs and MOSFETs in high-power switching circuits. Indeed, there are numerous technical problems in using them in the audio amplifiers.

But the last word seems to be the sound quality.

It is true fact: IGBT, SIT, vertical and lateral MOSFETS, all of these devices are inherently inferior in this respect to those old BIPOLAR TRANSISTORS!
The IGBT's used in the application note from AD IS MADE FOR AUDIO USE. It is the types mentioned :
GT20D101 and GT20D201 from Toshiba if i recall Farnell should have them.
By the way High Fidelity in denmark did make a project some years ago with the CFB design from AD but with more output devices.
This design has a highend soundQ. if it has a big PS and slightly higher current limit.
I did a design myself with a pair of GT20D101/D201 per channel and they did do thermal runaway. Use them as multiple.


toshiba igbts


we stock the toshiba igbts and they are splendid for audio.

the amplifier you refer to is a bad rebuild of mark alexanders current feedback amplifier.

the original use these igbts and is a very good amplifier indeed , for several reasons the high fidelity project turned into a power oscillator (and an expensive one you could add) , its not recommended!

mail me if you want more info or want to buy some of the igbts.

k madsen - [email protected]
I think the problem with the High Fidelity design Was to long wires from the driverstage to the igbt. Afterall it is a 1 MHz design even if you try to lower its bandwith by compensating in the feedback loop (Be carefull with CFB design .. Check some papers from AD about designing with CFB).

My own design is also a CFB but without the opamp.
I don't like the idear in the Alexander design. We always learn that opamp should have a low impedance supply even by 100MHz. That can't be achived by the alexanderdesign.
But the alexander amp works. who cares!?


I am on to build a new amp with a lot of currentsources. But it is still CFB design ... i think .. .lets see!?

Regards ... Have fun designing...

alexander cfb amp


>I don't like the idear in the Alexander design.>


>But the alexander amp works. who cares!? >

it not only works , its a splendid amplifier and it compares very well to other designs , you could add it has a very fine grained hi freq reproduction , it plays very good midrange and quite acceptable bass.

it has very good transient response and in the same time it has quite low bias current.

i think its a combination of very good specs and very good performance.

make a better one :) (before you dis the alexander amp) :)

k madsen - cadaudio.dk
I didn't say that i make a better one!

:) hahahaha

I still think it is a good amp.
Because i don't like the idear doen't mean that it is a bad design.

We all have to compare with other design's and who says that other people think that my amp sounds better!????? ... Just because i think it is good! :( :) :) ... My living is not designing amps but hydraulic valve controllers.

I think that designing amp in the sparetime is a kind of fun. It does not have to be cheaper!!!!!!



PS : KM is you also from denmark.
Alexander design

I would very much like to hear KMs opinion on the differences between the original MA design and the High Fidelity rebuild. I have a couple of unfinished boards of this design ready to be completed, when I get around to it.
I started this just for kicks, mainly to test a CFB design- but as I can recall right off hand, the differences are in minor component changes.
Maybe the wiring length was the crucial point... ???


2001-02-04 4:23 am
"The IGBT's used in the application note from AD IS MADE FOR AUDIO USE. It is the types mentioned :
GT20D101 and GT20D201 from Toshiba if i recall Farnell should have them.
By the way High Fidelity in denmark did make a project some years ago with the CFB design from AD but with more output devices.
This design has a highend soundQ. if it has a big PS and slightly higher current limit.
I did a design myself with a pair of GT20D101/D201 per channel and they did do thermal runaway. Use them as multiple.


Sonny" Hey Sonny, can you read a spec sheet? The SOA on the Toshiba GT20D101/201 is only 60W at 100V.And you have to de-rate it to zero at 150*C(junction temperature) .A MJ21193/94 has 200W safe area at 100V and maximum junction temperature is 200*C.That means you need at least six of the Toshiba outputs to do what one Motorola does.An IGBT is a FET driving a BJT.Why don't you just use an FET as the driver and a BJT for the outputs.Cost you a lot less and give you better reliability and performance.You can select the best FET for the job of driver and the best BJT for the job of output.The commercial IGBT ar optimized for switching.They are an expensive joke for a linear amp.


