It seems to me that this thread has drifted (from time to time) into how transistor selection can alter the sound of a device.
Bill Waslow (Liberty Instruments) has published a free run Windows application (Audio DiffMaker) that claims to be able to isolate audible differences produced by changes that do not also change frequency response.
I planned to use Audio DiffMaker for a different project but could not because of the applications sensitivity to frequency response differences. Since transistor changes in complex amplifier circuits should produce little difference in frequency response, the Audio DiffMaker application should work in this application.
Just another possible tool to use for your experiments.
Mark
http://www.libinst.com/Audio DiffMaker.htm
Bill Waslow (Liberty Instruments) has published a free run Windows application (Audio DiffMaker) that claims to be able to isolate audible differences produced by changes that do not also change frequency response.
I planned to use Audio DiffMaker for a different project but could not because of the applications sensitivity to frequency response differences. Since transistor changes in complex amplifier circuits should produce little difference in frequency response, the Audio DiffMaker application should work in this application.
Just another possible tool to use for your experiments.
Mark
http://www.libinst.com/Audio DiffMaker.htm
jacco vermeulen said:
Hi Jacco,
I'm a bit confused by your comment. I just bought 25 Fairchild SFH9240 vertical MOSFETs for less than $3.00 each. I'm certainly not under the impression that vertical MOSFETs are more expensive than decent BJTs. Is it possible that you are confusing this with the cost of Lateral MOSFETs?
Cheers,
Bob
MarkMcK said:
Hello Mark,
That's a VERY interesting little gem! Thanks for the heads-up.
Jan Didden
Mr Cordell,
think paradox.
At the time when i assembled your EC-amp, the devices were more costly per watt overhere than BJT power transistors. I'm supposing the same applied in the US region, in my familiar naive fashion.
Practically everyone seems to be using them nowadays because the vertical mosfets are cheap compared to big BJT devices, but most are using them in the wrong application, from my perspective anyway.
The way Mr P has chosen to use them from day1 eliminates nearly all items of your list, in high bias Class AB all the measures you mentioned apply.
Which boils down to: not for everything, not for everybody.
Can't help myself from continueing to think in market and monetary perspectives, mea culpa.
2nd to head-nod for the Diffmaker SW-tip
think paradox.
At the time when i assembled your EC-amp, the devices were more costly per watt overhere than BJT power transistors. I'm supposing the same applied in the US region, in my familiar naive fashion.
Practically everyone seems to be using them nowadays because the vertical mosfets are cheap compared to big BJT devices, but most are using them in the wrong application, from my perspective anyway.
The way Mr P has chosen to use them from day1 eliminates nearly all items of your list, in high bias Class AB all the measures you mentioned apply.
Which boils down to: not for everything, not for everybody.
Can't help myself from continueing to think in market and monetary perspectives, mea culpa.
2nd to head-nod for the Diffmaker SW-tip
lumanauw said:
Here is one approach that I use. I pick a peak current beyond which I will not allow the MOSFET to perform. In a 100W, 8-ohm amplifier using two pairs of MOSFETs this might be 12 Amperes for 1 ms (per MOSFET). I use source resistors with each MOSFET. Depending on application details, these may be on the order of 0.22 to 0.6 ohms. I sense the current across the source resistor. If the current exceeds 12 A for 1 ms (well within the SOA for an IRFP240 at 50V drop from rail), I trigger a latching circuit that kills gate drive to both top and bottom MOSFETs. This disables the amplifier and its output stage until power is cycled. The latching circuit clamps both ends of the VAS bias spreader to the amp output node. The VAS idle current passes through the latch circuit and keeps it energized. This circuit can be implemented with six or fewer transistors, or may incorporate opto-isolators or Photo-MOS switches. There are many ways to implement the function electronically in a non-intrusive way.
Note that this is NOT an I-V SOA limiter. When it triggers, the amplifier shuts down. However, the circuit can incorporate elements that modify the current trigger point in accordance with output voltage to advantage.
Obviously, there are a lot of details that must be considered for any particular amplifier design and architecture, but this is a simple description of the general approach.
Cheers,
Bob
Hi Mark,
"Audio DiffMaker" looks like a very interesting application. It may be able to answer some questions.
A big Thank You!
Is Bill Waslow a member here? He might enjoy some of the goings on in this place. It would also increase exposure to his products.
-Chris
"Audio DiffMaker" looks like a very interesting application. It may be able to answer some questions.
A big Thank You!
Is Bill Waslow a member here? He might enjoy some of the goings on in this place. It would also increase exposure to his products.
