Valve Shunt Regulator Design Walk-Through

[Cressy Snr]

Ok so gee'd up by the comment from Phil that some of us would not have an inkling of how to design shunt regulated power supplies
I have decided to do a step by step walk-through of how to design and get one working.

Why am I doing this?
Well you see, with the ability to design shunt regulators comes the facility to
be able to use valves such as the modern production KT120 I'm using at the moment and make an amp whose resulting sound can easily compete with
old exotica that is fast becoming rare and very expensive.

Now this is a very simple, basic walk-through and of course there are series regulators too, but shunts are simpler to understand, use less components and
can both source and sink current. They are also arguably better sounding, but we'll leave that argument out of it for now.

The biggest step change in sound quality I got was when I applied regulators to a power supply and I sincerely believe that they are
the single most important improvement you can make to a valve amp, both in terms of the finished sound and almost total
absence of noise buzz and hum, which we all know is the enemy of good sound.

Some folk can put up with it and good luck to them but why suffer!

Again this is going to be very basic stuff, but it could change your outlook on designing your amps forever

I'll be back soon.

[Cressy Snr]

Right so where to start?

At the beginning I suppose

Remember what I said in the other thread, that everything clicked when I realised that current flows top to bottom in a valve but the electrons go the other way? Well first thing we need to know is what current flows and where it goes? (catchy eh?)

OK below is a diagram of a simple feedback shunt regulator of the type I used in the 801A amp so we'll analyse that.



For example I want to build a mono amplifier that uses a choke input supply, 300V HT and the thing draws 55mA. So the first question is why is 110mA coming out of the end of the passive bit? (No it is not because I made a mistake and we are really building a stereo amplifier).

[Cressy Snr]

So moving on.
The rule of thumb with shunt regs is that the regulator has to be drawing the same current as the rest of the amp draws at idle, so if the amp draws 55mA quiescent then so the shunt reg has to do the same.

So looking at what current flows and where it goes, we come to Kirchoff's current law, which is represented by the green junction diagram at the top.

Kirchoffs law states that basically if a specific current goes in to a junction then it all has to come out of the ends.

So applying a bit of reverse logic to that statement, if 55mA is passing through the shunt regulator valve and 55mA is passing through the amplifier circuit itself, then 110mA has to go in at the start.

So looking at the diagram we can see the relationships.
55mA flows into the load and 55mA comes out at the bottom of the shunt reg.

OK so now we have some solid figures to work from.
We know the current the amp circuit wants, therefore we also know what we want the shunt reg to be drawing, hence the design process can begin.

[Cressy Snr]

We need a shunt valve that will pass 55mA with a 300V HT supply without melting so in the case of a power amp we obviously will need a power valve to do the work not an ECC83

In this example I have chosen a triode strapped EL34 to be the shunt element.



Now the first thing we need to consider is the need to provide a voltage reference for the EL34 to sit on top of.
This locks the cathode down to a specific voltage, and provides the reference upon which the regulator is fixed so it can't flap about in the breeze.
Without this reference the regulator would not be much use.The VR tube does does two jobs:

1 it provides the voltege reference that enables the regulator to regulate

2 it keeps the voltage across the EL34 within the window that keeps it within its plate dissipation limits. Let's choose a 75C1 neon gas tube stabiliser.

The important thing here is that the VR tube must be able to pass that 55mA to ground without exceeding its design specs, otherwise it will simply burn out.
The 75C1 has a maximum rating of 60mA so we are within the spec.

The 75C1 VR tube will hold a steady voltage of around 75V on its anode when the rated current is passing through it.

So we have now jacked up the cathode voltage at the bottom of the EL34 to 75V above ground.

A bit of subtraction follows.

300V is at the top of the EL34, 75 volts is sat at the bottom so 300 - 75 gives us 225V across the valve.....

[Cressy Snr]

Now perusing the data sheet, the triode strapped curves tell us that if 225V is across the valve and 55mA is flowing through it then the grid bias needs to be -15V.



Looking at the max dissipation curve we are just cruising here so we can carry on designing.

So how are we going to provide those conditions on our EL34?

[Nick]

Good and helpful description Steve. Just one thing being the ever cautious, its worth considering what happens to the shunt valve if the load fails, heaters go out, or the valve is not plugged in. The shunt valve then has to take all the current, so if possible aim for a 100% overload margin in the shunt valve, which I think you have in the examples.

(you may have said that, in which case sorry, its Sunday morning and it was a quick read).

[Cressy Snr]

Below we have the shunt regulator with component values added so let's explain.



There is a rock solid 75V sitting on the cathode of the EL34 so we can't bias it from there.

The bias then is provided by the potential divider string to the right of the valve.

Using the potential divider formula or an online calculator which is how idle folk like me do it we can calculate the voltage that needs to appear on the grid of the EL34 to provide the right conditions.

So if we have 75V on the cathode and we need a grid bias of -15V, this means that the grid must be 15V below that of the cathode.
A bit of subtraction 75V - 15V gives us 60V required on the grid.

