[Daygloworange]

I'm just curious as to the explanation of how and why power supplies can make the audible changes I've experienced in audio gear.

I have, at best, a mediocre understanding of electronics in the pure technical sense. Ohms, voltage vs current, impedances, and so forth.

Over the last little bit, I have had 4 different power supplies powering my modded SB 3. The changes to the sound are significant by simply using a different power supply. Improvements in micro and macro dynamics, resolution, blackness of background, and overall refinement are marked.

Years ago, as a musician, I experienced the difference between effects boxes being powered by batteries vs AC adaptors in my guitar rig. Batteries ruled in terms of sonics, but the practicality of AC adaptors won out in live situations. In the studio however, batteries won out, when sound mattered most.

Can anyone elaborate on how and why these effects work?

Is there a cutoff point to the refinement of power supplies in commercial audio to achieve a competitive price point holding back the current audio gear in terms of it's true potential?

How important do you feel the power supply is in circuit topology?

I'd like to hear all thoughts and opinions, and personal experiences.

Cheers

[DanTheMan]

Years ago, as a musician, I experienced the difference between effects boxes being powered by batteries vs AC adaptors in my guitar rig. Batteries ruled in terms of sonics, but the practicality of AC adaptors won out in live situations. In the studio however, batteries won out, when sound mattered most.

Very true.  Interesting subject.  Has anyone used quartz-locked power supplies?  I wonder why these things are not in heavier rotation.

[Freo-1]

Although this a bit of a oversimplification, an amp is nothing more than a modulating power supply. It only stands to reason then, that as goes the power supply, so goes the sound of the amplifier.  The more reserve the power supply has, the better job it will perform modulating the input signal without distortion, or running out of steam.

[JoshK]

Although this a bit of a oversimplification, an amp is nothing more than a modulating power supply. It only stands to reason then, that as goes the power supply, so goes the sound of the amplifier. 

Exactly.  The second sentence I don't quite agree with, but understand what you were trying to say.  Unfortunately its far more complicated than that and more reserve doesn't equate to better sound with a lot of gear.

Anyway, in an amp essentially the power supply is the power driving your speakers.  The transistor, tube or whatever else is controlling how much of the power from the power supply is being delivered to the driver.  The better the power in the power supply the better the amp sounds. 

[Audiovista]

There was an excellent article in audioXpress years ago - I think it showed real-life measurements of THD, isolating effects of power supplies. I'll look for it and post more details once I find it.

[Paul Hynes]

Hi Daygloworange,

How are you keeping?

I am writing this on the hop as it’s a public holiday over here and I have family issues to deal with very soon. So please forgive any lack of coherency on my part, with any of the following. If anything is less than clear, ask, and I will try again when I am less busy.

The majority of signal processing circuits don’t have particularly good power supply rejection (PSRR), especially as you move up the frequency bandwidth. Any noise or interference on the supply line will therefore find it’s way into your signal. The level of power supply interference injected into the signal is directly proportional to the signal circuit’s PSRR, which often varies with frequency. The supply interference will modulate the signal causing distortion of the waveform and masking of the low level information (the bit gives life to the music)

The better the power supply performance the less signal corruption you get. The ideal power supply would have zero impedance from DC to light frequencies. This is because you can’t generate any voltage fluctuation across zero impedance (Ohm’s Law) so no matter what load you put on the output it will not cause a voltage fluctuation. It should also have either no connection to the mains supply, which is notoriously full of interference, or have infinite supply line rejection to keep the mains interference out of the power supply output feed. Zero self-generated noise would also be a good attribute for the ideal power supply.

In practice many designers assume that their power supplies are perfect (possibly because a text book at college told them so) and just use industry standard power supply chips, commonly known as three terminal regulators for their power supplies. Some designers also use simple switching power supplies, which are generally worse than linear regulators. Both of these types of power supply were originally designed for low-resolution industrial applications, and they suck for audio power supplies, whether analogue or digital. There are literally hundreds of different regulator chips on the market and their performance varies from type to type and manufacturer to manufacturer. Designers may differ in their applications of these chips and you are hearing the different effects of the overall application when you change power supplies.

So what to look for in a power supply? Batteries are reasonably quiet and don’t have any mains interference issues. Unfortunately, unless you use a rather large car battery, like the Optima Red Top battery, the power supply impedance will be fairly high and possibly frequency dependant depending on the battery construction. This means that any load current fluctuations across the battery impedance will generate a voltage on the battery terminals. This will modulate the audio signal as explained above. That said the overall performance of batteries is usually better than the performance that simple industrial regulators can muster.

It is certainly possible to design mains driven power supplies that exceed even the performance of the Optima Red Top battery.

