• 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.

CFB Issues

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
After some successful experiments with PL504 SE and CFB I feel that it´s time to take it one step further and use real triodes instead of TV tubes.
I´m thinking 6AS7, one system per channel running at 150V, 80mA and 2k with about 33% CFB.
(I´m well aware of the driver stage problems but that does not belong here)

The question is: If I want 2k plate load and 33% CFB, should I wind an 2k OPT and "cut" the winding at 33%, using 2/3 on the plate and 1/3 on the cathode or should I wind a 2k winding plus an extra cathode winding that is 33% of the primary?
 

Attachments

  • 6as7opt.jpg
    6as7opt.jpg
    6.7 KB · Views: 478

PRR

Member
Joined 2003
Paid Member
> use real triodes instead of TV tubes.

TV tubes ARE real tubes.

> I´m thinking 6AS7

And the hasty-design 6AS7 is, IMO, not as good/useful/linear a tube as many of the TV V-sweep triodes. In 1949, it was 6AS7 or nothing. When TV production got big, they had to use sharper pencils.

> 6AS7... with about 33% CFB.

What the heck are you going to drive that with? Must be a lot of grid volts there. ~35V bias? 45V peak cathode swing? 35+45= 80V peak, need like 400V B+ in a resistance-coupled driver.

And I'm wondering how much leverage you get with Mu=2.7 or whatever it is. Plate-loaded, Zout is about 280 ohms; cathode-loaded about 100 ohms. Plate loaded, THD will push 10%; cathode loaded about 3%. So 33% in between gives Zout like 200 ohms, THD about 8%. Not a lot different from simple plate-loading. However I may be quite wrong; I often screw-up calculations for very-small NFB. 6AS7 does not have enough gain to give much NFB.

> If I want 2k plate load and 33% CFB, should I wind an 2k OPT and "cut" the winding at 33%, using 2/3 on the plate and 1/3 on the cathode...

I think I might call that 50% CFB; not sure; splitting hairs here. However, the cathode sure provides some output power. The total P+K winding should be like you would use for simple plate-loading, then break-off a part of that for the cathode's use.

> wind a 2k winding plus an extra cathode winding that is 33% of the primary?

That gives a 3.5K load on the tube. If you think 2K is best, then this is wrong. Probably does not make a bit of real difference, since triode loading is so uncritical. Either 2K or 3.5K plot about as well as you can do with that tube working at just 150V.

If you have enough drive voltage, you might try it both ways: big winding in plate as proposed, and big winding in cathode with the small winding in the plate. The latter gets close to cathode-follower operation without going all the way, and not needing such brutal drive voltages.
 
Well, I had another look at the tubes I had in mind for driver stage and it seems that none of the is capable of the swing that is needed, so let´s forget about 6AS7...:(

The total P+K winding should be like you would use for simple plate-loading, then break-off a part of that for the cathode's use.

Thanks, that what I wanted to know!

Gotta take a look in my tube inventory to see if I can find a more suitable piece for this job.
From the top of my head:
807
PL519
PL504 (Been there, done that)
PL36
EL90

Maybe 807 UL with ~25% cfb?
 
Ex-Moderator
Joined 2003
My experience with cathode feedback is that it works better with the higher gain valves (even if the feedback has been adjusted to suit a lowish-mu valve). Just my $0.02.

6AS7 has such a low mu that you're barely going to achieve any feedback. You could probably obtain a better reduction of distortion by selecting (from the very wide spread).
 
Let´s forget that I mentioned 6AS7, I realize now that it would not be a great choice here.
807 seems like a more reasonable tube, even though the OPT will be a bit more difficult to make. (~5k instead of 2k)

I´ve been thinking about Ultra Linear matched with cfb and it seems to me that if I use, say 33% cfb and connect the screen grids to B+ (or any other voltage which is constant WRT ground) I will also get 33% UL operation, since the screen to cathode voltage changes with 33% of the plate signal voltage, right?

Fo true pentode operation the screen voltage must remain constant WRT the cathode, which would be a bit complicated when the cathode swings perhaps +-70V or something.
 
Fuling said:

. . .
I´ve been thinking about Ultra Linear matched with cfb and it seems to me that if I use, say 33% cfb and connect the screen grids to B+ (or any other voltage which is constant WRT ground) I will also get 33% UL operation, since the screen to cathode voltage changes with 33% of the plate signal voltage, right?

Fo true pentode operation the screen voltage must remain constant WRT the cathode, which would be a bit complicated when the cathode swings perhaps +-70V or something.

Right, just returns screen by pass caps either to ground or to their respective cathode to select the mode you like.


Yves.
 

PRR

Member
Joined 2003
Paid Member
> of the swing that is needed

Not a huge problem, an annoyance. The 6AG7's 100K grid resistor does not help though.

80V peak into 100K can be done with like 12AT7 or 6J4, 500V supply, 50K plate resistor. Fiddle the cathode resistor to get the plate around 250V. This is within rating for the general purpose triodes. Plate voltage will be 500V at turn-on and maybe 350V on positive swing; good "300V" tubes will stand that abuse fine, and the average abuse is just 250V.

