Beta droop
I think this unlikely a source of additional distortion. The propely designed transistor circuit does not rely on beta too much. Nobody else but Duglas Self said this.
Let me add, that if an amplifier uses multiple pairs of output transistors, it is even less significant. The output transistors in good circuits never run at their full rated current except for the short pulses.
I started with Toshiba 2SC3281/2SA1302, but I found them far less reliable than Motorola ones. Maximal die temperature of 200C means something! Indeed, when I switched from Toshiba to Motorola devices in the same amplifier circuit, I noticed no change in both value and spectrum of distortions for any frequency.
By the way, there are motorola's transistors designated MJ3281 and 1302! They have TO-3 cases and excellent SOA.
Indeed, I saw them only in databook.
I think this unlikely a source of additional distortion. The propely designed transistor circuit does not rely on beta too much. Nobody else but Duglas Self said this.
Let me add, that if an amplifier uses multiple pairs of output transistors, it is even less significant. The output transistors in good circuits never run at their full rated current except for the short pulses.
I started with Toshiba 2SC3281/2SA1302, but I found them far less reliable than Motorola ones. Maximal die temperature of 200C means something! Indeed, when I switched from Toshiba to Motorola devices in the same amplifier circuit, I noticed no change in both value and spectrum of distortions for any frequency.
By the way, there are motorola's transistors designated MJ3281 and 1302! They have TO-3 cases and excellent SOA.
Indeed, I saw them only in databook.
This is interesting discussion to me! Has anyone experienced any stability issues in their designs due to the much higher than typical (for a power bipolar transistor) bandwidth of the the xxC3281/xxA1302? I have read in Randy Sloan's book on amplifier design (and also somewhere else; I just can't remember the other source) that the much higher than typical Ft for causes some less than desirable anomalies if you're not careful.
Michael
Michael
Stability
I think, the troubles begin when more than two-stage Darlingtons are used. If the output stage consists of a standard two-stage Darlington, an unwanted oscillation may occur if the front-end transistor has too much an input capacitance. I always use the small transistors like 2SB861/2SD1138(several pairs if needed) and small anti-oscillation resistors in their base circuits. Of course, the wiring layout is critical.
Indeed, now I see no advantage in using the output transistors with Ft>2MHz from the sound quality viewpoint.
I think, the troubles begin when more than two-stage Darlingtons are used. If the output stage consists of a standard two-stage Darlington, an unwanted oscillation may occur if the front-end transistor has too much an input capacitance. I always use the small transistors like 2SB861/2SD1138(several pairs if needed) and small anti-oscillation resistors in their base circuits. Of course, the wiring layout is critical.
Indeed, now I see no advantage in using the output transistors with Ft>2MHz from the sound quality viewpoint.
Did you ever look closely at a 3281? The gain bandwith product at 6A is only 3Mhz! A 21194 is 4Mhz at 6A! The SOA of the 21194 at 100V is double that of the 3281! As for beta droop, 400W at 8 ohms is 10A peak.With three pair that is only 3.3A per device.The beta is pretty good out to about 8A or so.Even if you are driving 600W at 4 ohms you are under 6A peak per device.
Agreed, the 2SC3281 bandwidth is as you say, but the datasheet for the MJE3281A indicates 7.5MHz at a Vce of 5V and 45MHz at a Vce of 10V, for an Ic of 6A.
The benefits of the MJL3281A and MJL1302A are not confined to the lack of gain droop at higher current levels. I have recently measured ten of each of these devices and found that the gain varies by less than 6% for the 3281A and 3% for the 1302A over a collector current range of 0.33mA to 100mA. After that, the datasheet graph is realistic (unlike some). Compare this to the wide variatons in gain for some other devices. For example, the measured gain for a TIP35C varies from 1055 to 230 over a collector current range of 100mA to 2.5A, a wide variation with actual gains well in excess of the datasheet graph for a typical device.
The benefits of the MJL3281A and MJL1302A are not confined to the lack of gain droop at higher current levels. I have recently measured ten of each of these devices and found that the gain varies by less than 6% for the 3281A and 3% for the 1302A over a collector current range of 0.33mA to 100mA. After that, the datasheet graph is realistic (unlike some). Compare this to the wide variatons in gain for some other devices. For example, the measured gain for a TIP35C varies from 1055 to 230 over a collector current range of 100mA to 2.5A, a wide variation with actual gains well in excess of the datasheet graph for a typical device.
Toshiba transistors and SITs
The 2SC3281/2SA1302 transistors are actually now obsoleted by Toshiba. They have a whole new family of audio power output transistors:
2SA1943/2SC5200 - 150W
2SA1942/2SC5199 - 120W
2SA1941/2SC5198 - 100W
2SA1940/2SC5197 - 80W
2SA1939/2SC5196 - 60W
All have very linear gain, and virtually no beta droop. Their transistion frequency is only specified at 1A collector current, but it is quite high at 30MHz.
You can get the datasheets off their website, and I believe MCM stocks them.
You might also want to look at the Sanken transistors.
To answer Denis's question about static induction transistors, there are a couple of manufacturers supposedly still making them for military applications, but I've never found a source or even datasheets.
The Japanese audio magazine MJ has run a series of articles over the past year discussing amps using SITs from Tokin. Unfortunately, they have been discontinued. There is a DIY electronics store in Tokyo (Hino Audio) that bought up the last remaining stock and is selling them for about $600/pair.
