[tangent]

Quote:

The resistor between the output and the adjust allows the regulator to maintain a voltage reference, or benchmark, and the resistor from adjust to ground determines the output. This resistor is tied to ground, so it allows the regulator to monitor the ground status relative to both the input and output.

An LM317 type regulator doesn't care about "ground". If you move the "ground" reference that a regulator of this type is connected to, OUT will move the same distance. If the reference voltage is bouncing, OUT will also bounce.

The NatSemi datasheet describes the regulator's operation in terms of current. Since OUT to ADJ is always 1.25V, putting a resistor between them gives a constant current (I=1.25/R). That constant current then goes across the lower resistor, which is usually a lower value, so the voltage drop across it is higher, so it raises the regulator up above whatever the resistor is connected to. If the lower resistor is connected to ground, so be it, but that's not the important thing. Indeed, Walt Jung has recommended putting a zener from ADJ to ground instead of a resistor. It does the same thing: the regulator's 1.25V floats above the zener's reverse voltage drop, instead of floating above a resistor's voltage drop.

I prefer to think of the operation in terms of a voltage divider. It's inside-out-and-backwards to think of it this way, but it works for me. Let's say you want 24V out. What do you multiply 24 by to get 1.25? About 0.052. You can also just divide 1.25 by 24 to get the exact answer. (These are just algebraic reorganizations of the same problem.) A 5% resistor pair that gives this division factor is 120 and 2.2K. If you run these values through the "official" equation (1.25 * (1 + R2/R1)) you get ~24V. I just got there by a different route.

 

[jeffreyj]

Quote:

Originally posted by tangent ...Indeed, Walt Jung has recommended putting a zener from ADJ to ground instead of a resistor. It does the same thing: the regulator's 1.25V floats above the zener's reverse voltage drop, instead of floating above a resistor's voltage drop....

Ol' Walt must've taken one two many hits off the bong that day if he suggested using a zener to lift the reference above ground!?! Zener's have terrible noise specs - many fuzz boxes for guitars use one for the white noise source! - and exhibit even worse voltage stability. Since the bandgap reference internal to the 317 is supplied by a 50uA constant current source, there is not only no advantage to using a zener (or even another bandgap) to lift the ADJ terminal above ground, there's actually quite a few good reasons not to!

But hey, who am I to argue with ol' Walt

 

[Voodoochile]

Quote:

That constant current then goes across the lower resistor, which is usually a lower value, so the voltage drop across it is higher, so it raises the regulator up above whatever the resistor is connected to.

How can the lower resistor (the one that is not the current source, but rather connecting adj to ground) be of lower value and still have the calculation {1.25 * (1 + R2/R1)} work? Maybe it's a typo, or perhaps I misconstrued your example.

I typically use a 909 for the current source R, and something in the neighborhood of 10k to ground, depending what I'm looking for in terms of output voltage (15v in this example).

I understand what you are saying about if the "ground" were to bounce up and down, so would the output. (I suppose the pole vaulter analogy still applys... if the ground rose 6", so would the pole rise 6"). But what would cause the ground to move if it's coming off the centertap of a transformer's secondary?

 

[tangent]

Quote:

Zener's have terrible noise specs

No argument there. I got this second-hand. I can't point you to anything Walt wrote that says this, which would presumably have some justification for this.

Quote:

even worse voltage stability

A zener's voltage stability is good when current and temperature are stable, which they are, in this case.

Quote:

who am I to argue with ol' Walt

The loyal opposition?

Quote:

How can the lower resistor be of lower value and still have the calculation {1.25 * (1 + R2/R1)} work?

Sorry, I didn't proofread my message carefully enough before posting it. V=IR If current is constant, a higher resistor value gives a higher voltage drop.

Quote:

I typically use a 909 for the current source R,

It should be in the 120 to 240 ohm neighborhood, so that the device's minimum operating current always flows from the OUT pin. The reason your regulator is working with such a high value must be that you're not running it unloaded, or you have some small draws downstream on the power supply board like an LED.

Quote:

what would cause the ground to move if it's coming off the centertap of a transformer's secondary?

In that case, probably nothing. But, wasn't this in response to my point about using a dual-voltage supply with something like a META42? In that case, I can assure you, the ground can indeed "bounce".

 

[Voodoochile]

Quote:

It should be in the 120 to 240 ohm neighborhood, so that the device's minimum operating current always flows from the OUT pin. The reason your regulator is working with such a high value must be that you're not running it unloaded, or you have some small draws downstream on the power supply board like an LED.

Yes, another 1K across the rails.


Quote:

In that case, probably nothing. But, wasn't this in response to my point about using a dual-voltage supply with something like a META42? In that case, I can assure you, the ground can indeed "bounce".

Actually, that wasn't my intention... I would not suggest using a dual PSU's Pos and Neg rails only on the META, either. If that's what you mean. I think I heard what I think you thought I heard, I hope!

 

[jeffreyj]

Quote:

Originally posted by tangent A zener's voltage stability is good when current and temperature are stable, which they are, in this case.

Good point about the current being stable - that does erase the dynamic resistance contribution from the voltage stability equation - but relying on the temperature to stay constant is a bit wishful.

Quote:

The loyal opposition?

Who? Me???

 

[Chipko]

Using a dual voltage supply and omitting the ground you get in between would be bad as you're superimposing 2 errors on top of each other. Good intellectual trick though and if someone wants to measure stability it will double peak to peak error without affecting average noise over time.

The polevault analogy is very good as it explains the importance of a fixed reference.

 

[tbla]

as i recall its "wellknown" that r1 should be 1k and r2 should be bypassed with 100uF......the r5 bleeder is there to make shure that the reg allways is loaded with minimum 10-15 mA.
that combination gives the best sound and noise specs....

 

[ppl]

Jeffery J Your average Zener Diode will have lots of noise. Oh yes they do make great Fuzz Boxes. Note that all Zeners are not created equal and that the band gap references you referred to are quite useful at improving the Performance of the standard Adjustable regulator. If you referrer to the Data sheet for the LTC LT-1033 on Pg.7 top of page under Typical Applications you will find just such a Circuit and the text states clearly that most performance parameters like Noise, Thermal drift, and long term stability is improved by over 6 times that of a standard regulator. What are they using? A band gap reference!


Oh BTW your comment quoted below really made my day LOL. Got to hand it to you Jeffery j you are very entertaining


“Ol' Walt must've taken one two many hits off the bong that day if he suggested using a zener to lift the reference above ground!?! Zener's have terrible noise specs - many fuzz boxes for guitars use one for the white noise source! - and exhibit even worse voltage stability. Since the bandgap reference internal to the 317 is supplied by a 50uA constant current source, there is not only no advantage to using a zener (or even another bandgap) to lift the ADJ terminal above ground, there's actually quite a few good reasons not to!” Jeffery J

The below link should take you to Pg 7 of the LT-1033 data sheet
http://www.linear.com/pdf/1033fc.pdf

 

[Voodoochile]

I must confess having a huge grin when I read that one, too!

I don't know Walt, or know that much about him. But if he is an audio nut, the possibility surely exists! Nelson? Even more probable

.