What is power supply output impedance?

[Kingston]

Hello,

this is a topic which has eluded my understanding. What is meant by impedance when talking about PSU output?

Occasionally one hears things mentioned, like "too large filtering capacitors at power supply output increase impedance".

Does this have something to do with how fast (or clean?) the power supply reacts to changes in current draw?

Or is this a PSU regulator related terminology?

Thank you,
Mike

 

[Twenty Log]

The Effective Series Resistance (the built in value of "parasitic" resistance) of the capacitors are also an issue and can make voltage regulator chips oscillate...

In the LM2991, for example, on page 4 (http://www.national.com/profile/snip.cgi/openDS=LM2991) as the output impedance changes, ripple rejection changes in the opposite direction...  It would seem that it is a tradeoff between transient response and stability amongst other things coupled through the feedback methodologies employed...  I am not a voltage regulator doctor (but I play one on T.V.)....

I suppose it could be also related to vacuum tube power supplies... If the supply is a pi filter style (2 caps and a choke) to filter out 60Hz or its harmonics, then there is a point where the filter is not effective when graphed over frequency and thusly has a larger impedance at the higher frequencies... The wrong value (or too large) of capacitors will skew the frequency response and thusly the impdeance at higher frequencies...

The ripple rejection up at higher freqs are way important for designs with A/D and D/A since any noise on the PSU lines near the sample rate gets modulated into the signal...  Mostly with switching PSUs or stray and errant RF buggery getting in the mix....

 

[Kingston]

If the supply is a pi filter style (2 caps and a choke) to filter out 60Hz or its harmonics, then there is a point where the filter is not effective when graphed over frequency and thusly has a larger impedance at the higher frequencies... The wrong value (or too large) of capacitors will skew the frequency response and thusly the impdeance at higher frequencies...

Makes sense.

I wonder about PSU design in general. Often one sees very long crcrc etc. networks especially in high power hifi amps. Then again, every once in a while this brute force method is discouraged. But why is this, and is it related to the output impedance somehow?

"fast load transient response"... Is that a feature of very low, or very high output impedance?

 

[Twenty Log]

The CRCRC is a filter of course and may be used in this example to also isolate different stages of transistors and keep the AC impedance low (although the R's in the circuit will increase the DC impedance; probably why it is not ideal).  I think the "fast transient response" also has to do with the feedback loop inside of the voltage regulators and any poles in the feedback loop to keep the PSU stable (to compensate for external ringing on long PCB traces etc.)... 

Of course, page 4 of the LM2991 also shows the transient response albeit for a single frequency, but this is probably sufficient for most usual applications...

I know that the larger the transistor (pass transistor) device, typically the larger the stray (parasitic) capacitance that affects the overall circuit, but can be "compesated" for within the feedback loop to keep things stable without oscillation, especially whence the circuit is connected to the real world loads which are inductive (long PCB traces) and capacitive (decoupling) and resistive in a sordid network of complex impedance at the load pin of the voltage regulator ...

I would submit that transient response is an abrupt change in load requirements, but that can perturb the complex impedance network, since there is a frequency component of this transient demand on the system...  The large PSU caps can make up some of the current demand, but they need to be recharged quickly, which may not be conducive to an "overcompensated" feedback loop....

 

[JohnRoberts]

Hello,

this is a topic which has eluded my understanding. What is meant by impedance when talking about PSU output?

Occasionally one hears things mentioned, like "too large filtering capacitors at power supply output increase impedance".

Does this have something to do with how fast (or clean?) the power supply reacts to changes in current draw?

Or is this a PSU regulator related terminology?

Thank you,
Mike

At the output of a typical VR IC the source impedance will be the parallel sum, of the active electronics impedance and any output caps. The typical (inexpensive) 3 terminal VR use 741 grade opamps inside so will have a rising source impedance at higher freqeuncy due to falling loop gain margin.

I suspect the advice you mention is talking about very large electrolytic capacitors having higher series inductance which will degrade their effective mpedance at very high frequency.

I did some research back in the '80s and found the falling impedance of a 1000uF electrolytic was a nice match to the rising impedance of inexpensive 3 terminal regulators to deliver a low flat source impedance at least another octave above the audio bandpass.

Of course it is good practice to also use smaller disc caps (.01-.1uF) at chip leads to provide low Z at uber-uber frequency.

If you want to be picky about PS you can roll your own from low noise opamps, but this is generally overkill for reasonable circuitry.

JR

PS: Note some here do not agree with my approach but that's life... I used it for many years without incident.
 