It's too bad they are so expensive, since they look like an ideal device -- triode characteristics but transistor level currents and power handling. They would be nearly perfect for OTL like applications, but at those prices, I'd be afraid to experiment with them.
If anyone is interested, I can provide copies of the Tokin SIT datasheets and the MJ articles. Unfortunately, its all in Japanese, but the graphs and schematics are fairly self explanatory.
-Jon
The 2SC3281/2SA1302 transistors are actually now obsoleted by Toshiba. They have a whole new family of audio power output transistors:
2SA1943/2SC5200 - 150W
2SA1942/2SC5199 - 120W
2SA1941/2SC5198 - 100W
2SA1940/2SC5197 - 80W
2SA1939/2SC5196 - 60W
All have very linear gain, and virtually no beta droop. Their transistion frequency is only specified at 1A collector current, but it is quite high at 30MHz.
You can get the datasheets off their website, and I believe MCM stocks them.
You might also want to look at the Sanken transistors.
To answer Denis's question about static induction transistors, there are a couple of manufacturers supposedly still making them for military applications, but I've never found a source or even datasheets.
The Japanese audio magazine MJ has run a series of articles over the past year discussing amps using SITs from Tokin. Unfortunately, they have been discontinued. There is a DIY electronics store in Tokyo (Hino Audio) that bought up the last remaining stock and is selling them for about $600/pair.
It's too bad they are so expensive, since they look like an ideal device -- triode characteristics but transistor level currents and power handling. They would be nearly perfect for OTL like applications, but at those prices, I'd be afraid to experiment with them.
If anyone is interested, I can provide copies of the Tokin SIT datasheets and the MJ articles. Unfortunately, its all in Japanese, but the graphs and schematics are fairly self explanatory.
-Jon
Motorola/ON Semi catalog
A check of the ON Semiconductor site shows there is an apparently more rugged version of the MJL1302/3281 in the
MJW series. There is no metal TO-3 device, which sucks for me because I can't substitute them into my Leach amp without
going to new heatsinks.
Their bipolar device catalog is available in .pdf format:
http://www.onsemi.com/pub/Collateral/DL111-D.PDF
It's something like 10 megabytes...
A check of the ON Semiconductor site shows there is an apparently more rugged version of the MJL1302/3281 in the
MJW series. There is no metal TO-3 device, which sucks for me because I can't substitute them into my Leach amp without
going to new heatsinks.
Their bipolar device catalog is available in .pdf format:
http://www.onsemi.com/pub/Collateral/DL111-D.PDF
It's something like 10 megabytes...
"Agreed, the 2SC3281 bandwidth is as you say, but the datasheet for the MJE3281A indicates 7.5MHz at a Vce of 5V and 45MHz at a Vce of 10V, for an Ic of 6A. " The Motorola data book for the MJ3281A, a device built under a technology transfer between Motorola and Toshiba, has a fig.2 showing the current gain-bandwidth product.At 1A it is 30Mhz, it peaks at 50Mhz at 2A, beyond it crashes.With 5V CE at 6A it is 3Mhz. At an operating temperature that would be typical in a large amplifier the current gain is flat as a ruler to about four amps where it begins to fall.At 7A the gain is about half what it is below 4A.The SOA changes to secondary breakdown limited at about 70V.The gain of the MJ21194 from 4A to 7A changes less than 20% and is still good out to 10A.I have about 50pcs of the 1302/3281 on hand.They make a great driver transistor.I think the 21193/21194 makes a better output.I should mention that I am using six pair at +/- 95V.
Motorola, Toshiba and SIT
I too can add that Motorola's datasheet are those I trust most. And the Motorola versions of various widespread bipolar transistors are much improved over their original versions. Let me take for example the famous 3055.
I always prefer to use the metal TO-3 output devices even for the replacements of the plastic ones! Indeed, such substitution is not always possible. In these cases it seems better to replace even the modern Toshiba transistors with Motorola MJL types.
Ft of 30 MHz is not essential for good sound. It is not difficult to get a full-power bandwidth of more than 100 kHz with any type of modern linear bipolar transistors. Otherwise, the nonlinear input capacitance is to be managed first. Neither the bandwidth nor the slew rate are now the limiting factors.
I think SITs are the dead branch of silicon technology. They have virtually no advantages over other silicon devices,and they have poor reliability. They draw current from the signal source just like bipolars, and they are vulnerable to second breakdown. From other hand, they cannot withstand a die temperature of 200C.
Finally, their output characteristics look very undesirable for audio applications, yes!
Even the vacuum triode cannot be considered the ideal audio amplification device by definition. The triode tube has internal negative feedback. It is this feedback that makes the triode a voltage amplifier. Of course, this is a simple way to build a voltage amplifier, because no external feedback is necessary. Indeed, this internal feedback is nonlinear, what is reflected by variation of triode gain factor mu over a whole range of plate currents and voltages. When a triode is used as an output device, this nonlinearity cannot be omitted. The most significant manifestation of triode's gain nonlinearity takes place when a triode amplifier, especially a single-ended one sees a reactive load like a loudspeaker. There is quite simple explanation why those SET amps like so much the sensitive loudspeakers, they operate in small-signal region far from the voltage and current clipping. Othewise, some more ideal amplifiers will make good dynamics at just 10W of 8 Ohm power rating even with 'speakers of medium sensitivity, say, 86 dB/1W.
