Bob,
On page 74, a cascode is added to the VAS. It appears that there is a very minor improvement in distortion and it also appears to not reduce the size of the Miller capacitor. I'm assuming that if a Zobel is used on the driver transistor, that it would probably also be unchanged. So, just what is the advantage of the cascode? Would it slightly improve square wave response?
Thanks.
On page 74, a cascode is added to the VAS. It appears that there is a very minor improvement in distortion and it also appears to not reduce the size of the Miller capacitor. I'm assuming that if a Zobel is used on the driver transistor, that it would probably also be unchanged. So, just what is the advantage of the cascode? Would it slightly improve square wave response?
Thanks.
As I understand it both the Early effect and the large variation of base-collector capacitance with voltage conspire to add distortion to a VAS using a single transistor - cascoding removes both of these issues by removing the voltage swing from the collector of the VAS transistor.
The "Miller" cap is (or is part of) the compensation network required for stable negative feedback, its value is dependent on the feedback loop, not the nature of the VAS stage, although it also serves to provide local negative feedback at high frequency and reduce VAS distortion accordingly.
The "Miller" cap is (or is part of) the compensation network required for stable negative feedback, its value is dependent on the feedback loop, not the nature of the VAS stage, although it also serves to provide local negative feedback at high frequency and reduce VAS distortion accordingly.
Excellent answer, Marc.
I would only add, as you implied, that a 2-transistor VAS, where the first transistor is an emitter follower, mitigates the non-linear Miller effect due to collector-base capacitance of the main VAS transistor.
If I had to choose between a non-cascoded 2T VAS and a cascoded 1T VAS, I would generally choose the former.
Cheers,
Bob
I would only add, as you implied, that a 2-transistor VAS, where the first transistor is an emitter follower, mitigates the non-linear Miller effect due to collector-base capacitance of the main VAS transistor.
If I had to choose between a non-cascoded 2T VAS and a cascoded 1T VAS, I would generally choose the former.
Cheers,
Bob
The Miller capacitor may well reduce local VAS distortion but it does little or nothing to reduce crossover distortion.
A stabilisation scheme which uses global feedback is difficult to implement, but can be done with additional time constants where crossover distortion can be reduced.
The best option is to use a two transistor plus cascode, probably with transitional Miller feedback. That seems to combine the low crossover distortion while maintaining stability.
A stabilisation scheme which uses global feedback is difficult to implement, but can be done with additional time constants where crossover distortion can be reduced.
The best option is to use a two transistor plus cascode, probably with transitional Miller feedback. That seems to combine the low crossover distortion while maintaining stability.
For quite some time I have tested VAS circuits with a single transistor and cascode and traditional VAS circuits with a single transistor without a cascode. The latter although theoretically worse in terms of distortion specs and probably bandwidth, the traditional simple VAS circuit with a good transistor tends to like the sound better, and not only me, but many friends who participated in the tests.
I respect Mr. Cordell very much, I have read his book and many of his opinions. At first I was obsessed with designing or achieving circuits that obtained better results in terms of distortion and bandwidth, then I only look for better sound results, I am only able to do it based on trial and error ...
I respect Mr. Cordell very much, I have read his book and many of his opinions. At first I was obsessed with designing or achieving circuits that obtained better results in terms of distortion and bandwidth, then I only look for better sound results, I am only able to do it based on trial and error ...
I ditched the cascode VAS and went for the two transistor ie beta helper + main VAS transistor a while back. Cascoded VAS have some issues that to my mind don’t seem to be worth the trouble. They’re good in LTP’s and CFA diamond buffer input stages where it enables the use of high hFE devices and in the case of a CFA, to keep the power dissipation in all 4 input devices the same (= very low DC drift).
This is an interesting point. These days, I always use the low Cob video transistors for the main VAS transistor. Those are wonderful transistors. Indeed, back in the early 80s when I did my MOSFET power amplifier with error correction, I did use a cascoded 2T VAS (which I referred to at the time as a Darlington cascode).
It is a fair question whether more ordinary VAS transistors would benefit significantly in a cascode arrangement. It may be worth a closer look, but I am not yet convinced that a 2T VAS using ordinary VAS transistors would be helped much. This is especially a question if the VAS transistor has significant emitter degeneration (I always use 10:1).
In the 2T VAS, nonlinear Miller compensation capacitance effects are essentially killed by the EF. However, nonlinear Cob of the VAS transistor still presents a nonlinear load to the VAS. This may be the remaining influence. However, even this influence is mitigated somewhat by the shunt feedback of the Miller compensation greatly reducing the effective output impedance of the VAS.