2001-10-13 2:25 pm

DJK,you are quite rigt!

Motorola makes excellent bipolar transistors specially for linear applications. And they have correct datasheets.

Indeed, I dislike an idea to drive the bipolars by MOSFETS.
The all-bipolar circuit will sound far better and it is possible to get the distortion spectrum as soft as that of the best tube amps! I already have an all-bipolar design which generates just 2nd harmonic for the most part of it's dynamic range. The higher orded harmonics appear just approaching the clipping. And this is class AB design.

When it comes to combine different types of active devices, I rather think of hybrid vacuum tube/bipolar design.
P.S. I saw Alexander amp in AD Technical Bulettins. My opinion: Nothing close to real high-end sound!
My amplifier isn't runing with IGBT right now. 3 years ago i was designing my CFB amplifier. It was designed to run with 2SC3281 and 2SA1302 wich is bipolar.
At the same time i got samples of GT20D101/D201. I did read SOA. My Supply rail's was +/- 35Volt ... But i did think i will try mount them (There was some recalculation of the bias and driverstage) ... Even if i was going to run near the the limit. It did sound good. But i turned the volume a bit high up. And then they died.... Nothing special about it. That's why i say "use them multiple" ...
And they are still made for audio.

My amplifier now run's with 2SC3281 and 2SA1302 and i will in the near future try MJ21193/94...

Why dis MOSFET or IGBT!?

Firms like GamutAudio makes some fine highend amplifiers with mosfets.

I think KM will agree in that!?

or even go for class T or class D as long as the sound is good!?

Why not compare the sound before judge or dis the technology. :)

Problems with IGBT's

I have never designed an amplifier using IGBT's since all my amplifier designs at the present time use MOSFETS. I have looked into potentially using IGBT's but I have always been dissatisfied with the specifications and the number/types of devices available for audio use. They appear to be made for primarily for motor and power control applications. Therefore I have never given much thought to using IGBT's in audio amplifiers. Straight bipolar devices appear to be much more attractive that IGBT's and in some cases even MOSFETS. You can always drive a bipolar device with a MOSFET if you are looking for high input impedance. The performance of this combination will likely be much better than with an IGBT. Also the choices of potential device combinations are increased by the thousands.

I recall reading a article somewhere that mentioned a latch up problem with IGBT's if they are driven close to or into saturation, and that therefore they must be current limited to avoid this. Does anyone have any information or experience with such a latch up problem?

John Fassotte
Alaskan Audio

[Edited by alaskanaudio on 10-27-2001 at 01:47 PM]
igbt / cfb amps


basically i think the high fidelity magazine amp was a crude rip-off from the original analog devices design (bear in mind a.d. had an igbt and a very similar mosfet design).

i think there were several problems with the magazine design , one of them was long wires and another was using the ibgts in multiples (like just suggested here :)

i agree that listening to the sound is a lot more important than talking specs or topologies, however its not that easu when we live in different parts of the world ;)

im from copenhagen - denmark , any danes here are welcome to contact me for a demo or whatever.

i do not agree that the sirius/gamut is a very good amplifier (sorry , but perhaps a matter of taste) , i have noticed they use a lot of effort eplaining the idea of one giant fet in the output, funny when we know its an industrial fet with many fets in one house :)
(btw. are they still in business?)

i still think the alexander cfb amp (with one set of output devices , igbts or fets) is a very good amplifier indeed and that it will give plenty of power even for not-so-efficient speakers.

class-d? yes in fact its possible to make very good hifi quality amplifiers using class-d topology , thats another story and perhaps worth another thread.

k madsen - [email protected]