The key issues to address are :- 
The regulator output impedance should be as low as possible over as wide a bandwidth as possible to reduce and output voltage fluctuations caused by load current changes.
The power supply line rejection should be as high as possible over as wide a frequency as possible to keep mains hash at bay.
Self-generated regulator circuit noise should also be low enough to be inaudible on the sound system.

I hope this helps a little.

My time has run out now. Got to go do the family stuff.

Regards
Paul

[Steve Eddy]

So what to look for in a power supply? Batteries are reasonably quiet and don’t have any mains interference issues. Unfortunately, unless you use a rather large car battery, like the Optima Red Top battery, the power supply impedance will be fairly high and possibly frequency dependant depending on the battery construction. This means that any load current fluctuations across the battery impedance will generate a voltage on the battery terminals. This will modulate the audio signal as explained above.

The battery's internal impedance can be addressed by adding capacitance across the battery.

se

[Paul Hynes]

Hi Steve,

The capacitor impedance characteristics will be in parallel with the battery impedance characteristics and the resultant impedance will be lower. However, the sonic signature of the capacitor has to be considered in the equation. All capacitors I have looked at have non-linear impedance characteristics and most have bad dielectric absorbsion problems. This makes the whole issue of matching and countering battery impedance deviations rather tricky. If you know of a capacitor with a flat impedance curve in the milliohm range and with low delectric absorbsion factor, I would be pleased to know about it. It would certainly make my life easier.

Regards
Paul

[ebag4]

Paul:

What type of regulator or DC to DC converter would you recommend to mantain the qualities of the Red Top Battery if 5VDC 1.5 amps is needed?

Thanks,
Ed

[DaveC113]

I've found using caps on my Optima Yellow Top battery makes a positive difference powering a 6 wpc Trends T-amp. I'm using 4 2200 uF Nichicon electrolytics and 3 180 uF tantalum caps, my understanding being that the tantalum caps are faster. Its also what I could find at the surplus store... I'm guessing this helps because the caps can release energy faster than the battery. How would you also match the impedence of the caps to the battery and what effects would it have vs. the random assortment of caps I've used?

Dave   

[Daygloworange]

Hi Daygloworange,

How are you keeping?

Hi Paul,

Keepin' on, keepin' on, y'know?

Just absorbing the knowledge. Thanks for the insight.

Keep it comin' 

Cheers

[Steve Eddy]

The capacitor impedance characteristics will be in parallel with the battery impedance characteristics and the resultant impedance will be lower. However, the sonic signature of the capacitor has to be considered in the equation. All capacitors I have looked at have non-linear impedance characteristics and most have bad dielectric absorbsion problems. This makes the whole issue of matching and countering battery impedance deviations rather tricky. If you know of a capacitor with a flat impedance curve in the milliohm range and with low delectric absorbsion factor, I would be pleased to know about it. It would certainly make my life easier.

Not sure what you mean by "non-linear impedance characteristics." The impedance curve not being flat doesn't mean it's non-linear.

If you look at a typical impedance plot for a large value electrolytic capacitor:



The downward slope on the left is basic capacitive reactance (X_c). It dominates until its value starts to approach the  resistance due to ESR, then it starts to flatten out as ESR becomes dominant and establishes the lowest point of the impedance plot. But there's also inductive reactance (X_l) due to ESL. And where the curve starts sloping upward is where X_l starts overtaking ESR and it becomes dominant.

If you modeled this in SPICE using ideal resistance, capacitance and inductance, you'd get the same type of curve, and the circuit would be perfectly linear.

Also, even if you had a perfect capacitor with no ESR or ESL, you still wouldn't have a flat impedance curve. The impedance curve would look like the left side of the curve above except it would keep going downward as frequency increased. To have a flat impedance curve, you'd essentially need a resistor.

ESL, which is largely determined by lead spacing, can be kept to a minimum by avoiding axial leaded capacitors and using radial capacitors with the smallest lead spacing. Paralleling capacitors will also reduce both ESR and ESL.

I also don't quite know what you mean by "bad dielectric absorption problems." Exactly how does dielectric absorption manifest itself as a problem?

se

[Paul Hynes]

Hi Ed,

Optima specify their Red Top impedance as 0.003 ohms. This is an admirable specification, as most 3 terminal regulators and switching power supplies can’t come close to this, particularly over a reasonably wide bandwidth. However the red top battery can’t maintain this impedance at high frequencies, as the impedance will start to rise at a point dictated by the electrical characteristics of it’s physical construction. It’s pretty good over the audio band, but at typical digital operating frequencies it’s not so good. It is certainly possible to achieve lower impedance over a wider bandwidth with careful regulator design.

It’s possible to achieve low impedance over a wide bandwidth with series regulators. They do need application of additional techniques, for instance bootstrapping, like the Jung type regulator, or pre-regulation to achieve good enough supply rejection to reduce line noise interference to the level of battery noise.

