JazzPeter,
NO WAY YOU'RE GOING TO SELL THOSE IGBT's!!!!!
If you are, I'm coming over to you (yes, thats about 850 kilometres) to make sure you build this amp and don't stop before it sounds good (that means, I'll be in Danmark at the beginning of march; february was impossible for me).
I didn't write all those email's for nothing! 🙂 🙂 🙂
BTW: Welcome kramer!
Grtz, Joris
NO WAY YOU'RE GOING TO SELL THOSE IGBT's!!!!!
If you are, I'm coming over to you (yes, thats about 850 kilometres) to make sure you build this amp and don't stop before it sounds good (that means, I'll be in Danmark at the beginning of march; february was impossible for me).
I didn't write all those email's for nothing! 🙂 🙂 🙂
BTW: Welcome kramer!
Grtz, Joris
BTW kramer,
isn't it possible to exchange the GT20D101/D201 with a pair of 2SJ201/2SK1058 (I think these are the right numbers) ??
Grtz, Joris
isn't it possible to exchange the GT20D101/D201 with a pair of 2SJ201/2SK1058 (I think these are the right numbers) ??
Grtz, Joris
Re: IGBT? Perhaps so...
Then how did he achieve stable Iq when he switched to MOSFETs?
I think your CPA-2 circuit may work with carefully calculated resistor values. My reasoning is this:
Vbe multiplier is mounted to same heatsink as output devices.
Output devices heat up and Iq increases
-> Vbe transistor heats up
-> Vbe voltage decreases
-> Current through driver emitter resistors decrease
-> Current through driver collector resistors decrease
-> Gate-source voltage decreases
-> Iq decreases
What do you think?
The reason why it did not work with IGBT may be because it has different Vgs/Id versus temperature characterisitics than MOSFETs.
pjacobi said:
Then how did he achieve stable Iq when he switched to MOSFETs?
I think your CPA-2 circuit may work with carefully calculated resistor values. My reasoning is this:
Vbe multiplier is mounted to same heatsink as output devices.
Output devices heat up and Iq increases
-> Vbe transistor heats up
-> Vbe voltage decreases
-> Current through driver emitter resistors decrease
-> Current through driver collector resistors decrease
-> Gate-source voltage decreases
-> Iq decreases
What do you think?
The reason why it did not work with IGBT may be because it has different Vgs/Id versus temperature characterisitics than MOSFETs.
Long life Janneman! You are a true "Current VS Voltage" Crusader. The last battle I see in http://www.diyaudio.com/forums/showthread.php?threadid=26170goto=newpost
continues here
continues here
Commentable Thoughts
Hey Guys
This Elektor amp is not a good design .
It has Stability problems indeed.
The mosfets are more stable in linear amps than IGBTs.
IGBTs are well suited to Switching Applications only.
Regards
Ampman
Hey Guys
This Elektor amp is not a good design .
It has Stability problems indeed.
The mosfets are more stable in linear amps than IGBTs.
IGBTs are well suited to Switching Applications only.
Regards
Ampman
Re: Re: IGBT? Perhaps so...
- Much lower gm.
- Inherent stability due to decreasing current with increasing temperature
Regards,
Peter Jacobi
ojg said:
- Much lower gm.
- Inherent stability due to decreasing current with increasing temperature
Regards,
Peter Jacobi
Is there anyway to make the Elektor IGBT design better and more stable?
Should I keep my new IGBT's or sell them?
Is ther anyone who has good exsperience withe the Elektor amp?
To Vigier: Your are welcome in march. If you can show me how to make the Elektor amp stable I'll keep them...;-)
- Peter
Should I keep my new IGBT's or sell them?
Is ther anyone who has good exsperience withe the Elektor amp?
To Vigier: Your are welcome in march. If you can show me how to make the Elektor amp stable I'll keep them...;-)
- Peter
Hi ojg, All,
Got the point with MOSFets tempco. This once more let's me wonder, how this amp is expected to achieve stability - and also its HexFet cousin.
Compare with a standard complementary feedback pair output stage (like http://www.linearaudio.de/scratch/CPA-1.pdf with R3 and R4 removed). Here Iq is under control of the local feedback loop and so the general advice is, to only mount the driver and Vbe multiplier together, but not on the output devices' heatsink.
