Unconditional stability, phase shift and feedback

[Don Ericksen]

I was reading a bit of Morgan Jones and his comments on stability and amplifier designs. He compares the Quad amplifier with the Williamson for stability and explains that the Quad with just two stages was easy to stabilize whereas the Williamson with its preamp, concertina splitter and drivers was a considerable challenge.

Let's just say I have decided to build an amplifier with a triode voltage amplifier, a longtailed pair phase splitter and follow that with cathode follower drivers-or even regular plate loaded ones-and that I propose to put feedback across the whole mess. First, is there any reason this is particularly bad and secondly, is there an iterative procedure i should follow to determine how much feedback I may use and how to figure out all the poles and zeroes to stabilize it? I do want to accomplish a couple of objectives, I want a capacitor between the front end and the first input tube grid for DC blocking, and I want of course a flat bandpass between 20 and 20 kHz but I would particularly like to roll off the lows below 20 Hz.

[ray_moth]

You can improve the chances of stability considerably by using direct coupling between the first stage and the LTP and again between the cathode followers and the OP tube grids.

[Yvesm]

You can also simply remove the first stage by using penthode LTP

http://www.dissident-audio.com/PP_6L6/Page.html

Drastic !

Yves.

[EC8010]

As Ray and Yves have both said, reducing the number of LF time constants by DC coupling and minimizing the number of stages is a Good Thing. You will have to adjust/optimize HF stability on test because it will largely be determined by your particular output transformer. Don't forget that HT supplies can form unwanted LF feedback paths - sometimes a regulator is a good thing. Since you don't need much current for the LTP (because it has cathode followers to buffer it from the output stage), regulation to that stage could be as simple as a couple of 150V neons.

Yves: Those are impressive 20kHz square waves on your 6L6 amplifier.

[Eli Duttman]

Don,

I want a capacitor between the front end and the first input tube grid for DC blocking, and I want of course a flat bandpass between 20 and 20 kHz but I would particularly like to roll off the lows below 20 Hz.

Blocking DC is important. Even more important is protecting the O/P trafo, which is inside a NFB loop, from core saturation. A RC high pass filter tuned to approx. 19 Hz., at the I/P grid, gets both jobs done.

Let's just say I have decided to build an amplifier with a triode voltage amplifier, a longtailed pair phase splitter and follow that with cathode follower drivers-or even regular plate loaded ones-and that I propose to put feedback across the whole mess.

What sort of O/P tubes do you have in mind? How much gain do you need? FWIW, my hunch is that the Mullard circuit using a 12AT7 in the LTP position will do the job for you. High gm is indicated for a LTP, especially in a circuit that employs loop NFB. High gm is protection against slew limiting.

If you decide to use voltage followers that are DC coupled to the "finals", I suggest you consider IRFBC20 MOSFETs for the job. The FET sounds good, requires no heater power, and (obviously) there is no heater to cathode potential issue. MOSFET Follies illustrates the technique.

[Don Ericksen]

Originally posted by Eli Duttman
Don,



Blocking DC is important. Even more important is protecting the O/P trafo, which is inside a NFB loop, from core saturation. A RC high pass filter tuned to approx. 19 Hz., at the I/P grid, gets both jobs done.



What sort of O/P tubes do you have in mind? How much gain do you need? FWIW, my hunch is that the Mullard circuit using a 12AT7 in the LTP position will do the job for you. High gm is indicated for a LTP, especially in a circuit that employs loop NFB. High gm is protection against slew limiting.

If you decide to use voltage followers that are DC coupled to the "finals", I suggest you consider IRFBC20 MOSFETs for the job. The FET sounds good, requires no heater power, and (obviously) there is no heater to cathode potential issue. MOSFET Follies illustrates the technique.

I was wanting to build an all triode amp just as a design exercise and not because I have anything against other devices. Would those FETs work in the form of a "Fetron" to replace the cathode follower triodes in the McIntosh 75, where there are harsh requirements?

[Don Ericksen]

Originally posted by Yvesm
You can also simply remove the first stage by using pentode LTP

http://www.dissident-audio.com/PP_6L6/Page.html

Well, yes.

[tubetvr]

is there an iterative procedure i should follow to determine how much feedback I may use and how to figure out all the poles and zeroes to stabilize it?

It seems to me that no one has answered this part of your question, as the answer is complicated I will not try to give you a complete answer but rather push you in the right direction. Yes, there are methods how to determine how much feedback to use and how to place poles and zeros in order to make an amp stable. In general an amp is unstable if the loop gain at some point has 180 degree phase shift because at that point the negative feedback will be positive and the amp will oscillate. The solution is either to use less feedback or to change the phase response by either moving poles or by adding zeros.

The easiest method on how to see if there is a risk of oscillation and what effect you get when changing poles and zeros is by a Baude diagram where you graphically plot amplitude and phase for each pole.

You can find how to do this in many basic text books on feedback systems. Another more modern method is to use a simulation tool like spice which can give you very quick results but it doesn't give you the same feel for what's going on.

As an example the original Wlliamson amplifier has quite low stability margin at low frequencies but it is also easy to rectify by introducing a phase stabilisation network, see here http://www.tubetvr.com/Williamson.pdf and here http://www.tubetvr.com/Williamson_comp2.pdf

Regards Hans