Regulators

[carlmart]

Even if they have probably been around for a long time, I recently got acquainted with TL783 regulators.

Has anyone used to good effect?

My concern here is for how they can influence audio quality, when compared with other types like 3X7 types or LT1085/1033.

To start with their impedance is a bit worst than the 3X7 types, as you can see on the graphs below, the 783's being on the right.

On the other side they look ideal to use as voltage regulators in low current stages in a power amp. Opinions?


Carlos

[peranders]

These curves shows the output impedance with a rather "naked" regulator. If you add caps you will get lower impedance but your step response will probely be worse but if you have a rather constant load, this doesn't matter so much. One trick can be that you add a small resistor (1-10 ohms) between the load and the regulator, then you can add caps as many you want.

[jam]

Has anyone run impedence plots on feedback regulators (including ic's), non-feedback regulators as well as shunt regulators.

I think the results should be quite interesting, as I believe the sonics of any circuit is highly dependant on the power supply, and this is one area worth more investigation. There is probably a signifiacnt correlation batween supply impedence and sonics. A new thread maybe?

Mr. Pass or Jonathan, maybe you would care to start.

Jam

[Chris]

Originally posted by peranders
These curves shows the output impedance with a rather "naked" regulator. If you add caps you will get lower impedance but your step response will probely be worse but if you have a rather constant load, this doesn't matter so much. One trick can be that you add a small resistor (1-10 ohms) between the load and the regulator, then you can add caps as many you want.

Hi peranders.

Is this the same trick you talk about ??
http://www.daisy-laser.nl/homeoptics/page32.html

[carlmart]

Originally posted by peranders
These curves shows the output impedance with a rather "naked" regulator. If you add caps you will get lower impedance but your step response will probely be worse but if you have a rather constant load, this doesn't matter so much. One trick can be that you add a small resistor (1-10 ohms) between the load and the regulator, then you can add caps as many you want.

A tricky question, or one which is dealt with according to different opinions is how much capacitance to add after the regulator.

Many seem to find it's better to put a lot, like 4700uF or more, but some people have reported HF oscillations when doing so. Apparently you should put small caps at the output, up to 47uF or so, and put a large capacitor bypassing the adjusting leg. In fact this is what the manufacturers advise, and I don't see why we shouldn't follow.

I think the resistor trick works much better if you put it before the regulator, in two or three stages, in Pi fashion.


Carlos

[peranders]

Originally posted by carlmart
I think the resistor trick works much better if you put it before the regulator, in two or three stages, in Pi fashion.

No, AFTER! The purpose of the resistor is to isolate the regulator from too heavy capacitive load but you don't need to this if you think the step response is OK and the regulator not oscillates.

Putting resistors before the regulator and the smooting caps will reduce overtones and make the unregulated voltage smoother. This is a good idea for a preamp power supply, not good for power amps!

[ALW]

I think the results should be quite interesting, as I believe the sonics of any circuit is highly dependant on the power supply, and this is one area worth more investigation. There is probably a signifiacnt correlation batween supply impedence and sonics.

There most certainly is a correlation between many PSU characteristics and sonics - irrespective of circuit topology.

No circuit is immune to PSU's no matter how high it's PSRR.

Andy.

[jcarr]

The attached circuit (by Takashi Kubota) has a fairly flat impedance curve across the audio bandwidth. Up to 20kHz it measures about 40mohms.

But in my experience, the impedance curve of a regulator is not the key to everything. Noise, intermodulation, step response, what happens to the ground returns, and other factors are also very important.

One worthy idea that I don't see used too frequently is to make a regulator that is fed with a constant-current source (the value should be equal to the maximum amount of current drawn by the load circuit, plus some margin). No matter what the load circuit does, the regulator will always draw a constant amount of power from the power transformer. For the same reasons, a "Super-Shunt" regulator is also worth trying. See:

http://www.c3-net.ne.jp/~ufo/AB100shunt.jpg

A constant-current draw should be particularly good for CD players, where analog circuits intermingle with motor drivers, servos, digital circuits, and display circuits, yet you shouldn't allow the power drain of one type of circuit to modulate any of the other circuits.

Another kind of regulator that I think has a lot of potential is a charge-pump (flying-capacitor). It should (theoretically) make the audio component immune to the condition of the AC powerline. See:

http://www.maxim-ic.com/appnotes.cfm/appnote_number/725

regards, jonathan carr

[jam]

Jonathan,

The shunt regulator schematic you provided is similar to a design used by Stax which is a very good sounding regulator indeed, but slightly parts intensive.

I have found that, to my ears at least, shunt regulators give the best sonics but you have to burn at least as much current through the shunt as through the circuit, to get to work at their best.

The Kobota circuit seems pretty popular in Japan and I believe that Kaneda has used it on occasion.

What is your preference?

Regards,
Jam

[jcarr]

Jam: As far as I know, Stax invented the super-shunt regulator at the end of the 1970s, right around the time that Masao Noro (who was working for Stax at the time) invented the folded-cascode.

Japanese DIY magazines from the first half of the 1980s have lots of designs featuring some variation on the super-shunt theme. Super-shunt regulators tend to manifest themselves most frequently in the bottom end - with a really solid, powerful, and tuneful bass. And yes, they do run hot.

However, I don't know if a full super-shunt regulator is the most effective way to use componentry and board area resources.

Here's an alternative approach. Let's split up the amplifier circuit into the input, voltage amplification and output sections. And let us also design each section of the amplifier circuit to have a constant current draw.

Now you can use a fairly simple linear regulator like a Kubota variant, have flat impedance across the audible bandwidth, and also constant-current characteristics (due to the unvarying load). Because of the simplicity, you could build multiple linear regulators with flat impedance and constant-current for the same resources as a single super-shunt type. The Connoisseur 4.0 uses this multiple regulator approach (12 regulators per amplifier module), and both the measurements and sonics seem ok.

>What is your preference?<

The basic "building-block" approach would be as above.

I have made regulators somewhat like the Kubota in the past, and if I were to use such a circuit, I would at least feed the differentials with constant-current sources. I also have some doubts as to whether a capacitor directly across a zener is the best way to absorb its noise (due to the low dynamic impedance of the zener). Probably better to insert a series resistor after the zener, and connect the capacitor between the resistor-differential input node and ground. Loading the differental outputs with a summing current-mirror would also be worthwhile trying, as would a pre-regulator.

>The Kobota circuit seems pretty popular in Japan.<

It is.

>I believe that Kaneda has used it on occasion.<

This I sincerely doubt. Kaneda and Kubota have different approaches to designing, and they appeal to different types of DIYers. Kubota tries to keep things simple and easy to build, while Kaneda is definitely for the more advanced DIYers. Besides, remember that Kaneda and Kubota are rival writers for MJ. If nothing else, designer's egos would almost guarantee that Kaneda wouldn't want to touch anything designed by Kubota.

regards, jonathan carr