• 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.

CF biasing and load ???s

A couple of questions regarding the biasing and load impedance for CFs such as shown below.

1) Given that the OPT has a non-zero DC resistance it appears that the DC voltage drop would have to be considered in the selection of the grid bias voltage. In fact it would appear that with the right combination of DC resistance and idle current the circuit could even be self biased. For example for a 6CA7 I belive that -14V gives about 100ma. So if the transformer had a DC resistance of 140 ohms the grid could be connected to ground.

Is my understanding of this correct? I imagine that depending on your transformer to provide the right resistance might be kind of risky though wouldn't it?

2) I have seen recommendations (for CC amplifiers) for using the same or even lower impedance OPTs for pentode connection v.s. triode connection. Can anyone explain the relationship of rp to optimal load impedence? Does the relationship change when considering the CF v.s. CC?

The reason that I ask is that the higher transconductance of the pentode connection would seem to have advantages WRT output impedence. It seems just possible that in a CC configuration the triode connection is perferred but in the CF configuration the pentode might give better fidelity. Opinions?

mike
 

Attachments

  • pentodecf.jpg
    pentodecf.jpg
    92.4 KB · Views: 543
mashaffer said:


...Given that the OPT has a non-zero DC resistance it appears that the DC voltage drop would have to be considered in the selection of the grid bias voltage. In fact it would appear that with the right combination of DC resistance and idle current the circuit could even be self biased. For example for a 6CA7 I belive that -14V gives about 100ma....


...I have seen recommendations (for CC amplifiers) for using the same or even lower impedance OPTs for pentode connection v.s. triode connection. Can anyone explain the relationship of rp to optimal load impedence? Does the relationship change when considering the CF v.s. CC?...


First things first. -14 might or might not be the right value- you have to specify at what plate and screen voltage. In theory, the transformer DCR would work, but in practice, it's generally too low. So you do need to consider it, but you also must be prepared to pad it up with some resistance.

Loading for a CF is the same as if you were designing a CC.

The big disadvantage of the pentode CF is that you must swing the screen in synchrony with the cathode. It can be done, but it normally takes a special transformer with a bifilar primary. Then there's the drive voltage requirements....
 
Thanks SY.

The big disadvantage of the pentode CF is that you must swing the screen in synchrony with the cathode. It can be done, but it normally takes a special transformer with a bifilar primary. Then there's the drive voltage requirements....

Hadn't thought of that. With the cathode moving around like that the screen would in effect be getting a signal right? I imagine some sort of constant voltage source between the screen and cathode could fix it but probably not worth the bother.

Wouldn't the drive voltage requirements be the same for either triode or pentode mode in CF arrangement? Am planning on a multi-voltage B+ and two stage voltage amplifying stage.

mike
 
So it appears you are making a UNITY GAIN Single Ended current amplifier....
You are best off deriving a NEGATIVE bias and applying it to the input grid....Forget about DC resistance in the winding to self bias.....You create excessive copper losses and will generate a bunch of heat in the windings...all not good...
As for optimal loading....The CF is not as critical as Plate loading...
You need to derive the CATHODE curves from the PLATE curves...then you can view them and then apply a load-line....
Personally I would use a TRIODE connection for the 6CA7 and go with a CF....
Basically the optimum load is more dependent on Damping in this configuration... are you planning on Global feedback????
Remember your input range is HUGE on a follower, so what does your driver look like...

Chris
 
No global feedback is planned.

I haven't figured out the driver stage yet. I am starting at the output and working my way back. My first thought is a two stage triode voltage amplifier with the second stage working off of a higher voltage supply so that adequate output voltage swing can be achieved. Eventually I think I would like to try mu-stage for the voltage stages but I would probably implement standard resistive loads first.

My concern is that the first stage must have low enough gain to not over drive the grid bias of the second stage. So the gain of stage two may need to be higher than the first stage (which is not optimal from a noise standpoint I suppose).

I might end up with a small 2:1 - 3:1 step up transformer in there if needed.

Or I could use a small signal pentode for stage two but I would rather not.

BTW I have a couple of sockets in my parts bin that I think might be magnoval. Anyone know the exact dimensions?

mike
 
mashaffer said:
Thanks SY.



... I imagine some sort of constant voltage source between the screen and cathode could fix it but probably not worth the bother.

Wouldn't the drive voltage requirements be the same for either triode or pentode mode in CF arrangement? ...

There are other ways to get the screen to follow the cathode; a floating power voltage source could be rigged up. I wouldn't take that path just because of the nightmares of getting that voltage source to work correctly. But someone smarter undoubtedly could make it work.