Cheers,
Bob
It is a fair question whether more ordinary VAS transistors would benefit significantly in a cascode arrangement. It may be worth a closer look, but I am not yet convinced that a 2T VAS using ordinary VAS transistors would be helped much. This is especially a question if the VAS transistor has significant emitter degeneration (I always use 10:1).
In the 2T VAS, nonlinear Miller compensation capacitance effects are essentially killed by the EF. However, nonlinear Cob of the VAS transistor still presents a nonlinear load to the VAS. This may be the remaining influence. However, even this influence is mitigated somewhat by the shunt feedback of the Miller compensation greatly reducing the effective output impedance of the VAS.
Cheers,
Bob
Since my application has been retrofitting vintage equipment, using the existing power supply, headroom is an issue. In one case, the transformer does have a 6.3 V winding (measures close to 7 V) that is not in use. If there were a second such winding, they could be added in series with the main winding, adding almost 10 V to the rails.
Here is an interesting oddball idea: Connect the 7 V winding so the transformer is changed from 30-0-30 to 37-0-30. The result would be 1/2 wave rectification until voltage dropped to where the lower voltage winding would start picking up some of the load.
Hi Fred,
I have long advocated for boosted rails for the IPS/VAS in high-performance amplifiers to increase headroom for these critical circuits. That is exactly what you are talking about. I used boosted rails in the early 80s when I did my MOSFET power amplifier with error correction. Vertical MOSFETs eat more headroom because of their higher Vgs_on voltage. Hawksford error correction also usually eats some headroom.
As you are undoubtedly aware, boosted rails for just the IPS, VAS and maybe driver can be attractive and economical because the current they must supply is fairly low. In fact, even if you only use a separate bridge rectifier and reservoir capacitor with the existing windings of the transformer you will get somewhat boosted rails, and which are better filtered. This is especially the case for transient peak power situations, where these rails will sag very little during the duration of the transient.
However, for true boosted rails the most straightforward approach is to employ two extra windings on the transformer, creating a higher AC voltage for the boosted rail bridge rectifier. You can also create 2 low-voltage isolated supplies that ride on top of the main rails, but this approach is not as good because main-rail ripple and sag will be transferred to the boosted rails.
If your power transformer is a toroid, you are in luck. It is easy to add a couple of low-voltage windings to a toroid just by winding on the outside. You will often get about 0.6 to 0.8 volts per turn. I have done this many times.
Alternately, you can add a small dual secondary low voltage, low wattage transformer for this purpose.
There are also some auxilliary rectifier configurations where you can achieve boosted rails by AC coupling some of the main winding voltage to the auxilliary rectifier. In fact, such techniques can also be made to work from a single low-voltage winding, like you have. Think about how voltage doublers, voltage triplers and voltage multipliers work.
Cheers,
Bob
I have long advocated for boosted rails for the IPS/VAS in high-performance amplifiers to increase headroom for these critical circuits. That is exactly what you are talking about. I used boosted rails in the early 80s when I did my MOSFET power amplifier with error correction. Vertical MOSFETs eat more headroom because of their higher Vgs_on voltage. Hawksford error correction also usually eats some headroom.
As you are undoubtedly aware, boosted rails for just the IPS, VAS and maybe driver can be attractive and economical because the current they must supply is fairly low. In fact, even if you only use a separate bridge rectifier and reservoir capacitor with the existing windings of the transformer you will get somewhat boosted rails, and which are better filtered. This is especially the case for transient peak power situations, where these rails will sag very little during the duration of the transient.
However, for true boosted rails the most straightforward approach is to employ two extra windings on the transformer, creating a higher AC voltage for the boosted rail bridge rectifier. You can also create 2 low-voltage isolated supplies that ride on top of the main rails, but this approach is not as good because main-rail ripple and sag will be transferred to the boosted rails.
If your power transformer is a toroid, you are in luck. It is easy to add a couple of low-voltage windings to a toroid just by winding on the outside. You will often get about 0.6 to 0.8 volts per turn. I have done this many times.
Alternately, you can add a small dual secondary low voltage, low wattage transformer for this purpose.
There are also some auxilliary rectifier configurations where you can achieve boosted rails by AC coupling some of the main winding voltage to the auxilliary rectifier. In fact, such techniques can also be made to work from a single low-voltage winding, like you have. Think about how voltage doublers, voltage triplers and voltage multipliers work.
Cheers,
Bob
Or you can add a DC-to-DC converter or two. You could put them on the amplifier PCB (as done in Winfield Hill's 1000 V/usec amp (here), or also as done on the DIY VFET front end cards Marauder and Dreadnought (there)).
Or, if you prefer, you can put the DC/DC converter(s) on the power supply PCB instead.
For the first and second stages of a power amp, you can get all the current you need from a little DC/DC converter module, no larger than a DIP relay.
