I did get an answer from Rod Elliot on this, quite some time ago. He says, and I quote:
Doug Self is a protagonist of the "Optimum Class-B" theory. In my experience, increasing bias current reduces distortion - I have *never* seen it increase again as bias is increased.
If the amp is any good to start with, you won't hear any variation in distortion at all. A sinewave is the best test, but most amps have distortion below the audible threshold.
traderbam,
Sorry for the glib reply. This is a great city. You should come visit your brother and spend a day or two hiking or skiing in the Rockies.
Doug Self is a protagonist of the "Optimum Class-B" theory. In my experience, increasing bias current reduces distortion - I have *never* seen it increase again as bias is increased.
If the amp is any good to start with, you won't hear any variation in distortion at all. A sinewave is the best test, but most amps have distortion below the audible threshold.
traderbam,
Sorry for the glib reply. This is a great city. You should come visit your brother and spend a day or two hiking or skiing in the Rockies.
Paul: glib? I didn't think so. I have been to Calgary and Banff several times and it is a beautiful place. You are very lucky.
I wonder what Rod Elliot means by "A sinewave is the best test, but most amps have distortion below the audible threshold". How can it be the best test if most amps cannot be distinguished by using it? Or have I missed something?
I wonder what Rod Elliot means by "A sinewave is the best test, but most amps have distortion below the audible threshold". How can it be the best test if most amps cannot be distinguished by using it? Or have I missed something?
Keep in mind that Rod's comment was a quick email reply. I'm not sure exactly what he meant by sinewave being the best test.
I think Self is correct if the criterion is THD measurement. Measured distortion often increases at higher bias on BJT Class AB amps.
It might be fair to speculate that Rod Elliot's statement refers to subjectively experienced distortion, and of course the statement "often increases" might not apply to his designs.
It might be fair to speculate that Rod Elliot's statement refers to subjectively experienced distortion, and of course the statement "often increases" might not apply to his designs.
Optimum bias point!
Self has measured the optimum bias level in form of emitter voltage in a BJT stage and as was mentioned earlier by MarelvdGit depends on the BJT transfer function. Self comes to the conclusion that about 23mV per emitter resistor is the optimum bias point but it is sligthly dependant on the resistance value.
The best test tone for an amplifier should be a signal in which the level varies all the time and in which the highest frequency is 20kHz. A sine wave does just that. An intermodulation test signal seems even better as it consists of multiple sine waves at different frequencies but the result when measuring IM-distorsion is allmost exclusively related to the THD-distorsion.
Self has measured the optimum bias level in form of emitter voltage in a BJT stage and as was mentioned earlier by MarelvdGit depends on the BJT transfer function. Self comes to the conclusion that about 23mV per emitter resistor is the optimum bias point but it is sligthly dependant on the resistance value.
The best test tone for an amplifier should be a signal in which the level varies all the time and in which the highest frequency is 20kHz. A sine wave does just that. An intermodulation test signal seems even better as it consists of multiple sine waves at different frequencies but the result when measuring IM-distorsion is allmost exclusively related to the THD-distorsion.
Re: Re: The sweet spot!!!
mikek:
would you say more about this:
Does your comment include the bias circuit he uses for the class a design? That one seemed more stable than necessary to me. Maybe I missed something?
mlloyd1
mikek:
would you say more about this:
mikek said:
Does your comment include the bias circuit he uses for the class a design? That one seemed more stable than necessary to me. Maybe I missed something?
mlloyd1
Re: Re: Re: The sweet spot!!!
Hi mlloyd..
Self's class A design does not require thermal compensation for bias stability, as it uses negative feedback for quiescent current control.......
I was refering to his thermal compensation methods for class B amps., which evidently ameliorate the problem of thermal tracking and longterm bias-drift, but cannot completely eliminate it, as this would only be satisfactorily accomplished by the transistor manufacturers incoorparating such a thermal compensating transistor directly on the same silicon real estate as the power transistor....
