Lateral FETs cannot be reasonably driven by a current source type class A driver. In switching applications one has to provide
around 12 amperes peak to charge or discharge the very large gate source capacitance of paralleled lateral FETs.
Such an amp would only produce dynamic distortion as it would act as an integrating amplifier.
Current-driven BJTs are certainly the better choice.
around 12 amperes peak to charge or discharge the very large gate source capacitance of paralleled lateral FETs.
Such an amp would only produce dynamic distortion as it would act as an integrating amplifier.
Current-driven BJTs are certainly the better choice.
Lumba Ogir said:
Certainly there is a relation otherwise it were random ...
FETs are only faster if driven from a voltage source having any
fast enough current delivering capability. In this case of current dumping BJTs are clearly superior to FETs because the driver is
class A. and hence we have current source. As said the Quad equations hold only for static case they do not cover dynamic behavior.
Perhaps I'm misunderstanding the context of the comments about vertical or lateral mosfets not being able to be driven by a current source. Certainly, in a switching application, when the object is to switch as fast as possible and avoid the linear region, large drive currents are desired. In those applications it is the gate drain miller capacitance that is the determining factor. This is well documented in switchmode power supply design notes.
In audio applications as a source follower, it is the smaller gate source capacitance that needs to be considered, and miller capacitance is not a factor. Erno Borbely submitted a current driven 60 watt amplifier in Audio Amateur magazine many years ago that featured vertical MOSFETs with a current drive stage. His article had a very detailed analysis of the FET drive requirements in a linear application. I have built that amplifier and several variations of my own design that feature current drive and can verify that with a sufficient drive current, the MOSFETs operate without issue and sound great too.
The drive stage must be capable of providing sufficient drive current for the required slew rate which can be determined by the capacitance, the voltage swing and the frequency.
i drive = 2* pi * freq * C in * V drive 0-pk
The p channel devices have an input capacitance of 900 pF for a net of 3600 pF with 4 devices. In this application the current sink provides 55 mA of drive current. Therefore, if i drive = 55mA, and c in = 3600 pF, and you use a 60 volt peak value for the drive voltage, the output stage will encounter some slew limiting at 40.5 kHz. That is certainly sufficient for the current dumping application, but increasing the class A stage sink current will push that frequency even higher. The n channel devices require a lower drive current due to their lower input capacitance. In this application, it is not an issue since the source transistor is driven and can supply whatever current is needed up to its' own failure.
The QUAD 405 current dumping approach provides two important design features, a freedom from the tricky temperature dependent bias setting on the output devices, and the ability to sum the error with the output and cancel the amplifier distortion completely. I have heard the original amplifier and I have built and auditioned several variations. The original 405 suffered from a poorly designed triac/diac output protection circuit that tended to trip too often, but save that it was an outstanding performer and a very accurate amplifier to boot.
I intend to build this QUASAR and experiment with the laterals. The base drive current on bipolar devices presents problems of it's own and that was why I set out to use MOSFETs many years ago. I had used vertical devices with no standing bias and had less than ideal results. I think the laterals, with their lower drive voltage and lower threshold, combined with a small standing bias will be just what my output needed.
In audio applications as a source follower, it is the smaller gate source capacitance that needs to be considered, and miller capacitance is not a factor. Erno Borbely submitted a current driven 60 watt amplifier in Audio Amateur magazine many years ago that featured vertical MOSFETs with a current drive stage. His article had a very detailed analysis of the FET drive requirements in a linear application. I have built that amplifier and several variations of my own design that feature current drive and can verify that with a sufficient drive current, the MOSFETs operate without issue and sound great too.
The drive stage must be capable of providing sufficient drive current for the required slew rate which can be determined by the capacitance, the voltage swing and the frequency.
i drive = 2* pi * freq * C in * V drive 0-pk
The p channel devices have an input capacitance of 900 pF for a net of 3600 pF with 4 devices. In this application the current sink provides 55 mA of drive current. Therefore, if i drive = 55mA, and c in = 3600 pF, and you use a 60 volt peak value for the drive voltage, the output stage will encounter some slew limiting at 40.5 kHz. That is certainly sufficient for the current dumping application, but increasing the class A stage sink current will push that frequency even higher. The n channel devices require a lower drive current due to their lower input capacitance. In this application, it is not an issue since the source transistor is driven and can supply whatever current is needed up to its' own failure.
