I have an array of IGBT output transistors on a heatsink (four pair), and in setting the bias for that side, I noticed that one particular transistor reads a little more than double what all the rest average. Considering that Toshiba stopped making these transistors more than ten years ago, is there any problem with doubling the emitter resistor from .47 ohm to 1 ohm to control the current flow from that one transistor? Obviously, the voltage drop over that transistor would increase, so we wouldn't use that resistor to calibrate that side, but the current should be more in line with the other transistors, so I could bias the whole group up without burning out the emitter resistor for that one wild card.
Does that make sense? Anything I'm not aware of here, like distortion effects or instability?
Thanks for your input. 👍
Does that make sense? Anything I'm not aware of here, like distortion effects or instability?
Thanks for your input. 👍
I think that transistor is simply quite unmatched to the rest, you won´t "solve" it by doubling its emitter resistor.
Besides, you would be "matching" it at very low, idle current ... and introducing a gross 2:1 mismatch at higher currents.
In any case, we need some numbers.
Besides, you would be "matching" it at very low, idle current ... and introducing a gross 2:1 mismatch at higher currents.
In any case, we need some numbers.
Agree with JMF. Being that far off, I wonder if the gate took a little ESD hit?
A bit sideways here, but I had never heard of using an IGBT in a linear audio amplifier. I thought their linearity was not that great and N vs P are a poor match. Usually used only as a switch.
A bit sideways here, but I had never heard of using an IGBT in a linear audio amplifier. I thought their linearity was not that great and N vs P are a poor match. Usually used only as a switch.
Hey you guys,
Thanks for taking time to help out. Here's a little more detail:
The Forte amps were built around the Toshiba GT20D201 P-channel and GT20D101 N-channel IGBT transistors, which, according to Mr Nelson, were the only matched pair of IGBT transistors ever made specifically for audio use - certainly, by Toshiba anyway. These transistors were introduced in the mid 1980s, and were removed from the market by 2004.
So in an ideal world I would buy a couple of tubes of 201s at $3.00 each and work with them until I found one that would fit in the circuit, but unfortunately, that is no longer an option.
Numbers:
Here's our circuit. This schematic is from a later version of the board with .33 ohm emitter resistors and 79VDC rails; mine has .47 ohm emitters and I am using a different tap from the transformer for 58VDC rails:
The board is in a test rig with 19VDC applied. The actual rail voltage when the board is installed is about 58VDC, so these numbers will spread out a bit in actual use.
DC millivolts across the N-channel emitter resistors (in this configuration) is: 21.7, 21.3, 21.7, 20.7. P-channel is 21.6, 21.6, 22, and 56.5.
Remember, we measure voltage across an emitter resistor just as a convenient means to an end - given that all the transistors are aligned in parallel across the same rail voltage and center tap, and if the transistors were all matched, the DC voltage across the emitter resistors would just tell us how much current was flowing though that particular transistor without having to interrupt (unsolder) the circuit and connect an ammeter. And the current flow is simply limited by the ability to disperse heat, so in the end, its all about heat transfer. It's like eight different streams flowing into a river (the center tap) and out to the speaker. It's the current we want to match, and if I have one transistor flowing a lot more current than the others, I need to limit the overall bias current to that one high transistor to prevent it from overheating and running away - which now limits the current that can flow through the other three.
IGBTs are voltage-activated at the gate, like a MOSFET, not current activated like a BJT, so changing the gate resistor has no affect.
So here's the options I can think of:
1. Replacing the unmatched transistor is probably not reasonable - I would have to buy every Forte amplifier I found, remove and test all the transistors until I found a matching P-channel. That sounds expensive.
2. Replace all the transistors with matching MOSFETs - I'm actually working on that project with a model 5 right now but it will take a while.
3. Leave it alone and bias that side around the high-current Q20 - but the whole point of getting model 6 boards is that they have double the transistors/current output of the model 5, and if I now lower the bias to about half of what it could be, am I not limiting the current output? Maybe not. Not sure about that one. All I really know about that is that it will transfer to A?B operation earlier.