P.S: Still haven'nt received those error-correction schematics you promised?
mlloyd1 said:mikek:
would you say more about this:
Does your comment include the bias circuit he uses for the class a design? That one seemed more stable than necessary to me. Maybe I missed something?
mlloyd1
Hi mlloyd..
Self's class A design does not require thermal compensation for bias stability, as it uses negative feedback for quiescent current control.......
I was refering to his thermal compensation methods for class B amps., which evidently ameliorate the problem of thermal tracking and longterm bias-drift, but cannot completely eliminate it, as this would only be satisfactorily accomplished by the transistor manufacturers incoorparating such a thermal compensating transistor directly on the same silicon real estate as the power transistor....
P.S: Still haven'nt received those error-correction schematics you promised?
Bias correction circuitry
As mikek mentioned the thermal stability is poor in Selfs designs as well as in any other designs except for some Kenwood amplifier ten years ago where they actually integrated a sensing diode in the transistorchip.
Audio Amateur presented a couple of amplifiers about five years ago where they used an OPAMP to measure and correct the bias current. The problem as we all know is that the bias current can only be measured in the overlap region in which the total voltage over the emitter resistors is the lowest. The circuit was based on a OPAMP which measured the voltage over the two resistors and discharged a capacitor down to the lowest detectable voltage plus a diode drop. The capacitor was then charged through a high value resistor until the next time the voltage came down to this low level again.
When I saw the circuit I had recently read Selfs book and I therefore saw great opportunities in getting the critical bias current stable.
After having worked on the circuit for a while I tested it at lower frequencies and discovered that it didn't work. The time constant had to be so long that the voltage over the capacitor did't increase more than a few millivolts between each discharge pulse which also ment that I was better of with thermal compensation
As I see it the only way of doing this is to use a sample and hold circuit which is triggered with a zero detector. Has anyone seen such a circuit?
As mikek mentioned the thermal stability is poor in Selfs designs as well as in any other designs except for some Kenwood amplifier ten years ago where they actually integrated a sensing diode in the transistorchip.
Audio Amateur presented a couple of amplifiers about five years ago where they used an OPAMP to measure and correct the bias current. The problem as we all know is that the bias current can only be measured in the overlap region in which the total voltage over the emitter resistors is the lowest. The circuit was based on a OPAMP which measured the voltage over the two resistors and discharged a capacitor down to the lowest detectable voltage plus a diode drop. The capacitor was then charged through a high value resistor until the next time the voltage came down to this low level again.
When I saw the circuit I had recently read Selfs book and I therefore saw great opportunities in getting the critical bias current stable.
After having worked on the circuit for a while I tested it at lower frequencies and discovered that it didn't work. The time constant had to be so long that the voltage over the capacitor did't increase more than a few millivolts between each discharge pulse which also ment that I was better of with thermal compensation
As I see it the only way of doing this is to use a sample and hold circuit which is triggered with a zero detector. Has anyone seen such a circuit?
The easiest way to measure bias in a complementary follower during operation is to measure the emitter to emitter voltage (the voltage across both emitter resistors in series) and take the minimum value. A minimum-hold circuit (opposite of a peak-hold) will put out a usable value for driving an opto-isolator used as a bias transistor.
If the circuit is not a complementary follower, this still can be done, but becomes slightly more complicated in the sensing.
I produced such a circuit for Threshold on my way out the door, and it was shown at CES, but never produced.
If the circuit is not a complementary follower, this still can be done, but becomes slightly more complicated in the sensing.
I produced such a circuit for Threshold on my way out the door, and it was shown at CES, but never produced.
There are at least two ways to prevent quiescent current stability
problems without sample-and-hold-like circuits and without a sensing diode on the same die as the output transistor.
One way is current dumping, where a low-current class A and a high-current class C stage are combined in a very ingenious error correction scheme, which has been used commercially in the QUAD solid-state amplifiers since 1976 or so. The patent on this technique has passed its due by date, so everyone is free to use it, also commercially.