The QUAD 405 current dumping approach provides two important design features, a freedom from the tricky temperature dependent bias setting on the output devices, and the ability to sum the error with the output and cancel the amplifier distortion completely. I have heard the original amplifier and I have built and auditioned several variations. The original 405 suffered from a poorly designed triac/diac output protection circuit that tended to trip too often, but save that it was an outstanding performer and a very accurate amplifier to boot.
I intend to build this QUASAR and experiment with the laterals. The base drive current on bipolar devices presents problems of it's own and that was why I set out to use MOSFETs many years ago. I had used vertical devices with no standing bias and had less than ideal results. I think the laterals, with their lower drive voltage and lower threshold, combined with a small standing bias will be just what my output needed.
My consideration is somewhat different.
First the gate source capacity is not constant, it depends in a non linear fashion on gate source voltage. Unfortunately the "nonlinearity" has its max near the gate threshold.
The max rise or fall speed of a sine wave is at zero cross.
In a "standard" MosFET amp the idling current is set to 100 mA
for a couple of good reasons one is to reduce the crossover distortion in two ways, this includes the gate source capacity variation which are also prone to device process tolerances.
But in Quad topology the power transistors are set to 0 mA idling.
When ring emitter BJTs are used in connection with a current source there are no difficulties at all because the variations of Vbe on are eliminated and the IB/Ic charcteristic is perfectly - well almost perfectly- linear.
The only difficulty is to keep the idling current at 0 with respect to junction temperature.
The older Quad 303 had a perfect idling current stability. It is still a wonderful "musical" amp. It could well be that the "sweet sound" of triode tube amps is due to stability of idling current
although the THD can be 10%.
I take some care of such design objectives and others, I won't use differential input, won't use current sources, and no ICs in the audio path.
First the gate source capacity is not constant, it depends in a non linear fashion on gate source voltage. Unfortunately the "nonlinearity" has its max near the gate threshold.
The max rise or fall speed of a sine wave is at zero cross.
In a "standard" MosFET amp the idling current is set to 100 mA
for a couple of good reasons one is to reduce the crossover distortion in two ways, this includes the gate source capacity variation which are also prone to device process tolerances.
But in Quad topology the power transistors are set to 0 mA idling.
When ring emitter BJTs are used in connection with a current source there are no difficulties at all because the variations of Vbe on are eliminated and the IB/Ic charcteristic is perfectly - well almost perfectly- linear.
The only difficulty is to keep the idling current at 0 with respect to junction temperature.
The older Quad 303 had a perfect idling current stability. It is still a wonderful "musical" amp. It could well be that the "sweet sound" of triode tube amps is due to stability of idling current
although the THD can be 10%.
I take some care of such design objectives and others, I won't use differential input, won't use current sources, and no ICs in the audio path.
hahfran said:
I have been looking for this 'Musical meter' and the ubiquitous 'sweetness measuring system' but have been unable to find them as yet. Have you managed to source one?
Please let me know the details so that I can get one two......Damn! every one I meet who claims to have something between their ears seems to be using these instruments.
Good idea! Throw out all the rest of the accepted industry standards as well. Why not just go back to the diaphram and horn system. This was really 'warm' sounding and had wonderfull 'smooth tone'! (Tested on the appropriate 'warmth' and 'smoothness' measuring systems of course.)
Cheers.....Meshuganah....
Hum...well as long as credit crunch doesn't bar my bank account
i.e. the bank runs out of cash I can go out and buy a $10000 amp
but...it is a waste of money as that amp does not produce the sound that I expect. It is all subjective and only therefore DIY makes sense. One can agree on numbers which encode measurements but cannot easily agree that the moon exists
whether or not there is at least one observer.
Anyway I revisited Walker's and Albinsons's 1975 paper and concluded there should not be bias diodes.
i.e. the bank runs out of cash I can go out and buy a $10000 amp
but...it is a waste of money as that amp does not produce the sound that I expect. It is all subjective and only therefore DIY makes sense. One can agree on numbers which encode measurements but cannot easily agree that the moon exists
whether or not there is at least one observer.