The SITs have the same drawback of nonlinear internal feedback plus their thermal unstability. Need one say more?
From other hand, a bipolar transistor is, in fact, a voltage driven current source. (Again Douglas Self is right.) This device is described by such parameter as transconductance. What is essential, the small-signal transconductance of any bipolar transistor is solely defined by collector current and chip temperature, and is independent of particular transistor type! What makes the real difference between power transistors, is their SOA. The beta is a parameter of least importance. It is just necessary to have the beta high enough. There are simple and effective circuit topologies which make the transistor circuits virtually insensitive to beta variations. At least, most modern linear transistors have quite narrow spread of this parameter, and it is not the beta variation what is responcible for high-order nonlinearity.
I too can add that Motorola's datasheet are those I trust most. And the Motorola versions of various widespread bipolar transistors are much improved over their original versions. Let me take for example the famous 3055.
I always prefer to use the metal TO-3 output devices even for the replacements of the plastic ones! Indeed, such substitution is not always possible. In these cases it seems better to replace even the modern Toshiba transistors with Motorola MJL types.
Ft of 30 MHz is not essential for good sound. It is not difficult to get a full-power bandwidth of more than 100 kHz with any type of modern linear bipolar transistors. Otherwise, the nonlinear input capacitance is to be managed first. Neither the bandwidth nor the slew rate are now the limiting factors.
I think SITs are the dead branch of silicon technology. They have virtually no advantages over other silicon devices,and they have poor reliability. They draw current from the signal source just like bipolars, and they are vulnerable to second breakdown. From other hand, they cannot withstand a die temperature of 200C.
Finally, their output characteristics look very undesirable for audio applications, yes!
Even the vacuum triode cannot be considered the ideal audio amplification device by definition. The triode tube has internal negative feedback. It is this feedback that makes the triode a voltage amplifier. Of course, this is a simple way to build a voltage amplifier, because no external feedback is necessary. Indeed, this internal feedback is nonlinear, what is reflected by variation of triode gain factor mu over a whole range of plate currents and voltages. When a triode is used as an output device, this nonlinearity cannot be omitted. The most significant manifestation of triode's gain nonlinearity takes place when a triode amplifier, especially a single-ended one sees a reactive load like a loudspeaker. There is quite simple explanation why those SET amps like so much the sensitive loudspeakers, they operate in small-signal region far from the voltage and current clipping. Othewise, some more ideal amplifiers will make good dynamics at just 10W of 8 Ohm power rating even with 'speakers of medium sensitivity, say, 86 dB/1W.
The SITs have the same drawback of nonlinear internal feedback plus their thermal unstability. Need one say more?
From other hand, a bipolar transistor is, in fact, a voltage driven current source. (Again Douglas Self is right.) This device is described by such parameter as transconductance. What is essential, the small-signal transconductance of any bipolar transistor is solely defined by collector current and chip temperature, and is independent of particular transistor type! What makes the real difference between power transistors, is their SOA. The beta is a parameter of least importance. It is just necessary to have the beta high enough. There are simple and effective circuit topologies which make the transistor circuits virtually insensitive to beta variations. At least, most modern linear transistors have quite narrow spread of this parameter, and it is not the beta variation what is responcible for high-order nonlinearity.
IGBT / Forte 4A
Forte 4A was a Threshold product that successfully used IGBTs. Forte 4a's are very reliable, very stable and sound particularly good. With Soderberg's mods a Forte 4A can challenge just about any 50 watt/ ch amp made today in terms of sound quality. I bought a Soderberg'ed Forte 4a for $650 recently. Show me another amp that sounds substantially better at anywhere near this price..... so, Nelson Pass' comments notwithstanding, it apparently IS possible to build a good-sounding amp with certain IGBT's. Bladelius accomplished this with the Forte 4a.
But, of course, the point is well taken: Using IGBTs requires a lot of extra matching to get good sound, so why go through all the trouble of building a good amp with IGBT's when you can use prefectly fine bipolar or FET devices to build a good amp without all that fuss. You don't gain anything by using IGBT's except some sort of "exoticity" - which maybe was seen by some marketing guys in the 1980's as a plus re: IGBTs.
Forte 4A was a Threshold product that successfully used IGBTs. Forte 4a's are very reliable, very stable and sound particularly good. With Soderberg's mods a Forte 4A can challenge just about any 50 watt/ ch amp made today in terms of sound quality. I bought a Soderberg'ed Forte 4a for $650 recently. Show me another amp that sounds substantially better at anywhere near this price..... so, Nelson Pass' comments notwithstanding, it apparently IS possible to build a good-sounding amp with certain IGBT's. Bladelius accomplished this with the Forte 4a.
But, of course, the point is well taken: Using IGBTs requires a lot of extra matching to get good sound, so why go through all the trouble of building a good amp with IGBT's when you can use prefectly fine bipolar or FET devices to build a good amp without all that fuss. You don't gain anything by using IGBT's except some sort of "exoticity" - which maybe was seen by some marketing guys in the 1980's as a plus re: IGBTs.
IGBT Amplifier
The speaker cone moving in the opposite direction is probably caused by speakers being wired 180 degrees out of phase i.e. + to the - and - to the +.