4. Remove the high transistor and it's N-channel match, and run that side on 3 pairs instead of four.
5. Limit the current from the high transistor with a larger emitter resistor , which should match the current flow at idle. Will this create a significant disparity as we flow more current through the entire output section, as in, playing music?
Hope that's enough info to make sense, and thank you for taking time to share your thoughts and knowledge.
Thanks for taking time to help out. Here's a little more detail:
The Forte amps were built around the Toshiba GT20D201 P-channel and GT20D101 N-channel IGBT transistors, which, according to Mr Nelson, were the only matched pair of IGBT transistors ever made specifically for audio use - certainly, by Toshiba anyway. These transistors were introduced in the mid 1980s, and were removed from the market by 2004.
So in an ideal world I would buy a couple of tubes of 201s at $3.00 each and work with them until I found one that would fit in the circuit, but unfortunately, that is no longer an option.
Numbers:
Here's our circuit. This schematic is from a later version of the board with .33 ohm emitter resistors and 79VDC rails; mine has .47 ohm emitters and I am using a different tap from the transformer for 58VDC rails:
The board is in a test rig with 19VDC applied. The actual rail voltage when the board is installed is about 58VDC, so these numbers will spread out a bit in actual use.
DC millivolts across the N-channel emitter resistors (in this configuration) is: 21.7, 21.3, 21.7, 20.7. P-channel is 21.6, 21.6, 22, and 56.5.
Remember, we measure voltage across an emitter resistor just as a convenient means to an end - given that all the transistors are aligned in parallel across the same rail voltage and center tap, and if the transistors were all matched, the DC voltage across the emitter resistors would just tell us how much current was flowing though that particular transistor without having to interrupt (unsolder) the circuit and connect an ammeter. And the current flow is simply limited by the ability to disperse heat, so in the end, its all about heat transfer. It's like eight different streams flowing into a river (the center tap) and out to the speaker. It's the current we want to match, and if I have one transistor flowing a lot more current than the others, I need to limit the overall bias current to that one high transistor to prevent it from overheating and running away - which now limits the current that can flow through the other three.
IGBTs are voltage-activated at the gate, like a MOSFET, not current activated like a BJT, so changing the gate resistor has no affect.
So here's the options I can think of:
1. Replacing the unmatched transistor is probably not reasonable - I would have to buy every Forte amplifier I found, remove and test all the transistors until I found a matching P-channel. That sounds expensive.
2. Replace all the transistors with matching MOSFETs - I'm actually working on that project with a model 5 right now but it will take a while.
3. Leave it alone and bias that side around the high-current Q20 - but the whole point of getting model 6 boards is that they have double the transistors/current output of the model 5, and if I now lower the bias to about half of what it could be, am I not limiting the current output? Maybe not. Not sure about that one. All I really know about that is that it will transfer to A?B operation earlier.
4. Remove the high transistor and it's N-channel match, and run that side on 3 pairs instead of four.
5. Limit the current from the high transistor with a larger emitter resistor , which should match the current flow at idle. Will this create a significant disparity as we flow more current through the entire output section, as in, playing music?
Hope that's enough info to make sense, and thank you for taking time to share your thoughts and knowledge.
Gate resistor effects the stability as it controls the TC of the gate capacitance.
Reducing the bias, i.e. spreader setting, changes the A to B transition point, not the available current.
I think it is a hint, " only IGBT's..." and "removed from market"
As the IPS and VAS are pretty nice, and you have the chassis and power supply, a bank of Exicon's would sound like a great plan to me! In the mean time, delete one pair and don't try to drive 4 Ohm speakers. With 54V, I seriously doubt you could measure or hear the difference. Although I am a fan of multiple pairs to gain an advantage in linearity in a BJT, with 57V, 2 pairs of MOSFETS would be overkill. Well, I like MOSFETS for their linearity. Seems over history, every amp I ever liked with the exception of the Aragon was MOSFET output.
Reducing the bias, i.e. spreader setting, changes the A to B transition point, not the available current.