A good wide band shunt regulator with current source drive can achieve very low impedance over a wide bandwidth and very high supply rejection with relatively simple structure.

The simple answer is check out the impedance and PSRR over your required operating bandwidth, the transient response and settling time and the noise specification. The regulator that offers the best specification with these parameters is going to give you the best audio performance.

Regards
Paul

[Paul Hynes]

Hi dave,

Lowering the impedance of the power supply by adding mix and match capacitors means less interaction with the load. However you could end up with an impedance curve like the contour of a hind leg of a donkey. You might like the sonic effects of the resultant impedance curve, or you might not.

Regards
Paul

[Paul Hynes]

Hi Steve,

As far as I am concerned “non linear impedance characteristics” of a power supply means it does not have a flat impedance curve over the required operating bandwidth. I’m well aware of the impedance curve variations of capacitors and my statement “If you know of a capacitor with a flat impedance curve in the milliohm range and with low dielectric absorbsion factor, I would be pleased to know about it. It would certainly make my life easier.” was a joke. If it did exist I wouldn’t need to design regulated power supplies.

If you add the capacitor shown in your example to any power supply it would have a non-linear effect on the resultant impedance curve. This would mean that the power supply would have different levels of interaction with the load at different frequencies. With a high resolution system this is audible and it will colour the tonality of your sound system. This may be beneficial in some circumstances, as you may like the resultant colourations.

Dielectric absorbsion in electrolytic capacitors is quite high. For those who are not aware of this phenomenon, ac signal current fluctuations across the capacitor electrodes induce a variable charge storage in the insulation (dielectric), which is released after the signal current has passed. This changes the envelope of the waveform corrupting the signal. Walt Jung, Dick Marsh and others have written a number of articles in this field. This is very useful information for choosing capacitors for a variety of electronic functions. It’s also applicable to cable insulation and printed circuit material selection.

Personally, in my domestic sound system, I prefer to reduce power supply interaction to the lowest possible levels. This reduces signal/power supply inter-modulation and masking of low level information allowing more of the music through. For musical instrument amplification it’s a different matter. If you want an accurate acoustic sound go for high performance power supplies, if you want to rock, power supply inter-modulation can add a little extra tonal colouration for the musician looking for a different sound, and here the world is your oyster.

Regards
Paul

[ebag4]

Thanks for the response Paul.

The simple answer is check out the impedance and PSRR over your required operating bandwidth, the transient response and settling time and the noise specification. The regulator that offers the best specification with these parameters is going to give you the best audio performance.

If this is the simple answer then I am in trouble

Thanks again Paul.

Best,
Ed

[Steve Eddy]

As far as I am concerned “non linear impedance characteristics” of a power supply means it does not have a flat impedance curve over the required operating bandwidth.

As you wish. But I think it's erroneous and misleading to refer to the results of a linear function as non linear.

I’m well aware of the impedance curve variations of capacitors and my statement “If you know of a capacitor with a flat impedance curve in the milliohm range and with low dielectric absorbsion factor, I would be pleased to know about it. It would certainly make my life easier.” was a joke. If it did exist I wouldn’t need to design regulated power supplies.

Fair 'nuff.

If you add the capacitor shown in your example to any power supply it would have a non-linear effect on the resultant impedance curve.

I still have problems with the use of the term "non linear" but I'll leave it at that for now.

This would mean that the power supply would have different levels of interaction with the load at different frequencies.

Technically, yes. Though one isn't limited to using just one capacitor. Smaller capacitors can be added in parallel to ameliorate the rising impedance due to inductance.

And even if you have the perfect regulator, the output of that regulator still has to be fed to the circuit it's powering. This creates a loop between the output of the regulator, the device it's feeding, and the path back to ground. That loop adds inductance and depending on circumstances can easily be significantly greater than the ESL of the capacitor given in the example.

This is why even with the best regulated power supplies, bypass caps are used as close as physically possible to the point where the power is actually being delivered in order to counteract the effects of lead inductance.

Dielectric absorbsion in electrolytic capacitors is quite high. For those who are not aware of this phenomenon, ac signal current fluctuations across the capacitor electrodes induce a variable charge storage in the insulation (dielectric), which is released after the signal current has passed. This changes the envelope of the waveform corrupting the signal.

No, it doesn't. This confusion comes about due to the typical means in which dielectric absorption is measured which leads people to believe that there's some "ghost" signal which follows along behind the real signal.

It's measured by charging a capacitor and holding that charge for a long period of time. The capacitor is then shorted for a relatively short period of time compared to charge time, then open circuited for a while and then the residual voltage is read.