The question of drive is a deep one. On the surface, you'd think, sure, it ought to be about the same. And, all things being equal, it would be pretty close. The problem is that if you want a circuit optimized for, say, a 6CA7 in triode, it will have a different load and a different B+ than a circuit optimized for pentode. In the analyses I've seen, the pentode mode ends up needing considerably more drive than the triode mode. In trade, the distortion and source impedance are lower, too.
 
Here is a quick schematic I threw together...It is just a idea you can use...
The values are simple to compute...once you selected the valves you want to use...
This is an old concept of "bootstraping" the driver valve to the screen ...the AC impedance of the 12K resistor will look much higher to the AC signal current, thus obtaining high voltage gain..

Chris
 
Cerrem... missing picture.

One solution to driver problem might be to use a small power tube which could accomodate a relatively large input voltage swing. Any thoughts on that approach?

Also I was wondering if the unfavorable distortion spectrum of pentodes also applies when used as SE voltage amps (as opposed to use for power amp). A second stage pentode might otherwise be advantageous since it has high gain. The high plate resistance should not be a big problem since the CF output stage has high input impedence and low effective Miller capacitance (I believe). I am just afraid of large HD3 since the only planned feedback in this design is possibly unbypassed Rk. Could possibly use one of the combined triode-pentode bottles if 2nd stage pentode is OK.

mike
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
> it would appear that with the right combination of DC resistance and idle current the circuit could even be self biased.

I have seen it done. As SY says, a "good" transformer's DCR is usually too low. If you are going to make a million of them, you custom-order a transformer with controlled DCR, higher than a "good" (low-loss) transformer would have. If you are just doing one or two, you are stuck with stock low-DCR iron, a pad-up resistor, and a bypass cap.

> using the same or even lower impedance OPTs for pentode connection v.s. triode

Triodes have a large output resistance, large voltage drop at large current. Fool around with Ohms Law, and you will find that the Power versus Load Z curve is very broad: large changes in load cause modest changes in power. With simple DC analysis, max power happens at Rl=Rp. Because a triode is a DC device and we only extract the AC from its swings, max audio power is Rl=2*Rp. But 5:1 change in Rl gives only about 3dB drop in power, it isn't fussy that way.

Gm and gain varies with current. For low THD, we want a high standing current and a low signal current swing. That gives terrible efficiency and economy. A Power Amp has to have large signal current swing, large gain change, high THD.

However, a triode has Voltage Feedback plate to grid. Hi-Z loads imply large plate voltage swing, large feedback, lower THD. When Rl<=Rp, THD is high. When Rl>5*Rp, THD is low. The sweet-spot tends to be Rl=2*Rp. Higher Rl, up to around 5*Rp, is also sweet, though your Watts number declines.

The pentode has no plate-grid feedback. We are stuck with signal current swing distortion.

The pentode has two "plate resistances". Up to ~50V, Rp is very low. Over ~50V, Rp is very high. This gives a "knee" in the plate curves. Maximum power (for given voltages) happens when you stick one end of the load-line in the crook of the knee, and take the other end out until you hit plate dissipation or voltage maximum ratings.

In addition to the simple signal current swing distortion, a pentode has a second curvature as you get into the cramped curves near the knee. If you stay well out of the knee, THD is the same as a heavily-loaded triode. But for maximum power you have to swing into the knee. You can play some games by hitting a little above or below the knee, and try to cancel some of the second-order curvature with a reverse-bend near the knee, giving a lower THD number (which however has more annoying harmonics). In real life with real speakers, anywhere around the knee is much the same until you get into gross overload.

So yeah. For similar tubes worked triode or pentode, the pentode makes more power with a lower Rl, the triode makes low power in any Rl but lower distortion with higher Rl than you pick for a pentode.

As a DIY audiophile, you do have to ask if you can fool yourself with a THD/Watts numbers game, or if you want to overbuild and loaf the amp far below distress.

> Does the relationship change when considering the CF v.s. CC?

As a power amp: no. With qualifications: in any worthwhile case, the CF has SO much feedback that all this fretting over THD tends to drop out of the picture. In triode, for good power, just set Rl=2*Rp and be done with it. In pentode, draw the loadline into the knee, except:

As SY says, the G2-K voltage has to stay constant or it is not a Pentode. Pushing G2 around is Hard Work. It can be 10% of the output power. If you try to do it with an R-C bootstrap, you may find that the bootstrap is sucking an absurd amount of power, enough so you may as well run Triode. Tertiary windings and floating supplies are less inefficient, but more trouble. I have seen very few Pentode CFs except as hobbyist toys.

