My take on a discrete shunt voltage regulator

[keantoken]

I know I'm a little late, but...

Would it be possible to make an AC-mode version of this that idles at a specific current and that you can hang off a normal series reg to improve performance? I think this would be a great way to bump up performance of say a LM7812 without messing up an existing board much. I'm thinking it would also improve noise and transient suppression. (AC mode meaning it acts like a capacitor so you don't have to adjust it to precisely match output voltage)

A necessity would be the ability to decrease BW to emulate ESL, so you don't drive the master series reg crazy.

- keantoken

[ikoflexer]

Hey kt, it's never too late for tech talk.

I'm just not sure what you'd be after. On the one hand you say it would act like a capacitor, but on the other hand you say it would emulate ESL. A filter?

[keantoken]

Not quite a filter.

This circuit has the not insignificant advantage of very low ESR. This is what gives it the advantage over simply loading a reg down with huge capacitance. A capacitor large enough to compare would either have plenty ESR or it would be expensive. It also has the advantage over a capacitor that it won't draw huge amounts of current as it charges up.

Hanging it off of a chip regulator would improve audio range output impedance as well as transient response, and takes the AC load off the regulator.

Again, depending on the series reg it would probably have to be BW limited (BW limit in a regulator "looks" to the load as if it has an inductive supply, hence ESL). If it has too little impedance beyond the series reg's BW, it will overstress it and cause them both to go into oscillation (then again, perhaps the opposite can happen - the shunt being so fast it keeps the series reg from oscillating. But I think it would have to be quite a bit faster. Maybe we could slow down the series reg to achieve this affect?).

Another advantage is that it could be wired onto an existing board to "upgrade" the present regulating circuitry, probably a chip reg of some sort.

- keantoken

[ikoflexer]

I don't think it would oscillate, as long as the shunt reg has enough voltage drop to do it's thing (the CCS part needs it). Say for an overall output voltage of 9V, use a LM7815 before the shunt reg, and set Vref to 9V. Apparently John Curl uses the likes of lm317 to pre-filter, then some other stuff, and finally, a shunt reg. I don't recall the exact details though and I'm too tired to look it up now.

To me it definitely makes sense to lower the output impedance of the supply via a shunt reg. I know you're fond of simulations (I use a fair bit the simulator too) but I think you'd enjoy comparing a regular lm317 regulator to a shunt reg. It's really quite a difference. You don't even need a pcb, I often prototype new circuits as point-to-point construction.

[keantoken]

Actually, my setup is near perfect for testing regulators, my oscilloscope can measure down to 10uV, providing I get the circuit grounded right... BW at that point is only 1MHz though. I can measure down to 5mV upwards of 20MHz.

Yes I am fond of simulations, but I have been trying to get my workbench to the level where I can experiment as often as I want. I now, with the addition of my (self) repaired Tektronix 465B, have in terms of equipment everything I need. So after some much thought-out mouser orders to stock up on miniature plastic explosives (transistors, in other words), I'll realize I no longer have the right to procrastinate or complain about my equipment. (I have to thank Kenpeter for the heavy, bulging abuses of the flat-rate postage system he's been sending, I would be able to procrastinate longer were it not for his valiant efforts on my behalf).

I'll have to do as you say, might buy a few boards myself.

- keantoken

[ikoflexer]

Ooh, that's interesting! A scope with the 10uV/div? Which one is that? BTW, the 465B is nice.

The best my scope can do is 100uV/cm. I've also built a low noise amp for regulator testing, with 20db, 40db, and 60db gain. But yeah, low noise measurements aren't trivial, especially one has to be careful about grounding.

[keantoken]

Tek 561B, from the late 60's I believe. It uses plugins, so I've got the 3B4 timebase and 3A9 diff amp. It's not the cream of the crop but I must say I like the style it brings to a bench. Everything nowadays has corners sharp enough to cut off a wandering finger...

Internal noise at 10uV covers approximately 1.2 div at that point, so I can't say it's much better than 100uV per div. An interesting thought is to replace the old components with newer, quieter, faster types. Of course, that is to be done with some caution... A curve tracer would be helpful to find exactly the right types, and most of the datasheets are not to be found (not to mention Tek commissioned many semiconductors which bear special undecipherable Tek numbers). Another semi issue is drift. It's not that bad but it can be bothering. Here are some pictures I took about a week ago when I finished repairing my Tek, pardon the mess.

Needless to say, I'm proud of my Teks.

- keantoken

[ikoflexer]

That's nice. I like those with modules. Mine's a really old one, a Tek 502a, tubes but can't complain about the 100uV/cm, differential input, and dual beam. I don't even dare think of starting to replace anything. As long as it works I won't touch it

Oh, and your mess looks very familiar

[marcus1]

Great thread - I discovered it today but don't find it too simplistic as I only understand about 50% of what's written !!
A while back, someone asked "What is max/min voltage range & max/min current range?"
I've got a kit-set 55W amp still waiting to be built (requires +36 and -36 volts per channel from the power supply) and wonder if your regulator might have the potential to supply this?

Thanks

[ikoflexer]

Hey marcus, thanks!

I think it would work. You'd need huge heat sinks though, to dissipate 55W continuously. This is one of the things about a shunt regulator, it needs to burn all this energy to have the current supply when needed by the load. Basically, if I can draw an analogy, think of it as a pipe that splits in two, one pipe goes to the load, which in your case is the power amp, and the other pipe goes to the shunt mosfet. The total "water" from the wall gets distributed to the load if it is needed; when the load doesn't "consume" this water, it gets redirected to the shunt mosfet, the other pipe. As soon as the load demands it again, the circuit redirects the water from the shunt mosfet to the load. But all the time, whether your load/amp consumes or not, the water flows. Wasteful, for sure. But able to supply on demand better than other circuits.

So, it's really like building another amp.