As seen in http://www.linearaudio.de/scratch/CPA-1.pdf, most of this feedback can still be achieved when the output stage has to produce some gain (but it works best, when using only the minimum gain needed).
The Elektor circuit (like http://www.linearaudio.de/scratch/CPA-1.pdf) only relies on the "thermal feedback", which is an unreliable co-worker. So I expect thermal runaway, at least for high gm devices.
I did some further sims with the circuits shown and made the the bias voltage dependant on the actual Vbe multiplier used in the Elektor circuit. Then I did two tests:
a) stepping global temperature from 25°C to 65°C
- Iq of CPA-1 moved from 350mA to 650mA
- Iq of CPA-2 moved from 350mA to 1100mA
b) moving the output devices' emitters by 100mV to simulate a temperatur induced turn on voltage change:
- Iq of CPA-1 moved from 350mA to 430mA
- Iq of CPA-2 moved from 350mA to 600mA
So in CPA-1 the Iq instability is already signifantly reduced. Lowering the gain somewhat and doing some additional tweaking of the resistors should give even better numbers.
Regards,
Peter Jacobi
Got the point with MOSFets tempco. This once more let's me wonder, how this amp is expected to achieve stability - and also its HexFet cousin.
Compare with a standard complementary feedback pair output stage (like http://www.linearaudio.de/scratch/CPA-1.pdf with R3 and R4 removed). Here Iq is under control of the local feedback loop and so the general advice is, to only mount the driver and Vbe multiplier together, but not on the output devices' heatsink.
As seen in http://www.linearaudio.de/scratch/CPA-1.pdf, most of this feedback can still be achieved when the output stage has to produce some gain (but it works best, when using only the minimum gain needed).
The Elektor circuit (like http://www.linearaudio.de/scratch/CPA-1.pdf) only relies on the "thermal feedback", which is an unreliable co-worker. So I expect thermal runaway, at least for high gm devices.
I did some further sims with the circuits shown and made the the bias voltage dependant on the actual Vbe multiplier used in the Elektor circuit. Then I did two tests:
a) stepping global temperature from 25°C to 65°C
- Iq of CPA-1 moved from 350mA to 650mA
- Iq of CPA-2 moved from 350mA to 1100mA
b) moving the output devices' emitters by 100mV to simulate a temperatur induced turn on voltage change:
- Iq of CPA-1 moved from 350mA to 430mA
- Iq of CPA-2 moved from 350mA to 600mA
So in CPA-1 the Iq instability is already signifantly reduced. Lowering the gain somewhat and doing some additional tweaking of the resistors should give even better numbers.
Regards,
Peter Jacobi
Re: IGBT? Nee!!
First of all, I'm happy my message reopened discussion about IGBT amplifiers.
Unfortunately, my knowledge of electronics is unsufficient to understand all of your comments, some of them I think I understand, and I'll react in my own, " down to earth" way.
I understand some of you blame instability of idle current (is that Iq?) on thermal phenomena. I don't think so. I'll describe more precisely my experiences:
- the very first time I put the amp under current, I very carefully adjusted Iq (?) to 400 mA. Current decreased very rapidly (beause of temperature raise of the Tr's), after 1 minute it lowered to 300 mA or so. I readjusted it. Next 3/4 our, again, I adjusted (increased) Iq several times to 400 mA.
I thought the job was done. I let the amp cool down, and the next day turned it on again. Because I wasn't focussed on instable Iq, only after some time (half a minute) I saw, to my alarm, ammeters having run out of scale. Heat sink appeared to be at boiling temperature after this short period. I turned off the amp, and began some experiments with adjusting Iq. Then I discovered this "instantly jumping" of Iq to a high value (5A or so?). So this did not happen after warming up, but suddenly: while slowly adjusting current it raised slowly, to some point (400-500mA) then, in a split second, jumped to a high value.
Because I didn't want my precious amp to run on Iq of only 50 mA or so, I experimented with an NTC in the adjusting circuit. I thermally connected the NTC with one of the regulators on the PCB. This seemed to work quite well: after "cold start" Iq started on e.g. 300 mA, then decreased a bit and, after 10 min. raised to 300-400mA. There it stabilised. Again I thought I solved the problem and could turn to enjoying music.
