• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

phase splitter issue

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Altec phase splitter

"trick with such coupling of concertina to LTP was also used in one Altec amp, later they gave it up I suppose because such a way it works like APF on very low frequencies."

Altec used this trick in many of their PP amp models. I like that it keeps the coupling DC low at cathode level to the next stage but I often wonder if the two phases are truly symmetrical or if the loads are identical. Are they? Or are you better off just use a classic Williamson?

Altec 1568A schematic
9713B9A1-18D5-420B-AF1A-08D7D903AF8F.jpeg
 
I'll try to, but it might take up to a week to get to it as I've got other things on my plate right now. If no one else gets to it before that I definitely will do it.

Thanks, I think it will be interesting since you built a very solid model for the 6sn7, the 6sl7 takes less heater and current by comparison, in the same setup how would it perform?
 
Altec used this trick in many of their PP amp models. I like that it keeps the coupling DC low at cathode level to the next stage but I often wonder if the two phases are truly symmetrical or if the loads are identical. Are they? Or are you better off just use a classic Williamson?

Here is another version, with 2 equal RC time constants, and no APF effect on lowest frequencies.
 

Attachments

  • 33552613-7FD0-4715-9C1F-A9A5BAD576D3.jpeg
    33552613-7FD0-4715-9C1F-A9A5BAD576D3.jpeg
    84.8 KB · Views: 110
I decided to tackle the problem right away because I was interested also in the 6sl6 concertina. Here's the 1/4 3/4 version:
concertina 6sl7 one quarter 3 quarters.jpg

And here's the 1/3 2/3 version:
concertina 6sl7 one third 2 thirds.jpg

I've found that there's more linearity, at least in simulation, with current on the low side. So I went with 1 ma in both sims. Whether that's close to starving it I don't know, but distortion went way up with more current.
 
Here's the FFT for 1/4 3/4 biasing with the max output at 60volts out and 120 v P-P:
FFT at  60volts 120V P-P in.jpg

That's the max good signal out that I could get in sim. It's not that impressive compared to 6sn7 that I did which had a good signal out at 180 volts P-P. Of course that was based on a 500 v rail and this is only 400v. Still...

Here's the FFT at 124v P-P:
FFT at  62volts 124V P-P in.jpg

It only gets much worse at signals above that. It's standard clipping.
 
Here's the FFT at the 1/3 2/3 setting. It's pretty bad. Notice that even though the bias changes from 1/4 to 1/3 and 3/4 to 2/3 the high quality signal out changes from 120v P-P to 64v P-P.

View attachment 609862

Here's the FFT where things start to go bad:
View attachment 609863

I guess I have a slight mea culpa. This is hard clipping. But perhaps I can be excused for thinking it's not. It does not scale with the change in bias level as you would presume. It probably has to do with the fact that at the bias point in the 1/3 2/3 case there is only 130 volts on the tube's plate to cathode with no signal. In the 1/4 3/4 case it's 200 volts. So if you can't get close to the 1/4 3/4 point (I'm looking at those will adopt direct coupling) then change your topology.
 
So, a cathodyne likes high rail voltages if you intend no more amplification stages after it, besides the final. The more the better almost.

Also, they seem to have a dominant 2nd harmonic and little else until clipping. They should sound different than differential stage phase inverters. Maybe that was a reason people liked the Williamson so much, who knows. Stranger things have happened. People always think they know more than they do. It's not just me.:eek:

(But I admit it, unlike some):p
 
For what it's worth, here is a plot showing THD against Vrms for a 12AT7 cathodyne, 400V B+, 39k anode and cathode load. Lines for various bias points including 1/5, 1/4 and 1/3 B+. Lines extend to near the clipping point (grid current or running out of headroom). Log axis for THD to better show differences, data from LTSpice.

For this particular example it would seem that somewhere around 1/5 B+ gives the highest output but if a high voltage swing is not required (e.g. driving EL84 or 7591) then there is no problem with a higher bias voltage, and there may be some reduction in distortion. Obviously this would help if direct coupling from the previous stage.
 

Attachments

  • cathodyne distortion.jpg
    cathodyne distortion.jpg
    89 KB · Views: 324
As I'm sure you're aware, the large amount of local feedback in a cathodyne makes it quite insensitive to load, within reason. For 80V bias, varying the load resistors from 22k(3.8 mA) to 56k (1.5 mA) hardly affects distortion. The point was that there is an "optimum" bias (as a percentage of B+) if you need to maximise output swing (there is no getting away from that), but often the required swing is defined by the output stage, so the bias point can be chosen to suit the overall design rather than following rigid rules.

Of course using a higher gm tube will give better results. I used a mosfet in my latest design.
 