The speaker cone moving in the opposite direction is probably caused by speakers being wired 180 degrees out of phase i.e. + to the - and - to the +.
Forte 4a
By all accounts, a Forte 4a amplifier with the "Soderburg" mods is a 'giant killer' amp. sounding as good as the best, and better than most other amplifiers.
I have one and I LOVE the sound on my Quad ESL-57's. I've recently bought another to use on the midrange panels of my tri-amped Magneplanar MG-3.6's, although I have not put it into service yet.
Forte' was a 'sister brand' of Threshold, meant to offer excellent performance at prices lower than Threshold - as I understand it. The Forte' IGBT amps were designed at the time that Nelson Pass was in charge of engineering, or maybe just after he left, I do not know. I do know he did NOT design these IGBT amps, and in fact does not think that IGBTs are a good choice for audio designs. I think that, in general, his points are valid, but that in the case of the Forte 4a at least, the designer "got it right" - and in spades. (Nelson Pass comments: http://www.diyaudio.com/forums/showthread.php?postid=1153171#post1153171 )
Forte 4a is 50 watts at 8 ohms, and 100 watts at 4 ohms per channel, class A. I don't think it's excellent sound is owed to the presence of IGBT's, I think it stems from a very good design that just happens to use specific IGBT's.
I've never heard a Forte 4a that had any kind of problems in the bass. The bass from my amps is solid, taut, extended and tuneful; the mids are the closest to the 'liquid' mids of an EL-34 amp that I've ever heard from a solid state amp, and the treble is 'smoother' and less 'grainy' than just about any other solid-state amp I've heard, while still offering loads of detail and extension.
I have not heard a 'stock' Forte 4a amp, so I don't know if these good qualities are due solely to the modifications performed by Mr. Soderburg, or if his mods just provide 'more of a an already good thing.'
By all accounts, a Forte 4a amplifier with the "Soderburg" mods is a 'giant killer' amp. sounding as good as the best, and better than most other amplifiers.
I have one and I LOVE the sound on my Quad ESL-57's. I've recently bought another to use on the midrange panels of my tri-amped Magneplanar MG-3.6's, although I have not put it into service yet.
Forte' was a 'sister brand' of Threshold, meant to offer excellent performance at prices lower than Threshold - as I understand it. The Forte' IGBT amps were designed at the time that Nelson Pass was in charge of engineering, or maybe just after he left, I do not know. I do know he did NOT design these IGBT amps, and in fact does not think that IGBTs are a good choice for audio designs. I think that, in general, his points are valid, but that in the case of the Forte 4a at least, the designer "got it right" - and in spades. (Nelson Pass comments: http://www.diyaudio.com/forums/showthread.php?postid=1153171#post1153171 )
Forte 4a is 50 watts at 8 ohms, and 100 watts at 4 ohms per channel, class A. I don't think it's excellent sound is owed to the presence of IGBT's, I think it stems from a very good design that just happens to use specific IGBT's.
I've never heard a Forte 4a that had any kind of problems in the bass. The bass from my amps is solid, taut, extended and tuneful; the mids are the closest to the 'liquid' mids of an EL-34 amp that I've ever heard from a solid state amp, and the treble is 'smoother' and less 'grainy' than just about any other solid-state amp I've heard, while still offering loads of detail and extension.
I have not heard a 'stock' Forte 4a amp, so I don't know if these good qualities are due solely to the modifications performed by Mr. Soderburg, or if his mods just provide 'more of a an already good thing.'
Last edited:
Hi: i have a "stock" Forte 4 amplifier, since new (and stored until some weeks) and i agree with member Milosz about great liquid mids and smoother and grainless treble, but bass is not taut or litle loose compared with other two small power amplifiers like Levinson 29 and Adcom 545II driving a pair or B&w801 series 3. i thinkthat one of Mr. Soderburg mods that improve bass are the caps and fast bridge rectifiers.
IGBT vs MOSFET
For what its worth, I am going to drop a reply in this thread. Originally I was working in another thread I created a few days ago regarding IGBT replacement in an existing amplifier manufactured by a brand that is known for excellent products. Ofc as with everything ells, one can talk about brand attachment and such, but I would not say that is the case here. I just happened to like the spec.
My goal is to “replicate” this amp for my own amusement and for the fun of doing so and turn it into an old school style with through hole components and narrow tolerances.
I notice that this thread is very old, spanning over 15 years.
Mister Nelson Pass made his reply back in 2001, that is 15 years ago. Surely that must matter, for if not, then IGBT’s and MOSFET has seen no evolution since they came out.
Let’s look at an article from 1999
The Insulated Gate Bipolar Transistor (IGBT).
University of Glasgow - Schools - School of Engineering
Structure
Fig.1 shows the structure of a typical n-channel IGBT. All discussion here will be concerned with the n-channel type but p-channel IGBT's can be considered in just the same way.
Operation Blocking Operation
The on/off state of the device is controlled, as in a MOSFET, by the gate voltage VG. If the voltage applied to the gate contact, with respect to the emitter, is less than the threshold voltage Vth then no MOSFET inversion layer is created and the device is turned off. When this is the case, any applied forward voltage will fall across the reversed biased junction J2. The only current to flow will be a small leakage current.
