D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg.

The denoisator also has the advantage of not degrading the internal resistance, in fact it improves it. Is it important for you?


this may be the only area of improvement...but is it needed?

the supply is for an opamp based xover drawing around 35mA

with the 10R / 470uF filter, the noise/hiss at the outputs of the xover is at the same level when the circuit is powered by batteries.
(without the 10R / 470uF filter, noise/hiss at the outputs of the xover is higher)
 
this may be the only area of improvement...but is it needed?
Well that's for you to decide: is your present situation unbearable?
Does it really need fixing?
Will you actually be able to fix something using the D-noisator?


The good thing about Dnoisator is is that it is purely additive, and changes nothing to the original circuit: you can always revert (compare?) to the original situation by removing 2~3wires.
If it does nothing positive, leave it out, even more so if it breaks or degrades something.
Since the cost is negligible and the mods minimal, testing it is not a huge investment, neither in time nor money
 
Member
Joined 2011
Paid Member
Do you mean this R7 (30 million ohms)?

_
 

Attachments

  • p003003ooii.png
    p003003ooii.png
    31.8 KB · Views: 1,645
Let us begin with the PSRR, which is the weakest aspect of the NoNoiser: the measured value @100Hz is 125dB, 1 to 2dB better than the best regulator of the survey (Salas reg, fig. 5).
If a ripple compensation resistor is added (R7), it is possible to gain an additional 5dB.
As said above, R7 is a minor tweak pushing the PSRR from 125dB to 130dB, and it needs to be adapted to your build, ie. you need means of discerning between 125 and 130dB, because sim, etc are of no help.

As a wrong value for R7 could as easily degrade the PSRR by several dB or even several tens of dB, it is probably safer to leave it out.

BTW if you need more than 125dB, you should first question the circuit you are providing a supply for
 
The high spread for L1 (50nH ... 500nH) seems rather suspect to me. Why not omitting this inductance and provide stability by paralleling R4 with a 47p ... 100p capacitor instead, like in a standard VAS of an audio amplifier? Same question is regarding C4 in your deNoiser circuit. Why not 47p ... 100p between base an collector of Q1 instead of C4?

In this same analogy to an audio amplifier stage: Did you simulate the effect of increasing beta of Q1 by adding an emitter follower, the way (e.g. same principle) Douglas Self does it in his blameless amplifier with TR4/12: /http://douglas-self.com/ampins/dipa/dpafig33.gif ? I am very aware that this would make your simple and elegant circuit more complex, which might be not wanted.

Last question: What is the exact purpose of R8? Fine-tune V_Out?
 
Last edited:
I mention it again...
add a tracking 317 as a pre regulator, and a LED, a resistor, a cap and you get >100dB headroom...
I have used this circuit on DAC´s.
Using 3-pin regulators off-piste: part 4
Everyone has his pet circuits...
Compared to yours, the nonoisator has half the dropout, half the wasted power, a 15µΩ output impedance, and the 125dB PSRR is meaningful and useful, because its internal noise is commensurately low (down to 90pV), and does not drown the good PSRR.
It also supports Kelvin connections, which is essential at that level of performance.


The high spread for L1 (50nH ... 500nH) seems rather suspect to me.
The spread is in fact an advantage: it does not mean that you have to select a peculiar value in this range to suit each circuit, but exactly the opposite: locating the compensation there is so effective that it tolerates anything between 50n and 500n without adverse effects.
Why not omitting this inductance and provide stability by paralleling R4 with a 47p ... 100p capacitor instead, like in a standard VAS of an audio amplifier?
If you examine the schematic, you see that it would amount to put the capacitor in parallel with R2 (because of C2).
It would be a shunt compensation, doesn't work well and degrades the HF performance.
Same question is regarding C4 in your deNoiser circuit. Why not 47p ... 100p between base an collector of Q1 instead of C4?
Same reason
In this same analogy to an audio amplifier stage: Did you simulate the effect of increasing beta of Q1 by adding an emitter follower, the way (e.g. same principle) Douglas Self does it in his blameless amplifier with TR4/12: /http://douglas-self.com/ampins/dipa/dpafig33.gif ? I am very aware that this would make your simple and elegant circuit more complex, which might be not wanted.
For me, it is a dead link, but anyway an increase in beta is not what is needed: what is needed for an increased performance is a mu increase

Last question: What is the exact purpose of R8? Fine-tune V_Out?
Yes, it is a trimpot
 
Thank you for your explanations, I understand ... I overlooked some interactions!