But:
This is not a thermal phenomenon, gentlemen.
This is a question I would like to ask you, people who are more familiar with specifications of power Tr's. In Elektor's Crescendo Millennium I mentioned in my 1st mail, power Tr's are 2SJ201, indeed, and 2SK1530.
BTW, I also have some (7) IGBT's for sale. I need no further experiments with these bits.
Unless.... one of you comes with the ultimate solution?? (that is, a truly built and proven stable amp, not enough for me to do some simulations in your PC, however interesting this me be).
First of all, I'm happy my message reopened discussion about IGBT amplifiers.
Unfortunately, my knowledge of electronics is unsufficient to understand all of your comments, some of them I think I understand, and I'll react in my own, " down to earth" way.
I understand some of you blame instability of idle current (is that Iq?) on thermal phenomena. I don't think so. I'll describe more precisely my experiences:
- the very first time I put the amp under current, I very carefully adjusted Iq (?) to 400 mA. Current decreased very rapidly (beause of temperature raise of the Tr's), after 1 minute it lowered to 300 mA or so. I readjusted it. Next 3/4 our, again, I adjusted (increased) Iq several times to 400 mA.
I thought the job was done. I let the amp cool down, and the next day turned it on again. Because I wasn't focussed on instable Iq, only after some time (half a minute) I saw, to my alarm, ammeters having run out of scale. Heat sink appeared to be at boiling temperature after this short period. I turned off the amp, and began some experiments with adjusting Iq. Then I discovered this "instantly jumping" of Iq to a high value (5A or so?). So this did not happen after warming up, but suddenly: while slowly adjusting current it raised slowly, to some point (400-500mA) then, in a split second, jumped to a high value.
Because I didn't want my precious amp to run on Iq of only 50 mA or so, I experimented with an NTC in the adjusting circuit. I thermally connected the NTC with one of the regulators on the PCB. This seemed to work quite well: after "cold start" Iq started on e.g. 300 mA, then decreased a bit and, after 10 min. raised to 300-400mA. There it stabilised. Again I thought I solved the problem and could turn to enjoying music.
But:
kramer said:
This is not a thermal phenomenon, gentlemen.
Vigier said:
This is a question I would like to ask you, people who are more familiar with specifications of power Tr's. In Elektor's Crescendo Millennium I mentioned in my 1st mail, power Tr's are 2SJ201, indeed, and 2SK1530.
BTW, I also have some (7) IGBT's for sale. I need no further experiments with these bits.
Unless.... one of you comes with the ultimate solution?? (that is, a truly built and proven stable amp, not enough for me to do some simulations in your PC, however interesting this me be).
Re: Re: IGBT? Nee!!
Hi Kramer, All,
To give full disclosure, I'll admit that I'm a theoretician in extremum. If my physics diploma would have been any bit more theoretical, it would have been transferred to the philosophical faculty.
So of course, you still need some practical advice on your amp.
Nevertheless I have a theory what happened.
Thank you for the detailed problem description!
(BTW: Iq is short for idle or quiescent current)
IMHO it is a question of a missing (or insufficent) feedback path, which would control Iq. What happens under actual music load, is self-heating of the devices. The junction temperature increases - even when no change of the case temperature is seen, the junction temperature varies with the load. The thermal coupling between junction and case can be seen as some R/C combinations, delaying and filtering the temperature changes traveling from junction to case.
Due to this phenomenon, regulating Iq by thermal measures alone, works only in well-behaved circuits. Also mounting drivers, bias servo and output devices on the heatsink, gives the slowest possible coupling. Trying to make the response faster, it is sometimes advised to put the bias servo device directly on an output device.
Anyway, in the topology choosen by Elektor (complementary feedback pair), Iq is normally controlled by electric feedback and no problem (HF oscillations are the typical problem - do you have a scope for checking this?). For reasons completely beyond my understanding, the circuit's author has omitted this feedback.
My posted circuit modification was an attempt to restore it.
You can email details and price to me and Ill consider it.