-Chris
Hi all
having designed my best amps with bipolars, here's my list of things they could do better:
Input capacitance
Although they only need low drive voltages, the input capacitance can be huge. This makes it necessary to be able to heave current into and out of the base for high speed. Circuits using slow-mo MJ802's etc even with capacitors connected across the bases often dont have enough drive to switch them quickly enough. Cue transient distortion.
Gain linearity
The old 2N3055 from RCA was a popular tranny in the 70's. Todays device still has a high gain roll-off after about 500 mA, so distortion is never very low unless you have high open loop gain. I don't agree with those who think the 2N3055 is "past it" though because the latest devices are built on an epi process which is several times faster than RCA's. This allows more feedback and <.01% distortion is quite possible in a 3055 amp!
The new RET's are just brilliant until you fall off the ft cliff. Stability issues may arise if you don't use the Miller slugged slew rate because the ft changes so much with current. Small signal (a.c. analysis) simulations just don't get this.
Second breakdown
OK the RCA devices set the standard for a while. But no-one liked the sound for true hifi. Today's RETs exceed the original 3055 by at least a factor of 2 on SOA, have higher fT by a factor of 30 and gain linearity which is almost perfect. But they still have power limitations above about 80V. Even so the MJ21193/4 outperform RCA's 2N3773 which was a snail at 200 kHz ft - almost germanium class. Which means that MOSFETS may be better at the 1kW amps than bipolars.
When will we see silicon carbide BJT's?
Do we need thermal tracs?
Not sure. With high impedance drive (see my earlier remarks) it is possible to achieve virtually zero crossover distortion and you only need to make sure that transients never cause the Iq to cut off. So running at relatively high Iq - like MOSFETS in practice- means no crossover distortion., and the high Iq allows quite a bit of leeway in variability.
It's all down to how well you want to design. I've no doubt commercially TT's will be popular - it might make simpler circuits easier to achieve better performance.
cheers
John
having designed my best amps with bipolars, here's my list of things they could do better:
Input capacitance
Although they only need low drive voltages, the input capacitance can be huge. This makes it necessary to be able to heave current into and out of the base for high speed. Circuits using slow-mo MJ802's etc even with capacitors connected across the bases often dont have enough drive to switch them quickly enough. Cue transient distortion.
Gain linearity
The old 2N3055 from RCA was a popular tranny in the 70's. Todays device still has a high gain roll-off after about 500 mA, so distortion is never very low unless you have high open loop gain. I don't agree with those who think the 2N3055 is "past it" though because the latest devices are built on an epi process which is several times faster than RCA's. This allows more feedback and <.01% distortion is quite possible in a 3055 amp!
The new RET's are just brilliant until you fall off the ft cliff. Stability issues may arise if you don't use the Miller slugged slew rate because the ft changes so much with current. Small signal (a.c. analysis) simulations just don't get this.
Second breakdown
OK the RCA devices set the standard for a while. But no-one liked the sound for true hifi. Today's RETs exceed the original 3055 by at least a factor of 2 on SOA, have higher fT by a factor of 30 and gain linearity which is almost perfect. But they still have power limitations above about 80V. Even so the MJ21193/4 outperform RCA's 2N3773 which was a snail at 200 kHz ft - almost germanium class. Which means that MOSFETS may be better at the 1kW amps than bipolars.
When will we see silicon carbide BJT's?
Do we need thermal tracs?
Not sure. With high impedance drive (see my earlier remarks) it is possible to achieve virtually zero crossover distortion and you only need to make sure that transients never cause the Iq to cut off. So running at relatively high Iq - like MOSFETS in practice- means no crossover distortion., and the high Iq allows quite a bit of leeway in variability.
It's all down to how well you want to design. I've no doubt commercially TT's will be popular - it might make simpler circuits easier to achieve better performance.
cheers
John
Take a look:
http://www.transic.com/
And here for IGBTs:
http://www.cree.com/products/pdf/CID150660.A.pdf
The main focus is currently on high power switching devices, the target being something like 1200V/100A. Silicon carbide is good for power devices because has high intrinsic breakdown voltage, good thermal conductivity, low reverse currents (and a low thermal coefficient, therefore devices are able to run much hotter than the silicon counterparts), high speed. Except for the last, I doubt there's any major advantage for audio applications. Silicon carbide bipolar devices can be built to have a negative temperature coefficient of the current gain, although this is normally at high currents.