The second way is by using a non-linear common-mode loop. As
far as I know, it was invented by Johan H. Huijsing and Frans Tol in 1976, it is frequently used in operational amplifiers, but rarely in audio power amplifiers. The Philips TDA1514A power amplifier IC is an exception; it uses a harmonic mean loop.
Build a non-linear circuit which senses the currents through both output transistors and which generates an output signal that depends mainly on the smallest of the two output device currents. That is, the non-linear circuit should be a smooth approximation to a minimum selector. Then make an additional feedback loop that increases or decreases the currents
through both output devices until the output of the non-linear network equals a reference value. This kind of loop is known as a class AB bias loop or non-linear common-mode loop.
For example, if the non-linear network calculates the harmonic mean of the output device currents (2*I1*I2/(I1+I2)), and the feedback loop makes this equal to Iref, then in the quiescent point, I1=I2=Iref. When the output current of the amplifier is much greater than the quiescent current, one output device will conduct a large current defined mainly by the output voltage and load impedance, while the non-linear common-mode loop makes
the current through the other output transistor approach 0.5*Iref.
Therefore, the output stage automatically becomes non-switching.
Making a harmonic mean circuit with discrete transistors or with
cheap transistor arrays is difficult, but there are other non-linear
functions which give similar performance. For example, you can easily approximate an e^(-I1*R*q/kT)+e^(-I2*R*q/kT) function with a couple of small resistors and two matched small-signal transistor pairs. This is actually what I use in my power amplifier.
Coming back to the conventional complementary emitter follower,
the factor ln(2) in my previous post is a mistake. It can be shown
that when the bias voltage across each emitter resistor equals kT/q in the quiescent point, and the transistors behave according to the exponential law Ic=Is*exp(Vbe*q/(kT)), then the transconductance in the quiescent point equals the transconductance when one output device is switched off and the other conducts a large (infinite, in theory) current. This should be pretty close to the optimum. Indeed, kT/q~=26mV at room temperature, which is close to Self's 23mV.
problems without sample-and-hold-like circuits and without a sensing diode on the same die as the output transistor.
One way is current dumping, where a low-current class A and a high-current class C stage are combined in a very ingenious error correction scheme, which has been used commercially in the QUAD solid-state amplifiers since 1976 or so. The patent on this technique has passed its due by date, so everyone is free to use it, also commercially.
The second way is by using a non-linear common-mode loop. As
far as I know, it was invented by Johan H. Huijsing and Frans Tol in 1976, it is frequently used in operational amplifiers, but rarely in audio power amplifiers. The Philips TDA1514A power amplifier IC is an exception; it uses a harmonic mean loop.
Build a non-linear circuit which senses the currents through both output transistors and which generates an output signal that depends mainly on the smallest of the two output device currents. That is, the non-linear circuit should be a smooth approximation to a minimum selector. Then make an additional feedback loop that increases or decreases the currents
through both output devices until the output of the non-linear network equals a reference value. This kind of loop is known as a class AB bias loop or non-linear common-mode loop.
For example, if the non-linear network calculates the harmonic mean of the output device currents (2*I1*I2/(I1+I2)), and the feedback loop makes this equal to Iref, then in the quiescent point, I1=I2=Iref. When the output current of the amplifier is much greater than the quiescent current, one output device will conduct a large current defined mainly by the output voltage and load impedance, while the non-linear common-mode loop makes
the current through the other output transistor approach 0.5*Iref.
Therefore, the output stage automatically becomes non-switching.
Making a harmonic mean circuit with discrete transistors or with
cheap transistor arrays is difficult, but there are other non-linear
functions which give similar performance. For example, you can easily approximate an e^(-I1*R*q/kT)+e^(-I2*R*q/kT) function with a couple of small resistors and two matched small-signal transistor pairs. This is actually what I use in my power amplifier.