> Wouldn't the drive voltage requirements be the same for either triode or pentode mode in CF arrangement?

At a glance, I agree, except that pesky G2 drive can really clobber things. As SY says: "deep".

> the higher transconductance of the pentode connection

Plate Gm is higher in Triode, for the same tube near the same operating point. Around 10% higher, because G2 current is about 10% of Plate current, and both are modulated by grid voltage.

The reasons we have a lot of medium-high Gm pentodes, not so many triodes, are commercial and historical. But the highest-Gm tubes are little UHF radio Triodes.

> the first stage must have low enough gain to not over drive the grid bias of the second stage.

Huh? No! You have a volume control. You will set it so nothing overloads. Or if it must overload, the most expensive stage (the output) overloads first. If not for the (IMHO foolish) CF output, it is simple. 400V supply, driver can swing about 0.2*Vs or 50V, find an output tube that will saturate with 40V input. The driver input and output overload points are the same: you can't have a clean output and an overloaded input or vice-versa. If the driver has gain of 400, you adjust the volume control for up-to 0.1V to the driver; if it has gain of 4 you adjust the volume control for up-to 10V to the driver. If you have another stage in front, you adjust the volume control so it makes up-to what the driver needs to push the output to the limit.

So how much swing do you need? Assuming Rl=2*Rp, the peak output signal swing will be 1/3rd of the supply voltage. Lazy-math, assume Vs is 300V, output swing is 200V.

Already the CF looks awkward. A reasonable rule of thumb for resistance-coupled voltage-amps is that peak output swing for low THD is 20% of supply. If we need 200V swing, we need 1,000V supply! And this rule-o-thumb's "low THD" is 3%, 5% depending. So offhand, we have reduced power-stage THD from 5% to 0.5%, but need a heroic driver to slap it around with even 5% THD.

And I have neglected grid-cathode voltage. This tends to be 0.6*(Vs/Mu). A good Mu for an audio power tube is 5. 0.6*300/5= 36V peak on top of the 200V peak output swing. A maximum-power Mu is very low, like 2 (6080): 0.6*300/2= 90V peak on top of 200V peak, ouch. You like 6CA7? Mu is ~11, 0.6*300/11= 16V extra, negligible.

So we need 216V to 290V peak grid drive, albeit at super high impedance. R-C coupled we want more than 1,100V to 1,600V supply on the driver. Oy. We could use a very fat tube to reduce this a bit, but R-C coupling kinda forces large current-swing or excess supply voltage for low THD. Given that we have a cathode follower, bootstrapping offers some hope of less-insane driver supply voltage; that has to be worked out in detail and judged.

Choke-coupled has promise. As a rough-guess, we could swing that ~230V peak with the same 300V supply. One thing is, stray coil capacity tends to limit effective impedance over the wide hi-fi audio band to maybe 10K. That means we need a tube with Rp much less than 3K to get a 230V swing out of a 300V supply. THD may rise at the top and bottom of the audio band; inductance biting the bass and capacitance biting the treble, but midband the choke Z is very high, tube works semi-constant current with high linearity. For power bandwidth, we need to drive the choke's lowest Z, 10K, to over 200V peak, over 20mA of bias current. This is a 6 Watt job, so we are looking at small power tubes. At 20mA we are probably not going to get Gm higher than 4,000uMhos, we want Rp less than 3K, so our Mu should be around 10 or less. Looks like 2A3 to me, though a spare 6CA7 is possible. Gain with hi-hi-Z load will be about 80% of Mu, say 8. Input voltage will be about 30V peak. We will need a stage in front, though many tubes will work here.

(Someone will suggest replacing the choke with a CCS. That is similar for small-signal work. But the available CCSs only waste voltage; they can't kick above the supply rail like a choke loves to do. For effective CSS work, we'd need double the supply voltage, which gets back to heroic designs.)

Of course the devil is in the details and if any of these rough approximations get away from us, choke-coupled may not have enough swing, we may not get the most out of our costly output tube(s) and power supply. That's a commercial disaster, but in DIY it may only be a shrug.

> I might end up with a small 2:1 - 3:1 step up transformer in there if needed.

