Has somebody used IGBT in power amp

The Crescendo Amplifier (Millenium Edition) article in a recent Elektor magazine compares the sound of the Crescendo Mosfet amplifier with that of their earlier published IGBT amp (Compact AF Amplifier - using Current Feedback instead of the usual Voltage Feedback). The report says that the sound of the MOSFET circuit is more laid back but less detailed than the IGBT circuit. This could be due to topology differences but....

I have heard an IGBT amplifier that goes by the name of Forte'. If I am not mistaken, this is a sister-concern of the famous Threshold brand. The amp sounded very detailed and authoritative. But on prolonged listening, the bass seemed to be "out-of-phase" with the rest of the sound spectrum; it seemed that the speaker cone was moving in the opposite direction than required. However, in the same set-up, the Ultra-fast amplifier design published in Electronics World, far out-performed the IGBT amplifier in every aspect of sound reproduction. Smaller amplifiers, both Bipolar and Mosfet output types, did not suffer from the afore-mentioned bass problem, but failed to exhibit the same authority and speed as the Ultra-fast or IGBT amplifiers.

This is only an observation that I could make as well as verify with other expert listeners but could not arrive at any conclusions, for lack of enough data regarding circuit details etc.,

I do have at hand a pair of Toshiba IGBTs and the PCBs for the Compact AF Amplifier (Elektor design) but have not yet assembled these and tested them.

John Linsley Hood who was asked by a Japanese manufacturer to test IGBTs for audio purposes, reported his findings in Electronics World, along with the circuit used in this comparative study of Bipolars, Mosfets and IGBTs. You could probably have a look at this informative article.

Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
Before I left Threshold, I delivered a white paper to
the "new management" entitled "IGBT's: Threat or Menace?"
and it was my opinion that the only advantage to them was

The amplifiers to which you refer were designed by
Michael Bladelius.

As you may be aware, IGBT's are essentially two devices
on a chip; a MOSFET whose output Drain goes into the Base
of a high power Bipolar device whose Collector attaches
back to the Source of the FET. In an N channel IGBT the
MOSFET is N channel and its output is boosted by a PNP

For high end linear applications there are a couple of
problems. First, there is no way of biasing the MOSFET
independently of the Bipolar. Normally we like to run
previous stages at significantly higher bias than the
needs of the following stage, which gives better linearity
and faster response. Since all the current from the Drain
of the MOSFET goes into the Base of the Bipolar and there
is no external access, this cannot be adjusted.

Secondly, matching is essential when using them in
parallel, but the match has to occur not only between
the MOSFET portion, but also the Bipolar device, and this
is very difficult to do, resulting in current hogging and
reliability problems.

The parts were a pain in the butt, both Threshold
and Counterpoint seemed to have abandoned them, and
the manufacturer discontinued offering complementary
parts for audio use.

IGBT's remain popular for switching applications, but
I haven't seen them in an audio amplifier since.

About 10 years ago, a design was published as an Analog Devices application note, by a guy called Mark Alexander. The amp was a current feedback design. This design was taken on by the Danish magazine " High Fidelity", and very slightly altered by Poul Ladegaard et.al. This amp used IGBTs in the output, and was loudly praised for its qualities, "authority " being one of them.

I'm quite sure NP has seen this app.note. It would be quite interesting to hear NPs comments on this, also on the subject of current feedback. Also since I have two fully populated PCBs of this amp in my drawer marked " eternal projects". ( BTW- anyone else but me that has such a drawer ?? )

Halgeir Wold

[Edited by AuroraB on 10-12-2001 at 05:24 AM]


2001-10-13 2:25 pm

What will you say about all-bipolar amplifier wihch generates only 2nd harmonic of distortion for almost all it's dynamic range? And this amp even runs in class AB.
I think the best amplification devices after the vacuum tubes are the bipolar transistors.


2001-02-04 4:23 am
"IGBT's remain popular for switching applications, but
I haven't seen them in an audio amplifier since." Hex-fets are poor choices for high voltage swiching due to the lossses from the Rds.The IGBT does not have this problem.On the other hand it has the same problem that all high voltage BJT switching transistors have.It is not suitable for linear operation due to very poor secondary breakdown characteristics.Apex makes some 10KW class D modules with IGBTs.You would be lucky to get 100W with the same parts in a class AB amplifier.
My only expirience with the IGBT´s is in ELEKTOR magazine´s amp. you can see that in my web site :http://download.tripod.de:81/Promitheus/950077.pdf
The original design was with small hexfets I think irf540 and irf9540. After that they used the same topology scaled up with the Toshiba IGBT´s GT20D101 and GT20D201. I wasn´t so satisfied with the sound and I dind´t like the idea after that they used the IGBT´s in place of the mosfets since IGBT´s are like two in one devices (bipolar driving a mosfet). When I changed the output devices back to hexfets like the original smaller version but with bigger devices, it sounded perfect. I used in the new version irfp250 and irfp9250. I prefer them a lot over igbt´s.

sorry, but any class AB amp using any type of output device will not be capable of generating "only 2nd harmonic distortion" over it's entire dynamic range. This is quite impossible. Hint: the measurement frequency plays a critical role in distortion measurement... it's easy to hide all the distortion spectra under the noise floor of your test setup when making a measurement at 100 Hz. Try doing this at 20kHz or 50kHz where your distortion products will be much larger.
I've got some (thrown-away) Semikron IGBT-modules intended for use in frequency converters (see for instance http://www.vacon.com). The module consists of three IGBT's rated at 800 V and 200 A (they have models for up to 3300 V and 500 A). They use a couple of modules when making frequencý converters from 0.5 kW up to several MegaWatt. I know that these modules are intended for switching, but it would be fun to try to make a linear class-A amp with them. They are so much overrated that it would be no problem with stabilisation.


