Hi Ray_moth,
It simply cannot make much of a (noticable) difference changing the driver stage more towards a differential amp operation by "lenghtening" the tail, because AC symmetry is excellent already due to the preceeding concertina stage.
There are several small tweaks possible on the Williamson that gain much more and better improvement (both measurable and soundwise) than giving the driver stage a CC sink at the cathodes.
Everbody eager to improve on the original Williamson first should carefully read and digest the arcticle "Improving the Williamson Amplifier" by Talbot M. Wright, Electronics World, 1961.
You can find scans of that article here.
Regards,
Tom
P.S.: In that article, don´t miss the additional small cap labeled C11 and think closely about its function - and what this in turn means regarding your question
ray_moth said:
It simply cannot make much of a (noticable) difference changing the driver stage more towards a differential amp operation by "lenghtening" the tail, because AC symmetry is excellent already due to the preceeding concertina stage.
There are several small tweaks possible on the Williamson that gain much more and better improvement (both measurable and soundwise) than giving the driver stage a CC sink at the cathodes.
Everbody eager to improve on the original Williamson first should carefully read and digest the arcticle "Improving the Williamson Amplifier" by Talbot M. Wright, Electronics World, 1961.
You can find scans of that article here.
Regards,
Tom
P.S.: In that article, don´t miss the additional small cap labeled C11 and think closely about its function - and what this in turn means regarding your question
ray_moth said:
It should do , to date I have never found a diff stage not benefitting from a high impedence tail . I have had very good results with 6SN7/6J5/7N7 with CCS loads for both the anode and the tail . For the top loads connect a high value resistor (2.2M etc) across the CCS and for the tail , use a cascode , adjust this in circuit for slightly higher current than that of the combined top loads
cheers
316a
Yes, I did read it and I didn't see anything about the balance of the driver stage becoming uncritical. I think I agree with his points about the various stages being under-biased, but even that seems to be controversial - I've asked about it before and received a mixture of opinions.
Two schools of thought seem to exist, namely, that
- a CCS is unnecessary in a differential stage with balanced inputs; or
- that a CCS improves the stage, giving better balance.
Either the problem is insoluble and both parties are guessing; or one of the parties is plain wrong; or, as I suspect, a CCS does help to overcome any imbalance in the gain.
Re: Re: Would a Williamson sound better with a CCS tail in the driver?
I don't think that this can work properly. A diffamp with a CCS in the common cathode tail needs to have a changing current in the anodes. A CCS in the anodes cannot work IMHO. Thats against the principle of such a stage.
Regards, Simon
316a said:
I don't think that this can work properly. A diffamp with a CCS in the common cathode tail needs to have a changing current in the anodes. A CCS in the anodes cannot work IMHO. Thats against the principle of such a stage.
Regards, Simon
Good article but to be fair it was written decades before commodity hardware permitted close examination of the distribution of distortion components. The circuit still relies on the 'let heavy feedback sort it out' philosophy. My suggestion would be to open the loop and measure the CCS vs. resistor tail spectra before deciding.
I also didn't see how it could work at first. Now, I think I do understand.
The way I see it is that a triode has three inter-related parameters that define its operating point-
* grid-cathode bias voltage;
* plate-cathode current; and
* plate-cathode voltage.
If you fix any two of these, the third will fall into line (within the limitations of the tube).
With a differential amp having both cathode and plate CCSs, the balanced input signals drive the two grids in anti-phase. The tail CCS ensures that the sum of their plate-cathode currents is constant. The CCS plate loads ensure that the individual currents are also constant. The only thing that can change, in response to the input signal, is the plate-cathode voltage - and that's all we're asking for. That is our output signal.
The only potential problem I see is conflict between the CCS requirements if they're not set up properly (addressed by 316A's post) or if they drift.
Somebody please tell me if my thinking is a**-backwards!