Now we do not need much current to provide this bias so we can use high value resistors in our potential divider.

Feeding the numbers into an online calculator gives us 390K at the top and 100K at the bottom of the divider. so 60V will appear at the middle.
But valves and VR tubes are not the precision devices we would like them to be so a 10K preset in the middle of the PD will move the mid point around a bit so that teh correct voltage can be dialled in.

For example our VR tube could be at 78V instead of its published voltage. The pot enables us to maintain the -15V relationship between the cathode and the grid, so that the right bias is applied.

The series resistor in the choke input power supply before the regulator sets the voltage drop from the output of the passive section to the 300V we want it to be and like the pot has to be adjusted on test.

I have given a 320V output needing about 200R to drop the volts to 300V

Once we have set the volts to 300 the shunt reg will keep it that way.

The cap on the output provides a low impedance AC path to ground for any remaining ripple that might have escaped..

The path through the pot to the grid provides a feedback loop that keeps things stable, and also provides an out of phase ripple signal that provides some cancellation of output ripple.

It is possible like Paul does to carefully design the regulation without caps so that ripple is cancelled out by using the individual characteristics of the shunt elements themselves, but it is fiendishly difficult to do.

Us numpties make sure that the ripple is smoothed by the choke input section and the output caps so that cancellation is an insignificant part of the ripple reduction technique.

It's a compromise, but it works and gives far better results than a passive approach. With a shunt regulator we can discard the power wasting resistor across the power supply normally used to keep the supply in choke input mode. You get a steady rising HT with no peaks. Neat eh?

It can also be seen that because we have an active element in the PSU chain the influence of anything before it on the sound of the amp is greatly reduced, because the signal return from the output stage is now virtually all going via the shunt valve rather than the PSU caps. However, don't use regulation as a band-aid for bad passive section design. The passive section and mains transformer still have to provide all of the juice to enable the amp to perform properly in terms of speed and dynamic envelope.

No doubt there will be comments (I hope) but that's it in a nutshell.
Hope it helps.

[Cressy Snr]

Good and helpful description Steve. Just one thing being the ever cautious, its worth considering what happens to the shunt valve if the load fails, heaters go out, or the valve is not plugged in. The shunt valve then has to take all the current, so if possible aim for a 100% overload margin in the shunt valve, which I think you have in the examples.

(you may have said that, in which case sorry, its Sunday morning and it was a quick read).

Yeah if you look at the operating point of the EL34 it is about half way to max dissipation, so giving plenty of headroom if anything were to go wrong.
You'd notice the music had stopped and would be able to turn off the amp before any damage occurred. The VR tube could end up toast but it's better than damaging anything more expensive.

[Cressy Snr]

One thing I forgot to mention is that if we are operating the cathode of a regulator valve significantly above ground (which we usually are) then it is likely that we will need to elevate the heater supply to said regulator valve to avoid exceeding Vh-k.

[Paul Barker]

Good and helpful description Steve. Just one thing being the ever cautious, its worth considering what happens to the shunt valve if the load fails, heaters go out, or the valve is not plugged in. The shunt valve then has to take all the current, so if possible aim for a 100% overload margin in the shunt valve, which I think you have in the examples.

(you may have said that, in which case sorry, its Sunday morning and it was a quick read).

Yeah if you look at the operating point of the EL34 it is about half way to max dissipation, so giving plenty of headroom if anything were to go wrong.
You'd notice the music had stopped and would be able to turn off the amp before any damage occurred. The VR tube could end up toast but it's better than damaging anything more expensive.

Yes VR will go very bright then go out and never light again if output valve fails in this circuit.

But the topology is so very easy and adaptable that it makes the whole thing worth the risk for someone like you and me Steve. It might be a problem implying the uninitiated can do it.

For safety we really need to come up with another method which doesn't rely on the Vr tube as the source of electrons. We could add to the impedance and put two reasonably matched VR's on groound, combine them with a pair of 100 ohm resistors so that the 60mA capability of each VR is not called upon unless an output valve fails. It does actually say on VR data sheets that this is the way to use them for more current, not my idea. But I suspect they would have to be matched or a hum pot device used to balance them.

Ideally we need to come up with a more reliable design. this is very simple and adaptable but it has this achilees heel.

Maybe a Zenner sounds no worse? I have no idea, someone needs to try that.

The other option is to use a valve in positive bias ground the cathode and use a voltage devider whith take off right near ground.

the way I was using which was very hard to implement was the string of VR's feeding the grid of the 572b from B+. you have to juggle with the grid current implications (ensuring sufficient grid current to strike the VR string) and the bias chosen was limited in acuracy due to available voltages of vR, so you are put into compromises. Then when you consider you have to do this in at least two seperate stages which entail therefore different voltages to juggle at each stage, and you are talking "not for the un-initiated".

There are other ways which are effective for ac ripple rejection but don't voltage stabilise, they will allow dc potential to rise and fall but they will reject ripple. These don't require a Vr tube.