That seems like the way to get more swing than a choke can offer. And it eases output grid bias connections. But that ~10K maximum winding impedance applies to the highest-Z winding. If the ratio is 1:2, and the output is 10K worst-case, the input is 10K/2^2 or 2.5K. If the driver plate resistance is still 3K, we have gained almost nothing. And we'll have to drive that 2.5K to ~230V/2= 115V peak, 115V/2,5K= 46mA peak output current, we need to idle the driver at 50mA, 15 Watts, a power-tube for sure.

We can fudge the 10K rule. A very good winding can hold 50K impedance over 20-20KHz. However we need to run big DC current in the winding, which makes things harder. And there are not a lot of good chokes/transformers, not at this impedance and power level. A common power choke can show resonances inside the audio band. Hi-fi output transformers with 10K and 50mA ratings exist, but are not cheap or "small", and you only get one hi-Z winding, the other is low-Z. Sowter has 10K line transformers and one of them may be rated for hundreds of volts of signal, but you will be working it to the limit. And while fairly priced and compact, it isn't easy on the budget.

So now you see why "nobody uses cathode follower outputs". Everything you gain in the output, you lose in the driver. The exceptions are Macintosh's half-follower affair ($$$ transformers) and line-level gear that only asks 10V-20V of signal from 300V supply (too inefficient for speaker work).

www.TubeCAD.com has some recent articles on the topic of power cathode followers.
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
> gain of stage two may need to be higher than the first stage (which is not optimal from a noise standpoint I suppose).

Noise is rarely a killer issue in a power amp. Gain distribution is an issue only if the first stage gain is quite low. Gain of 3 may be noticeable. If the second stage has high noise, you may want more first stage gain. But you kinda have to do something "dumb" to get into noise trouble. Like gain of 2, followed by tone controls with loss of 10, followed by a lot of gain to get back up to where you need to be. And even this "dumb" design has been used with happiness.
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
> www.TubeCAD.com has some recent articles on the topic of power cathode followers.

I had read these:
http://www.tubecad.com/2005/June/22June2005.pdf
http://www.tubecad.com/2005/June/29June2005.pdf

I had not read this, or I would have referred to it instead of typing so much:
http://www.tubecad.com/2005/July/07July2005.pdf

> bootstrapping offers some hope of less-insane driver supply voltage; that has to be worked out in detail and judged.

See 7 July 2005, first schematic.

This actually shows transformer DCR biasing the tube. Note he does not give a real production part. It does happen that many SE hi-fi transformers do have DCR in this range: SE transformers always have "a lot" of DCR because they need many-many turns to make up for softening of the iron with large DC current in the winding. You could get lucky.

From the implied voltage and current, we may assume the KT88 should drive about a 5K load. Follow the 22uFd bootstrap cap and see what else it drives: a 5K resistor up to the AC-grounded positive supply rail. So the bootstrap is eating half the tube's total output (and the total load is lower than optimum).

We can fiddle that some. The driver could use higher plate resistors. Half is absurd, but in most cases the driver bootstrap will be a heavy load on the output.

See the second schematic, from Elector. Here Broskie makes the subtle point that a bootstrapped driver and cathode follower is NOT a cathode follower. The driver output voltage is developed where? Not between output grid and ground, but between output grid and cathode. That's equivalent to grounded-cathode operation! The bootstrapping cancels the 90% negative feedback in a cathode follower.

Thirdly he shows choke/transformer coupled drivers. He makes the points I touched. I would argue that he does not seem to know just how difficult hi-Z transformers can be. He mentions tube input capacitance, but winding capacitance will usually be larger than tube capacitance. (Logical: a tube is a couple metal slats a millimeter apart, a transformer winding is many layers with larger area packed as tight as the varnish will allow.) He implies the old misconception that ratio is a problem: ratio is not the problem, maximum impedance is the problem. We can design fabulous 0.1Ω:200Ω or 1:45 transformers, we can't wind a hi-fi 50K:50K, and even 600:60K is marginal hi-fi (though they were common in recording and broadcasting well into the 1960s, they got lumpy at 15KHz which had to be trimmed-out, more or less).
 
I was struck by the coincidence between this discussion and Broskie's latest. He does a fine job of showing what horrible things one is asking of a driver for a CF output stage. In his latest, he shows a choke-loaded 6SN7- to get that to work, he has to run that poor driver balls to the wall. With 400 volts and 10mA per section, it will not last long. And even at that, it will not be a low distortion driver.

If I were to be so foolish as to tackle a CF amp, I would bite the bullet and go hybrid. A cascode with something like a 6DJ8 on the bottom and a 1200V MOSFET on top might squeeze out a bit more swing at an acceptable distortion level.
 
Wow!