Hi, John Ellis,
A bit off topic. I'm looking for RCA transistor manual from 1974, where there are examples of how to use RCA transistors for making audio gears. Do you have this book?
A bit off topic. I'm looking for RCA transistor manual from 1974, where there are examples of how to use RCA transistors for making audio gears. Do you have this book?
Hi lumanauw,
I've seen it. There were also project books on audio amplifiers that RCA put out to promote their products.
Check in the rear of the book, possibly they did like the tube manuals.
-Chris
I've seen it. There were also project books on audio amplifiers that RCA put out to promote their products.
Check in the rear of the book, possibly they did like the tube manuals.
-Chris
john_ellis said:
Thanks, John.
These ara all very good points about the limitations of BJTs. The ft drop-off of the RETs can be an especially nasty surprize, when you think you've bought a 30 MHz or 60 MHz ft device. It is especially concerning that it drops off at low collector-emitter voltage on top of high current. The Sanken spec sheet only shows ft at 12V, and its drop-off at high current is bad enough. But this is not the whole story. Now look at the On Semi ThermalTrak BJT data sheet. These are similar devices. They show ft vs Ic at both 10V and 5V. The 5V number is much worse, and I assume this will similarly be the case with the Sankens.
The answer here is to (1) run a number of devices in parallel to reduce the peak current seen by each device to the order of perhaps 3 amps. (2) don't count on decent performance within less than 5-10V of the rail; (3) use compensation that is conservative enough to ensure unconditional stability, even under the degraded output transistor ft conditions of simultaneous high current and low Vce; (4) use a high-current driver that can deal with the high total Ccb that you have now created by paralleling output devices; (5) use a high-current driver that can cope with the high dynamic base current demands for minority carrier sweep-out created under the degraded ft conditions.
Note that this underlines the wisdom of John Curl using fully nine pairs of output devices in his JC-1.
Cheers,
Bob
Bob Cordell said:
Now that you mention it, this is an issue with Mosfets also, as the
capacitance increases fairly dramatically at low D-S voltages.
😎
lumanauw said:
I studied those as a kid, earlier versions than 1974, I have the 1969 version here, and my brother has an ever older version with mainly Germanium based reference designs in the back.
The only good thing in the 1969 version is a 70W quasi comp amp that is clearly the basis for the Harman Kardon Citation 12 amp. What they mainly did was to increase the input coupling cap, and feedback cap to extend the LF response. They used some better transistors and a dual in the diff amp in the Citation. The Citation is a better design and you can find the schematic online.
I can understand you wanting it for historical reasons but I don't know where to get it.
Pete B.
Nelson Pass said:
Hi Nelson,
This is a good point, although it is more to the point of Ccb and Cgd than it is to ft. In MOSFETs, it is the Cgd that increases rather dramatically as Vgd goes toward zero. However, if you look at the Ccb of BJT RETs, it is pretty much just as bad as Vcb heads toward zero.
Cheers,
Bob
Overcurrent Protection Scheme
Here is one approach that I use. I pick a peak current beyond which I will not allow the MOSFET to perform. In a 100W, 8-ohm amplifier using two pairs of MOSFETs this might be 12 Amperes for 1 ms (per MOSFET). I use source resistors with each MOSFET. Depending on application details, these may be on the order of 0.22 to 0.6 ohms. I sense the current across the source resistor. If the current exceeds 12 A for 1 ms (well within the SOA for an IRFP240 at 50V drop from rail), I trigger a latching circuit that kills gate drive to both top and bottom MOSFETs. This disables the amplifier and its output stage until power is cycled. The latching circuit clamps both ends of the VAS bias spreader to the amp output node. The VAS idle current passes through the latch circuit and keeps it energized. This circuit can be implemented with six or fewer transistors, or may incorporate opto-isolators or Photo-MOS switches. There are many ways to implement the function electronically in a non-intrusive way
Bob,
This is very similar to the scheme I plan to incorporate in the EC amplifier I am designing. The only difference is the use 0.05 ohm current sense resistors on the drain side of the paralleled devices, so the current monitored represents the sum through all positive or negative output devices. The use of 0.2 ohm source resistors eliminates any current hogging, so monitoring the total positive and negative current is sufficient.
FET relays (turnoff time ~ 100 us) are used between the driver stage and the output stage. If an overcurrent condition (>60 amps) is encountered on either the positive or negative side the drive is removed from both the positive and negative output devices by turning off the relays. Hysteresis is built into the overcurrent circuits so that the outputs remain turned off until power is removed and reapplied. Any nonlinearity introduced by the FET relay is mitigated by use of error correction.