The forward breakdown voltage is therefore determined by the breakdown voltage of this junction. This is an important factor, particularly for power devices where large voltages and currents are being dealt with. The breakdown voltage of the one-sided junction is dependent on the doping of the lower-doped side of the junction, i.e. the n- side. This is because the lower doping results in a wider depletion region and thus a lower maximum electric field in the depletion region. It is for this reason that the n- drift region is doped much lighter than the p-type body region. The device that is being modelled is designed to have a breakdown voltage of 600V.
The n+ buffer layer is often present to prevent the depletion region of junction J2 from extending right to the p bipolar collector. The inclusion of this layer however drastically reduces the reverse blocking capability of the device as this is dependent on the breakdown voltage of junction J3, which is reverse biased under reverse voltage conditions. The benefit of this buffer layer is that it allows the thickness of the drift region to be reduced, thus reducing on-state losses.
On-state Operation
The turning on of the device is achieved by increasing the gate voltage VG so that it is greater than the threshold voltage Vth. This results in an inversion layer forming under the gate which provides a channel linking the source to the drift region of the device. Electrons are then injected from the source into the drift region while at the same time junction J3, which is forward biased, injects holes into the n- doped drift region (Fig.2).
This injection causes conductivity modulation of the drift region where both the electron and hole densities are several orders of magnitude higher than the original n- doping. It is this conductivity modulation which gives the IGBT its low on-state voltage because of the reduced resistance of the drift region. Some of the injected holes will recombine in the drift region, while others will cross the region via drift and diffusion and will reach the junction with the p-type region where they will be collected. The operation of the IGBT can therefore be considered like a wide-base pnp transistor whose base drive current is supplied by the MOSFET current through the channel. A simple equivalent circuit is therefore as shown in Fig.3(a)
Fig.3(b) shows a more complete equivalent circuit which includes the parasitic npn transistor formed by the n+-type MOSFET source, the p-type body region and the n--type drift region. Also shown is the lateral resistance of the p-type region. If the current flowing through this resistance is high enough it will produce a voltage drop that will forward bias the junction with the n+ region turning on the parasitic transistor which forms part of a parasitic thyristor. Once this happens there is a high injection of electrons from the n+ region into the p region and all gate control is lost. This is known as latch up and usually leads to device destruction.
- end of article -
Lets move forward.
http://www.infineon.com/dgdl/choosewisely.pdf?fileId=5546d462533600a40153574048b73edc
IOR article:
IGBT or MOSFET: Choose Wisely by Carl Blake and Chris Bull, International Rectifier
IGBT or MOSFET: Choose Wisely by Carl Blake and Chris Bull, International Rectifier
With the proliferation of choices between MOSFETs and IGBTs, it is becoming increasingly difficult for today’s designer to select the bes t device for their application. Here are a few basic guidelines that will help this decision - making process.
Device Evolution: Bipolar Transistors, MOSFETs and IGBTs
The bipolar transistor was the only “real” power transistor until the MOSFET came along in the 1970’s. The bipolar transistor requires a high base current to turn on, has relatively slow turn - off characteristics (known as current tail), and is liable for thermal runaway due to a negative temperature co - efficient. In addition, the lowest attainable on - state voltage or conduction loss is governed by the collector - emitter saturation voltage V CE(SAT).
The MOSFET, however, is a device that is voltage - and not current - controlled. MOSFETs have a positive temperature co-efficient, stopping thermal runaway. The on-state-resistance has no theoretical limit, hence on-state losses can be far lower. The MOSFET also has a body drain diode, which is particularly useful in dealing with limited free wheeling currents.
All these advantages and the comparative elimination of the current tail soon meant that the MOSFET became the device of choice for power switch designs.
Then in the 1980s the IGBT came along. The IGBT is a cross between the bipolar and MOSFET transistors (see figure 1). The IGBT has the output switching and conduction characteristics of a bipolar transistor but is voltage - controlled like a MOSFET. In general, this means it has the advantages of high-current handling capability of a bipolar with the ease of control of a MOSFET. However, the IGBT still has the disadvantages of a comparatively large current tail and no body drain diode. Early versions of the IGBT are also prone to latch up, but nowadays, this is pretty well eliminated. Another potential problem with some IGBT types is the negative temperature co-efficient, which could lead to thermal runaway and makes
the paralleling of devices hard to effectively achieve. This problem is now being addressed in the latest generations of IGBTs that are based on “non-punch through” (NPT) technology. This technology has the same basic IGBT structure (see Figure 1) but is based on bulk-diffused silicon, rather than the epitaxial material that both IGBTs and MOSFETs have historically used.
MOSFETs and IGBTs: Similar But Different.
When comparing Figures one and two, the MOSFET and IGBT structures look very similar. The basic difference is the add ition of a p substrate beneath the n substrate. The IGBT technology is certainly the device of choice for breakdown voltages above 1000V, while the MOSFET is certainly the device of choice for device breakdown voltages below 250V.
Between 250 to 1000V, there are many technical papers available from manufacturers of these devices, some preferring MOSFETs, some IGBTs. However, choosing between IGBTs and MOSFETs is very application-specific and cost, size, speed and thermal requirements should all be considered.