I think it is a hint, " only IGBT's..." and "removed from market"
As the IPS and VAS are pretty nice, and you have the chassis and power supply, a bank of Exicon's would sound like a great plan to me! In the mean time, delete one pair and don't try to drive 4 Ohm speakers. With 54V, I seriously doubt you could measure or hear the difference. Although I am a fan of multiple pairs to gain an advantage in linearity in a BJT, with 57V, 2 pairs of MOSFETS would be overkill. Well, I like MOSFETS for their linearity. Seems over history, every amp I ever liked with the exception of the Aragon was MOSFET output.
"I think it is a hint, " only IGBT's..." and "removed from market""
Ha ha... yup, that's a clue! But it's water over the dam now, and I love the sound of the earlier Forte amps when they're working right. 😉
"a bank of Exicon's would sound like a great plan to me!"
Okay, so that's where I've been spending some time and energy with the Model 5. I'm working with that one because I bought it cheap as a fixer upper, and it only has two pairs of IGBTs per side, so less expensive to experiment with. My stopping point right now has been trying to choose a set of MOSFETs (voltage driven) that will work in the existing circuit, and I am new to designing. Using Mouser and Digikey filters, I had boiled down to the Vishay IRFP9140PBF - 180w, 2Vgs, 200 mOhm Rds. But to be honest, I'm just guessing. Can you tell me more about the Exicons and why you thought of them?
"Although I am a fan of multiple pairs to gain an advantage in linearity in a BJT, with 57V, 2 pairs of MOSFETS would be overkill."
Very interesting. I'm trying to learn here - can you tell me a little more about this? Are you saying that with lower rail voltage additional transistors have a diminishing affect? Or is the voltage high enough that the additional current is not noticeable?
Thanks a lot for your time! 👍
Ha ha... yup, that's a clue! But it's water over the dam now, and I love the sound of the earlier Forte amps when they're working right. 😉
"a bank of Exicon's would sound like a great plan to me!"
Okay, so that's where I've been spending some time and energy with the Model 5. I'm working with that one because I bought it cheap as a fixer upper, and it only has two pairs of IGBTs per side, so less expensive to experiment with. My stopping point right now has been trying to choose a set of MOSFETs (voltage driven) that will work in the existing circuit, and I am new to designing. Using Mouser and Digikey filters, I had boiled down to the Vishay IRFP9140PBF - 180w, 2Vgs, 200 mOhm Rds. But to be honest, I'm just guessing. Can you tell me more about the Exicons and why you thought of them?
"Although I am a fan of multiple pairs to gain an advantage in linearity in a BJT, with 57V, 2 pairs of MOSFETS would be overkill."
Very interesting. I'm trying to learn here - can you tell me a little more about this? Are you saying that with lower rail voltage additional transistors have a diminishing affect? Or is the voltage high enough that the additional current is not noticeable?
Thanks a lot for your time! 👍
The Exicons ( have to order from England) are the de-facto replacements for the Toshiba MOSFETS. The gate resistor is a little different, but they work quite well. Mine are TO-3 but most go for the TO3-P case. I think there are several threads on them. Might search the Hafler threads.
One of the big advantaged of MOSFETS over BJT's is they are more linear to start with. Does that really matter with global feedback? I won't swear to it, but somehow, I always find them my preference when it comes to listening to music. My comment on the lower voltage is with 57V, usually one pair is all a OEM would use, so 2 or 3 pairs if your IGBT's is already distributing the load like the better amps that had many more devices. The lower the delta between idle and peak the more even the die temp etc. Tough answer, but as John Curl lead me on when I got into it "study the cures" He gave hints, not answers, so you had to actually understand it, not just copy.
I used ECF10N20 and ECF10P20. with 420 and 330 gate resistors respectively in my modified DH120 with 54V rails. Only one pair. I keep thinking about a second pair, but just can't bring myself to quit listening it for long enough. Besides the different outputs, it used Miller compensation and several other changes for compensation. Actually eliminated two compensation caps. 4 times the original rail storage and of course, bigger rectifier for surge current, changed the wireing more to full star topology, small change in IPS local feedback ( LTP emitter resistors) and global feedback ratio.
