Lowest noise power supply

[mawyatt]

I know of a much more sophisticated form of frequency compensation for difficult loads involving a feedback network with less than 90 degrees of phase lag, which increases the frequency margin, but it has not been required here.

Are you referring to fractional pole compensation? https://www.radioeng.cz/fulltexts/2016/16_02_0312_0320.pdf

I think its a really cool technique that is rarely discussed, despite the fact that I suspect most modern LDO regulators that are stable for wide ranges of capacitor ESR use it.

Very interesting technique, thanks for posting.

Around 2000 in the early stages of developing RF, Microwave, MMW Systems on Chip (SoC) we required multiple on-chip LDOs for various functions. We utilized traditional PMOS output LDO designs which suffered the usual noise, and capacitor stability issues. One particular on-chip need was an ultra low noise LDO for biasing Voltage Controlled Oscillators (VCO) for use with Frequency Synthesis.

We developed a new type LDO that utilized a Super Source Follower type architecture which allowed unique noise decoupling to help bypass noise generated from the Bandgap Reference and other regulator elements, see patent 8692529.

https://patents.justia.com/patent/8692529

Recall we were able to achieve a measured noise density 10~12nv/Hz @1KHz at the regulator output which included Bandgap Ref and all other internal regulator noise sources. At the time recall this was one of the lowest noise LDOs available, but surly there are better LDOs today. Back then a major microwave company became interested in a custom development LDO for their low Phase Noise VCOs as they were having to utilize custom discrete component LDOs at the time to meet the Phase Noise specification. Intel duplicated the concept (long story) for use within their processors for the on-chip DLL timing generators voltage regulators.

Anyway, the LDO Fractional Pole Compensation technique is indeed intriguing

Best

[David Hess]

I know of a much more sophisticated form of frequency compensation for difficult loads involving a feedback network with less than 90 degrees of phase lag, which increases the frequency margin, but it has not been required here.

Are you referring to fractional pole compensation? https://www.radioeng.cz/fulltexts/2016/16_02_0312_0320.pdf

That looks like the same idea although I do not know it under that name.  Basically the capacitor in the commonly used integrator with 6dB/octave roll-off and 90 degrees of phase lag is replaced with an RC network resulting in an integrator with 3dB/octave roll-off and 45 degrees of phase lag, which adds 45 degrees of phase margin making the regulator much more stable.

I think its a really cool technique that is rarely discussed, despite the fact that I suspect most modern LDO regulators that are stable for wide ranges of capacitor ESR use it.

There is a much simpler solution for modern low dropout regulators.  The collector/drain of the output transistor is tapped to produce an AC feedback signal.  This produces what is effectively a resistance in series with the low ESR output capacitor at high frequencies, which adds phase lead to make the regulator stable.  It is equivalent to adding a small series resistance to the low ESR capacitor to make a zero, but the series resistance is in a different spot.

Some of the early Linear Technology regulator datasheets discuss what is going on and how it is done.

[iMo]

I wonder whether a 723 + LT1021 combo would be "low noise" enough..
For smaller currents (up to 50mA let say) it may work without the transistor, imho..
PS: added 723+LM329 combo

[Gerhard_dk4xp]

Oh, 10 years old already!

I measured some stuff I had around:
<   http://www.hoffmann-hochfrequenz.de/downloads/Noise_Measurements_On_Some_Laboratory_Power_Supplies.pdf   >

<  http://www.hoffmann-hochfrequenz.de/downloads/NoiseMeasurementsOnChemicalBatteries.pdf   >
Very low frequencies may be shown worse than they are because of the preamplifier.

When it counts, I use Lithium batteries.

Gerhard

[GigaJoe]

I use 2DW232 ( metrology section ) as a reference, additionally trimmed  for a minimum tempco
another option IR LED , as lowest noise source in LED,  but 2DW232 at min. tempco  really great.  basic circuit opamp zener transistor  ;  zener fed from opamp , V output doubled .. 10.4V approx.

use capacitor multiplier to feed Ref. circuit and error opamp,   2-n order filter with 680 ohm resistors ,  with 5-10ma load - it does some filtering.

all ref / mgm  circuit in enclosed shielded box , just pins for external connection, shielded variable resistors and shielded wire to connect sensitive points.

and it fail .... ringing when inductance  load,   (  maybe not   ) 3 wirewound resistors ... oh well , seems it will be a new iteration ...