 

[Gordy]

Power supply output impedance is power supply output impedance. There are no tricks in the words.

If you were an active circuit drawing current and "looked back" into the power supply you would "see" the power supply output impedance.

In a regulated supply the output impedance is the sum of the regulator output impedance in parallel with the impedance of the output capacitors. It normally varies with frequency.

Current drawn by your active circuit develops a voltage across the power supply impedance. You have to determine how important that is and how low you want (/need) the power supply output impedance to be.

 

[ColinS]

What you need is these

http://en.wikipedia.org/wiki/Th%C3%A9venin%27s_theorem

http://en.wikipedia.org/wiki/Norton%27s_theorem

These Thevenin and Norton equivalents were amongst the first things I was taught as part of my Electronic Engineering degree. They are essential tools for designing and analysing electronic circuits. Understanding these theorems will help you answer your own question.

They basically reduce all the components down to a source (either voltage or current, depending on what theorem you work from) and either a series or a parallel impedance. Under DC conditions these impedances will be a plain old resistance, but for AC systems there will also be some reactive element.

Throw in some of this

http://en.wikipedia.org/wiki/Nodal_analysis

and this

http://en.wikipedia.org/wiki/Ohm%27s_Law

and you'll be sorted.

It is possible to derive a Thevenin or Norton equivalent circuit for a microphone or indeed any source of audio. Probably useful for designing pre-amps or line-drivers....

Hope this helps and isn't too maths heavy.

 

[CJ]

do not forget the transformer on the utility pole,

i guess you could analyze all the way back to a Toshiba turbine at a dam near you?

 

[shabtek]

thanks CJ.
it is not a 'telephone pole'
hope not to climb any tonight-- one more night of graveyard shift.
see yous in the mornin'

 

[Twenty Log]

The good news is that the utility pole is really a low frequency pole in the transformer's..... aw never mind.... 

You can also look up a Jung regulator, from Walt Jung, one of the gurus at Analog Devices... The Jung regulator is a lot of parts and there are various incarnations throughout the DIY world... It is a low noise audiophile grade throughout frequency range (impedance looking into the output a la Mr. Thevenin).  This helps in isolating the transformer on the pole from your audio circuits

 

[Kingston]

You can also look up a Jung regulator, from Walt Jung, one of the gurus at Analog Devices... The Jung regulator is a lot of parts and there are various incarnations throughout the DIY world... It is a low noise audiophile grade throughout frequency range (impedance looking into the output a la Mr. Thevenin).  This helps in isolating the transformer on the pole from your audio circuits

Hey thanks! This kind of regulator is the hidden undercurrent of my above questions. I want to learn what the most important parameters are for audio specific PSU regulators. For LM317 and other bulk friends the design seems "safety first", leaving noise issues, even if marginal. I've been wanting to move past that for academic reasons.

[edit]

I came across with this: http://tangentsoft.net/elec/opamp-linreg.html

What a well written and thoroughly informative article. It condenses the various sources, variations and publications on both Sultzer and Jung regulators with even history of both wrapped in. It also walks through the clever bits and short comings of each design in a relative short article. That author should be given an award! That's the best article I've read in months.

 

[JohnRoberts]

The simple answer is PS impedance is the change in output voltage for change in supply current.

There is no simple answer for what is required for good audio performance. Since the current drawn from each supply is not simple linear audio, but roughly half wave rectified, the voltage on these PS rails will be distortion if leaked into the audio path. The sensible way to mitigate this is a combination of both, designing the audio path to ignore PS noise "and" make the PS rails low impedance and clean.

The example below is a PS I designed for a phono preamp back in the early '80s. Since this preamp was the third one I published, by that time I was way past "good enough" and well into "over design". 

By posting this I am not suggesting this much effort is justified. I also designed good PSR into the rest of that design.

One could probably throw even more cost and complexity than I did into a PS but this is generally not a thoughtful "whole picture" analysis of a design, but more of a how good can I make this one little part of a system.  An interesting mental exercise perhaps.

Note: My PS design is lacking thermal protection, and fancy current limiting common in even cheap 3 terminal devices. There will be a crude upper limit to output current determined by resistors in series with the pass transistors, and available base drive. The resistor values were dialed in for a specific transformer. You could improve the input operating range by replacing R42 and R33 with zener diodes. The voltage of the zener isn't critical. Just a few volts drop so the opamp doesn't have to swing above the rail. By powering the opamp inside this requlator from it's own regulated output the PSRR of the regulator is much improved, compared to powering it from the unregulated voltage.  I used small to-92 pass transistors because my total current needs were < 100 mA. Opamp was a TL072 but there may be more modern choices to use today.