Regards,
Peter Jacobi
Hi Kramer, All,
To give full disclosure, I'll admit that I'm a theoretician in extremum. If my physics diploma would have been any bit more theoretical, it would have been transferred to the philosophical faculty.
So of course, you still need some practical advice on your amp.
Nevertheless I have a theory what happened.
kramer said:
Thank you for the detailed problem description!
(BTW: Iq is short for idle or quiescent current)
IMHO it is a question of a missing (or insufficent) feedback path, which would control Iq. What happens under actual music load, is self-heating of the devices. The junction temperature increases - even when no change of the case temperature is seen, the junction temperature varies with the load. The thermal coupling between junction and case can be seen as some R/C combinations, delaying and filtering the temperature changes traveling from junction to case.
Due to this phenomenon, regulating Iq by thermal measures alone, works only in well-behaved circuits. Also mounting drivers, bias servo and output devices on the heatsink, gives the slowest possible coupling. Trying to make the response faster, it is sometimes advised to put the bias servo device directly on an output device.
Anyway, in the topology choosen by Elektor (complementary feedback pair), Iq is normally controlled by electric feedback and no problem (HF oscillations are the typical problem - do you have a scope for checking this?). For reasons completely beyond my understanding, the circuit's author has omitted this feedback.
My posted circuit modification was an attempt to restore it.
kramer said:
You can email details and price to me and Ill consider it.
Regards,
Peter Jacobi
IGBT's
These transistors are used in amps of fa Copland - with the same results as was said here - poor temperature stability of bias ( I was measured one by the test for one magazine).
These transistors are used in amps of fa Copland - with the same results as was said here - poor temperature stability of bias ( I was measured one by the test for one magazine).
Re: IGBT? Nee!!
Actually this is to be expected. With a class-AB amplifier, your ammeters in the rail will measure not only Iq but also the RMS value of the current going to the speaker. So with a low Iq (50-100mA) most of the current is that going to the speaker, and they will certainly look like VU meters. When you turn up the volume, the RMS value of the output current will be out of scale of our meter.
Always hit the pause-button on your CD-player before measuring Iq 😉
kramer said:
Actually this is to be expected. With a class-AB amplifier, your ammeters in the rail will measure not only Iq but also the RMS value of the current going to the speaker. So with a low Iq (50-100mA) most of the current is that going to the speaker, and they will certainly look like VU meters. When you turn up the volume, the RMS value of the output current will be out of scale of our meter.
Always hit the pause-button on your CD-player before measuring Iq 😉
Re: IGBT? Nee!!
To Peter Jacobi:
seems I have to reconsider my ideas of speed of thermal processes. The way you describe it this might fall within the limits of my understanding of electronics (thermal processes on micro-level, changing temperatures on junction level within 10-th’s of seconds?).
After a second glance on your suggestion for modification this seems to me a relatively simple job to perform (except for the fact that I will have to partly dismantle my now proudly playing amp again).
Just in case I come to do this, I resume the changes you suggest, to prevent me from doing stupid things that are self-evident to you:
1. I remove R1 and R2 (orig) (orig = Elektors original circuit);
2. R3, R4 (orig) are changed from 22 Ohm to 100 Ohm;
3. R8, R12 (mod) (mod = your modification scheme) are added newly;
4. R9, R5 (orig) are relocated from emittor side to collector side (become R1 and R2 (mod)) (I suspect the R’s of 0.01 Ohm are there just for simulations sake);
N.B. gate resistors R33 and R37 of 22 Ohm (original Elektors drawing) stay where they are? (not seen in both of your schemes)
If I come to all this, I certainly will mail the results on this forum (might take some time)
BTW, don’t you confuse your schemes in your mail from 31 jan? (CPA-1 = your modification, CPA-2 is Elektors original)
This was confirmed by what I saw when I had changed to the Mosfets: playing on about the same soundlevel, I saw ammeters hardly moving ( perhaps some 10-20 mA increase on Iq of 250 mA).
In Elektors article they claim one of the advantages of an Iq of ca. 400 mA that the amp will run in classA to a level of 2 or 3 watts. So this amp would, if in order and played on normal listening level, most of the time run in classA. (I realise my use of dimensions, “a soudlevel of xxx Watts” , is far from correct, just hope you understand what I mean).