The annoyance is that now you need a fat 150V supply and a light 500V supply. Life really is simpler if the driver can eat power-stage supply.

> cathode feedback is that it works better with the higher gain valves

EC nailed it. Feedback works by trading gain. That implies that you have gain to spare. 6AG7 has such low gain, any trade-off is distressing. Tubes that have voltage gain are better bets; they have more to spare.

> find a more suitable piece for this job.

Note that already, over half your drive voltage is that 33% NFB, not the 6AS7's grid swing. For the same B+, a tube with infinite gain would still need 44V peak drive. And high-gain tubes tend to be lower current, but also rated for more voltage. This suggests working them at higher voltage and load impedance, but that increases the voltage swing at the cathode terminal. Put 500V on 6L6GC or 6550, swing 450V peak, that's 150V peak at the cathode. Add ~40V for grid-cathode swing, almost 200V peak instead of ~80V peak for the 6AG7 working at 150V and 33%.

While the 6AG7 is embarrassingly bad, few tubes are killer for CFB. The tube designer has a trade-off: higher current or higher gain. After the cathode size is picked, optimizing this balance is the major performance factor. In triodes, this balance is roughly "Mu": 12AX7 Mu=100 has lots of voltage gain but teeny power. 6AS7 Mu=2.7 has lots of current and some power but low-low voltage gain. (For a better comparison: there is a tube like 6AS7/6080 but with Mu more like 15; it is easier to drive but won't max-out with as much current as 6080.)

For a triode, this trade-off is very tough because a triode turns-off just when you want it to turn-on (when plate swings low). Lower Mu makes more power, but a Mu of less than 5 or 4 requires more grid swing than you can get with a tube driver working on the same power supply as the output tube. Hence you have the Classic Audio Triodes 2A3 and 300B. When the question was re-opened with improved technology for 1950s TV V-sweep work (V-sweep is just an audio amplifier), they came up with 6EM7 with Mu=5.3. The 6AG7 works because in a voltage regulator, the volt-amp can eat a larger voltage than the pass-tube drops. Roughly double, so we can use Mu=5/2 and get added current.

Pentodes break the relation between plate voltage and plate current. However the G1-G2 Mu has a similar effect on gain and maximum current (for a given cathode size). Since G2 typically does not swing, the tradeoff is not so tough. Pentodes have Mu(G1-G2) of 8 to 20 so that they can advertise less grid-drive and allow sloppier driver stages. Still, Mu=8 goes with current-monsters like 6L6/807, and Mu=20 goes with thin tubes like 6BQ5 that drive real easy but want hi-Z loads and give moderate power output.

With classic ultra-linear, what we are doing is bringing a fraction of plate voltage through the G2 Mu and reducing it. If Mu(G1-G2) is 20, and we apply 0.43 of plate voltage to G2, it now acts like a Mu of 20*(1-0.43)= 11.4; for Mu=8 and 43% we wind up at effective Mu of 8/(1-0.43)= 4.56. Mu=11.4 can be driven easily by a driver on the same supply voltage as the output; Mu=4.5 is very difficult to manage without an added higher supply rail. That's probably why there are more 6CA7/EL36 ultralinear amps than 6L6 ultralinears (though fashion plays a part too). Note that using cathode bias effectively raises the supply voltage available for the driver without increasing the output tube swings, so 6L6 and 6550 can be made to work fine with cathode-bias ultralinear.

Remember there is no free lunch. If there's "free" food on the bar, find out how much salt is on it and how much the beer costs. Remember too that designers have been looking for best-deal lunches for 80 years. You don't find much commercial use of cathode-feedback except (as in Macintosh) in conjunction with monster drivers (push-pull bootstrapped 450V 12AX7).
 
Ex-Moderator
Joined 2003
The other problem is maintaining stability once CFB has been applied. Despite that, CFB can work very well. This particular amplifier measured well and sounded great. But I built and tested twenty different circuits before I arrived at this one...
 

Attachments

  • n78.gif
    n78.gif
    53.1 KB · Views: 420
Forget this, my attempt to wind a new pair of OPT´s went straight to **** :bawling:
No. 1 is good enough to keep for some guitar amp project, but no. 2 didn´t make it...
Too many turns of too thin wire.

When I´ve recovered from this failure I´ll start over again with a project that requires a way smaller number of turns, perhaps PSE with 3 x PL504, 600ohm/8ohm OPT. Much more managable.
 
Hi PRR,

you wrote:

With classic ultra-linear, what we are doing is bringing a fraction of plate voltage through the G2 Mu and reducing it. If Mu(G1-G2) is 20, and we apply 0.43 of plate voltage to G2, it now acts like a Mu of 20*(1-0.43)= 11.4; for Mu=8 and 43% we wind up at effective Mu of 8/(1-0.43)= 4.56.

UL% is defined by the AC voltage (= winding) ratio of the tap being used, counted from B+ to plate. Do you agree?

Thus, 100% UL (G2 at plate) would mean (pseudo-) triode mode.

Now, according to your formula, the gain for (pseudo-) triode mode would be µ(G1-G2)= x (1-1) = 0 :bigeyes:

Where is my error?

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