JR

 

[Speedskater]

I recognize that power supply:

"The Audio Amateur" magazine #3 -1985

Preeminent Preamp
Phoenix Systems

 

[JohnRoberts]

yup.. I was a habitual phono preamp designer... I even had a few minor tweaks I could have put in a later version but I lost interest in designing for the audiophool market, that responded to things other than the accuracy or signal integrity of the product.

WRT PS, IMO you can get 99.99% there with a ua7815 and 1000uF cap in parallel, but this gives you something to talk about.

JR

 

[Kingston]

Thanks for the information John Roberts. I like the way that schematic is drawn. Very easy to understand ground paths from it.

Looking at your design a simplification popped into mind. Wouldn't it be possible to use a single opamp here to "share" the error amp for both rails? It would no longer see ground, but just the two rails. Same difference signal is fed to both pass transistors. I suppose it would only work if load is always perfectly equal for both rails (which is never).

This is just an academic question since dual opamp vs. single opamp cost is probably never an issue either.

 

[JohnRoberts]

Thanks for the information John Roberts. I like the way that schematic is drawn. Very easy to understand ground paths from it.

Looking at your design a simplification popped into mind. Wouldn't it be possible to use a single opamp here to "share" the error amp for both rails? It would no longer see ground, but just the two rails. Same difference signal is fed to both pass transistors. I suppose it would only work if load is always perfectly equal for both rails (which is never).

This is just an academic question since dual opamp vs. single opamp cost is probably never an issue either.

You probably already know the answer to that.. Since the current drawn from each supply will be significantly different, using a single feedback loop means the opamp can only manage to some average result. At best the correction will only be 1/2 of each open loop path error, with 1/2 the correction error cross coupled into each opposite output.
-------
I actually did a PS trick once where I had to deal with managing assymetrical PS draw from a non-centertapped transformer (external single secondary winding, wall wart). I needed a 5V supply for digital circuitry and +/- whatever I could get for the analog signal handling circuitry. Left to it's own behavior the extra digital supply current would have pulled down the + analog supply making the - analog supply more voltage than the +, reducing total headroom through the path. I added a simple shunt regulator to draw current from the - supply to center the two analog supplies around ground. Not very green but it worked to maximize headroom.

JR
 

 

[riggler]

What you need is these

http://en.wikipedia.org/wiki/Th%C3%A9venin%27s_theorem

http://en.wikipedia.org/wiki/Norton%27s_theorem

These Thevenin and Norton equivalents were amongst the first things I was taught as part of my Electronic Engineering degree. They are essential tools for designing and analysing electronic circuits. Understanding these theorems will help you answer your own question.

They basically reduce all the components down to a source (either voltage or current, depending on what theorem you work from) and either a series or a parallel impedance. Under DC conditions these impedances will be a plain old resistance, but for AC systems there will also be some reactive element.

Throw in some of this

http://en.wikipedia.org/wiki/Nodal_analysis

and this

http://en.wikipedia.org/wiki/Ohm%27s_Law

and you'll be sorted.

It is possible to derive a Thevenin or Norton equivalent circuit for a microphone or indeed any source of audio. Probably useful for designing pre-amps or line-drivers....

Hope this helps and isn't too maths heavy.

COLINS!!! This is the holy grail my friend, thank you!! 8)

 

[Kingston]

At best the correction will only be 1/2 of each open loop path error, with 1/2 the correction error cross coupled into each opposite output.

Doh! Going from error amp to error mixer. Not exactly desirable.

 

[ColinS]

COLINS!!! This is the holy grail my friend, thank you!! 8)

Glad they're of use.

Would it be worth starting a meta thread of electrical engineering tools like the stuff on the end of those links? Stuff like Thevenin and Norton, Ohm's Law etc. There's bound to be loads more I don't know or have forgotten about. I graduated about 6 years ago and have scarcely used any of that theory since.....

C

 

[Kingston]

Would it be worth starting a meta thread of electrical engineering tools like the stuff on the end of those links? Stuff like Thevenin and Norton, Ohm's Law etc.

The first three META links are filled with exactly these.

What the raw theory links (like the one's you posted) often miss is application. For a beginner, or even intermediate aspiring electronics engineer, the theory means very little with no real world examples. Just floats in the air connected to nothing. Hence threads like this one.