• 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.

More Tube Learning (for me) Cathode followers

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Ok todays lesson was tube buffers. I tried to work out a simple Cathode Follower and a White Cathode Follower. Here are my results. I still haven't figured out what the cathode bypass cap would be on the WCF.... The B+ and rk and tube choice for the regular cathode follower were lifted from one tubecad page and the idea to replace the cathode resistor with a choke was lifted from another. I selected a choke with a dcr equal to the recommended rk. So I believe the CF draws 10ma, not sure about the WCF. My hope would be that one would be worth building to as a buffer for a CDP and later as a active volume control for a tube amp, with the addition of a 1m dual ganged audio pot replacing the 1m resistor at the input. As usual I'm avoiding CCS and the like, I worked out a low noise PSU as well. So here they are, is one of them usable?
 

Attachments

  • cf.jpg
    cf.jpg
    38.7 KB · Views: 472
The plate resistor of 140R might have been correct, if Mu was 84.
Since your 5687's Mu is only 16, value 3K9 may be closer to right.

The top plate will multiply the input by Mu 16.
The 3K9 plate and 1M grid resistor divide by 256 (or 16^2).
The lower plate multiples again by 16, thus following the input.

I am not sure, the cathode bypass cap to GND might be large
enough to temp one to abuse an electrolytic for that purpose...

I'd be tempted to divide the output by 16 and regeneratively
(positive feedback) to that cathode, creating a virtual bypass.
The cap needed to bypass this way would be smaller by factor
of 15. The bypass (feedback) resistor needed would be equal
cathode bias resistor *15 or 4K3.

Since the plate resistor was wrong (possibly scaled for 12AX7?)
I'd doublecheck that your cathode bias resistor isn't similarly off.
If it changes, so does the value needed for bypass.
 
Roughly like this. Though I still havn't done anything to check
if the 287R suggested cathode bias was optimum for 5687.

-----------------------------------------------------------------------------

Occours to me that degeneration of the cathode resistor
reduces the lower triode's effective Mu by 1, would be
easy to compensate for that small difference in the upper
plate resistor value, and use no lower cathode bypass at
all. Less parts for the same result is almost always better.

And occours the me the upper Triode's plate is also Mu 15,
and the missing "16th Mu" appearing at the cathode as
the output of the follower. So my 3K9 was wrong anyway.

4K46 (4K7 might be close enough) at the upper plate and
eliminate the lower bypass altogether. Do retain the bias!
 

Attachments

  • wcf.gif
    wcf.gif
    25.1 KB · Views: 439
Your right about the values, I chopped this out of a circuit that had a gain stage in front of it using a 12AU7. I'm still in the process of learning how to calculate rp and rk for the WCF.

I had an easier time wrapping my head around the simple cathode follower, I just thought I'd try and learn the white cathode follower if I was going to use a twin triode.

I'll try and figure out the rest in the morning! :)

I see you posted an example while I first wrote this, Thanks! I'll take a look when I'm not falling asleep on the keyboard :)
 
From the Raytheon CK5687WA spec sheet, I am thinking that 287R
gives you about 16mA of cathode current and slightly more than 4V
of negative bias (assuming 120V per plate). But this is only half the
5687' recommended design center 36mA for that plate voltage.

Raytheon design center values seem to suggest maybe 55R5 for
36ma and cathode at 2V of negative bias. Maybe that doesn't
give us enough head room for our input signal? I dunno, its a
follower for cripesake...

Maybe next spin I fudge for 25ma idle current? And specs seem
to indicate slightly higher Mu, 16~18 depending on the loadline.
Like weevils in my Cheerios, Devil's always in the details...

----------------------------------------------------------------------------------

So I recalculate 120R for cathode and 3K1 for the plate, then
realize that (plus the cathode resistor) is gonna eat up more
than 80V! Not a problem that much drop, but does throw all
my earlier assumption about quiescent operating point back
into the drink. Up the power supply by another 80V, doesn't
seem necessary, and might hazard the maximum standoff of
the filament. I just need to keep playing with the resistors.
 
I am having difficulty getting idle current much above 15mA with
voltages these filaments will withstand. To run this tube full bore
would want higher power supply, and elevated filament voltage.

If you don't care running at only 1/4 of its true capability, then
15mA at 250V B+ with grounded filaments is certainly do-able...

I have to keep tellin myself, its only a follower. No huge swing...
Transconductance benefits of 5687 will apply to the output even
if I don't run both plates to the limits of dissipation.
 
I seem to remember reading somewhere that a CDP only need around 5v swing MAX, worst case. I'm still digesting your posts :).

And I was also thinking that I could reference the heaters to a portion of the B+ if I needed so I put the parts on the spread sheet.
 
If it ain't broke, don't fix it. Why can't I leave well enough alone?
Oh the misery... Well, it don't look much like it started anymore.
I am hating the parts count, I am not real pleased with where
this circuit variant has gone way too far the path of complexity.

I unstacked the totem to get more plate voltage (and current).
And also to keep the filament stand-off voltage within tolerance.
I am burning both candles as hot as they will go, at 3.75W ea.

Then I stated noticing a bunch of goofy things as I mucked with
the plate and cathode resistor, finally concluding that an equal
split (cathodyne) was perhaps the best compromise.

Then I notice I can opt to get an inverted output by tapping the
other plate and cathode... Its a cathodyne now too, after all?