Anyway I revisited Walker's and Albinsons's 1975 paper and concluded there should not be bias diodes.
Hi,
current dumping topology doesn’t like fast power devices due to C8 (capacitance in the bridge of Quad405) that limits the speed of the class A stage correcting the crossover distortion of the dumpers.
The impedance connected to the collector of the lower dumper (sometimes to the upper one too) has just the purpose to slow its switching time down.
Therefore the use of fast devices like Mosfets would not be a good idea here.
Regards,
Pintur
current dumping topology doesn’t like fast power devices due to C8 (capacitance in the bridge of Quad405) that limits the speed of the class A stage correcting the crossover distortion of the dumpers.
The impedance connected to the collector of the lower dumper (sometimes to the upper one too) has just the purpose to slow its switching time down.
Therefore the use of fast devices like Mosfets would not be a good idea here.
Regards,
Pintur
Yes the rise time of the original design is reportedly 1.5 V/usec.
MosFEts provide only disadvantage here. Fast BJTs are also not useful. Rugged standard such as BD 249/250 will do. However the distinct adavantage of ring emitter is the IB/IC linearity.
I think I'll get back to QUAD 303 reborn...for low and midrange I'll keep on considering current dumping.
MosFEts provide only disadvantage here. Fast BJTs are also not useful. Rugged standard such as BD 249/250 will do. However the distinct adavantage of ring emitter is the IB/IC linearity.
I think I'll get back to QUAD 303 reborn...for low and midrange I'll keep on considering current dumping.
There is nothing about the current dumping topology that precludes MOSFETs or fast output devices. Why would it? The output devices are unity gain followers and, unless in the case when MOSFETs were incorrectly applied with very large value gate resistors, the higher bandwidth of the output devices does not factor into the amplifier stability.
Ultimately, the output inductor and the integrating capacitor, typically 3 uH and 120 pf in the bridge, are the bandwidth limiting factors. I have a several breadboards of various current dumping amplifiers I have built over the past 20 plus years and I have replaced slower bipolar outputs with faster ones and measured (and heard I might add), no adverse effects. The faster devices require good lead dress and a small value resistor in series with the base to prevent parasitic oscillation.
The faster output devices, like the MJL3281A have a higher beta that is also more constant over 100mA compared to the originally specified devices, and that reduces the demand on the class A stage, which is a good thing.
With an 8 ohm load, and 100 watt output the original QUAD 405 performs amazingly well. The class A amplifier stage ultimately determines the output distortion. It is also called on to drive the output (through a 47 ohm resistor) until the "dumpers" are conducting, and to provide the drive current to them as well. It is quite informative to look at the output drive voltage on an oscilloscope with various output devices and see real time what changes.
If one replaces the original bootstrap load on the class A stage with a current sink, then the source current capability will remain quite high, but the sink current of the class A stage is limited to whatever the current sink is set to provide, in most cases 50 to 55 mA. If the output device has a beta of 100, the output current will be limited to roughly 5 amperes. That is fine for 8 ohms at 100 watts RMS, but not enough as the output load changes, which is the case with real loudspeakers. If you wish to increase the output power, the sink current must also be addressed, or the output devices.
There are a number of solutions including increasing the current sink (and providing sufficient heat sinking), darlington configurations, and MOSFETs. The problems with the darlington has to do with device turn off and is too complex to address here other than to say it is not without problem.
The approach taken in the QUASAR is elegant and involves the 22 ohm resistor between the gates. This allows the drive stage to develop a larger voltage (which is what the MOSFET requires) and largely takes the current load limit off the class A stage. I suspect the 22 ohm resistor would need to be increased slightly if this design were to go into full scale production to cover all lateral outputs as their threshold voltages are specified with a pretty wide tolerance, but to this point it appears to work quite well. I look forward to building one of my own.
