I came across this super simple tube-based shunt "regulator" circuit which I think is pretty cool. For use with line amps requiring a clean 15-30 ma supply. It's simplicity gets me wanting to try it using a DC Link for the final 220uf capacitor. The ability to trim the voltage is appealing to me as well as choke input, although I'd trim it then replace the trimmer with a resistor. Also not needing heatsinks, IC regulator chips, etc... is appealing.
Looking at the circuit below it doesn't seem to be a "regulator" at all as it has no solid reference voltage, it floats with mains variation. Is it more like a simple ripple-canceling shunt? Or is it indeed regulating the 2-4 volt drift my house mains might experience? For line amps I'm not really too concerned about the 2-4 volts my house mains may vary, ripple cancelation is more important.
I have a few questions (this is one reason I need to learn spice this year):
1) Is it too simple? (whatever that means) to be a reliable circuit.
2) What effect does the mu of the shunt tube have on the operation of this circuit?
3) I've never seen a 6SN7 used this way, I thought it was a "voltage triode" not a "power triode".
4) The vertical deflection dual triodes 6BX7 and 6BL7 are base-compatible with 6SN7 (6BX7 10u; 6BL7 15u) and considered "power triodes". Would those be more reliable long term? Do I need to have 20u? Would more amplification factor be even better?
5) Is the 2k5 trimming voltage or current?
Looking at the circuit below it doesn't seem to be a "regulator" at all as it has no solid reference voltage, it floats with mains variation. Is it more like a simple ripple-canceling shunt? Or is it indeed regulating the 2-4 volt drift my house mains might experience? For line amps I'm not really too concerned about the 2-4 volts my house mains may vary, ripple cancelation is more important.
I have a few questions (this is one reason I need to learn spice this year):
1) Is it too simple? (whatever that means) to be a reliable circuit.
2) What effect does the mu of the shunt tube have on the operation of this circuit?
3) I've never seen a 6SN7 used this way, I thought it was a "voltage triode" not a "power triode".
4) The vertical deflection dual triodes 6BX7 and 6BL7 are base-compatible with 6SN7 (6BX7 10u; 6BL7 15u) and considered "power triodes". Would those be more reliable long term? Do I need to have 20u? Would more amplification factor be even better?
5) Is the 2k5 trimming voltage or current?
Last edited:
Your analysis is correct, it's not a regulator as such, but it does lower the ripple.
The 2.5k trimmer trims booth current and voltage, and there lies its weakness.
For a shunt to work as intended, the shunt current at the desired output voltage and zero load must at least be as high as the expected max load current.
I'm not enough of a tube guy to see whether that is the case here.
Probably the 4.7k Rk can be used to set the current at the set voltage but the two adjustments will influence each other.
As far as the 6SN7 not being a power tube, that's all relative.
If the shunt current and output voltage stays within the max Pa and Ia of the shunt tube, it doesn't matter what it's 'official' designation is.
If you're a spice guy you could simulate it and explore the settings.
Jan
The 2.5k trimmer trims booth current and voltage, and there lies its weakness.
For a shunt to work as intended, the shunt current at the desired output voltage and zero load must at least be as high as the expected max load current.
I'm not enough of a tube guy to see whether that is the case here.
Probably the 4.7k Rk can be used to set the current at the set voltage but the two adjustments will influence each other.
As far as the 6SN7 not being a power tube, that's all relative.
If the shunt current and output voltage stays within the max Pa and Ia of the shunt tube, it doesn't matter what it's 'official' designation is.
If you're a spice guy you could simulate it and explore the settings.
Jan
Thanks. Wow this one sentence helps me understand how to optimize any shunt regulator on the breadboard, I think. I always was very confused by the current relationship that has to be set using a shunt regulator. So basically if I wanted to do my optimization experimentally on the breadboard, (old school), I'd do something like this?: (lets say in this experiment I'm building out the Salas 6V6 line amp):
1) Build the 6v6 line amp and get it working with any old random hefty power supply that can deliver 340 volts without drooping.
2) Now measure how much current it is eating up. Let's say that measure the Sala 6V6 is demanding a constant 21ma. So I write that number down, 21.
3) Now I have two numbers 21ma and 340V (the voltage Salas wants me to operate at)
4) Now I fire up my shunt regulated power supply and I simultaneously adjust its load resistor and one of the filter RC resistors until until I see about 25ma appearing across the shunt tube AND I'm also seeing 340V simultaneously. this may take a while.
5) Now I know my power supply with shunt is both delivering 340V and eating up 25ma all on its lonesome. 🙂
6) Last step, I hook the line amp to this optimized power supply.
7) So in effect Salas 6V6 amp is going to parallel my shunt once the two are married up? So they share the load? After marrying them up I should not see the 340V drop correct? Because the shunt itself is set at 25ma which is greater than the 21ma of the load.
8) At this point I'll know the ripple is being cancelled as best it could be?
The basic idea is that the shunt current goes up if Vo goes up and vice versa.
Within a reasonable approximation that means that the sum of Ishunt and Iload is constant.
Now look at the two extremes:
If Iload is max, Ishunt can be zero; if Iload is zero, Ishunt is max, and should be the max expected load current.
Normally you'd take a safety factor so set Ishunt a bit higher.
In this case there is only an AC drive for the shunt, as noted by analog_sa above.
In any case your steady-state adjustment method will not work because there's no DC drive to the shunt.
I am not familiar with such a setup and am not convinced it is working well.
Thus the suggestion to run it through LTspice.
Jan
Within a reasonable approximation that means that the sum of Ishunt and Iload is constant.
Now look at the two extremes:
If Iload is max, Ishunt can be zero; if Iload is zero, Ishunt is max, and should be the max expected load current.
