How to choke load for big voltage swings?

[SpreadSpectrum]

I've been told or seen it written by folks that are smarter/more experienced than me that choke loading is the way to go for big voltage swings, but I'm trying to understand how this can work well, so I figured I would pound out some math and you guys can tell me what I am doing wrong if anything.

My example was going to be an ECC99 biased at 12mA and 380V on the plate with a 60H choke.

Let's calculate the plate load impedance at the ends of the frequency band of interest. XL = 2*PI*f*L. At 20kHz, XL = about 7.5M. At the upper end of the frequency band, load line will be flat and gain will be about mu. However, at 20Hz XL = about 7.5k. This follows most folks recommendation that the load impedance be 3xRp. The problem is that if you draw the load line, the line crosses the I = 0 "x-axis" at a mere 90V above the quiescent state(V = IR = .012*7500). I was hoping to get 200Vrms out of this and am disappointed that it seems I will clip at low frequencies at levels much lower than this.

Have I made an error somewhere or is this just a drawback of choke loading?

 

[DualTriode]

Hello,
How is your tube biased? Do you have a capacitor bypassed cathode resistor? If you do choke or not the grid can only swing so far and there will be hard clipping. Increase the bias voltage and or remove the bypass resistor, there will be a bigger output voltage amplitude.
DT
All just for fun!

 

[SpreadSpectrum]

I haven't actually built it but for the sake of argument let's say it is biased by a perfectly bypassed CCS.

I don't think you understand my question. It is not a question of there being enough bias voltage. If I make the bias voltage more negative, my plate voltage will simply shift upward(it's already right next to the rated maximum) and I will still clip 90V from my idle point due to the slope of the load line and the idle current being at 12mA.

It looks to me that it is impossible to get more than 180Vpk-pk out of ANY tube idling at ANY plate voltage at 12mA with 60H load without clipping.

 

[Arnulf]

The problem is that if you draw the load line, the line crosses the I = 0 "x-axis" at a mere 90V above the quiescent state(V = IR = .012*7500). I was hoping to get 200Vrms out of this and am disappointed that it seems I will clip at low frequencies at levels much lower than this.

Have I made an error somewhere or is this just a drawback of choke loading?

And where would it go if you used a purely resistive load ?

Note that choke has relatively low DC resistance so if you neglect it (just like in the case of OPT) the loadline can be in any point above Va = B+ (at any current, determined by bias conditions).

If you take DC resistance into account the line from Va = B+ is still almost vertical and your quiescent point will be on that line. AC loadline will then go through quiescent point with predetermined slope.

There is no way this could ever lead to lower swing than resistive loading.

 

[SpreadSpectrum]

There is no way this could ever lead to lower swing than resistive loading.[/QUOTE]

"And where would it go if you used a purely resistive load ?"

If I used a purely resistive load, it would swing up to B+, which would need to be twice as high or so as the quiescent point (the price you pay for not using the choke) so it would swing up the full 380V if you drive it there.

"Note that choke has relatively low DC resistance so if you neglect it (just like in the case of OPT) the loadline can be in any point above Va = B+ (at any current, determined by bias conditions)."

At any point above Va = B+ AND lower in current since the choke will always have positive impedance, right? The loadline is frequency-dependent.

"If you take DC resistance into account the line from Va = B+ is still almost vertical and your quiescent point will be on that line. AC loadline will then go through quiescent point with predetermined slope."

Agreed. And again slope is frequency-dependent and as I see it altogether too steep at 20Hz to get the kind of swings I am looking for.

"There is no way this could ever lead to lower swing than resistive loading."

I think you may be painting with overly broad brushstrokes here. If I use several kilovolts of B+ and a large plate resistor I can get better swing. Maybe you meant with equal B+? Even then, I can get better swing with a resistor if inductance or frequency is too low. It will clip at Ia = 0, right? The electrons aren't going to start flowing from the anode to the cathode to help me get 200Vrms, are they?

My current experiments are with 6SN7 CCS loaded and 780V B+. I haven't tested down that low, but I presume I am getting 200Vrms all the way down to a few Hertz(limited by AC coupling at input). Load line is flat across audio band. I was looking for similar performance with a choke but am still unconvinced I can get it and don't want to spend the money on a Lundahl choke unless someone can point out some kind of error in my thinking.

 

[Arnulf]

It was my assumption that you meant that ceteris paribus. Your question (thread subject) would make no sense otherwise as the answer could be simple: use higher B+.

 

[dsavitsk]

Why are you running the tube at such an odd point? It isn't linear, and it will clip. So, bias it at, say, -7V running at ~225V on the plate. You have 14Vp-p on the input without driving the grid positive, and with a mu of ~20 you should get close to 280Vp-p.

Oh, and don't forget that the next stage is in || with the choke, so the choke is not a load, and the load line is not nearly as flat as you think.

Edit: and Electra-Print will sell you fine chokes for less than half the cost of the Lundahls.

 

[SpreadSpectrum]

It was my assumption that you meant that ceteris paribus. Your question (thread subject) would make no sense otherwise as the answer could be simple: use higher B+.

Sorry, this comes from discussions I had in other places where it was suggested that I would get better results with choke load than CCS and higher voltage. I presented my results (1.2%THD at 200Vrms) and asked if it could be improved upon.

