In most single or cascode valve circuits, swapping the Anode resistor for a CCS will muck up the biasing. The Anode voltage will be either too high or too low and this may give high THD, clipping and/or damage. A load inductor doesn't have this problem as the Anode voltage will be well defined.
To use a CCS on the Anode properly, you will have to arrange some way to stabilize the Anode voltage without losing the advantages of the CCS.
To use a CCS on the Anode properly, you will have to arrange some way to stabilize the Anode voltage without losing the advantages of the CCS.
With a resistive load, the output impedance is that of the internal resistance of the tube in parallel with the load resistor.
With a CCS load, the output impedance is that of the internal resistance of the tube in parallel with that of the CCS, usually a lot higher, so the BW is lower.
Regarding the noise, it depends: the CCS will inject its own noise at the OUTPUT, so if you think "equivalent input noise", the contribution of the CCS is divided by the gain.
Another often overlooked factor is the linearity of the CCS: these guys are usually used in static conditions, i.e. voltage regulators. Under dynamic conditions, i.e. music, well, they are... sand. I don't usually use them unless there's no other way. Yes, I confess I'm a diehard tube guy.
With a CCS load, the output impedance is that of the internal resistance of the tube in parallel with that of the CCS, usually a lot higher, so the BW is lower.
Regarding the noise, it depends: the CCS will inject its own noise at the OUTPUT, so if you think "equivalent input noise", the contribution of the CCS is divided by the gain.
Another often overlooked factor is the linearity of the CCS: these guys are usually used in static conditions, i.e. voltage regulators. Under dynamic conditions, i.e. music, well, they are... sand. I don't usually use them unless there's no other way. Yes, I confess I'm a diehard tube guy.
See my latest measurements on my thread - link above.
Theories are fine,but nothing beats real world measurement....
Theories are fine,but nothing beats real world measurement....
The same circuit parasitics and a higher plate impedance --> lower bandwidth.
And the CCS will add its own parasitic capacitance as well.
There is no extra noise using a 10M45 and it's said that a CCS greatly improves the PSRR so even less noise of mains origin.
There was also no drop in bandwidth (until I added a 4n7 capacitor across the feedback resistor)
There was also no drop in bandwidth (until I added a 4n7 capacitor across the feedback resistor)
Relative noise of the CCS and a resistor load depends on the specific circuit details.
Increasing the plate load impedance will always decrease the bandwidth, due to the higher gain.
See "gain-bandwidth product" in any EE text.
Increasing the plate load impedance will always decrease the bandwidth, due to the higher gain.
See "gain-bandwidth product" in any EE text.
Gain is not an issue for audio , tubes have very high bandwidth .
Other things kill the bandwidth much sooner than gain
Other things kill the bandwidth much sooner than gain
Me too,-diehard tube fanatic, plus I like Einsteins principle: As simple as possible, but no simpler.
Ok , then make your choice rayma ... gain-bandwidth or Miller
One 6SN7 triode with plate resistor has a gain of let's say 15 , with CCS max theoretical 20 , what a huge difference 🙂
And the bottom triode in cascode probably has the same gain as before
For driving next stage Miller capacitance ( the power tube in this case ) I bet the tube with CCS has an advantage over the tube with plate resistor
One 6SN7 triode with plate resistor has a gain of let's say 15 , with CCS max theoretical 20 , what a huge difference 🙂
And the bottom triode in cascode probably has the same gain as before
For driving next stage Miller capacitance ( the power tube in this case ) I bet the tube with CCS has an advantage over the tube with plate resistor
Last edited:
Gain-bandwidth is an abstract figure of merit of the circuit.
Miller effect is a specific aspect of the circuit operation.
The bottom tube of a cascode has little or no voltage gain, due to the low input impedance of the common grid stage
that it drives. The lower tube's gm is what sets the voltage gain of the entire cascode, along with the load.
This is why a high gm device in the lower position is helpful with a cascode.
Miller effect is a specific aspect of the circuit operation.
The bottom tube of a cascode has little or no voltage gain, due to the low input impedance of the common grid stage
that it drives. The lower tube's gm is what sets the voltage gain of the entire cascode, along with the load.
This is why a high gm device in the lower position is helpful with a cascode.
