Hi there,
imagine a transistor amplifier stage (common emitter) that is cascoded by another transistor.
Most often, the reference voltage that feed the base of the cascoding device taken from the positive rail with the aid of a voltage divider formed by two resistors.
Sometimes you see batteries here. Another option would be a local regulator.
How huge is the impact of the 'quality' of the reference voltage in your experience?
Did anyone ever tried different options (e.g. much decoupling capacitance, different kind of regs or even coils) and made comparisons soundwise?
Rüdiger
imagine a transistor amplifier stage (common emitter) that is cascoded by another transistor.
Most often, the reference voltage that feed the base of the cascoding device taken from the positive rail with the aid of a voltage divider formed by two resistors.
Sometimes you see batteries here. Another option would be a local regulator.
How huge is the impact of the 'quality' of the reference voltage in your experience?
Did anyone ever tried different options (e.g. much decoupling capacitance, different kind of regs or even coils) and made comparisons soundwise?
Rüdiger
if you keep Vce of common emitter transistor reasonably high (more than 0.1-0.2V) it doesn't matter much.
BTW it is strange to see how many designers forgot, that low Vce results in terrible parameters degradation.
BTW it is strange to see how many designers forgot, that low Vce results in terrible parameters degradation.
Onvinyl
I did evaluate the effect on THD of different types of input stage cascode reference with the pure resistive divider usually to ground with a bypass C consistently the best.
For what it's worth.
I did evaluate the effect on THD of different types of input stage cascode reference with the pure resistive divider usually to ground with a bypass C consistently the best.
For what it's worth.
amplifierguru said:Onvinyl
I did evaluate the effect on THD of different types of input stage cascode reference with the pure resistive divider usually to ground with a bypass C consistently the best.
For what it's worth.
Using a Zener Locked divider will certainly degrade the performance.
I guess that one wouldn't find the batterie option in commercial gear.
amplifierguru, resistive divider had lower THD than -- *what*?
Isn't noise an issue?
Jonathan Carr once stated, that he uses local regs for css, maybe someone tried that approach on cascodes? Or maybe it makes sense to have the same degenerations of V+ fed in the reference? I guess I'll try it on my own (soon, I hope...)
Rüdiger
amplifierguru, resistive divider had lower THD than -- *what*?
Isn't noise an issue?
zener is noisier than a resistor, is that the reason?Using a Zener Locked divider will certainly degrade the performance
Jonathan Carr once stated, that he uses local regs for css, maybe someone tried that approach on cascodes? Or maybe it makes sense to have the same degenerations of V+ fed in the reference? I guess I'll try it on my own (soon, I hope...)
Rüdiger
not in terms of THD, but I think it is vital to keep the reference voltage devider least inductive possible
Onvinyl - I was using an AP test set at the time measuring THD+N in 80KHz BW. Yes better than zeners. Admittedly low noise regulators were not tried, it was a final tweak of a commercial design and the board layout was final. LEDs gave no reason to change from the resistive divider.
If you picture the cascode transistor as a common emitter with the noise signal at the base and the collector impedance of the diff'l BJT/FET as it's emitter load then your noise gain is the load on the stage divided by the collector impedance at it's emitter. So sensitivity depends on circuit parameters.
I might add that this was for an input diff'l cascode - later stages like Vas are not so noise sensitive.
If you picture the cascode transistor as a common emitter with the noise signal at the base and the collector impedance of the diff'l BJT/FET as it's emitter load then your noise gain is the load on the stage divided by the collector impedance at it's emitter. So sensitivity depends on circuit parameters.
I might add that this was for an input diff'l cascode - later stages like Vas are not so noise sensitive.
Hi amplifierguru,
many thanks for clearing this up! And for your explanation of the way the signal might be affected as well.
Yes, I'm thinking of an input fet diff pair as well, here serving as a first stage in a phono amplifier. Much has been said about these stages here in diyaudio, so my focus is on the cascode and the ways to better it now.
Rüdiger
many thanks for clearing this up! And for your explanation of the way the signal might be affected as well.
Yes, I'm thinking of an input fet diff pair as well, here serving as a first stage in a phono amplifier. Much has been said about these stages here in diyaudio, so my focus is on the cascode and the ways to better it now.
Rüdiger
Rüdiger: At least in the Connoisseur 4-2SE and 5, I use independent regulators for literally everything, and this includes both ccs and cascodes. Would I recommend this approach for a normal (non-SOTA) design? No, for a number of reasons.