... For me, it is a dead link, but anyway an increase in beta is not what is needed: what is needed for an increased performance is a mu increase ...

I understand ... but not everything: What is a "mu" increase. I do not understand "mu" and what it's related to.

And sorry for the dead link. My fault. Correct would have been without the "http://"-prefix. Correct: douglas-self.com/ampins/dipa/dpafig33.gif
 
Mu is the theoretical voltage gain, related to the transconductance, but the transconductance being linked to the collector current, it cannot be increased arbitrarily unless the collector load is active, which brings other types of problems and complications
 
Kelvin = 4 wire connection: one pair supplying the power (force terminals), and the other going to the error amplifier (sense terminals).

More info on wikipedia or here: https://www.google.com/url?sa=t&rct...lc=eng&cc=IN&usg=AOvVaw2n4kJrg3UqjkeEr9KY7Gkh

With the denoizator, Kelvin connection are not required, but they can be implemented: just connect the terminals with the small arrows directly across the load.

With the nonoiser, 4-wire configuration is imperative if you want to have a chance to reach the advertised performance.
 
Here is another question, sent by PM.

I will answer it here, because other people will probably find it useful
RickTH said:
Hey LV,

Some things are not clear :

DeNoiser : do the values of the components change for different output voltages ?
No, the capacitors just need to have a sufficient working voltage for the output.
What do .ac dec 100 0.01 100k
:tran 0 100m 0.1u
SINE(0 100u 1K) mean ?
These are spice directives, telling the simulator what it has to do.
You can ignore them completely.

You state 0.66 uV, is this rms or pp ( or even sqrHz) ?
Am I right that 0.66 uVrms = 5nV/sqrHz ? (for 10hz to 20 kHz)
0.66µV rms, in a 10Hz to 10kHz bandwidth.
This would translate to 6.6nV/sqrtHz for a flat noise (which it is not), but in fact the 0.66µV was measured before I used my new LNA.

The actualized value (posted later) is in fact 0.3µV

Does R1 220 ohm and C3 220uF still give a slower output voltage rise at startup ?
Yes

Your NoNoiser :

again :
:dc temp 0 100 0.5
ac dec 100 2m 10e6
: noise V(out) V2 dec 100 1 1e6 even google translate doesn't know ;)
Ignore again

R8 2k2 ... is this a potentiometer ? Or why is there a connection dot in the middle of the resistor ?
If it's a pot , then you have to tune it ?
It is a trimmer.
You could have to adjust it, because the DC output won't be as accurately defined as in a regular, or denoiser 317 circuit.
In the nonoiser, you have to compute the output voltage slightly differently, compared to a regular 317 circuit.
I will provide a formula, if you cannot compute it by yourself.

Is here too a slow output voltage rise at startup , since the 220 uF is not exactly where it is in regular designs.
There will also be a slower rise
It is unclear which Resistors to change for different output voltage .
I'll give a detailed answer later
What are the values for a 15 V output ? Are the specs for noise and PSRR the same for different output voltages an different loads ?
The noise and PSRR vary slightly with the output voltage, and a bit more with the load, but the variations are much smaller than for a regular 317 regulator.
The nonoiser tends to dominate, and evens out the variations
 
^ Thanks LV.
Wow ... so an ordinary LM317 + the DeNoiser gives us 300nVrms ~ 3nV/rHz from 10Hz to 10kHz . That is as low as the ultra low noise LT3042 , and the best of low noise audio opamps! And this without kelvin connections .
Most of TI and LT's low noise regulators (LDO's) are around 40 to 60 uVrms , or 10 to 20 times more noisy . Very impressive . Not to speak about their average PSRR...
I hope someone verifies these very low numbers . If really as low as 300nVrms , TI and LT can close down some of their regulator production lines . Who wants these expensive chips that have 10 times or more noise than a simple LM317 and a transistor ?