Normally you'd take a safety factor so set Ishunt a bit higher.
In this case there is only an AC drive for the shunt, as noted by analog_sa above.
In any case your steady-state adjustment method will not work because there's no DC drive to the shunt.
I am not familiar with such a setup and am not convinced it is working well.
Thus the suggestion to run it through LTspice.
Jan
I see, this particular circuit is problematic to tune. I'm not married to this circuit, I just came across it, so disregard this circuit. Lets say I did use a different shunt circuit with DC drive. Then I should be able to use my steps above to optimize that shunt circuit for any given constant load pre-amp circuit (within a reasonable current limit)? Is the objective still to get the unloaded shunt to produce the correct voltage and be consuming current a little above what my load demands? then I'm good to go when I hook it up.
I would be arriving at the conditions of this rule in the end no? "For a shunt to work as intended, the shunt current at the desired output voltage and zero load must at least be as high as the expected max load current."
Basically I want to learn how to go about setting the optimal operating point of a shunt power supply to a given random load ma, not using Spice. I really thank you for this help, I couldn't begin to grasp this for the longest time.
For a really regulating shunt you need a reference voltage, and an amplifier stage that draws a shunt current, with two inputs, the reference and a fraction of Vo.
When the fraction of Vo goes above Vref, the current has to increase, when it falls below Vref, the current has to decreases.
That's the basic mechanism, and if the amplifier stage voltage-to-current gain is high enough you end up with a rock-steady clean Vo.
In your original circuit, you can use the tube as the voltage to current amplifier stage, it has two inputs, the cathode and the grid.
Put a couple of LEDs in the cathode instead of an R, divide down Vo and feed to the grid, and Bob's your uncle.
Jan
When the fraction of Vo goes above Vref, the current has to increase, when it falls below Vref, the current has to decreases.
That's the basic mechanism, and if the amplifier stage voltage-to-current gain is high enough you end up with a rock-steady clean Vo.
In your original circuit, you can use the tube as the voltage to current amplifier stage, it has two inputs, the cathode and the grid.
Put a couple of LEDs in the cathode instead of an R, divide down Vo and feed to the grid, and Bob's your uncle.
Jan
Last edited:
What is "power"? How long is a long rope? ALL tubes are power tubes. With 300B or 6L6 or 833, "power" is serious. With 12AX7, 'power' is minuscule, and yet I saw a 12AX7 mini-power driver today. 6SN7/12AU7 comes from "general purpose triode" in days when transformer loading was standard, and a tube needed enough grunt to carry transformer parasitics. You can find 6SN7 making 100mW very clean and half-watt not-too-dirty.
That plan is not a regulator. As drawn, it is not even a good smoother. Sit in the tube. It has <2.5k plate load and 4.7k cathode load. Gain is less than unity! It can't do much good like that! Maybe the cathode-cap fell off?
SPICE says it actually ADDs ripple as valued, and adding a fat cathode cap does not reduce ripple. However a simple cheap cool 100uFd cap, no tube, makes a big difference!
But does it need a 100u+10H+100u+10H+100u filter in front? That's twice as much as early guitar amps. They weren't perfect but "good" squared is really really good. Before you even SPICE, pencil out your noise budget. You don't really have 100dB dynamic range. 100dB down from 1V signal is 10 microVolts. While it is easy to stack parts in SPICE or PSUD and figure microVolts, in real life stray coupling and ESR (even wire resistance) will spoil any fantastic numbers your SPICE-cat computes.
The ripple killers I saw before have the adjustable resistor in the connection between the 0.1 uF (or an other value) capacitor and the plate connection.
The schematic on page 27 also shows the adjustable resistor at that position: Electronics June 1937
The schematic on page 27 also shows the adjustable resistor at that position: Electronics June 1937
Yep, the resistor between raw ripple voltage passed to control to grid and output is what eats up the ripple. But this really only works with a constant load, ie class-A circuit. If the output load is wildly varying, there will be signal related ripple on the rails (that signal is not passed to the control grid). But there is where local bypass capacitors come in, they should be large enough to provide all local current variation needs.
It is basically a ripple current reducer. The idea is to measure the ripple current with a small resistor and shunt the load with an equal an opposite current. There are plenty of solid state circuits for this on the web.
Cheers
Ian
Cheers
Ian
This competes well for the worst circuit I've ever seen. The only difference between the active portion and a current source is a 0u1F capacitor. Effectively, it's a variable resistor with this goofy active current source in parallel with a big capacitor. A huge improvement would be to remove the parasitic active bits., keep the series resistor and bypass cap, and get some sleep.
All good fortune,
Chris
All good fortune,
Chris
I've seen a manuscript with LOTs of better plans. To mis-quote: today a 30uFd capacitor is even better, and avoids tube cost and power.
The link in PFL200's post is interesting because the author could NOT use any practical (for 1937) bulk capacitor. But that cat can be skinned other ways.
In the article the author writes that he used the ripple killer instead of a filter capacitor because a filter capacitor would introduce phase shift.
That's a clear indication he hasn't a clue. That makes the whole circuit suspect.
I would discard it.
Jan
I would discard it.
Jan
Theres nothing wrong with a shunt ripple eater. Combined with a cap multiplier it can be very good. In MI amps you dont always want regulated rail. But you dont want ripple noise.
So the design by O.H. Schade (RCA) would not work like described in the article (see post #10) because Schade, or atleast the author of the article, hasn't got a clue?
The circuit that was posted has a lot of problems and wouldn't work as advertised. I do not concern myself with where it comes from, or it's alterations, 'improvements' or whatever it suffered on its journey.
I don't think that's relevant - what's relevant is the circuit.
Jan
I don't think that's relevant - what's relevant is the circuit.
Jan