IGBTs have been the preferred device under these conditions:
• Low duty cycle
• Low frequency (<20kHz)
• Narrow or small line or load variations
• High-voltage applications (>1000V)
• Operation at high junction temperature is allowed (>100°C)
• >5kW output power
MOSFETs are preferred in:
• High frequency applications (>200kHz)
• Wide line or load variations
• Long duty cycles
• Low-voltage applications (<250V)
• < 500W output power
The hard-switched measurements clearly show the lower losses of the MOSFET in their applications. Chart 1 shows that the losses of the IGBT are approximately equal to the losses of an IRFP460 if the switching speed is reduced to 50 kHz. This could allow a smaller IGBT to replace the larger MOSFET in some applications. This was the condition in 1997.
However, the newer lower charge MOSFETs now available lower the losses at high frequency and therefore re-asserted the dominance of MOSFETs in applications using hard switching above 50kHz.
Chart 2 shows the losses in an application using Zero Voltage Switching at 50kHz, 500W, the IGBT losses of 9.5W are higher than the MOSFET losses of 7W at room temperature. When the temperature is raised up to operating conditions however, the conduction losses of the MOSFET rise more quickly than the switching losses of the IGBT. The losses at elevated temperature increase 60% for the MOSFET while the total losses for the IGBT increase only 20%. At 300 watts this makes the power almost equal, while at 500 watts the advantage goes to the IGBT.
Chart 3 shows the losses at 134kHz, 500W, elevated temperature, the IGBT losses of 25.2W are slightly worse than the MOSFET, with total losses of 23.9W. At room temperature in this same application the losses were 17.8 and 15.1 watts respectively. The switching losses are higher at higher frequency which eliminates the advantage of the IGBT at high temperature
, when switching at the lower frequency. This illustrates the subject of this paper, namely there is no iron clad rule which can be used to determine which device will offer the best performance in a specific type of circuit. Depending upon the exact power level, devices being considered, the latest technology available for each type of transistor, the results will change slightly.
Some of the conclusion: Finally, there seems to be an industry wide perception that MOSFETs are a mature product, which will not offer significant performance improvements in applications and IGBTs are a new technology, which will replace MOSFETs in all applications above 300 volts. No such generalizations are ever true and the huge improvements in MOSFET performance over the last two years certainly confirms that MOSFETs are a very dynamic product, and continues a trend of rapid growth. In fact, new low-charge MOSFETs such as the International Rectifier IRFP460A and IRFP22N50A significantly move performance benchmarks just when the IGBTs appear to offer an alternative in hard-switched applications.
Infineon Showcases 650V TRENCHSTOP™ 5 – Performance of est-in-Class IGBT Gains High Customer Demand
Infineon Showcases 650V TRENCHSTOP 5 – Performance of Best-in-Class IGBT Gains High Customer Demand - Infineon Technologies
More:
http://www.infineon.com/dgdl/Infine...N.pdf?fileId=5546d462501ee6fd015023070b8b306d
http://www.infineon.com/dgdl/an-983.pdf?fileId=5546d462533600a40153559f8d921224
http://www.infineon.com/dgdl/an-990.pdf?fileId=5546d462533600a40153559fae19124e
IGBT generation 3,4,5 and 6 have surfaced since late 90's.
In the end what we are looking for is an amplifier that perform well as sound magnificent and the difficult part is what sounds magnificent? That will almost always be a subjective answer. Does a CD sound better than an vinyl record? perhaps, perhaps the CD (digital) is better when it comes to bass notes which vinyl cannot reproduce with the same dynamic range.
Is a tube amplifier worse than a JFET, VFEt or MOSFET amplifier... actually, its different. Efficiensy is only one aspect of things. Some swear by class A amplifiers while some swear by class D.
Some listen to music while some analyse the music. Some don't care about aesthetic aspects while some will pay $10 000 for something that looks good.
My conclusion is that in order to select what I enjoy, I must test both IGBT and MOSFETS, even a Bipolar Transistor.
For what its worth, I am going to drop a reply in this thread. Originally I was working in another thread I created a few days ago regarding IGBT replacement in an existing amplifier manufactured by a brand that is known for excellent products. Ofc as with everything ells, one can talk about brand attachment and such, but I would not say that is the case here. I just happened to like the spec.
My goal is to “replicate” this amp for my own amusement and for the fun of doing so and turn it into an old school style with through hole components and narrow tolerances.
I notice that this thread is very old, spanning over 15 years.
Mister Nelson Pass made his reply back in 2001, that is 15 years ago. Surely that must matter, for if not, then IGBT’s and MOSFET has seen no evolution since they came out.
Let’s look at an article from 1999
The Insulated Gate Bipolar Transistor (IGBT).
University of Glasgow - Schools - School of Engineering
Structure
Fig.1 shows the structure of a typical n-channel IGBT. All discussion here will be concerned with the n-channel type but p-channel IGBT's can be considered in just the same way.
Operation Blocking Operation
The on/off state of the device is controlled, as in a MOSFET, by the gate voltage VG. If the voltage applied to the gate contact, with respect to the emitter, is less than the threshold voltage Vth then no MOSFET inversion layer is created and the device is turned off. When this is the case, any applied forward voltage will fall across the reversed biased junction J2. The only current to flow will be a small leakage current.
