Linear voltage regulators and noise impact

[Vadim18]

Yes op amp will try to supress power line noise by common mode rejection, but power line instability also reflects on input bias current.
So in worst case scenario noisy regulator may increase fliker noise and produce additional chattering.
Before i never make attention to such noise parameters becouse i think power line is always filtered by big capacitances and thus noise is not a problem but
as it turned out not everything is so simple and 1/f noise will be present.
I know solution like RC filter or even beter isolation diode + RC filter is exist about half of centuary, it may be used, but it will turn out regulated power line again to unregulated.
One of possible choise is to use capacitance multiplier as post regulator but with voltage drop for base-emitter saturation.
Another choise is to choose an optimal LDO with an acceptable noise level.

When I spent time optimizing the Surfmaster, I did a little research on voltage regulators and found LD1117 superiority over 7805,
it gives only 0.9V voltage drop at 17mA load instead of 1.5V for 7805. Also LD1117 drawing itself less current than 7805.
I place here a list of availible LDO-s with specs in comparsion with 7805 as jellybean reference.

------------Vdrop(at 17mA)---------Idle current---------Noise in range 10Hz - 100kHz (datasheet)
MC7805........1.52V..........................4mA.. .....................50uV
LD1117.........0.99V.........................3mA.. ......................100uV
MIC2951.......0.15V.........................0.1mA. .....................430uV
AD3334.........0.05V........................0.09mA .....................45uV

So you see the picture.Surely noise beyond 10 Hz will rise exponentially until some levels.
3334 will be the best, but what about 2951?
I know that ground noise will always be dominant, but why not make the detector work well, at least on the table.

[bklein]

How does this one do?: https://www.ti.com/product/TPS7A94. Noise .46uV!

[Vadim18]

How does this one do?: https://www.ti.com/product/TPS7A94. Noise .46uV!

This is good one, moreover this is the first one ive seen with 1 Amp capability.
Well, If such low noise LDOs exist on the market with 0.46uV noise rank than such low noise are in demand (in sensitive aplications)
Yes it must be pricey, but good stuff have good price.
Have you use it? If so, any changes in noise performance?
Thanks for pointing for such beast.

[Vadim18]

I think in my case half a kilogram of aditional capacitors will solve the problem.

[bklein]

Hard to get. Did manage to get samples - for a work-related effort though.

[ivconic]

Hard to get. Did manage to get samples - for a work-related effort though.

Precisely!
The main reason for "giving up" on such solutions is because modern components are difficult to find in local stores.
And when they meet; they are super small SMD components, which are difficult to work with in amateur conditions.
It would be good to do a comparative analysis with "old school" noise suppression methods.
The first thing that comes to mind is capacitance multiplier with or without zener regulator.
Without zener, obviously it will be less noisy but unregulated.
With zener it will be regulated but with added zener noise... which btw is not negligible.
But a good selection of transistors with higher current gain and a low esr capacitor of
medium capacity (100-470uF) will largely filter out the added noise from the zener diode.
It would be very good to compare such a method with today's modern regulators.

[ivconic]

Look at this "madness":
https://www.ebay.com/itm/141709057916

[Vadim18]

Interesting future for 1117. I think in additional pcb placed error amplifier with shrinked passband.
Thanks for link.

[Skippy]

Resistor-capacitor filtering of opamp power rails has additional benefits:
It reduces coupling between other components sharing the same supplies. If you have a lot of gain in your system, the output stages can feed back signals to the early stages, via the power supplies. This can cause instability, even oscillation if it's bad, and there are phase shifts, eg. from low-pass filters.
And any digital devices, ADC's, micro's can also inject noise onto the power rails, too.

Electrolytic caps that I have found very good at eating noise are the Sanyo OsCon type, purple packaging.

[Vadim18]

Resistor-capacitor filtering of opamp power rails has additional benefits:
It reduces coupling between other components sharing the same supplies. If you have a lot of gain in your system, the output stages can feed back signals to the early stages, via the power supplies. This can cause instability, even oscillation if it's bad, and there are phase shifts, eg. from low-pass filters.
And any digital devices, ADC's, micro's can also inject noise onto the power rails, too.

Electrolytic caps that I have found very good at eating noise are the Sanyo OsCon type, purple packaging.

Yes using solid state capacitors instead of general type is the first step for denoising. Solid state have more than 2 times better ESR. I already use 2 caps in parallel in place of main capacitor. But again resistor makes power rail floating.

[Vadim18]

And for example i have 50 Ohm resistor in RC filter and my frontend amplifier draws abou 4mA so voltage drop will be 200 mV.
But in saturated condition front end amp can draw more current, say 4mA +0.5mA, if it draw +0.5mA than Vdrop may be 225mV and that 25mV voltage variations is much bigger than 100uV noise from 7805 stab.
A really interesting situation can be in a such state when electromagnetic noise or AC line noise will be recieved by our front end amplifier and as result power rail will be modulated.

[Skippy]

According to a bit of reading on the subject of 'average' linear IC regulators, the error amplifier contained inside is something like a 'UA741' opamp, hence the noise generation. The first step when making a discrete regulator is to use a decent low-noise opamp as the error amp. Then decouple/filter important internal points in the circuit, like the reference generator. Then tricks like powering the opamp and / or the reference from the output of the regulator, to give them a clean rail.
Some articles and stuff to browse:
https://github.com/AuteurAudio/elect...oise-regulator

https://tangentsoft.net/elec/opamp-linreg.html

pdf article:
https://www.diodes.com/assets/App-No...zetex/an51.pdf

[Vadim18]

According to a bit of reading on the subject of 'average' linear IC regulators, the error amplifier contained inside is something like a 'UA741' opamp, hence the noise generation. The first step when making a discrete regulator is to use a decent low-noise opamp as the error amp. Then decouple/filter important internal points in the circuit, like the reference generator. Then tricks like powering the opamp and / or the reference from the output of the regulator, to give them a clean rail.
Some articles and stuff to browse:
https://github.com/AuteurAudio/elect...oise-regulator

https://tangentsoft.net/elec/opamp-linreg.html

pdf article:
https://www.diodes.com/assets/App-No...zetex/an51.pdf

Big thanks for such useful information. In next step i will try make test measurements and compare regulators what i recently have .

[dbanner]

3 transistor voltage regulator +5V from Tektronix scheme.

[Davor]

Zeners below 7V should work well, and the best noise-wise is about 6V. Below that they have even less noise, but their inner resistance is getting high, and temperature dependence is more pronounced.
Band gap references are fine for temperature stability, but their noise is bad. Most of the regulators come with a band gap reference.
Red/yellow/green LEDs are fine, but also drift with temperature.
Diodes are fine, including Vbe ones, but same as LEDs need a few mA to run silent.
Best choice is to build your own regulator and use a 6.2V zener or a string of LEDs as reference.