First, having excessively low-impedance drive into the base of a cascode transistor can trigger oscillations at ultra high frequencies (100~300 MHz). If the circuit node loading the cascode output is low-impedance, this may not happen, but higher load impedances may cause the likelihood of HF oscillations to increase. You can usually quell the oscillations with stopper resistors on the cascode bases (or gates), but when you add the impedance of the stopper resistors to the output impedance of your local regulator, you may find that the local regulator no longer makes much sense.
Next, an all-too common failing for many aspects of circuitry design is a tendancy to look at the schematics without considering the physical design. The schematic is only a starting point, a crude representation of what may or may not be built (usually not). The real circuit includes, in addition to the omnipresent pitfalls of misleading device models, parasitic inductance and capacitance from the board structure, ground current errors, various RF and EM fields that are generated by the operation of the amplifier or the surrounding environment, and more. Local regulators may look snazzy, but before committing yourself, you need to ask yourself - "Where am I going to locate this in relation to the circuit node that it is serving, can I get power into the regulator without causing problems for the other circuits (amplifier and regulators), is the location close enough to the circuit node that it is feeding, what are the consequences on the total _physical_design_ of the amplifier?" And so on. If you don't have good answers to these questions, the regulators may be more of a liability than an asset.
Also, most amplifier designs (and audio systems) have too much distortion and too little resolution to make it a certainty that you will hear a subjective benefit from the extra cost, complexity and effort that you put into the regulator design. In the majority of cases, resistive dividers will be sufficient (particularly in a power amp). And I can verify that it is possible to make a very nice-sounding amplifier using resistive dividers only. This doesn't mean that you shouldn't try your hand at separate regulation, but there is a very real possibility that you will expend a great deal of effort (if you want to do the job properly) and not have much to show for it at the end of the day.
hth, jonathan carr
PS. As an aside, separate regulation succeeds in the Connoisseur 4-2SE and 5, because each amplifier board has a companion regulator board which sits directly below the amplifier PCB (the two boards were designed as a matching pair). Every regulator is in immediate proximity to the amplifier circuit node that it supplies power to - literally within millimeters, yet because the regulators all reside on a separate board, they don't compromise the layout and function of the main amplifier circuitry.
First, having excessively low-impedance drive into the base of a cascode transistor can trigger oscillations at ultra high frequencies (100~300 MHz). If the circuit node loading the cascode output is low-impedance, this may not happen, but higher load impedances may cause the likelihood of HF oscillations to increase. You can usually quell the oscillations with stopper resistors on the cascode bases (or gates), but when you add the impedance of the stopper resistors to the output impedance of your local regulator, you may find that the local regulator no longer makes much sense.
Next, an all-too common failing for many aspects of circuitry design is a tendancy to look at the schematics without considering the physical design. The schematic is only a starting point, a crude representation of what may or may not be built (usually not). The real circuit includes, in addition to the omnipresent pitfalls of misleading device models, parasitic inductance and capacitance from the board structure, ground current errors, various RF and EM fields that are generated by the operation of the amplifier or the surrounding environment, and more. Local regulators may look snazzy, but before committing yourself, you need to ask yourself - "Where am I going to locate this in relation to the circuit node that it is serving, can I get power into the regulator without causing problems for the other circuits (amplifier and regulators), is the location close enough to the circuit node that it is feeding, what are the consequences on the total _physical_design_ of the amplifier?" And so on. If you don't have good answers to these questions, the regulators may be more of a liability than an asset.
Also, most amplifier designs (and audio systems) have too much distortion and too little resolution to make it a certainty that you will hear a subjective benefit from the extra cost, complexity and effort that you put into the regulator design. In the majority of cases, resistive dividers will be sufficient (particularly in a power amp). And I can verify that it is possible to make a very nice-sounding amplifier using resistive dividers only. This doesn't mean that you shouldn't try your hand at separate regulation, but there is a very real possibility that you will expend a great deal of effort (if you want to do the job properly) and not have much to show for it at the end of the day.
hth, jonathan carr
PS. As an aside, separate regulation succeeds in the Connoisseur 4-2SE and 5, because each amplifier board has a companion regulator board which sits directly below the amplifier PCB (the two boards were designed as a matching pair). Every regulator is in immediate proximity to the amplifier circuit node that it supplies power to - literally within millimeters, yet because the regulators all reside on a separate board, they don't compromise the layout and function of the main amplifier circuitry.