The forward breakdown voltage is therefore determined by the breakdown voltage of this junction. This is an important factor, particularly for power devices where large voltages and currents are being dealt with. The breakdown voltage of the one-sided junction is dependent on the doping of the lower-doped side of the junction, i.e. the n- side. This is because the lower doping results in a wider depletion region and thus a lower maximum electric field in the depletion region. It is for this reason that the n- drift region is doped much lighter than the p-type body region. The device that is being modelled is designed to have a breakdown voltage of 600V.
The n+ buffer layer is often present to prevent the depletion region of junction J2 from extending right to the p bipolar collector. The inclusion of this layer however drastically reduces the reverse blocking capability of the device as this is dependent on the breakdown voltage of junction J3, which is reverse biased under reverse voltage conditions. The benefit of this buffer layer is that it allows the thickness of the drift region to be reduced, thus reducing on-state losses.
On-state Operation
The turning on of the device is achieved by increasing the gate voltage VG so that it is greater than the threshold voltage Vth. This results in an inversion layer forming under the gate which provides a channel linking the source to the drift region of the device. Electrons are then injected from the source into the drift region while at the same time junction J3, which is forward biased, injects holes into the n- doped drift region (Fig.2).
This injection causes conductivity modulation of the drift region where both the electron and hole densities are several orders of magnitude higher than the original n- doping. It is this conductivity modulation which gives the IGBT its low on-state voltage because of the reduced resistance of the drift region. Some of the injected holes will recombine in the drift region, while others will cross the region via drift and diffusion and will reach the junction with the p-type region where they will be collected. The operation of the IGBT can therefore be considered like a wide-base pnp transistor whose base drive current is supplied by the MOSFET current through the channel. A simple equivalent circuit is therefore as shown in Fig.3(a)
Fig.3(b) shows a more complete equivalent circuit which includes the parasitic npn transistor formed by the n+-type MOSFET source, the p-type body region and the n--type drift region. Also shown is the lateral resistance of the p-type region. If the current flowing through this resistance is high enough it will produce a voltage drop that will forward bias the junction with the n+ region turning on the parasitic transistor which forms part of a parasitic thyristor. Once this happens there is a high injection of electrons from the n+ region into the p region and all gate control is lost. This is known as latch up and usually leads to device destruction.
- end of article -
Lets move forward.
http://www.infineon.com/dgdl/choosewisely.pdf?fileId=5546d462533600a40153574048b73edc
IOR article:
IGBT or MOSFET: Choose Wisely by Carl Blake and Chris Bull, International Rectifier
IGBT or MOSFET: Choose Wisely by Carl Blake and Chris Bull, International Rectifier
With the proliferation of choices between MOSFETs and IGBTs, it is becoming increasingly difficult for today’s designer to select the bes t device for their application. Here are a few basic guidelines that will help this decision - making process.
Device Evolution: Bipolar Transistors, MOSFETs and IGBTs
The bipolar transistor was the only “real” power transistor until the MOSFET came along in the 1970’s. The bipolar transistor requires a high base current to turn on, has relatively slow turn - off characteristics (known as current tail), and is liable for thermal runaway due to a negative temperature co - efficient. In addition, the lowest attainable on - state voltage or conduction loss is governed by the collector - emitter saturation voltage V CE(SAT).
The MOSFET, however, is a device that is voltage - and not current - controlled. MOSFETs have a positive temperature co-efficient, stopping thermal runaway. The on-state-resistance has no theoretical limit, hence on-state losses can be far lower. The MOSFET also has a body drain diode, which is particularly useful in dealing with limited free wheeling currents.
All these advantages and the comparative elimination of the current tail soon meant that the MOSFET became the device of choice for power switch designs.
Then in the 1980s the IGBT came along. The IGBT is a cross between the bipolar and MOSFET transistors (see figure 1). The IGBT has the output switching and conduction characteristics of a bipolar transistor but is voltage - controlled like a MOSFET. In general, this means it has the advantages of high-current handling capability of a bipolar with the ease of control of a MOSFET. However, the IGBT still has the disadvantages of a comparatively large current tail and no body drain diode. Early versions of the IGBT are also prone to latch up, but nowadays, this is pretty well eliminated. Another potential problem with some IGBT types is the negative temperature co-efficient, which could lead to thermal runaway and makes
the paralleling of devices hard to effectively achieve. This problem is now being addressed in the latest generations of IGBTs that are based on “non-punch through” (NPT) technology. This technology has the same basic IGBT structure (see Figure 1) but is based on bulk-diffused silicon, rather than the epitaxial material that both IGBTs and MOSFETs have historically used.
MOSFETs and IGBTs: Similar But Different.
When comparing Figures one and two, the MOSFET and IGBT structures look very similar. The basic difference is the add ition of a p substrate beneath the n substrate. The IGBT technology is certainly the device of choice for breakdown voltages above 1000V, while the MOSFET is certainly the device of choice for device breakdown voltages below 250V.
Between 250 to 1000V, there are many technical papers available from manufacturers of these devices, some preferring MOSFETs, some IGBTs. However, choosing between IGBTs and MOSFETs is very application-specific and cost, size, speed and thermal requirements should all be considered.