To all: apart from all this technical talk, I wonder if changing only power Tr’s in a given circuit has much influence on “how the amp sounds”. In my case, would Hexfets, in a furthermore unaltered circuit, sound differently from IGBT’s, or other Fet-like devices?
To Peter Jacobi:
seems I have to reconsider my ideas of speed of thermal processes. The way you describe it this might fall within the limits of my understanding of electronics (thermal processes on micro-level, changing temperatures on junction level within 10-th’s of seconds?).
After a second glance on your suggestion for modification this seems to me a relatively simple job to perform (except for the fact that I will have to partly dismantle my now proudly playing amp again).
Just in case I come to do this, I resume the changes you suggest, to prevent me from doing stupid things that are self-evident to you:
1. I remove R1 and R2 (orig) (orig = Elektors original circuit);
2. R3, R4 (orig) are changed from 22 Ohm to 100 Ohm;
3. R8, R12 (mod) (mod = your modification scheme) are added newly;
4. R9, R5 (orig) are relocated from emittor side to collector side (become R1 and R2 (mod)) (I suspect the R’s of 0.01 Ohm are there just for simulations sake);
N.B. gate resistors R33 and R37 of 22 Ohm (original Elektors drawing) stay where they are? (not seen in both of your schemes)
If I come to all this, I certainly will mail the results on this forum (might take some time)
BTW, don’t you confuse your schemes in your mail from 31 jan? (CPA-1 = your modification, CPA-2 is Elektors original)
I do not expect this with an AB amp: I agree current I observed was total current through my amp. But values went up to more than 3 amp (higher I couldn’t witness, out of scale). This would mean I played my poor loudspeaker on a level of some 180 Watt or more!! Altough I played loudly, average level, I guess, would be 2 or 3 Watts or so (this is always be said to be an average listening level).ojg said:
This was confirmed by what I saw when I had changed to the Mosfets: playing on about the same soundlevel, I saw ammeters hardly moving ( perhaps some 10-20 mA increase on Iq of 250 mA).
In Elektors article they claim one of the advantages of an Iq of ca. 400 mA that the amp will run in classA to a level of 2 or 3 watts. So this amp would, if in order and played on normal listening level, most of the time run in classA. (I realise my use of dimensions, “a soudlevel of xxx Watts” , is far from correct, just hope you understand what I mean).
To all: apart from all this technical talk, I wonder if changing only power Tr’s in a given circuit has much influence on “how the amp sounds”. In my case, would Hexfets, in a furthermore unaltered circuit, sound differently from IGBT’s, or other Fet-like devices?
Re: Re: IGBT? Nee!!
Hi Kramer, All,
For an example, look at this datasheet:
http://www.infineon.com/cmc_upload/documents/014/330/SKW20N60_1.pdf
(page 7)
The thermal response of this IGBT is given as series of five R/C elements. The timeconstant of the fastest one given as 94uS.
Microseconds!
I'll double-check and will give you confirmation of this.
Its somewhat confusingly formulated, but I read it as the right way around. I'll check this too.
Regards,
Peter Jacobi
Hi Kramer, All,
kramer said:
For an example, look at this datasheet:
http://www.infineon.com/cmc_upload/documents/014/330/SKW20N60_1.pdf
(page 7)
The thermal response of this IGBT is given as series of five R/C elements. The timeconstant of the fastest one given as 94uS.
Microseconds!
kramer said:
I'll double-check and will give you confirmation of this.
kramer said:
Its somewhat confusingly formulated, but I read it as the right way around. I'll check this too.
Regards,
Peter Jacobi
Re: Re: IGBT? Nee!!
This is more normal, but currents may be larger than you think:
0.5W into 8ohm => 0.25A rms
2W into 8ohm => 0.5A rms
3W into 8ohm => 0.6A rms
Assuming "normal" speakers, 3W is quite loud and then the ammeters would act as VU meters, even more so if you use 4ohm speakers.
But as I said before, to measure Iq reliably: Play some music for a while then presss the pause button and look at the ammeter. This will be Iq and it should change very little over time.