Ultimately, the output inductor and the integrating capacitor, typically 3 uH and 120 pf in the bridge, are the bandwidth limiting factors. I have a several breadboards of various current dumping amplifiers I have built over the past 20 plus years and I have replaced slower bipolar outputs with faster ones and measured (and heard I might add), no adverse effects. The faster devices require good lead dress and a small value resistor in series with the base to prevent parasitic oscillation.
The faster output devices, like the MJL3281A have a higher beta that is also more constant over 100mA compared to the originally specified devices, and that reduces the demand on the class A stage, which is a good thing.
With an 8 ohm load, and 100 watt output the original QUAD 405 performs amazingly well. The class A amplifier stage ultimately determines the output distortion. It is also called on to drive the output (through a 47 ohm resistor) until the "dumpers" are conducting, and to provide the drive current to them as well. It is quite informative to look at the output drive voltage on an oscilloscope with various output devices and see real time what changes.
If one replaces the original bootstrap load on the class A stage with a current sink, then the source current capability will remain quite high, but the sink current of the class A stage is limited to whatever the current sink is set to provide, in most cases 50 to 55 mA. If the output device has a beta of 100, the output current will be limited to roughly 5 amperes. That is fine for 8 ohms at 100 watts RMS, but not enough as the output load changes, which is the case with real loudspeakers. If you wish to increase the output power, the sink current must also be addressed, or the output devices.
There are a number of solutions including increasing the current sink (and providing sufficient heat sinking), darlington configurations, and MOSFETs. The problems with the darlington has to do with device turn off and is too complex to address here other than to say it is not without problem.
The approach taken in the QUASAR is elegant and involves the 22 ohm resistor between the gates. This allows the drive stage to develop a larger voltage (which is what the MOSFET requires) and largely takes the current load limit off the class A stage. I suspect the 22 ohm resistor would need to be increased slightly if this design were to go into full scale production to cover all lateral outputs as their threshold voltages are specified with a pretty wide tolerance, but to this point it appears to work quite well. I look forward to building one of my own.
This misses the point: one cannot choose closed loop bandwidth
independently. The current dumping condition x1x2=x3x4 or
R1R2=L/C does not in any way allow to choose any x free.
There is an optimum solution and that depends on the characteristics of the output devices. In fact higher transition frequency of the output transistors does not affect power bandwidth hence using faster FETs or BJTs is not an advantage
in terms of bandwidth or slew rate but it is an advantage in terms of class A driver stage.
In other words the current dumping condition reduces degrees of design variables of choice such as lead/lag compensation amongst others.
independently. The current dumping condition x1x2=x3x4 or
R1R2=L/C does not in any way allow to choose any x free.
There is an optimum solution and that depends on the characteristics of the output devices. In fact higher transition frequency of the output transistors does not affect power bandwidth hence using faster FETs or BJTs is not an advantage
in terms of bandwidth or slew rate but it is an advantage in terms of class A driver stage.
In other words the current dumping condition reduces degrees of design variables of choice such as lead/lag compensation amongst others.
Now well I am a theorist...first I look at mathematical description
and try to find a flaw...almost always I am successfull.
Nevertheless the point remains: optimization of a current dumping amp meets more restrictions than a conventional negative feedback ( favorably nested feedback loops ) amp.
There is another limitation: actually one has to figure in the factor
(1+1/A0) where A0 is open loop gain of the class A driver. But A0
is frequency- dependent. Thus, conventionally one has to calculate C and L first in order to determine stability vs frequency of the forward correction.
It is less easy than it appears at first glance.
and try to find a flaw...almost always I am successfull.
Nevertheless the point remains: optimization of a current dumping amp meets more restrictions than a conventional negative feedback ( favorably nested feedback loops ) amp.
There is another limitation: actually one has to figure in the factor
(1+1/A0) where A0 is open loop gain of the class A driver. But A0
is frequency- dependent. Thus, conventionally one has to calculate C and L first in order to determine stability vs frequency of the forward correction.
It is less easy than it appears at first glance.
jonusgrumby said:
Also Class A should have enough gain to make error correction working.Bridge resistors will determine class A gain as well.
I think is good to read and digest John Vanderkooy and Stanley P. Lipshitz document.
http://quad405.com/jaes.pdf
Jonus !Tibi