If I were building an amp from scratch, I would do th boards so I had no wire leads tot eh transistors to have better management of the parasitics. My old Sanyo Plus 55 did that by using a heat-pipe heat sink. Sounded very nice but you could hear the Freon gurgling in the pipe!
One of the big advantaged of MOSFETS over BJT's is they are more linear to start with. Does that really matter with global feedback? I won't swear to it, but somehow, I always find them my preference when it comes to listening to music. My comment on the lower voltage is with 57V, usually one pair is all a OEM would use, so 2 or 3 pairs if your IGBT's is already distributing the load like the better amps that had many more devices. The lower the delta between idle and peak the more even the die temp etc. Tough answer, but as John Curl lead me on when I got into it "study the cures" He gave hints, not answers, so you had to actually understand it, not just copy.
I used ECF10N20 and ECF10P20. with 420 and 330 gate resistors respectively in my modified DH120 with 54V rails. Only one pair. I keep thinking about a second pair, but just can't bring myself to quit listening it for long enough. Besides the different outputs, it used Miller compensation and several other changes for compensation. Actually eliminated two compensation caps. 4 times the original rail storage and of course, bigger rectifier for surge current, changed the wireing more to full star topology, small change in IPS local feedback ( LTP emitter resistors) and global feedback ratio.
If I were building an amp from scratch, I would do th boards so I had no wire leads tot eh transistors to have better management of the parasitics. My old Sanyo Plus 55 did that by using a heat-pipe heat sink. Sounded very nice but you could hear the Freon gurgling in the pipe!
LOTS of useful information here, thank you for taking time to write that out! It is true, in these forums we seem to get a lot more opinions and hints than well presented hard-core information! 😉
In reading this, a couple of things come to mind:
The Forte power supply uses a center tap, so its probably more accurate for me to say the rails are +58 - 0 - -58. Is it correct for me to say the rail voltage is 58VDC, or 116VDC?
One of the significant characteristics of the Forte amps was their minimal-to-none (none in the early versions, minimal in the later versions of the boards) negative feedback, which is a typical Nelson Pass design. This is why transistor matching was so important, and why I agree with you about the MOSFETs.
"2 or 3 pairs if your IGBT's is already distributing the load like the better amps that had many more devices." bing <light bulb> Of course. All this time I have been thinking of the transistors as supplying more current to the load (speaker) but it sounds like what you are saying is that in fact, they are just distributing the load so that each flows less current. (?) It's kind of like saying a 600VA transformer supplies more power than a 300VA transformer - the transformer doesn't supply power, it's just capable of supplying more current before it overheats.
"I used ECF10N20 and ECF10P20. with 420 and 330 gate resistors" - would the gate resistor calculation be based on (and vary based on) the existing Vgs at the gate now?
Thanks for your help!
In reading this, a couple of things come to mind:
The Forte power supply uses a center tap, so its probably more accurate for me to say the rails are +58 - 0 - -58. Is it correct for me to say the rail voltage is 58VDC, or 116VDC?
One of the significant characteristics of the Forte amps was their minimal-to-none (none in the early versions, minimal in the later versions of the boards) negative feedback, which is a typical Nelson Pass design. This is why transistor matching was so important, and why I agree with you about the MOSFETs.
"2 or 3 pairs if your IGBT's is already distributing the load like the better amps that had many more devices." bing <light bulb> Of course. All this time I have been thinking of the transistors as supplying more current to the load (speaker) but it sounds like what you are saying is that in fact, they are just distributing the load so that each flows less current. (?) It's kind of like saying a 600VA transformer supplies more power than a 300VA transformer - the transformer doesn't supply power, it's just capable of supplying more current before it overheats.
"I used ECF10N20 and ECF10P20. with 420 and 330 gate resistors" - would the gate resistor calculation be based on (and vary based on) the existing Vgs at the gate now?
Thanks for your help!