Thanks for the detailed education and links. I think I learned a lot more by looking at something off of the wall. It does sound as if there is good reason why this is not often done. :)

Interestingly enough I was also coming to the conclusion that a power tube would be needed for the driver (I was looking at the 6CA7 driving a KT88 or EL509) just because of the input voltage swing but it is obviously more than that which makes a high voltage driver so hard.

Thanks again.

mike

P.S. Sy, Judging from the new avitar the surgury must have worked. ;)
 
Now this looks familiar

So something like this has been done. Lizzy CF Power Amp

I do note that he claims a tetrode connection on the output stage but if I read his schematic right it is a psuedo tetrode connection as mentioned here before since the cathode-screen potential is not constant.

In my brief look at what he is doing here it is not clear to me how he got around the driver problems that we discussed. I also don't see any mention of power output.

My interest in something like this would primarily be for a subwoofer amplifier (where low Zout is most imporatant). It occurs to me however that a CC configuration with some NF on the output stage might also provide adequate damping.

PRR, SY etal. I would interested in your thoughts on the Lizzy design and even more interested in what direction you would head in designing a tube sub amplifier.

mike
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
> not clear to me how he got around the driver problems

Using a choke-coupled Power Tube.

Even so, I don't think he is swinging the 6L6 to full advantage.

I don't see where the +250V for the screens comes from? (I'm getting java errors on some of those links.)

> don't see any mention of power output.

Yeah, long on good impressions and short on hard numbers. He's happy, I'm happy for him. Can I be happy? Hard to know without more info.

> subwoofer amplifier (where low Zout is most imporatant)

Oh, so? How low?

DF=4 will be less response error than the average room. "Worse" DF will give more sub-woof. Don't fall too far into the infinite-DF myth. DF ain't bad, but that does not mean you need humongous DF numbers.

> a CC configuration with some NF on the output stage might also provide adequate damping.

It might. Several tube amps meet the arbitrary standard "DF>40". On paper, this gives worst-case rise at resonance of less than 0.25dB. But you can't hear 1dB errors in bass, no woofer is made to 1dB precision, woofers designers know they don't have zero source impedance, and real rooms have +/-10dB modes and nodes.

> what direction you would head in designing a tube sub amplifier.

Four or six P-P parallel 6550, plate-loaded pentode mode, with 10dB to 20dB negative feedbag.

Or in winter: a larger number of 6550 triode-strapped.

Unless space is cheap. Then I would use multiple 15"-18" drivers in a room-sized box, and any modest-power amp (P-P 6L6 or 2A3 etc). But space for a GOOD subwoofer is usually expensive, so bass efficiency is low and a heroic amplifier is required.
 
Just a quick thought: The issues facing a design with a CF output are comparable to designing for electrostatic headphones.
The difference is that electrostatic headphones are inherently push-pull, where the amp has no such constraints.
High voltage swing at low distortion, Capacitive load and high gain. I started 3 years ago trying to design one, and still revisit the idea.

Observations that many of the ESL amps use bi-polar supplies, relatively high voltage output tubes and often some feedback.
Tubes like the 6S4, 6SN7GT and EL34 seem to appear most often.

Just my $.02

Doug
 
Thank you for your continued patient explinations PRR.

Unless space is cheap. Then I would use multiple 15"-18" drivers in a room-sized box, and any modest-power amp (P-P 6L6 or 2A3 etc). But space for a GOOD subwoofer is usually expensive, so bass efficiency is low and a heroic amplifier is required.

Initially I will have single IB 15" (probably just sand powered at first), but the eventual project involves an archetectual horn where mega-power will be unnecessary.

mike
 
I should be ashamed of myself for saying this, but... there's no justification for using a tube amp on a sub. Output transformers are horrible things in the bass, and OTLs are unhappy with typical subwoofer loads, not to mention their overall twitchiness.

As vacuum-crazy as I am, I use high-power silicon on my high-efficiency subs and save the tubes for midbass on up..
 
It's always worth looking at peak power requirements- I hate clipping amplifiers, especially when the recovery-from-overload characteristics are dicey. You might average a watt, but need 10 or more watts to prevent clipping on peaks.

For a high-efficiency horn, class A silicon can be very, very nice- I had a chance to hear Nelson Pass's Kleinhorns driven by the First Watt units and was very impressed. My preference would be for tubes, but that's my own little brand of irrationality which I would not expect Nelson to share.

I remain unconvinced of the sonic merits of class A except insofar as it relaxes some of the design constraints on the power supply at the expense of efficiency.