2001-10-13 2:25 pm
2nd harmonic

I said " for almost all of the dynamic range". Of course, 3rd and higher order harmonics do appear when the clipping approaches. No sign of something higher than 2nd harminic is visible below -3dB in respect to full output level. Of course, I took all precautions against noise and mains interference! The measurements are similar at any frequency up to 60 kHz. At the output voltage of 2.83 vrms, what corresponds to 1 watt/8Ohms, there is certainly pure 2nd harmonic and THD<0.1%.
By the way, my PP pentode amplifiers using EL509 valves behave the same way. Indeed, they operate in pure class A for 8 or even 4 Ohms, but transit to class AB operation at certain level for lower load resistances. Only those people, who visited my laboratory, now beleive!
The phenomenon of odd harmonisc cancellation in pentodes was first observed in 1950s. The same approach is applicable to bipolar transistors too.
Some details of my solid-state design:
The output stage utilises from 2 to 6 pairs of Motorola TO-3 transistors running at 250-450 mA of total idle current. There are also several paralleled pairs of drivind transistors which run also at considerably large idle current. The front end consists of just three transistors, and there is no differential pair at the input. This topology allowed me to apply moderate global NFB of just 12-14 dB.
If you have more questions, please write.


2001-02-04 4:23 am
You guys just don't get it.Would you like to buy a bridge I own in Brooklyn? Better yet buy my switching transistors. The original MRO wholesale price from Motorola was $44 each.I have about 600 of them.Yep, $25,0000 worth.They are rated at 150V 50A continuous 100A peak 250W and have an Ft of 30Mhz with a typical gain of 100 at 10A.THEY ARE WORTHLESS FOR AUDIO.At 50V they can only handle 15W because of secondary breakdown.And that is at 25*C, you have to derate that over temperature.An IGBT is this kind of a transistor with an FET hung on the base.If anybody is stupid enough to want these they can have them for $1 each.I may hand them out to the kids that come to trick-or-treat this year.


2001-10-13 2:25 pm

To make a good solid-state amp one has to have the best LINEAR transistors. The best ones are Motorola (OnSemi) MJ15024 and MJ15025! These TO-3 devices handle 250W at 25C, and they are extremely robust against second breakdown. It is possible to build a 400W/8 Ohms amp on just 6 pairs of them, and this amp will withstand continuous operation into 2 Ohms, and the real world short-output current of 40A during 10ms. These transistors are the real workhorses of professional audio manufacturers, and their price is usually $3/each! What is remarkable, they are linear to such an extent that it is possible to make an amplifier without global NFB and not obviously it must be a pure class A one. Other types as MJ21193 and 21194 look even better on their datasheet, but I have no experience with them.
No MOSFET or IGBT is comparable with these brilliant bipolars.
Indeed, I see an advantage of using large MOSFETs in high-power amps where the output devices of more than 250V rated Vceo are needed.
Theres a good description of IGBTS <a href="http://www.elec.gla.ac.uk/groups/dev_mod/papers/igbt/igbt.html"><b>here</b></a>

The idea of a parasitic thyristor and possible latch-up doesn't inspire confidence.

With so few commercial audio products using IGBTs it makes me wonder about their sound/relability.

If you are looking for a novel output device what about <a href="http://www.ne.jp/asahi/evo/amp/SIT/SITaf.htm">Static Induction Transistors (SIT)</a>
and another <a href="http://www.mobara.ne.jp/~yamada/sit/index.html">here</a>.
Both pages are in Japanes.

Motorola transistors and beta droop

This should maybe be a thread on it's own , because the subject is quite interesting,----but....

One of the few references I have found in the audio community that deals with the phenomenon of "beta droop", i.e. a reduction of beta as function of load current, is Douglas Self in his E&WW articles about distortion mechanisms. Beta droop definately seems to have an impact on distortion figures, as can well be imagined..

He also has a small survey of relevant parts, where in his test the MJ 15024/15025 comes in third, with MJ21193/21194 second and the often used japanese 2SA1302/2SC3281 as the very best of the ones tested.


2001-10-13 2:25 pm

MJ21193 and 21194 really look excellent. In fact, from the viewpoint of second breakdown, they have much more than twofold advantage at Vc=150-200 volts, that allows one to operate them even at +-110Vcc. Indeed both 15024/25 and 21193/94 in proper circuits operate at their thermal limitations rather than second breakdown. So, they must be used in equal quantities to ensure reliable service.