I initially designed an output protection circuit with foldback limiting. However, it tended to oscillate in the presence of a low frequency overload signal. So, to be on the safe side, I opted for the more conservative approach of simply disabling the output if an overcurrent condition is detected. As such, the amplifier can drive full swing transient currents into loads as low as 1.5 ohm loads without tripping the overcurrent circuit.
Here is one approach that I use. I pick a peak current beyond which I will not allow the MOSFET to perform. In a 100W, 8-ohm amplifier using two pairs of MOSFETs this might be 12 Amperes for 1 ms (per MOSFET). I use source resistors with each MOSFET. Depending on application details, these may be on the order of 0.22 to 0.6 ohms. I sense the current across the source resistor. If the current exceeds 12 A for 1 ms (well within the SOA for an IRFP240 at 50V drop from rail), I trigger a latching circuit that kills gate drive to both top and bottom MOSFETs. This disables the amplifier and its output stage until power is cycled. The latching circuit clamps both ends of the VAS bias spreader to the amp output node. The VAS idle current passes through the latch circuit and keeps it energized. This circuit can be implemented with six or fewer transistors, or may incorporate opto-isolators or Photo-MOS switches. There are many ways to implement the function electronically in a non-intrusive way
Bob,
This is very similar to the scheme I plan to incorporate in the EC amplifier I am designing. The only difference is the use 0.05 ohm current sense resistors on the drain side of the paralleled devices, so the current monitored represents the sum through all positive or negative output devices. The use of 0.2 ohm source resistors eliminates any current hogging, so monitoring the total positive and negative current is sufficient.
FET relays (turnoff time ~ 100 us) are used between the driver stage and the output stage. If an overcurrent condition (>60 amps) is encountered on either the positive or negative side the drive is removed from both the positive and negative output devices by turning off the relays. Hysteresis is built into the overcurrent circuits so that the outputs remain turned off until power is removed and reapplied. Any nonlinearity introduced by the FET relay is mitigated by use of error correction.
I initially designed an output protection circuit with foldback limiting. However, it tended to oscillate in the presence of a low frequency overload signal. So, to be on the safe side, I opted for the more conservative approach of simply disabling the output if an overcurrent condition is detected. As such, the amplifier can drive full swing transient currents into loads as low as 1.5 ohm loads without tripping the overcurrent circuit.
Re: Overcurrent Protection Scheme
Some nice ideas here!
I was being a bit lazy in looking at the source current of just one device.
Cheers,
Bob
analog_guy said:
Some nice ideas here!
I was being a bit lazy in looking at the source current of just one device.
Cheers,
Bob
Well, at least you folks are finally thinking strongly about Mosfet output protection. Zener diodes are nice for gate protection, but they are not enough. Circuit breakers that 'might' work for bipolar transistors, aren't fast enough. And so it goes!
This was my original premise about using vertical mosfet output devices. They need essentially about the same amount of protection as bipolars.
I believe this 'myth' was started semiconductor manufacturers who wanted to open a new market and did not have enough experience with audio power amps. Later, they all quietly backed off and left us with the 'myth'.
Now please don't get me wrong. I LOVE FETS, and I would make each and every one of my products with 100% fets, if it were practical and possible. I just know better.
This was my original premise about using vertical mosfet output devices. They need essentially about the same amount of protection as bipolars.
I believe this 'myth' was started semiconductor manufacturers who wanted to open a new market and did not have enough experience with audio power amps. Later, they all quietly backed off and left us with the 'myth'.
Now please don't get me wrong. I LOVE FETS, and I would make each and every one of my products with 100% fets, if it were practical and possible. I just know better.
john curl said:
Hi John,
I really hate it when you say "you folks". It is so condescending. Most of the people on this board know a heck of a lot more than you give them credit for.
No, I think you are just finally listening. The need for MOSFET protection has been discussed before. I have always pointed out the need for MOSFET output short circuit protection. The technique I described I have been using for many years.
The "myth" you refer to probably had legitimate origins with the use of Lateral MOSFETs. They are fundamentally more forgiving of short circuits, since their performance essentially collapses. Vertical MOSFETs, since they will readily conduct humongous amounts of current when asked to do so, are no so forgiving.
It IS practical and possible to build outstanding amplifiers with MOSFETs (including at high power), and there are numerous examples. I frankly don't think you "know better".
As I have said many times, you have made many many excellent bipolar amplifiers, but I do not see you as an expert on the application of MOSFETs to audio power amplifiers.
Bob