_
Or, if you prefer, you can put the DC/DC converter(s) on the power supply PCB instead.
For the first and second stages of a power amp, you can get all the current you need from a little DC/DC converter module, no larger than a DIP relay.
_
Last edited:
Yes, it is amazing how small the switching DC-DC converters are. One caveat - for this application they need to be isolated DC-DC converters if they are sitting on top of the main rails. I have some isolated ones, but they are a bit less prevalent and a bit larger in my experience. They usually need a 4V - 18V DC input supply.
Caution: I have seen some of these little supplies be unable to start up if they are loaded with significant reservoir capacitance and fed with a finite-impedance input supply (especially a current-limited bench supply). The reason for this is that the switching converters often present a negative impedance to the supply source. Since they are largely constant-power devices, at low source voltage they draw more current for a given fixed load current.
Cheers,
Bob
Caution: I have seen some of these little supplies be unable to start up if they are loaded with significant reservoir capacitance and fed with a finite-impedance input supply (especially a current-limited bench supply). The reason for this is that the switching converters often present a negative impedance to the supply source. Since they are largely constant-power devices, at low source voltage they draw more current for a given fixed load current.
Cheers,
Bob
Again, this amplifier is an upgrade to a vintage receiver. Transformer is conventional, not a toroid. Very little space to add circuitry. Boards are already stacked. It is already good for slightly over 50 watts per channel. Any significant improvement (100 watts) would require about 12 volts more on the rails.
Attachments
Last edited:
Hi Fred,
... and more output transistors, bigger heat sinks and a larger power transformer core.
Its designed for 50 watts per channel and I'll assume it is reliable. Pushing it to 100 watts per channel will make it unreliable. I think that plan is still-born. You started with a modest receiver and should really be satisfied with the huge quality increase. Going for more power on top of that is asking for trouble.
-Chris
... and more output transistors, bigger heat sinks and a larger power transformer core.
Its designed for 50 watts per channel and I'll assume it is reliable. Pushing it to 100 watts per channel will make it unreliable. I think that plan is still-born. You started with a modest receiver and should really be satisfied with the huge quality increase. Going for more power on top of that is asking for trouble.
-Chris
Anatech, You are correct and any discussion of increased power is only discussion. The original unit is a Fisher 600-T and the original power amplifier was 4 PNP germanium transistors per channel in series to handle the voltage. The circuit was essentially from an early RCA transistor manual. What makes this good for the upgrade is that there are 8 positions on the back panel heat sink for TO-3 transistors, so parallel devices was an obvious choice.
There are 2 ways to design a power amplifier: Full power for an indefinite period, or reproduction of music. In the music case, the amplifier only needs to provide full power for short periods as the average is 10 dB below that. In fact, Fisher even stated that full power measurements be limited to 10 minutes for one channel and 5 minutes for both channels. So, if the amplifier could be made to provide 100 watts per channel, it would only need to sustain 10 watts per over the long term.
The other method of design, full power indefinitely, would be something like a McIntosh. Some of those are used for non audio industrial applications where long term full power is required.
There are 2 ways to design a power amplifier: Full power for an indefinite period, or reproduction of music. In the music case, the amplifier only needs to provide full power for short periods as the average is 10 dB below that. In fact, Fisher even stated that full power measurements be limited to 10 minutes for one channel and 5 minutes for both channels. So, if the amplifier could be made to provide 100 watts per channel, it would only need to sustain 10 watts per over the long term.
The other method of design, full power indefinitely, would be something like a McIntosh. Some of those are used for non audio industrial applications where long term full power is required.
I was not referring to music power but honest continuous power for some reasonable period of time, maybe a few minutes. In simulation of a power amp, I also simulate the power supply. For any power above maximum continuous (such as IHF music power) the filter capacitors are down to final value in just a few cycles, far shorter than any musical peak.
The difference here is that this will produce 50 watts continuous for a long period of time but the heat sink and transformer will eventually be overheated.
The difference here is that this will produce 50 watts continuous for a long period of time but the heat sink and transformer will eventually be overheated.
The global feedback is what squashes any OS distortions (large signal and crossover, and sometimes fuse distortion too!).
Sometimes things are complicated when the compensation scheme is more involved, but a simple Cdom is there to both linearize the VAS and provide stability.
Simple (no emitter degeneration) common-emitter circuits are not very linear at all (for instance a VAS at <=20Hz is probably the major source of distortion in the amp), but with local feedback they can be great performers.
At 20Hz the loop gain is high enough to linearize both VAS and OS. At 20kHz the loop gain is struggling to linearize just the OS, but the VAS is very linear at HF due to Cdom local feedback.