I couldn't see a choke would be an improvement as plate load impedance would be low at low frequencies. In fact, I can't see any way to get 200Vrms out of any smallish (<10W) tube at 20Hz. I just have to start at too high of a voltage and too high of a current. Plus, the CCS is just a lot cheaper than a choke. I am willing to do whatever it takes within reason to get good performance.

So, back to my original question: Did I mess up conceptually here? This will clip at 90V swing in the positive direction, right?

 

[SpreadSpectrum]

Why are you running the tube at such an odd point? It isn't linear, and it will clip. So, bias it at, say, -7V running at ~225V on the plate. You have 14Vp-p on the input without driving the grid positive, and with a mu of ~20 you should get close to 280Vp-p.

Oh, and don't forget that the next stage is in || with the choke, so the choke is not a load, and the load line is not nearly as flat as you think.

Edit: and Electra-Print will sell you fine chokes for less than half the cost of the Lundahls.

The answer is that I want 200Vrms(570Vpk-pk) or better with less distortion than I am getting with the 6SN7 if possible.

My following stage is 1M || 5.5pF(source follower), not much of a load, but it will swamp the choke's impedance on the very HF side. Load line should be pretty flat compared to how others would use a choke, driving a power tube grid resistor directly.

This was a hypothetical last voltage amplifier for a circlotron amp that may be built some day. If I were going to build a circlotron (a complex and costly endeavor) I would spare no expense on the driver as it will be the most significant source of distortion in the amp. If I have to go A2 on a driver stage, so be it. Mosfets are cheap.

 

[dsavitsk]

However, at 20Hz XL = about 7.5k.

I was hoping to get 200Vrms out of this and am disappointed that it seems I will clip at low frequencies at levels much lower than this.

In fact, I can't see any way to get 200Vrms out of any smallish (<10W) tube at 20Hz. I just have to start at too high of a voltage and too high of a current.

W=V^2/R

V is 200 and R, at 20Hz, is 7.5K.

You are asking your tube that can dissipate 5W per side to output 5.3W at 20Hz.

 

[Michael Koster]

I don't see how a choke could possibly give you a larger or more linear swing than a CCS + higher B+. The choke will limit the low frequency swing due the load line tilting (actually becoming elliptical at low frequency). SE gapped chokes don't have much inductance or much AC voltage capability relative to what you want to do.

Sounds like you need a low impedance triode like a 300B if you want to go with a choke loaded driver.

 

[SpreadSpectrum]

I don't see how a choke could possibly give you a larger or more linear swing than a CCS + higher B+. The choke will limit the low frequency swing due the load line tilting (actually becoming elliptical at low frequency). SE gapped chokes don't have much inductance or much AC voltage capability relative to what you want to do.

Sounds like you need a low impedance triode like a 300B if you want to go with a choke loaded driver.

Thanks, Michael. It is nice to hear that I am not insane.

I can get a whole lot of mosfets and 10m90s for the price of one 300B(or a choke for that matter), so I think I know what to do here.

 

[tinitus]

Gary Pimm's DIY audio pages

worth a study

 

[DougL]

Two comments. 3 times Rp is a compromise for output transformers balancing distortion, power output. Probably not appropriate for a driver.

Second, while a CSS will have better low frequency output and similar swing if you have twice the power supply, cost for the power supply is estimated at 4 times the lower voltage one.. Valid trade, but its not choke vs silicone. its choke vs 4 times the caps, more expensive transformer, etc.

Pay your money and make your tradeoffs. That's just good engineering.

Doug

 

[kevinkr]

Try again assuming 100H and 200H chokes and you will see a very significant improvement particularly with the 200H choke, the caveat of course is that the higher inductance choke will be much larger and will probably have much higher stray capacitance as well - negatively impacting the load line at the high end of the desired range. (My 100H chokes exhibit the problem you mention at 20Hz using the D3A to drive them and the capacitance is sufficient that they are -3dB around 60kHz, IMHO a bit disappointing.)

Lower rp tubes help quite a lot, the 5842/417A being an example of a tube that works well with choke loading. (You will not be able to achieve your 200Vrms swing with this tube however.)

Might be achievable with a 6CK4 or 6AH4..

 

[SpreadSpectrum]

Two comments. 3 times Rp is a compromise for output transformers balancing distortion, power output. Probably not appropriate for a driver.

Second, while a CSS will have better low frequency output and similar swing if you have twice the power supply, cost for the power supply is estimated at 4 times the lower voltage one.. Valid trade, but its not choke vs silicone. its choke vs 4 times the caps, more expensive transformer, etc.

Pay your money and make your tradeoffs. That's just good engineering.

Doug

A lot of people are afraid of a complex power supply. The thing is, you already have to have a negative supply for the bias and I wanted to have a source follower driver on the grids of the power tubes. I just bought a Hammond 800VCT transformer and used a bridge rectifier configuration for + and - 560V raw. After that, all kinds of HV possibilities opened up, thus the current experimentation.

 

[Andrew Eckhardt]

A pretty well pat trick to use for the capacitance problem with big inductances is to divide the choke into two or more units, with at least one of the chokes having all the high frequency performance you need, preferrably the one wired closest to the fast signal. Probably not good to purposefully set up a high Q series resonance between components, but if you watch that at all you can get somewhere helpful with it.