Best low noise regulator?

[analog_sa]

Originally posted by janneman



The error amp works open loop.

Jan

Any comments on topologies where the error amp has a defined closed loop gain such as Manfred Huber's designs?

I wonder whether this really can be called open loop as it obviously closes through the power rail.

[syn08]

Originally posted by housing
How about LT1027?

Looks good on paper, although the frequency response doesn't seem that great - unfortunately I don't have any. LT is not very generous with samples

[jan.didden]

Originally posted by analog_sa



Jan

Any comments on topologies where the error amp has a defined closed loop gain such as Manfred Huber's designs?

I wonder whether this really can be called open loop as it obviously closes through the power rail.

That's not a regulator, It's a DC amp that amplifies the ref voltage a number of times with a common source follower buffer.

Jan Didden

[analog_sa]

Not a regulator?! Are you talking about the circuit in post #10? Yes, the amp is there but it amplifies the difference between the ref voltage and a fraction of the output. How can this not be a regulator?

[kaos]

Post 10 looks like a regulator to me (with global feedback), a HV one at that. The nested loop around the OPA637 helps to control overshoot and ring to small signal deviations on the rail, but does so at a cost of overall line/load performance. It will also tend to flatten out the regulator’s impedance (ignoring the effects of the bypass cap), in the case of the OPA637, to a bit over a 1MHz. But it does so by bringing up the minimum impedance value. Looks like a pretty decent design for a tube supply (does a tube working at 350V really need all that??).

But what do I know, I don’t understand shunt regulators.

[analog_sa]

Thank you for your reply kaos. It confirms my understanding of these circuits which was really a bit shaky and growing even shakier with syn08's responses. And yes, tube circuits really need a good power supply, maybe even more than opamps

[KSTR]

Originally posted by syn08
LT is not very generous with samples

I remember times (late 90ies) when they would pamper us engineers with a box full of samples and ev-boards of their newest stuff almost every week. Even stuff I never requested! Now the cat got fat and lazy, it seems.

BTW, I also don't seem to understand the role of those caps accross the LED feed resistor. Is this an empirical thing only or something conceptual I'm missing?

- Klaus

[Terry Demol]

Originally posted by kaos
I’m sorry, I can’t see the regulator in post 21’s schematic working at all. The gain of the error amp is set by the feedback loop ratio of R4 & 5, which as pictured is around 26. The LED reference should be around a volt or so for that LED (looking at the data sheet), so that would yield an output voltage of around 26V. The input is only 15 volts in the schematic. What’s the output voltage supposed to be?? The addition of C1 & 2 across R5 is a positive feedback path and should lead to worse noise, if not outright instability. Perhaps C3 mitigates that to some degree (lag vs. lead)?? Sounds tricky at best. Commonly, despite the pictures, C1 & 2 would be across D1 (C1 would be made somewhat larger in that case), with a counter diode across R5 to protect U13’s input from the charge stored in C1 & 2 at power down. To reduce noise it’s common to place a cap across the feedback resistor, in this case R4, though stability issues need to be considered when that’s done. I’m at a loss as to how this regulator can be expected to work optimally as pictured. Am I missing something?? Sorry, don’t mean to come across harsh, just at a loss …

Originally posted by syn08


Yes. You miss the understanding on how parallel regulators are working Sorry, don’t mean to come across harsh, etc...

Syn,

Unfortunately I didn't have time to follow the earlier thread where
your reg came up so apologies if you have already covered this.

There are some interesting but somewhat, at first glance, odd
arrangements in your reg.

Is it possible that you can explain a little more clearly how the
(effectively) OP driven LED works?

If this reg is re-drawn as a simple opamp circuit sans OP fet, with
the +/- IP's swapped it appears not optimal noise wise.

It would appear to me that the noise of R243 is multiplied by the
gain of that resistor string R242 & R243.

Thanks,

Terry

[jan.didden]

Originally posted by analog_sa
Not a regulator?! Are you talking about the circuit in post #10? Yes, the amp is there but it amplifies the difference between the ref voltage and a fraction of the output. How can this not be a regulator?

You are right, it *is* a regulator. My bad.
I wonder why that design has the error amp limited in gain? One would expect that running it open loop would make for a much better regulator. Maybe he had stability issues he tried to fix by lowering the loop gain?

Jan Didden

[jan.didden]

Originally posted by kaos
Post 10 looks like a regulator to me (with global feedback), a HV one at that. The nested loop around the OPA637 helps to control overshoot and ring to small signal deviations on the rail, but does so at a cost of overall line/load performance. It will also tend to flatten out the regulator’s impedance (ignoring the effects of the bypass cap), in the case of the OPA637, to a bit over a 1MHz. But it does so by bringing up the minimum impedance value. Looks like a pretty decent design for a tube supply (does a tube working at 350V really need all that??).

But what do I know, I don’t understand shunt regulators.

That maybe the reason for the limited loop gain (trying to minimize ringing/overshoot). But proper compensation should be used for that, and then getting the lowest possible output impedance and best PSRR. A flat impedance curve in itself is no virtue; I'd rather have a Zout of 100u Ohms at 1kHz rising to 10m Ohms at 1Meg than a flat one of 10m Ohms from 1k to 1Meg.

Jan Didden