As to whether it will sound different with different output devices: The major source of distortion in an amp comes from the output stage especially in a class-AB amplifier. So changing devices will change the distortion spectrum. Whether this is audible or not is always debateable 😉
D. Self's book covers this in detail, some info is also on his (poorly designed) webpage
Actually it would mean Irms*Irms*R = 3*3*8=72W, but I agree that this is much more than normal and indicates Iq runaway.kramer said:
This is more normal, but currents may be larger than you think:
0.5W into 8ohm => 0.25A rms
2W into 8ohm => 0.5A rms
3W into 8ohm => 0.6A rms
Assuming "normal" speakers, 3W is quite loud and then the ammeters would act as VU meters, even more so if you use 4ohm speakers.
But as I said before, to measure Iq reliably: Play some music for a while then presss the pause button and look at the ammeter. This will be Iq and it should change very little over time.
As to whether it will sound different with different output devices: The major source of distortion in an amp comes from the output stage especially in a class-AB amplifier. So changing devices will change the distortion spectrum. Whether this is audible or not is always debateable 😉
D. Self's book covers this in detail, some info is also on his (poorly designed) webpage
Re: Re: IGBT? Nee!!
Modifcation steps detailed
All devicemames in parentheses refer to the Elektor circuit in the magazine (also seen in post #1), without parentheses to my PDFs.
The 0.22 Ohm emitter resistors R9=(R34) and R5=(R38) are moved and replaced by direct connection (or an ammeter, or a fuse [not that it would be fast enough...]). The R5/R9 shown as 0.01 Ohm are dummies, as you already guessed.
The 0.22 Ohm Rs are now used as R1/R2 at IGBT's collectors.
Feedback network:
R3 (R31), R4 (R35) are changed from 22 Ohm to 100 Ohm, 5W.
R1 (R30) is removed.
R2 (R29) is removed and replaced by a wire.
R8 and R12 are added (100Ohm, 5W, low inductance)
(C18) is also removed, and you may optionally use 2n2 parallel to R8/R12 (I would leave them out, for a first try).
Yes, (R33) and (R37) stay the same.
It doesn't look easy on Elektor's reference photo, but if your arrangement would allow for it:
Put (T14), (T15) and (T16) on a small separate 'heatsink' (some scrap aluminum would be enough).
The thermal concept of the modified version is:
(T14), (T15) and (T16) should stay on the same temperature and the temperature of the IGBTs doesn't matter.
Regards,
Peter Jacobi
Modifcation steps detailed
kramer said:
All devicemames in parentheses refer to the Elektor circuit in the magazine (also seen in post #1), without parentheses to my PDFs.
The 0.22 Ohm emitter resistors R9=(R34) and R5=(R38) are moved and replaced by direct connection (or an ammeter, or a fuse [not that it would be fast enough...]). The R5/R9 shown as 0.01 Ohm are dummies, as you already guessed.
The 0.22 Ohm Rs are now used as R1/R2 at IGBT's collectors.
Feedback network:
R3 (R31), R4 (R35) are changed from 22 Ohm to 100 Ohm, 5W.
R1 (R30) is removed.
R2 (R29) is removed and replaced by a wire.
R8 and R12 are added (100Ohm, 5W, low inductance)
(C18) is also removed, and you may optionally use 2n2 parallel to R8/R12 (I would leave them out, for a first try).
Yes, (R33) and (R37) stay the same.
It doesn't look easy on Elektor's reference photo, but if your arrangement would allow for it:
Put (T14), (T15) and (T16) on a small separate 'heatsink' (some scrap aluminum would be enough).
The thermal concept of the modified version is:
(T14), (T15) and (T16) should stay on the same temperature and the temperature of the IGBTs doesn't matter.
Regards,
Peter Jacobi
IGBT: maybe some time ever?
Hi Peter,
thank you for your reply. I think all's clear to me.
Only question about putting T14/T15/T16 on the same (small) heatsink: in the present situation these 3 Tr's are already mounted on the same (big) heatsink, together with the IGBT's (now Hexfet's). Looking to the lay-out, from left to right:
T15-T16-T14-T17-T18, all with 3 or 4 mm distance between them.