Then noticing that impedance at the cathodes are not equal the
impedance of the plates, and the coupling cap might be kinda big.
So I made em equal by adding a series resistor to the cathodes.
That also helped resolve an issue about cathode feedback that
was buggin me. It might have been better just use the big caps
and screw making the push and pull impedances be equal?

I think the impedance at either output sum is close to 1K?
And enough headroom to follow and/or split an input signal
maybe as large as 80V Peak to Peak?
 

Attachments

  • wcf3.gif
    wcf3.gif
    33 KB · Views: 389
Ok now you've lost me :) Is this still a 0 gain tube buffer?

I'm not sure what I'm doing wrong. I draw my load line, one end is at 250v the other is at 56ma (250/4k46). 20ma falls on the -6 grid voltage, 6/20 gives me a cathode resistor of 300 ohms and a plate voltage of around 160v.
 
hemgjord said:
Just an observation. Raytheon datasheet states a max gird circuit resistance of 0.1M not 1M. I never really figured out if this is that crucial in CF application. Anyone?


/Olof

You had to go and mention that.... Agh. Starting over...

Well, that correction has little or no effect on the cathodyne variant.
But totally mucks with the classic version, still trying to work out how
this would necessitate changes in the other component values.
 
athos56 said:

I'm not sure what I'm doing wrong. I draw my load line, one end is at 250v the other is at 56ma (250/4k46). 20ma falls on the -6 grid voltage, 6/20 gives me a cathode resistor of 300 ohms and a plate voltage of around 160v.

Two tubes stacked, each seeing half (80v?) the total plate voltage...
Bias voltage divider to the top grid wasn't scaled correctly anyway.
Where the heck did you get 56mA from?
 
I read a page where a guy went through the process of drawing a loadline for his pre-amp, he said that you find the points were current is 0ma, or (0ma,250v) and where voltage is 0v, or (x,0v) he said that x=B+/Rp so 250/4.47=56 (rounded off), I'm beginning to think this page was wrong though...

Maybe I shouldn't get my info from "a page" written by "a guy" :)
 
Yeah, that works where the plate resistor is the only other load
in series with the tube. And the cathode resistor is small enough
to be insignificant, or well bypassed.

--------------------------------------------------------------------------------

Take the cathodyne variant for example, there is 2K above and
below each triode. Thus each triode here sees a 4K loadine.

Fortunately there isn't much bleed across to parallel resistors
over on the other side of the push pull pair to muck this up.
AC voltages on the other side have already been forced equal,
the resistance between nodes of equal voltage will appear to
be infinate, because the current between them would be zero.

The impedance at the cathode(s) is 1/gm, or about 87 ohms.
In parallel with 2K to ground thats about 83 ohms impedance.
I add a 1K8 in series, so the output sees 1K883 impedance.
I am not entirely sure this was a good thing to have done,
but keeps the coupling cap value a reasonable 4.7uF and
more closely matches what impedance the plates can drive.

The impedance of the plates would be 2K if the cathodes were
grounded, but with this much cathode feedback its probably
much closer to infinate. In parallel with a 2K plate resistor to
B+, we have just slightly less than 2K impedance.

2K pulling from the cathode in parallel with 2K pushing from
the plate, we see roughly 1K at each output.

Or at least thats how its supposed to work, I don't actually
know how to run a spice sim to confirm anything. Definately
not quick to build a real circuit to measure, as I still have other
neglected irons in the fire.
 
Administrator
Joined 2004
Paid Member
I think the circuit shown in the second post may not be that far off the mark in terms of component values. I use white followers as outputs in a headphone amplifier with the 5842s which does have significantly higher mu so perhaps I am off the mark.

Bearing in mind that the cathode resistor for the lower tube in the pair sets the dc operating point for the pair I would adjust this value until you achieve the operating current you desire.

The signal voltage at the upper plate should be slightly less than 1/mu x the input signal level - this isn't strictly correct but will get you very close provided there is a bypass cap across that cathode resistor.

I used fixed bias on the lower tube and a very tiny resistor to measure the bias current.

Do use 100K grid resistors with the 5687.. (Although I usually get away with twice this without problems.)

Use a large coupling cap in the lower tube grid circuit if you want to maintain low Z all the way to the LF cutoff.

Morgan Jones Valve Amplifiers 3rd edition has some interesting comments and some math analysis of the White CF. It is not so easy to analyze.. :D
 
athos56 said:
I read a page where a guy went through the process of drawing a loadline for his pre-amp, he said that you find the points were current is 0ma, or (0ma,250v) and where voltage is 0v, or (x,0v) he said that x=B+/Rp so 250/4.47=56 (rounded off), I'm beginning to think this page was wrong though...

Maybe I shouldn't get my info from "a page" written by "a guy" :)

I think I should have said Ra. I found some more info while reading. (http://www.tubecad.com/october99/page7.html) It says that Ra=Rp/mu. So according to the spec sheet Rp at 250v is 3k, divide that by mu (16), and I get 187.5 for the Ra.

KevinKr, I should use a value higher then 1uf? Something more like 2.2uf, 3.3uf, or 4.7uf?

I'm still reading for hints on the other values, thanks for the replies, even if I don't understand them the questions the pose lead me in the right direction. :)
 

Attachments

  • wcfb.jpg
    wcfb.jpg
    48.6 KB · Views: 259
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.