However if you make this a shunt ripple reduction valve, you get the benefit of ac signal loop and you get the smoothing benefit. If it is preceeded by a series reg and or a choke input power supply the dc stability may not be an issue.

For myself I am happy to risk the vr tubes. I killed plenty already a few more won't matter.

[Paul Barker]

Thinking aloud.

What the data sheet for Barrister's says when expecting the Barrister to pass more current than it's design current is to parallel it with a resistor to bypas excess current.

I can't just remember the ratio, but there is a percentage of current which is allowed to bypass the barrister. It may be 20%. Obviously constant current effect is compromised but barrister survives.

It may well be that a threefold approach for us would involve. Two VR's combined via a hum pot (each capable of the normal current demand), and the whole lot bypassed with a resistor to take away 20% of the current anyway (but do not bypass it with a cap due to Vr tubes adjacent).

this would marginally compromise dc regulation, but not alter ac regulation or ac signal path.

[Cressy Snr]

the topology is so very easy and adaptable that it makes the whole thing worth the risk for someone like you and me Steve. It might be a problem implying the uninitiated can do it.
For myself I am happy to risk the vr tubes. I killed plenty already a few more won't matter.

As you say Paul, the circuit is very simple to build and easy to calculate.
Maybe it should come with a warning.

Don't fire up the regulator without having ensured that the rest of the circuit is operational.

Maybe a 10 point first fire-up guide is in order maybe like this:

1. Test your amplifier circuit as you normally would
carry this out with no regulator valves or VRs inserted.
Ensure the amp is working as it should be.

2. Switch off and allow charges to dissipate.

3. Insert the shunt valve and VR tube/s

4 Switch on again.
After around 20 seconds the VR tubes should strike and settle to a steady glow.
There may be a whooshing noise as the tubes strike but after that there should be no buzzes, hums or any unusual noises from the speakers.
If you hear anything untoward, switch off and investigate.

5. Check your shunt valve for signs of red plating,

6. If no red plating, the shunt circuit is probably functioning as it should.

8. Put a voltmeter between the top of the VR tube and the grid of the shunt valve. Use the preset pot to adjust for the desired bias voltage. The design
current through the regulator should now be correct.

9. Switch off and allow caps to discharge.

10. Switch on again making sure everything still comes up and stabilises as before. Now enjoy superior sound from your amplifier

I mean valve amp voltages are lethal under any circumstances
so users used to building valve amps I am sure would take all the necessary precautions to avoid bangs.

I'd rather have the pzzzzt of a VR tube going west, than suffer the almighty crack of a solid state HV regulator's chip being shorted to ground with a test probe.
I don't think my nerves would stand that.

[Mike H]

The next logical improvement to this is a a series regulator of some description, followed by a shunt regulator.

When the current drain increases (amplifier load) the shunt reduces its drain but the series reg increases its output; when the reverse is true, the shunt is drawing more, and the series reg is only supplying mostly that.

Safety features might be added, such as supposing the shunt reg current goes too high because of amp heater failure / valves missing , a current sensor in its cathode path say could trip off the mains input, for example.

In fact Nick's power controller module is readily adaptable to this idea, I'm using 490VAC choke input for about 395V DC after E-choke series reg, so you might imagine what would happen if the o/p valves went AWOL.

So I gots a MOSFET series shunt type thing but doesn't do much actually unless the reservoir side goes up to around 500VDC, or slightly under ~ however the MOSFET and its ballast resistor can't handle all that on their own hence the the shunted current drain operates a relay that breaks the controller's interlock loop. Controller then shuts off the mains

[Cressy Snr]

The next logical improvement to this is a a series regulator of some description, followed by a shunt regulator.

That's exactly what I've got on the KT120 monoblocks but I thought newbies to the regulator scene would appreciate a simple straightforward approach first.

Admittedly there are slight risks to the VR tubes holding up the cathodes by taking this simple approach but
an effective reg can be built with the minimum of fuss, provided a reasonable amount of care and the correct test sequence is followed when bringing it online.

That's not to say we can't complicate things at this point in the thread.
Gimme some circuits!

[Paul Barker]

You could use a cheap 811a like this.

What you have to juggle is the grid current and voltage.

Say for instance you have a 300v B+ and you want to pull 50mA your grid current is about 10mA. Your bias about +10v.

So you ground the filament, you use a resistive devider from B+ to ground which consumes 15mA of current (10 watts capability though dissipation 4.5 watts). but rather than using a simple pot because the grid current would fry it, you have to chop and change power resistors to arrive at what you need. You can play with the dc potential of the filament with a dc heater supply and a hum pot. With the hum pot you infinately vary the bias by +/- 3v, so fine adjustment is here, and use fixed resistors on grid to set the scene.

This would be much more reliable (than the valves and vr tubes used at the moment) and would deal with all output valves removed for a while, would withstand one removed for ever.

It is very similar to my 572b approach but muted down to suit requirements and pocket.