Last edited:
The voltage gain is A1xA2 like 2 triode stages in series . The bottom triode gain is low because the plate resistor is the upper triode low cathode impedance . This is why the input Miller capacitance of the cascode is small like a pentode , but gain of the two multiplied is still pretty high
The CCS for the top triode has virtually no effect for Miller , gain-bandwidth or whatever you want , only maybe in a high abstraction form , not reality . Especially not for audio
The CCS for the top triode has virtually no effect for Miller , gain-bandwidth or whatever you want , only maybe in a high abstraction form , not reality . Especially not for audio
Last edited:
As in post #34, voltage gain for a cascode is ~ gm x R
where gm is for lower tube section, and R is the total top section anode resistor load.
If the circuit is specifically arranged to have significant voltage gain for the bottom tube,
then the advantage of the cascode, which is low Miller capacitance, will be taken away.
where gm is for lower tube section, and R is the total top section anode resistor load.
If the circuit is specifically arranged to have significant voltage gain for the bottom tube,
then the advantage of the cascode, which is low Miller capacitance, will be taken away.
Last edited:
That is the simplification of the formula , the 2 stages can be treated independently , both triode have loadlines and both gains
As you know changing the anode load for a CCS can increase open loop gain. This can lead to stability problems when NFB is applied. The NFB component may need to be adjusted.
What I remember reading everywhere in EE texts is that resistive loading is the simplest but very performance-limited method to construct a gain stage. Maybe my memory is not what it used to be.
I am sure if someone was proposing R loading, there would be the same vocal group insisting the CCS is much better and cite good-ol' EE-texts.
Joe Roberts should add corollary to his fundamental law "One of the fundamental laws of audio system-building is that given enough space and money, a real audio nut could use up all the space and all the money, then take out a loan to get more space and money and use all that, too."
- 'no matter what is proposed about tube circuitry, the tube nut will always argue that the opposite is better' 🤓
I am sure if someone was proposing R loading, there would be the same vocal group insisting the CCS is much better and cite good-ol' EE-texts.
Joe Roberts should add corollary to his fundamental law "One of the fundamental laws of audio system-building is that given enough space and money, a real audio nut could use up all the space and all the money, then take out a loan to get more space and money and use all that, too."
- 'no matter what is proposed about tube circuitry, the tube nut will always argue that the opposite is better' 🤓
Here's a typical curves for 6SN7. The orange line shows the load line for resistor loaded while the red one shows the load line for a 2 terminal CCS loaded.
Noise could be from different sources. The tube itself, the CCS or resistor, and the power supply section.
A CCS which used as an anode load is usually constructed with Depletion MOSFET cascode with JFETs. A single resistor is used to set the current. All of these parts has extremely low noise figure.
https://refsnregs.waltjung.org/ax_High_Perf_Current_Regs_Revisited_0409_052921.pdf
This is a Cascode CCS research report from Walt Jung.
Data from this page tells that the current noise of such type of CCS is about 14pA√Hz. It will generate 98 nV/√Hz through the Rp of 6SN7 which is about 7Kohms. That will be about 3μVrms when bandwidth is limited to 100KHz. In most of situations, 3μVrms at the output is acceptable in a tube amplifier.
But when talking about noise from the power supply section, the situation get reversed.
A high performance 2 terminal CCS can provide extremely high PSRR which will block any noise from the power supply, while the resistor will just let the noise go into the output signal.
In conclution, which one has lower noise should be discussed in detailed design. We need to figure out which is the mainly element of the total output noise.
Miller effect is actually a R-C Low Pass Filter. The output impedance of this stage and the input capacitance of the next stage. Let's just assume the input capacitance is 100pF. The output impedance is the parallel of resistor or CCS together with Rp of the tube. In a resistor loaded circuit, the output impedance is 7K // 25K = 5.47K. That will form a 290KHz LPF. In a CCS loaded circuit, the output impedance is 7K // ∞ = 7K. That will form a 227KHz LPF.
It's little difference in a tube amplifier.
The bandwidth of the CCS itself is not an issue. You can see from the article that the CCS could maintain over 100dB of PSRR at 100KHz.
Converting the gain difference into log scale. It's still not a big difference in a tube amplifier. Just like changing the tube from a wore out one to a brand new one.
The flaws are merely ignorable and the advantages are obvious. Use it or not, you can listen to your heart. Someone doesn't like these silicons in a tube amplifier, while someone is seeking new semiconductor devices helping vacuum tubes to work in a "better" condition. Convincing each other through words instead of trying in the flesh is meaningless.