It's not a**-backwards - at least not according to LTspice. I tried modeling the following simple all-balanced amp:
Stage 1: 6SL7 LTP splitter, with cascoded transistor CCS in the tail and single transistor CCS in each plate load, bypassed by 2.2 Meg resistor as 316A suggests. This is pseudo-direct coupled (using step networks to get sensible operating points and to provide stability for global NFB) to-
Stage 2: 6SL7 differential driver, again with cascoded transistor CCS in the tail and single transistor CCS in each plate load, bypassed by 2.2 Meg resistor. I could obviously have used 6SN7 for Stage 2 but I needed good open-loop gain for global NFB. Stage 2 is indirectly coupled to-
Stage 3: 6SN7 PP cathode followers, with negative grid bias and a cascoded transistor CCS as the cathode load to the negative line of -115v for each triode. This is directly coupled (to avoid the risk of blocking distortion) to-
Stage 4: PP EL34s in true pentode mode.
There is cross-coupled NFB from the EL34 plates back to the LTP splitter plates and global NFB from the OPT secondary to the grounded grid side of the LTP splitter. There are a couple of HF correction networks across the plate loads of Stage 1 and across the NFB resistor, and staggering of the time constants due to the coupling caps between Stage 1/Stage 2 and Stage 2/ Stage 3, to keep the whole thing stable at HF and LF respectively.
It models beautifully - good operating points, perfect balance and very low distortion. Of course, in the real world it would not be so perfect and I'd be inclined to use 6SU7s instead of the 6SL7s for balance purposes, but I feel it's a good indication that it is at least feasible.
I know this is exteme (crazy?) use of CCSs, but I've seen recommendations for every one of these applications at one time or another and I wanted to try modeling with them all in the same amp. And transistors are very inexpensive!
Stage 1: 6SL7 LTP splitter, with cascoded transistor CCS in the tail and single transistor CCS in each plate load, bypassed by 2.2 Meg resistor as 316A suggests. This is pseudo-direct coupled (using step networks to get sensible operating points and to provide stability for global NFB) to-
Stage 2: 6SL7 differential driver, again with cascoded transistor CCS in the tail and single transistor CCS in each plate load, bypassed by 2.2 Meg resistor. I could obviously have used 6SN7 for Stage 2 but I needed good open-loop gain for global NFB. Stage 2 is indirectly coupled to-
Stage 3: 6SN7 PP cathode followers, with negative grid bias and a cascoded transistor CCS as the cathode load to the negative line of -115v for each triode. This is directly coupled (to avoid the risk of blocking distortion) to-
Stage 4: PP EL34s in true pentode mode.
There is cross-coupled NFB from the EL34 plates back to the LTP splitter plates and global NFB from the OPT secondary to the grounded grid side of the LTP splitter. There are a couple of HF correction networks across the plate loads of Stage 1 and across the NFB resistor, and staggering of the time constants due to the coupling caps between Stage 1/Stage 2 and Stage 2/ Stage 3, to keep the whole thing stable at HF and LF respectively.
It models beautifully - good operating points, perfect balance and very low distortion. Of course, in the real world it would not be so perfect and I'd be inclined to use 6SU7s instead of the 6SL7s for balance purposes, but I feel it's a good indication that it is at least feasible.
I know this is exteme (crazy?) use of CCSs, but I've seen recommendations for every one of these applications at one time or another and I wanted to try modeling with them all in the same amp. And transistors are very inexpensive!
Ok ok ok ok ok...
Besides this practical experiences (which are quite good as I see) there is theoretical aspect which I will face now.
A differential amplifier works because of something called current distribution in the two arms. Lets assume that the Tubes share a common current of 10mA. Each tube has a quiescent current of 5mA. With the grid, we can change this current. When we change the current of one Tube to 3mA, the other will automatically conduct 7mA. The higher the Resistance at the common cathode, the better this regulation. So a CCS in the tail is definitely a good thing.
As I told this works with current sharing in the anodes. When you now place a CCS in each anode, the current will always remain 5mA in each arm. We can change the anode voltage of the Tube whose grid voltage we change, but the other tube won't do anything, because there is noch changing current anymore which would cause the other Tube to conduct in anti phase.
