Best low noise regulator?

[jwb]

Originally posted by Bonsai
Its no good simulating the output load with a sine wave - you need to use a square wave to see how the regulator responds to fast transients. This will let you get an idea of how your loop is compensated and whether it needs tweaking (you won't get the absolutely correct answer from the sim, but the sim will help you understand what is going on in the circuit and where th e best place is to tweak). The last thing needed on a high performance reg whether shunt or series pass is overshoot or ringing on load or line transients.

That's true, but it's also true that the regulator is relied upon only in the region < 100kHz. The large bypass capacitor is in use from 100kHz to several MHz, and the small bypass capacitors handle frequencies from a few MHz up to 1GHz, where the mutual capacitance of the power planes handles the remainder. In my opinion, it's irrelevant how a regulator performs for load steps with rise times less than 10µs.

Also if you have large input transients you have a major architectural problem with your circuit!

That said, I think the Jung has creditable transient performance. Here, for instance, is the response to a 1V transient with 1ns rise time on the input.

[jwb]

Response to load transient with 1ns rise time.

[jwb]

And finally, response to load transient with a 10µs rise time. All fairly good. The trick is to have a good PCB layout so you achieve these performances in reality :-)

[Bonsai]

Looking at your last post, if I'm reading it correcly I'd say you have problems. There should be no overshoot or ringing.

Using a rise time of 1ns is not practical.

Use a few u seconds.

[Bonsai]

The reason to test with a step or square wave input is to se e how th e circuit behaves wrt to line noise - and you do get spikes and straight HF noise coming through much more than you would believe. How does your cuircuit behave unde r thes e conditions - does it pass th e nois e straigh t through? does it oscillate? Its easy to deal with DC or low frequncy input voltage shifts, but at higher frequncies this is another matter.

[jwb]

Originally posted by Bonsai
Looking at your last post, if I'm reading it correcly I'd say you have problems. There should be no overshoot or ringing.

Using a rise time of 1ns is not practical.

Use a few u seconds.

The last post is with a 10us rise (fall) time.

[MRupp]

There are zeners with substantially lower dynamic Z than LEDs these days. I just bought a box full of 6.2V zeners with 9Ω dynamic impedance at 5mA.

FWIW: According to the book - and here I actually mean "Valve Amplifiers" by Morgan Jones - cheap red LEDs and infrared LEDs should have impedances in the range of 5 - 6 Ohms, and the new and brighter LEDs of varying colours obvioulsy have higher impedances. I think LEDs like healthy currents of 10 - 20 mA to be at their optimum (Zeners btw. too). Off course, this is all written lore and not my personal measurements.

[keantoken]

Hi all. This is a very interesting thread.

First of all,

http://www.diyaudio.com/forums/showt...05#post1759105

The 2N6387 is a power darlington! Is this an error?

Its no good simulating the output load with a sine wave - you need to use a square wave to see how the regulator responds to fast transients. This will let you get an idea of how your loop is compensated and whether it needs tweaking (you won't get the absolutely correct answer from the sim, but the sim will help you understand what is going on in the circuit and where th e best place is to tweak). The last thing needed on a high performance reg whether shunt or series pass is overshoot or ringing on load or line transients.

I have a simple and brutal circuit here which should suffice.



Lastly, Does anyone want to take a poke at a Rush Cascode based Shunt or Linear regulator? Here is my little stab at a shunt regulator (attached). Input regulation is very good but not as good as my linear version, which I will post if anyone wants to see it... I have not tested this circuit in real life.

I know someone will complain about thermal drift but really, is a perfect 12V output voltage at all times that important for an amplifier?

- keantoken

[jan.didden]

Originally posted by Bonsai
The reason to test with a step or square wave input is to se e how th e circuit behaves wrt to line noise - and you do get spikes and straight HF noise coming through much more than you would believe. How does your cuircuit behave unde r thes e conditions - does it pass th e nois e straigh t through? does it oscillate? Its easy to deal with DC or low frequncy input voltage shifts, but at higher frequncies this is another matter.

Bonzai,

A test with a sine sweep to some high frequency showing curves of magnitude and phase against frequency gives you the same info and is often easier to interprete. I can see for instance that PSRR decreases to say 20dB at say 100kHz, or that pphase shift reaches 180 degr at say 200kHz. Looking at a ringing sq wave doesn't give me so much quantitative info. But I guess it's also a matter of personal preference.

Jan Didden

[Bonsai]

Jan,

for the measurements you mention I do not disagree (you can alsodo a straight AC gain and phase plot for both th eoutput and input). I do these to get this info. For loop stability and feedthrough, I prefer to use a squarewave load or a squarewave on the input.