Coming back to the conventional complementary emitter follower,
the factor ln(2) in my previous post is a mistake. It can be shown
that when the bias voltage across each emitter resistor equals kT/q in the quiescent point, and the transistors behave according to the exponential law Ic=Is*exp(Vbe*q/(kT)), then the transconductance in the quiescent point equals the transconductance when one output device is switched off and the other conducts a large (infinite, in theory) current. This should be pretty close to the optimum. Indeed, kT/q~=26mV at room temperature, which is close to Self's 23mV.
thermal tracking in class a AB MODE
this is a quote from the authers of the class AB amp.An important problem concerns the bias control loop . often acomplementary common-collecter output stage is used and the power transistors are included in the bias control loop. This can easily cause thermal instability due to the larte temperature variations in the ouput transistors.
Thermal coupling of all diodes and transistors in the class-AB control loop can improve the thermal stability of the quiescent current in the output stag, bu this is in most cases too slow to react to burst signals. As a result, emitter resistors are usually added to the power transistors to improve thermal stability. However, the voltage drop across the emitter resistors can switch off the transistor that is conducting the resisdual current.
Because of the limited bandwidth of the distortion reduction by means of using negative feedback, transistor switching can be a source of high-frequency distortion. Additional circuitry is ncessary to prevent this. Moreover a common-collector stage is not able to reach a rail to rail voltage output swing due to the base-emitter voltages.
Common-emitter output stages are usually based on aa complementary feedback pari. However the local feedback loop around the pair can be a sources of HF oscillation.
In order to achieve thermal stability without switching problems, and to allow maximum output voltage swing, we designed a common-emitter power amplifer based on a new current-mode class-AB driver circuit. Due to the absence of local feedback at the output, the stability of theamplifier is only dependent on the global feedback-loop. end quote. I hope this is of use this c omes from electronics world dec 1999, the schematic is in the thread I quoted anyone interested in the full article let me know.
this is a quote from the authers of the class AB amp.An important problem concerns the bias control loop . often acomplementary common-collecter output stage is used and the power transistors are included in the bias control loop. This can easily cause thermal instability due to the larte temperature variations in the ouput transistors.
Thermal coupling of all diodes and transistors in the class-AB control loop can improve the thermal stability of the quiescent current in the output stag, bu this is in most cases too slow to react to burst signals. As a result, emitter resistors are usually added to the power transistors to improve thermal stability. However, the voltage drop across the emitter resistors can switch off the transistor that is conducting the resisdual current.
Because of the limited bandwidth of the distortion reduction by means of using negative feedback, transistor switching can be a source of high-frequency distortion. Additional circuitry is ncessary to prevent this. Moreover a common-collector stage is not able to reach a rail to rail voltage output swing due to the base-emitter voltages.
Common-emitter output stages are usually based on aa complementary feedback pari. However the local feedback loop around the pair can be a sources of HF oscillation.
In order to achieve thermal stability without switching problems, and to allow maximum output voltage swing, we designed a common-emitter power amplifer based on a new current-mode class-AB driver circuit. Due to the absence of local feedback at the output, the stability of theamplifier is only dependent on the global feedback-loop. end quote. I hope this is of use this c omes from electronics world dec 1999, the schematic is in the thread I quoted anyone interested in the full article let me know.
where does the kT/e formula come from?
can someone explain it to me?
I've never seen a such formula for transistors, but something like this for diodes (but BJT are in fact made of diodes...)
Is there something leki this for mosfets?
Bricolo - that likes designing amps according to maths 🙂
can someone explain it to me?
I've never seen a such formula for transistors, but something like this for diodes (but BJT are in fact made of diodes...)
Is there something leki this for mosfets?
Bricolo - that likes designing amps according to maths 🙂
Temperature compensation
Is anyone aware of the commercial use of the Sanken output darlingtons which incorporate a on-die diode for the purpose of temperature sensing? This should allow a near perfect bias circuit to be constructed without all of the thermal lag issues that Self investigated and modelled over several pages of his book.
A (short!) datasheet can be found here:
http://www.sanken-ele.co.jp/CGI/en/...pdf_name=sap16ne.pdf&s_name=SAP16N&gen_big=TR
James
Is anyone aware of the commercial use of the Sanken output darlingtons which incorporate a on-die diode for the purpose of temperature sensing? This should allow a near perfect bias circuit to be constructed without all of the thermal lag issues that Self investigated and modelled over several pages of his book.