Is it important they are thermally separated from the IGBT's, so are mounted on a different heatsink, or doesn't this matter. Only difference: they now heat up a little by IGBT's. From earlier experiences I would guess temperature of the heatsink raises not over 40 oC. Off course in the near distance of IGBT's temperature will be higher.
Again, thanks for your time. As I said before: carrying out these modifications needs me to got caught by a sudden impuls. When (if) this happens, I surely will let you know.
Hi Peter,
thank you for your reply. I think all's clear to me.
Only question about putting T14/T15/T16 on the same (small) heatsink: in the present situation these 3 Tr's are already mounted on the same (big) heatsink, together with the IGBT's (now Hexfet's). Looking to the lay-out, from left to right:
T15-T16-T14-T17-T18, all with 3 or 4 mm distance between them.
Is it important they are thermally separated from the IGBT's, so are mounted on a different heatsink, or doesn't this matter. Only difference: they now heat up a little by IGBT's. From earlier experiences I would guess temperature of the heatsink raises not over 40 oC. Off course in the near distance of IGBT's temperature will be higher.
Again, thanks for your time. As I said before: carrying out these modifications needs me to got caught by a sudden impuls. When (if) this happens, I surely will let you know.
Leaving all together on the big heatsink should work. Especially as T14 is nearest to the heat.
Peter
Peter
IGBT amps...
I dont have experiences especially with IGBT for an audio amplifier but as far as I know about IGBT's there's also a problem with the current tail when "switching off" the IGBT, eg. their switching off capability is far from ideal which is the reason that they haven't been used even for SMPS at 100 kHz.
Those I have worked with are proportionately huge 6 or 7 pack for inverters and were switched with a frequency of 5 kHz.
Moreover these IGBT's in switching application do also need a dead-time generator in some applications just because of the current tail phenomenom otherwise the start to "burn".
I checked up those Toshiba IGBT's mentioned in this thread and I can't say I'm thrilled over when reading the datashet, IGBT's are known for their current tail and not specifying it is IMO like hiding a disadvantage!
I found a datasheet over GT20G101 which I dont know how close it is to GT20D101 besides from difference of the letters D and G, but this IGBT have a typical fall time of 4 uS and turn off of 4,5 uS and it don't "sound" good to me.
Eg. I suspect the current tail phenomenom with IGBT's can give some crossoverdistorsion at higher audio frequencies and more else warm up IGBT's, but this should be checked wheter it could be so I don't want to make any 100% claim here at present before the issue is investigated further.
I think it would be much better to mimic an IGBT by using discrete BJT's and FET's and thereby controlling the built-in FET's drain connected to the base inside the IGBT etc if IGBT's is a must...
Anyhow here's an article of another IGBT amp with also a simple latch-up protection circuit: http://www.audiodesignguide.com/tra/igbtsch.jpg
I dont have experiences especially with IGBT for an audio amplifier but as far as I know about IGBT's there's also a problem with the current tail when "switching off" the IGBT, eg. their switching off capability is far from ideal which is the reason that they haven't been used even for SMPS at 100 kHz.
Those I have worked with are proportionately huge 6 or 7 pack for inverters and were switched with a frequency of 5 kHz.
Moreover these IGBT's in switching application do also need a dead-time generator in some applications just because of the current tail phenomenom otherwise the start to "burn".
I checked up those Toshiba IGBT's mentioned in this thread and I can't say I'm thrilled over when reading the datashet, IGBT's are known for their current tail and not specifying it is IMO like hiding a disadvantage!
I found a datasheet over GT20G101 which I dont know how close it is to GT20D101 besides from difference of the letters D and G, but this IGBT have a typical fall time of 4 uS and turn off of 4,5 uS and it don't "sound" good to me.
Eg. I suspect the current tail phenomenom with IGBT's can give some crossoverdistorsion at higher audio frequencies and more else warm up IGBT's, but this should be checked wheter it could be so I don't want to make any 100% claim here at present before the issue is investigated further.
I think it would be much better to mimic an IGBT by using discrete BJT's and FET's and thereby controlling the built-in FET's drain connected to the base inside the IGBT etc if IGBT's is a must...
Anyhow here's an article of another IGBT amp with also a simple latch-up protection circuit: http://www.audiodesignguide.com/tra/igbtsch.jpg
- Status
- Not open for further replies.