IGBTs have been the preferred device under these conditions:
• Low duty cycle
• Low frequency (<20kHz)
• Narrow or small line or load variations
• High-voltage applications (>1000V)
• Operation at high junction temperature is allowed (>100°C)
• >5kW output power
MOSFETs are preferred in:
• High frequency applications (>200kHz)
• Wide line or load variations
• Long duty cycles
• Low-voltage applications (<250V)
• < 500W output power
The hard-switched measurements clearly show the lower losses of the MOSFET in their applications. Chart 1 shows that the losses of the IGBT are approximately equal to the losses of an IRFP460 if the switching speed is reduced to 50 kHz. This could allow a smaller IGBT to replace the larger MOSFET in some applications. This was the condition in 1997.
However, the newer lower charge MOSFETs now available lower the losses at high frequency and therefore re-asserted the dominance of MOSFETs in applications using hard switching above 50kHz.
Chart 2 shows the losses in an application using Zero Voltage Switching at 50kHz, 500W, the IGBT losses of 9.5W are higher than the MOSFET losses of 7W at room temperature. When the temperature is raised up to operating conditions however, the conduction losses of the MOSFET rise more quickly than the switching losses of the IGBT. The losses at elevated temperature increase 60% for the MOSFET while the total losses for the IGBT increase only 20%. At 300 watts this makes the power almost equal, while at 500 watts the advantage goes to the IGBT.
Chart 3 shows the losses at 134kHz, 500W, elevated temperature, the IGBT losses of 25.2W are slightly worse than the MOSFET, with total losses of 23.9W. At room temperature in this same application the losses were 17.8 and 15.1 watts respectively. The switching losses are higher at higher frequency which eliminates the advantage of the IGBT at high temperature
, when switching at the lower frequency. This illustrates the subject of this paper, namely there is no iron clad rule which can be used to determine which device will offer the best performance in a specific type of circuit. Depending upon the exact power level, devices being considered, the latest technology available for each type of transistor, the results will change slightly.
Some of the conclusion: Finally, there seems to be an industry wide perception that MOSFETs are a mature product, which will not offer significant performance improvements in applications and IGBTs are a new technology, which will replace MOSFETs in all applications above 300 volts. No such generalizations are ever true and the huge improvements in MOSFET performance over the last two years certainly confirms that MOSFETs are a very dynamic product, and continues a trend of rapid growth. In fact, new low-charge MOSFETs such as the International Rectifier IRFP460A and IRFP22N50A significantly move performance benchmarks just when the IGBTs appear to offer an alternative in hard-switched applications.
Infineon Showcases 650V TRENCHSTOP™ 5 – Performance of est-in-Class IGBT Gains High Customer Demand
Infineon Showcases 650V TRENCHSTOP 5 – Performance of Best-in-Class IGBT Gains High Customer Demand - Infineon Technologies
More:
http://www.infineon.com/dgdl/Infine...N.pdf?fileId=5546d462501ee6fd015023070b8b306d
http://www.infineon.com/dgdl/an-983.pdf?fileId=5546d462533600a40153559f8d921224
http://www.infineon.com/dgdl/an-990.pdf?fileId=5546d462533600a40153559fae19124e
IGBT generation 3,4,5 and 6 have surfaced since late 90's.
In the end what we are looking for is an amplifier that perform well as sound magnificent and the difficult part is what sounds magnificent? That will almost always be a subjective answer. Does a CD sound better than an vinyl record? perhaps, perhaps the CD (digital) is better when it comes to bass notes which vinyl cannot reproduce with the same dynamic range.
Is a tube amplifier worse than a JFET, VFEt or MOSFET amplifier... actually, its different. Efficiensy is only one aspect of things. Some swear by class A amplifiers while some swear by class D.
Some listen to music while some analyse the music. Some don't care about aesthetic aspects while some will pay $10 000 for something that looks good.
My conclusion is that in order to select what I enjoy, I must test both IGBT and MOSFETS, even a Bipolar Transistor.
My Elektro 50W IGBT amp
Yes I build the Elektor one, best sounding amp I ever heard as of today.
Check out this thread: https://www.diyaudio.com/forums/solid-state/97551-igbt-amplifiers-6.html#post5978199
And have a look to my project: Public - Google Drive
Now agree this setup is especially sensitive to correct surrounding "micro-architecture": that is power supply, decoupling, Zobel networks, and so much more it works reliably. Once you mastered this result really is astonishing! Others say that.
Chris
Yes I build the Elektor one, best sounding amp I ever heard as of today.
Check out this thread: https://www.diyaudio.com/forums/solid-state/97551-igbt-amplifiers-6.html#post5978199
And have a look to my project: Public - Google Drive
Now agree this setup is especially sensitive to correct surrounding "micro-architecture": that is power supply, decoupling, Zobel networks, and so much more it works reliably. Once you mastered this result really is astonishing! Others say that.
Chris
My general understanding is that IGBTs are inefficient compared to MOSFETs, but much more rugged and reliable at high voltage (where the efficiencies are more comparable anyway), due to the vulnerable gate oxide being shielded from the high voltages of the drain. Think tetrode v. triode. Or it may be more a dV/dt sensitivity issue - high voltages and rapid switching produce large currents through stray capacitances, and there's a significant drain-gate capacitance in a power MOSFET.
I'd suggest MOSFETs have the edge unless you are into exotic voltages, and they have more variants (n/p channel, lateral/verticle, Si/GaN/SiC/??)
I'd suggest MOSFETs have the edge unless you are into exotic voltages, and they have more variants (n/p channel, lateral/verticle, Si/GaN/SiC/??)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.