Jonathan, what can I say ... thanks a lot for your thorough advice and the sharing of your findings!
How low is low enough, is there a rule of thumb? At the moment the cascode is fed with a NiMh Batterie that has a Ri of 2R max. (quite high I think,) a good regulator will be several magnitudes lower.
Oh, I will certainly try, it is a (very-) long-term diy project that shall teach me how to do those things properly. I will for sure take your advice serious about the physical layout.
Rüdiger
If the circuit node loading the cascode output is low-impedance
How low is low enough, is there a rule of thumb? At the moment the cascode is fed with a NiMh Batterie that has a Ri of 2R max. (quite high I think,) a good regulator will be several magnitudes lower.
This doesn't mean that you shouldn't try your hand at separate regulation, but there is a very real possibility that you will expend a great deal of effort (if you want to do the job properly) and not have much to show for it at the end of the day.
Oh, I will certainly try, it is a (very-) long-term diy project that shall teach me how to do those things properly. I will for sure take your advice serious about the physical layout.
Rüdiger
I did some early tests last night. It is a diff stage with paralled 2sk369, cascoded and CCS'ed in a phono amp. Main PSU is a bank of lead batteries with some cap filtering. I tried different ways for the voltage reference of the cascode. In order of sonics:
1) Worst: 12V lead battery. Sound is comparably grey and grainy.
2) not good as well: 12V lead battery and mosfet follower. Better, but same footprint.
3) Nimh-batterie 9V: Quite good, no greyness anymore, round and musical, maybe a bit soft (but not to much)
4) 12v lead battery, simplest 431-shunt, heavily biased (actually over-biased...). Amazing! Much as 3), but more slam, more stable bass, and even better in trebles. Has *not* more ground noise than previous (jugded by ear), not so sure about stability (some strange noises sometimes).
What I plan to try also is a borbely-like fet regulator, didn't make it work last night.
And definitly a more advanced 'local' shunt. (Circuits or ideas, anyone?)
Very educative task!
Rüdiger
edit: spelling
1) Worst: 12V lead battery. Sound is comparably grey and grainy.
2) not good as well: 12V lead battery and mosfet follower. Better, but same footprint.
3) Nimh-batterie 9V: Quite good, no greyness anymore, round and musical, maybe a bit soft (but not to much)
4) 12v lead battery, simplest 431-shunt, heavily biased (actually over-biased...). Amazing! Much as 3), but more slam, more stable bass, and even better in trebles. Has *not* more ground noise than previous (jugded by ear), not so sure about stability (some strange noises sometimes).
What I plan to try also is a borbely-like fet regulator, didn't make it work last night.
And definitly a more advanced 'local' shunt. (Circuits or ideas, anyone?)
Very educative task!
Rüdiger
edit: spelling
I was at least trying to say that the use of the different approaches did not change the noise floor of the whole amp. This with fully opened volume pot and ears in the speakers, not measured
Rüdiger
Rüdiger
Sorry, my english is poor and leaves me here. What are you trying to tell me? I just said that I can't hear differences in the noise floor.
I think that to be a good thing, since the LM431 is often said to be far too noisy.
Rüdiger
I think that to be a good thing, since the LM431 is often said to be far too noisy.
Rüdiger
Hi mastertech,
ok, I see. I admit that I will repeat this tests to check if the results are the same. And yes, it is far away from a tip-top double-blind test set-up. Just trying to trust my ears...
Rüdiger
ok, I see. I admit that I will repeat this tests to check if the results are the same. And yes, it is far away from a tip-top double-blind test set-up. Just trying to trust my ears...
Rüdiger
My home brew beer talking (6pm +) . Your English is better than my Deutshe everytime.
No criticism intended. I like TL431 ( I think the same) but careful where I use it.
No criticism intended. I like TL431 ( I think the same) but careful where I use it.
Cascodes ? Use search and user "Jocko Homo" 😉 .
Vref for cascode ? Here you are:
http://www.diyaudio.com/forums/showthread.php?postid=618169#post618169
🙂
best regards
Lenny
Vref for cascode ? Here you are:
http://www.diyaudio.com/forums/showthread.php?postid=618169#post618169
🙂
best regards
Lenny
- Status
- Not open for further replies.
- Home
- Amplifiers
- Solid State
- Quality of reference voltage in a cascode