So this can't work.
Now, this refers to ideal constant current sources which have an R=∞.
Because normal CCS never can achieve such a condition, the circuit will work anyway.
Another thing because this would even work with ideal CCS is, that there is also the input resistance of the next stage at the anodes of the tubes.
If I'm wrong in my theory, please correct me
Regards, simon
Besides this practical experiences (which are quite good as I see) there is theoretical aspect which I will face now.
A differential amplifier works because of something called current distribution in the two arms. Lets assume that the Tubes share a common current of 10mA. Each tube has a quiescent current of 5mA. With the grid, we can change this current. When we change the current of one Tube to 3mA, the other will automatically conduct 7mA. The higher the Resistance at the common cathode, the better this regulation. So a CCS in the tail is definitely a good thing.
As I told this works with current sharing in the anodes. When you now place a CCS in each anode, the current will always remain 5mA in each arm. We can change the anode voltage of the Tube whose grid voltage we change, but the other tube won't do anything, because there is noch changing current anymore which would cause the other Tube to conduct in anti phase.
So this can't work.
Now, this refers to ideal constant current sources which have an R=∞.
Because normal CCS never can achieve such a condition, the circuit will work anyway.
Another thing because this would even work with ideal CCS is, that there is also the input resistance of the next stage at the anodes of the tubes.
If I'm wrong in my theory, please correct me
Regards, simon
Simon,
Well I cannot speak of others' circuits, but of course part of the idea of my pentode cathode followers was to ensure a nice high input resistance. It is well known that diff pairs do not work well into a stage where input resistance is reduced on one side, e.g. AB2 or class B.
As for the issue of constant current, well, you make a reasonable point I agree, but then it worked! I expect someone cleverer than I might be able to explain this apparent paradox.
7N7
Well I cannot speak of others' circuits, but of course part of the idea of my pentode cathode followers was to ensure a nice high input resistance. It is well known that diff pairs do not work well into a stage where input resistance is reduced on one side, e.g. AB2 or class B.
As for the issue of constant current, well, you make a reasonable point I agree, but then it worked! I expect someone cleverer than I might be able to explain this apparent paradox.
7N7
Simon, it does work. Remember, for a triode the plate current versus plate voltage curves are not flat. That means that even if both sections are drawing the same current, the plate voltages will not be the same if the grid-to-cathode voltages are different.
Let's imagine a circuit where the zero-signal plate voltage is 150V, the mu (defined as the change in plate to cathode voltage divided by the change in grid to cathode voltage at constant current) is 10, and the current is 5mA per section. With one input grounded, apply 1V to the other grid. By symmetry, there will be 0.5V at the common cathode. So the grid to cathode voltage of one section is 0.5, the other is -0.5. See where this is going...? One plate will swing down by 5V, the other will swing up by 5V.
Let's imagine a circuit where the zero-signal plate voltage is 150V, the mu (defined as the change in plate to cathode voltage divided by the change in grid to cathode voltage at constant current) is 10, and the current is 5mA per section. With one input grounded, apply 1V to the other grid. By symmetry, there will be 0.5V at the common cathode. So the grid to cathode voltage of one section is 0.5, the other is -0.5. See where this is going...? One plate will swing down by 5V, the other will swing up by 5V.
Ahh I only considered the current distribution. But didn't consider that the cathode always "follows" the grid (in a single stage with a CCS in the cathode). Of course, with 2 tubes and the second grid grounded the cathode voltage will be constant and also the plates will swing symmetrically. Even if the current is constat in both sections...
Thanks for your explanation !
Thanks for your explanation !
the_manta said:
The reason it works is because the cathode voltage is *not* constant, but follows the driven grid. Actually, it follows with a gain of 1/2. This is true whether there are CCS's in the plates or not. It's an important detail of the LTP. In fact, it's the most important detail in understanding how they work, no matter how the plates are loaded. It's how signal gets 'transfered' to the undriven side.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.