4-part Harmony, a 9 watt class A amp using a voltage regulator IC.

[Circlotron]

Just something I threw together the other night to see how well it would work... The name comes from the fact that it has only four parts fundamentally. With a 16 volt supply the power output should be about 9 watts given a 12v peak signal into 8 ohms. I didn't have enough input signal to drive it far enough though because the gain is only +1. Distortion-wise, I haven't measured it but using a 24 ohm resistive load and a 40v supply the output tracks the input (with a 1.25v offset) within +/- 0.5mV to 30v. That's a 0.00166% error. Hopefully distortion will be similar, but looking at the waveform, the scope of the choke loaded version is pretty clean.

The 1.2 ohm resistance of the choke manages to set the quiescent current at almost exactly 1 amp because of the 1.25 v output of the voltage regulator chip when the Adj terminal is held at ground potential. There is approx 50uA flowing in this lead hence the 1k input resistor. Depending on the layout, you may also need a few tens of nanofarads across this resistor as well to keep the thing from oscillating.

Why use a voltage regulator chip? Because they're not supposed to be used that way; that's why. Anyway, they are a linear power IC and they don't really know whether they are processing music or power supply noise. Hard to tell the difference with some people's music though. They look just the ticket for a class A follower amp. I'll have to look at some of the larger types of reg chip now.

[psarin]

Hi

I was also intrigued by the use of LM317. I tried the following circuit from the current article on http://www.tubecad.com (a very interesting read, and a good analysis of the Zen topology etc, including a tube input stage etc.)

I breadboarded this one together using a few sample chips, and was quite pleasantly surprised at the result, vis-a-vis my six-channel Gainclone Kind of makes you think that all op-amps are more or less alike.

Cheers -

Pradeep

[jcarr]

Circlotron: As someone who also enjoys dabbling in stuff like this, I recommend that you add the TL431 to your list of guinea pigs, er, _candidates_.

The 431 has been marketed and popularized as a 3-pin shunt regulator device, but it appears that another valid way of looking at it is as a discrete NPN transistor with inordinately high Vbe.

In fact, if you use two them, you can make a differential amplifier with tolerably low distortion figures. Applying all of the normal design stuff like degenerative local feedback is no problem whatsoever. The package is a standard TO-92, so if you conceal the markings, visually it looks just like any other small-signal plastic transistor.

We are already have transimpedance circuits and transconductance circuits. Maybe we can now also have "transvestite circuits."

best, jonathan carr

[Circlotron]

Originally posted by jcarr
Circlotron: As someone who also enjoys dabbling in stuff like this, I recommend that you add the TL431 to your list of guinea pigs, er, _candidates_. best, jonathan carr

Actually, I have a few dozen of them and I have often thought about doing this but was too slack. In the TI databook they show an app for one as a class A amp using an output tranny and it gives about 400mW. I think a '431 would be very suitable as an addition to this amp, simply on principle if nothing else. Thanks.

[jan.didden]

Originally posted by Circlotron
Just something I threw together the other night to see how well it would work... The name comes from the fact that it has only four parts fundamentally. [snip]
Why use a voltage regulator chip? Because they're not supposed to be used that way; that's why. [snip]

Circlotron: I once read that "Creativity is the capability to re-order" (Askes). If that is true, you are one hell of a creative guy!

Jan Didden

[Circlotron]

Here's something else. I don't know if anyone has ever tried this combination before. Two LM317's in what I call a Tetrahedron amplifier. More details in this thread -> Tetrahedron output stage topology.
With the limited drive voltage I had available it made a nice clean 10.1 watts into 6 ohms resistive. I just happened to have a big 6 ohm resistor so that's what I used. The circuit is biased more or less class A with about 600mA quiescent per '317. The 1ohm resistor sets the bias because it has a constant 1.25 volts across it. It also serves to make the circuit into a long tailed pair so it has better balance, but with about 80dB internal feedback on the 317's it probably doesn't do much in this department. I'll have to try backing off the bias and up the drive a bit (I'll just *have* to get a proper oscillator ) to see just how much I can get out of it. Actually, if each '317 doesn't conduct for the full 360 degrees the resistor voltage may move around a bit and it may distort. I might just be stuck with p/p class A. But the main thing is - it works. Another good thing about these IC's is that you just set and forget the bias because they are very temperature stable compared to a normal transistor or mosfet. They have to be, to give a steady output voltage in their normal mode of use.

Next trick (seriously) is to use two LM317's and two LM337's (the girl version of the '317) to make an "X" amp. That'll be fun.

[Circlotron]

Originally posted by Circlotron
The 1ohm resistor sets the bias because it has a constant 1.25 volts across it. It also serves to make the circuit into a long tailed pair so it has better balance...

A few hours after I posted this I realised this is not quite true. It would only work as a LTP if the 4700uF caps were removed. Otherwise the "lower" winding fed by that 1ohm resistor cannot move in common mode because of unbalance therefore send a corrective signal to the opposite device.

[Dirk]

Circlotron,

just let me tell this circuit (first post) is the most beautiful amplifier I have ever seen!

... pity it's class A (I like the additional constraint of minimal power consumption :-)

Dirk

[Circlotron]

Well thank you Dirk. You do realise of course that saying this obligates you to spend five minutes making one and reporting back your findings.

I did say I was going to make an X amplifier next but after thinking about it a bit, I am first going to make a complementary symmetry buffer using an LM317 / LM337 combination. Maybe even a bridge amp.

Also the other day I made a little amplifier stage from a '317 with 12 ohms from "output" pin to ground, 170 ohms from "input" pin to +40v, and 1k from "adj" pin to ground. Signal fed into "adj" pin as before. Output is from the "input" pin. Gain is 170/12 = x14.16. Frequency response is flat to 50kHz at least, and the output swing was at least 25v p/p, probably more; I can't remember now. Distortion is not visible till about 30khz by which time it is about several percent by eye. On a square wave the rise time was ~300nS and fall time ~2.9uS. The value of the top resistor should drop about half the supply given that the current is 1.25 / (bottom resistor ohms).

With equal value top and bottom resistors and the adj pin bias set at 1/4 rail voltage - 1.25v (e.g for 40v rail set bias at 10v - 1.25v = 8.75) the setup should make a nifty little split load phase splitter.

[Circlotron]

Originally posted by psarin
I breadboarded this one together using a few sample chips, and was quite pleasantly surprised at the result, vis-a-vis my six-channel Gainclone Cheers - Pradeep

Hi Pradeep. Yes, the constant current sink at the bottom is good value for this kind of cct, particularly if the load is realtively high impedance. It makes the top ic *very* linear. I measures a '317 the other day set up as a ccs and it measured about 260k slope resistance. (change in voltage / change in current). If you really want to go overboard like any serious diy'er try this ccs; it has a slope resistance of *at least* 400 Megohms. http://www.diyaudio.com/forums/showt...7225#post67225

Actually, come to think about it, it should work very well at the top half of the amp as well if you drive the gate and adj pin together, with the gate about a fixed 10v above adj. Works best with high current fets because the lower you are on the characteristic curve the more horizontal the "curve" is. Also high voltage ones may be better IIRC, but it gets all a bit academic really.

Thanks for your reply psarin.