My take on a discrete shunt voltage regulator

[HiFiNutNut]

This regulator gives a clean, detailed and neutral sound. Impressive is that you chose RF parts yet I can't hear any oscillation so I hope it does not have any.

If you are into a bit of warmth, tone, slam, musicality, etc, you may stick with Salas v1 or v1.1, as it is very good in that department. This v1.5 may not be the one for you as those may be lost. Of course, when we talk about these subjective terms, it is highly system dependant.

If you look for details, transparency, accuracy, then this v1.5 is superb. As those characteristics are what I am looking for, in my personal opinion, it is the best I have had and I am finally settled on this one.

Thanks, Iko. Well done.

I may not have a lot of time these days checking into the forum. I usually stay in the Multi-way forum these days and don't check other forums. If your circuit further evolves in the future please DO send me an updated schematic to my personal email address (please keep it!).

I guess there may be a chance that the performance can be improved by using all FETs. You may also select the parts to make it more temporature stable.

[ikoflexer]

Thanks for the report! BTW, the RF parts are chosen with the sole purpose of avoiding oscillation.

At the moment I don't think that using fets everywhere would improve anything. The circuit will stay like this for a while because I'm super busy with the pcb and moving forward to offer it to people who aren't as brave as you

You well done!

[Salas]

Congratulations Iko. Good work.

NutNut: Thanks for the experiments and the explanations about how they worked in your system. I am sure people will notice your experiences as good indicators.

[HiFiNutNut]

I am using 2sk170. Would the 2n3819 and J201 give better drive, more temporature stability, less chance of oscillation, lower noise?

[ikoflexer]

Using a particular resistor between G and S in current sources will make it more temperature stable. There should be an improvement using 2n3819 and J201. I don't know if it would be audible.

[HiFiNutNut]

There is nothing to tweak other than the output capacitor. In my case, the "input" capacitor is already the best I have, i.e. a quite transparent Vishay Blue Box MKP 1.1uF.

Surprisingly, the output capacitor still has a strong sonic signature.

I know your reference is to make the uF as small as possible.

Initially I had a 10uF Nichison Gold. An old capacitor taken from a NAD CD player which may have been played for a while. The reason I used it was because that was the lowest value I could find in my parts bin.

The sound was generally good but dry. I have done this sort of listening (output caps on regulators) a lot so I guess I am fairly experienced with it. I guessed that either the ESR was too high or the uF was not large enough.

I then replaced it with a Rubycon ZL 22uF/50V 0.34R ESR. The music was a lot more fuller, smoother and more live. However, just like every other Rubycon ZLs, the treble became a bit thick / exaggerated, a touch of brightness.

I was shocked by the fact that the output capacitor can influence the sound so much, given that you said the output capacitor only serves for stablity, not for the close loop. No matter how low the impedance of the capacitor, the regulator still has far lower impedance so in theory the capacitor should not influence the sound much. But reality is different from our theory.

[HiFiNutNut]

You should surely be able to model it using LTSpice to find the optimal output capacitor value and ESR?

[HiFiNutNut]

I found in my parts bin Rubycon ZL 15uF/100V ESR=0.54, which may be better than the 22uF/50V ESR=0.34.

But before I try that, I have been thinking about that if the output indeed needs only small uF to be stable, as indicated in the schematic of 5uF only, then I may use film capacitor as well.

I don't want to consume my 1.1uF Vishay though, as they are too valuable to me and too many are needed to be consumed in this case. They should be used as input/output caps and in filter circuits. But if they do give a better sound then I may have to use them.

I have a plenty of ICW SA Clarity caps. I don't mind using them in modelling XOs. I don't like them in line level circuit. Solen is a lot worse. In fact, I am not sure if any large bulky MKP made for XOs sound better than ordinary MKP made for lower voltage use.

An alternative is to use 2.2uF Polyester MKT capacitors, I have at least two dozens of them on hands. They are very small so I can even parallel 5 of them to make up to 11uF. However, in my previous experiments using them in critical locations, I found that the Vishy blue box sounds better, and I prefer MKP to MKT.

In whichever case, I guess I need to pad some resistence to the film capacitors to prevent oscillation. Perhaps 0.5R? 1R?

It would be good if you could provide us an answer with your LTSpice modelling, not just "you have to experiment yourself".

[HiFiNutNut]

Originally Posted by ikoflexer

Using a particular resistor between G and S in current sources will make it more temperature stable. There should be an improvement using 2n3819 and J201. I don't know if it would be audible.

Would you please explain more? What resistor value with which JFET gives the best stability?

I was thinking about whether the lowish current level was related to temporature instability.

[HiFiNutNut]

Opps, your new schematic already has the resistor values for the JFETs. So I guess they are the optimal values. It looks like I will have to order those parts from Farnell and RS Components soon.