Attachments

  • cathodyne distortion 2.jpg
    cathodyne distortion 2.jpg
    74.5 KB · Views: 301
I agree with most of what you said. I disagree that a cathodyne is insensitive to load. The upper load is very sensitive to any current being driven though its load and will become unbalanced with the lower load when there is current. The simulation I did was without any external load attached to either upper or lower resistors. That's why I described it as the "theoretical" capability of a cathodyne. You would be safe in your statement if use mosfets as loads as you are planning, either source followers or straight mosfet amplification. Driving cathode followers also keep current out of the loads and will keep it balanced.
 
Last edited:
Yes, I meant varying upper and lower resistors. All the simulations were done driving 180k resistors after the coupling capacitors. Others have shown that uneven current draw is what upsets the cathodyne, not current draw itself, at least at low levels.

I used a mosfet instead of a triode in the cathodyne itself (a "sourcodyne" if you will) as I didn't have a spare hole in the top plate. Allowed a higher anode voltage in the preceding direct coupled driver stage. I made sure the output stage (pentode 7591) would be clipping well before the phase splitter. Works well.
 
With the passing of time, it has become apparent to me that a more complete and correct Improved model of the Cathodyne exists than has been presented to date. Here’s a schematic of a generic Cathodyne followed by that of the Traditional model which has been discussed extensively in this thread.


attachment.php

Cathodyne Schematic​

attachment.php


Traditional Model​

The Traditional model has several limitations and deficiencies:
  1. It does not account for B+ supply noise.

  2. It excludes ground and all connections to it. Therefore, it can’t make justifiable claims as to the anode (a) and cathode (k) voltages with respect to ground – only with respect to one another. Nor can it account for the ground-referenced a and k impedances Zag and Zkg, which explain the differing susceptibilities of those terminals to stray electromagnetic power line fields.

  3. Its form falsely implies that the Cathodyne is a balanced circuit. But if the Cathodyne were truly balanced, its outputs’ B+ supply noise voltages would be identical, as would its Zak and Zag impedances. But a simple ‘scope measurement confirms different noise voltages. And the Improved model, as well as Thevenin’s Theorem (see the “Impedance Controversy” attachment), reveals different impedances. For an example of a truly balanced differential driver, consider a long tailed pair whose inputs are driven from equal and opposite signals.

  4. A common version of this model claims that e = eg · u / (u+1) and z = 1/gm. But simulations (see the LTSpice attachment) reveal that this is a not terribly good approximation of the actual circuit. The exact expressions are a bit more complicated:

2 · e = eb · [ 1 – ( Za + Zk) / Ztot ] - u·eg · ( Za + Zk ) / Ztot

and

2 · z = ra || [ ( 1 + Za/Zk ) / gm ] || ( Za + Zk )

where

u = ra · gm, and Ztot = ra + Za + ( 1 + u )·Zk​

But even the use of the exact expressions does not mitigate the first three issues listed above. For that, the Improved model of the Cathodyne is required:

attachment.php


Improved Model​

The Improved model offers the following:
  1. Its simulations are identical to those of the Cathodyne from which it is derived (LTSpice attachment again).

  2. It applies to either a Cathodyne (Za = Zk), a Cathode Follower (Za = 0) or a Voltage Amplifier (Za > Zk).

  3. It clearly depicts the Cathodyne outputs’ unbalanced nature: unequal single-ended impedances, and unequal (supply noise) voltage magnitudes. Of course, the outputs’ portions of grid-excited voltages are still equal and opposite. The Traditional models are incapable of demonstrating either of the unbalanced features.

The development of the Improved model requires only the application of Kirchoff’s Conservation of Current Law, Thevenin’s theorem, and a little bit of algebra. It’s accomplished without employing any test loads, rendering moot a prior objection to such (see the “Impedance Controversy” attachment.) The “Model Development” attachment to this post is self-descriptive. The final attachment is an LTSpice model which supports simulations of the Cathodyne schematic and all of the models discussed.

In conclusion, the Improved Model corrects certain approximations and flawed, misleading and/or incomplete aspects of the Traditional Models. It accounts for the effects of grid excitation as well as B+ supply noise, develops exact (and unequal) single-ended output voltages and impedances, and demonstrates that the Cathodyne is not a balanced circuit. The attached derivation also gives us something else that has been missing from many Traditional Models which are simply declared without supporting documentation – a clear path of development from a schematic to its model.

Comments are of course welcomed. But even if there are none, the Improved model is now out there for future reference, and maybe I can finally stop thinking about this darn thing!
 

Attachments

  • Cath_sch_resize.jpg
    Cath_sch_resize.jpg
    69.7 KB · Views: 228
  • Y_Model_resized.jpg
    Y_Model_resized.jpg
    33.7 KB · Views: 239
  • resized solved_thev.jpg
    resized solved_thev.jpg
    121.5 KB · Views: 223
  • Model Derivations.pdf
    175.1 KB · Views: 46
  • The Impedance Controversy.pdf
    99.5 KB · Views: 84
  • diyaudio_body.asc
    12.8 KB · Views: 40
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.