A (short!) datasheet can be found here:
http://www.sanken-ele.co.jp/CGI/en/...pdf_name=sap16ne.pdf&s_name=SAP16N&gen_big=TR
James
No need for temperature compensation
May I propose to your attention our last work on class AB amplifiers, which does not required temperature compensation? Being returned from AES convention, I have a small amount of cd-roms with our preprint “Error Correction in Class AB Power Amplifiers”. Could send by snail-mail to interested parties. Comments are welcomed
🙂
May I mention to anyone who shows an interest that our design ideas are free from patent protection and free for anyone to use.
May I propose to your attention our last work on class AB amplifiers, which does not required temperature compensation? Being returned from AES convention, I have a small amount of cd-roms with our preprint “Error Correction in Class AB Power Amplifiers”. Could send by snail-mail to interested parties. Comments are welcomed
🙂
May I mention to anyone who shows an interest that our design ideas are free from patent protection and free for anyone to use.
Re: No need for temperature compensation
ampman...where is this hallowed schematic...please?
Hi dimitri...would love copy of preprint...can you send pdf by e-mail? get in touch.....cheers.
AMPMAN said:
ampman...where is this hallowed schematic...please?
dimitri said:
Hi dimitri...would love copy of preprint...can you send pdf by e-mail? get in touch.....cheers.
Nice transistor!
nemestra
That was a nice component, it is wonderful how information spreads throughout the world by enthusiasts😉
It seems as the easiest way of making the amplifier bias stable is to use such a component.
Error correction has though got other advantages, we are working on a transparent amplifier on a forum in Sweden. You can see some of the results here.output stages
I designed an error correction circuit which only corrects for AC-errors. It is much easier getting it to work properly without lowering the gain if it is AC-coupled.
nemestra
That was a nice component, it is wonderful how information spreads throughout the world by enthusiasts😉
It seems as the easiest way of making the amplifier bias stable is to use such a component.
Error correction has though got other advantages, we are working on a transparent amplifier on a forum in Sweden. You can see some of the results here.output stages
I designed an error correction circuit which only corrects for AC-errors. It is much easier getting it to work properly without lowering the gain if it is AC-coupled.
mikek
the schematic is in the solid state section, go into a search for ampman or alaskan audio where we discussed the merits of output impedance.If you need the full article showing the osciliscope wave forms and a far more detailed description ,email me and Iwill send them. you will have trouble finding the output transistors but you can substitute toshiba 2sc5200 and 2sa1943
the schematic is in the solid state section, go into a search for ampman or alaskan audio where we discussed the merits of output impedance.If you need the full article showing the osciliscope wave forms and a far more detailed description ,email me and Iwill send them. you will have trouble finding the output transistors but you can substitute toshiba 2sc5200 and 2sa1943
No need for temperature compensation Post #35
No need for temperature compensation Post #35
May I propose to your attention our last work on class AB amplifiers, which does not required temperature compensation? Being returned from AES convention, I have a small amount of cd-roms with our preprint “Error Correction in Class AB Power Amplifiers”. Could send by snail-mail to interested parties. Comments are welcomed
May I mention to anyone who shows an interest that our design ideas are free from patent protection and free for anyone to use.
I'm interested 😎 I'd love to know more about it, perhaps the article can be e-mailed ?
No need for temperature compensation Post #35
May I propose to your attention our last work on class AB amplifiers, which does not required temperature compensation? Being returned from AES convention, I have a small amount of cd-roms with our preprint “Error Correction in Class AB Power Amplifiers”. Could send by snail-mail to interested parties. Comments are welcomed
May I mention to anyone who shows an interest that our design ideas are free from patent protection and free for anyone to use.
I'm interested 😎 I'd love to know more about it, perhaps the article can be e-mailed ?
- Status
- Not open for further replies.