Where does this situation exist in an audio system? It doesn't. That's why dielectric absorption is primarily a concern with things such as sample and hold circuits where a capacitor is used to hold the voltage that's to be sampled.

Dielectric absorption is a linear phenomenon and is quite accurately modeled using a number of very low value capacitances in series with very high value resistances in parallel with the main capacitance. Now how would this somehow create any "ghost" signals?

Personally, in my domestic sound system, I prefer to reduce power supply interaction to the lowest possible levels.

That's fine with me. You should do whatever it is you prefer. I just think sometimes we tend to get a little too obsessed with numbers.

This reduces signal/power supply inter-modulation and masking of low level information allowing more of the music through.

Ok. But if such a microscopically small amount of power supply modulation is able to mask low level information, how does the low level information manage to not be masked by the signal's high level information?

se

[DanTheMan]

Steve, is dielectric absorbtion a factor a factor when you design something?  If so, could you explain a little bit of the why for my peabrain absorb?

Thanks,  Dan

[Paul Hynes]

Hi Steve,

Perhaps I didn’t explain my thoughts as clearly as I could have done. It was late when I replied to your post (around 1.30am) after an extended weekend of partying at a family gathering. I’m still suffering the aftermath due to lack of sleep, but I will try to be a little more lucid but as briefly as possible as I have a backlog of work to catch up with.

Maybe we differ over our definitions of linear but I see it as meaning a straight line. If the impedance plot varies from a straight line I feel justified in describing it as non-linear.

I have tried the multiple bypass approach with a wide range of capacitors and, whilst it can make very audible differences, I have never been happy with the tonal aberrations this technique can cause.

I would be the first to admit that there is no such thing as a perfect regulator, but that is no reason not to aim for perfection.

I agree that having a regulator remote from the load is not ideal. I prefer to locate my regulators right next to the load, which I do in my own amplifier designs. Using a high quality shunt regulator keeps the load circulating currents local, to minimise inductive loop effects around the power supply system.

Sometimes it is not possible to apply local regulation easily, for instance, when modifying equipment. Under these circumstances I like to construct my power feeds in a manner that minimises inductance around the load current loop. Usually, local decoupling has been installed, by the equipment manufacturer, to ensure stability, I leave well alone here, but still gain very useful sonic improvements by utilising better regulation.

I have to admit that my knowledge of Dielectric Absorption (DA also known as capacitor soakage) is not up to degree level, however, I have read much about it’s effects on audio signals by highly regarded authors in the audio industry. My own tests, with a variety of capacitor types, back up the information presented by the authors. Rather than get into a long discourse on this subject, where perhaps my qualifications are lacking, I would suggest that, those who are interested, check out the following example [http://www.national.com/rap/Application/0,1570,28,00.html], titled Understand Capacitor Soakage to Optimize Analog Systems, for a general appraisal of Dielectric absorption and it’s effect on analogue circuitry. This is just one of many articles that all draw similar conclusions. On the weight of the evidence of these articles and of my own capacitor trials I’m afraid I have to disagree with your comments on DA.

I do agree that we can sometimes get a little obsessed with numbers, however every time I move regulator performance closer to the ideal, I hear profound improvements in the sonic performance of my music system. With results like this I will gladly get a little obsessed.

An electrical signal representing music is an aggregate of waveforms of different frequency and amplitude giving a resultant envelope that is the sum of all the waveform potentials at any given point. In other words the waveforms all ride on and form part of the signal envelope. Add some power supply noise and interference into the mix and it becomes part of the signal envelope making changes to the envelope vector. It doesn’t take a great deal of vector change to significantly change the shape of the low level information. Once again it’s all a question of degree. How much change can be tolerated before the results become audible. In my experience it’s not a lot and I will do whatever it takes to minimise the problem.

Regards
Paul

[Steve Eddy]

Perhaps I didn’t explain my thoughts as clearly as I could have done. It was late when I replied to your post (around 1.30am) after an extended weekend of partying at a family gathering. I’m still suffering the aftermath due to lack of sleep, but I will try to be a little more lucid but as briefly as possible as I have a backlog of work to catch up with.

Yeah, yeah. Excuses, excuses.

Maybe we differ over our definitions of linear but I see it as meaning a straight line. If the impedance plot varies from a straight line I feel justified in describing it as non-linear.

Well, I'm coming at it from the fact that if a system is composed of linear elements, that system is linear, no matter how many elements you have or how you combine those elements.

But since we both seem to know what each other is meaning to say, no use arguing about the terminology any further.

I have tried the multiple bypass approach with a wide range of capacitors and, whilst it can make very audible differences, I have never been happy with the tonal aberrations this technique can cause.

How exactly did you establish that it makes an actual audible difference?

On second thought, forget I asked that. I think I'm just going to bow out of this discussion and the cable discussion.

se