Bunch of questions regarding plate CCS and some else.

[Nevod]

Well, there is some more questions beside I would like to get a high output Z, high gain amp - an one suitable for ESL (there seems a possibility that they would actually need a current input for flat response). I'm not planning to make one, not in near future - it'd be an expensive project, so I'm just trying to put my brains to the limits.

I've once read that an ESL amp must be similar to 'scope vertical deflection amplifier, and from what i've heard, they were often using cascodes with plate CCS to achieve high bandwidth, so i'm quite intrested in VT scope circuits, but haven't found pretty much any circuit, and have decided to try and come up with comething myself.

Reading some old books(particularly MIT's Radiation Laboratory's #18, Vacuum Tube Amplifiers) and the forum, I've come to think that plate CCS is, simply speaking, good. Recently, I revisioned them somewhat more deeply, and found that actually, there is a problem with loading a plate CCS'ed cascade.

The problem is that common tube CCS, actually, only keeps current constant through itself. If a tube beneath is loaded, current will split between tube and load, making loadline non-flat and reducing gain. There is a circuit which seems to handle that - the active plate load, where output is taken from top tube's cathode, but now, current through the top tube is non-constant, plus there are some side effects. And I don't have enough expirience, knowledge, or inagination to think up which one would yield better results, be more linear/fast/stable.. That is the primary question - the configuration of CCS - which is the best?

And well, the second question is a little less theory: While trying to find a suitable tube(s), i've found a quite interesting one - SV572-160. While not exacly a gem, at least in my opinion - being mostly designed for grid currents, and being directly heated - it shows pretty enough plate voltage - up to nearly 3kV, 125W dissipation, and mu of 160 - there are actually several variants with different mu. Set them into cascode, and you get staggering mu of about 25k. Of course, mu is not gain itself, but with a plate CCS gain could be made close to that. However, so much gain is not needed - about 1k would be quite enough. (though it doesn't seem to be very correct to speak about voltage gain in current-out circuit, but, as output voltage swing will be present, i mention the gain) Is it possible to use the gain to linearise the circuit, to make it more tolerant to tube characteristics without applying global NFB? There are degen resistors, but from what i've learned, their use is quite limited. Plus, as input capacitance seems to be low - about 7pF - is it possible to make a single-stabe amp this way? Provided that there's no grid current, of course.

Sorry for sumburous post, i'm a little in haste writing it. Will try to explain what i wanted to know later.

[exeric]

Hi Nevod,
I think we are on a similar path here about trying to implement a CCS and basically for the same reason, though I won't be using it for ESL. I think you might find these two web pages very useful.

The first is a tutorial about CCS and their implementation by Gary Pimm. If I'm not mistaken he actually worked for Tektronics but doesn't reference that or make any mention of scopes in the article. The second web site is about feedback and its use in stabilizing amplifier circuits to the point that even electrostatic speakers are stable. I think you'll find them interesting.

http://www.pacifier.com/~gpimm/Activ...t_control.html

http://www.normankoren.com/Audio/FeedbackFidelity.html

[SY]

Is the 572-160 still available? I thought that series was discontinued.

In any event, if the load (other than CCS) is more than about ten times the plate resistance, the load line will not be disturbed too badly off horizontal. And if a CCS-loaded stage is direct coupled to a cathode follower, its operation will be hardly affected.

Gary Pimm's CCS are excellent, true overkill engineering. I've had very satisfactory results with bipolar cascodes and depletion-mode MOS cascodes.

[Nevod]

exeric
Thanks, it's useful, though have to adapt it to tubes.
SY
Don't know about wheter 572-160 are produced, but they are being sold by New Sensor. Where you've got info that they are not produced?
As for load - that is one of the problems - I want a high output Z amp, so no cascode after CCS, only the panel..

[SY]

I can't remember where I read that, but you're right, they're still on New Sensor's catalog page.

[Tweeker]

How would you drive the grid of the top valve in a 572-160 cascode? Grid current is required with such mu.

[Nevod]

Tweeker
Well, I suppose it is still possible to apply negative bias ang get no grid current, though judging from curves, it won't be linear that way. I don't really see ony other way than a resistive divider to apply bias.

Anyways, as it seems that, really, SV572 (probably the whole series) is relly no longer produced as unprofitable, and old similar tubes like G-811 (= just 811) are hard to find, an HV amp would be restricted to low-mu RF triodes like various 3CX250s and the like. With these tubes it's not possible to obtain a high Z w/o feedback, So it seems that I have to really dive into current feedback. At least around output stage, the feedback is needed.

While I'm not into the esoterical approach to sound, I tend to listen to arguments of both sides and consider what may really be damaging to sound. Quite a lot of tubers say that NFB damages sound, one of the explanations I've heard is that it leaves residues of "previous" signals in the current signal. That seemed logical, but I didn't investigate further, as I didn't want to mess with NFB then.
Recently, I've read on this forum about measurements done on distortion created by NFB, read about TIM, and its relationship with amp's bandwidth. Then i found that most modern amps have no problems with TIM, having bandwidth good enough. And just now I've came to thought that most tube audio amps seem to have quite low bandwidth (not completely sure about that, but I've read about falloffs at 15khz and even less), and that very well may be the cause of tubers dislike of NFB.

So, in a fast, high-bandwidth amp there will be nearly no phase lag in audio domain, hence no "residual" signals. Sum of all these residues will anyway be finite, like THD is finite.

If you think that this post has been written with aim of self-justification, you're right.

[DougL]

Several random thoughts for you.

ESL speakers are a nearly perfect capacitive load. It is not uncommon for them to reach 2 uF.

In order to drive them, you need enough current to charge / discharge the current at the highest frequency and the highest voltage swing.

http://www.tubecad.com/november99/page3.html
Has a formula for the peak current.
In the example, it took 8 ma to swing 500 volts into 130 pF at 20K hz.

If my math is correct, to swing 1000 V into 2 uF at 20KHz, it would require 251 mA

If you are interested in tube electrostatic amps, look at the designs at Headwise Headwise
as a starting point.

Good luck.

Doug

[Bandersnatch]

hey-Hey!!!,
On the cascode, it is plate curves past 2x the Ec2 is that of a pentode, where gain is gm*load. Get below that number and you're into the 'here be monsters' region. The cascode creates a pentode that delivers a *VERY* high impedance g2( effective that is ), instead of one that requires significant input current.
cheers,
Douglas

[Nevod]

DougL
My point is that, by Walker's equation, ESL's SPL depend on current, not voltage. While in Baxandall's book there are additional corrections to this formula, most measurements show that this is more or less the case - with voltage drive, sensitivity rises on HF. On the other side, of course is the fact that these ESLs are transformer-coupled, and as such, obtained parameters are not readily applicable for an direct-driven panel.

Still, to get current drive, you need high output Z, and besides, a current-driven ESL consumes most at LF - where impedance is high, and same current produces more voltage. With headphones, it is different, as directivity is not applicable to them, and essentially, all the acoustic power emitted by them goes into ear.

Moreover, i've once read an article which stated that current-driving ESL would produce more phase - correct sound, even allowing to reproduce square wave.

Bandersnatch
Not really got your thought.. By Ec2 you mean potential at lowe tube's cathode? What exactly is in the "monster" region?