Audio Power Amplifier Design book- Douglas Self wants your opinions

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
well by objective measures like word count, dedicated forums, ads, shelf space... the "pure subjectivists" are doing a bit more slinging - and by boycotting the Scientific Method, refusing Blind Tests, Controls, claiming literally incredible memory abilities they aren't likely to progress audio reproduction any more than they haven't in the last 1/2 century

If Doug, Rod, Peter have leaned too far, occasionally miss some things that do have decent Psychoacoustic evidence – the balance is still in their favor over that of the vastly more publicized “Audiophile” press' “Golden Eared” reviewers who praise just about any well packaged, appropriately bejeweled, properly audio guru “storied” objects of conspicuous consumption
 
JCX,
Being of the scientific persuasion the only problem I have leaving out the subjective is that we don't seem to be able to correlate really low distortion numbers with improved sound quality. Otherwise I would have to agree that straying to far from the truly objective analysis would be a fools errand.

I leave the Bejeweling to Apple and the apple fans who will buy anything that looks pretty whether a great product or just a better looking package.
 
How many SMPS specify the transient current capability?
If a 150W +-45Vdc smps can output 1.7A per polarity, is that a continuous 1.7A or 50% duty cycle,
Can that same 150W smps supply 200% (3.4Apk) for 100ms?
Can it supply 300% (5Apk) for 50ms?
Can it supply 900% (15Apk) for 1, or 5, or 10ms?
All of these current pulses are available from a +-20mF smoothing (the size I recommend for good bass into an 8ohms speaker) cap bank to feed a reactive 8ohms speaker.

The smoothing caps and the inductors (that I have seen) on the output of the SMPS certainly can't supply 15Apk.
Many SMPS will refuse to start up with a decent cap bank attached.

That is why I ask that question. Where does the current come from, when there is no smoothing cap bank as found on conventional PSUs?
It matters not how often the capacitors are recharged.

If one were to specify a +-45Vdc, +-15Apk SMPS then the question is answered for us.
That is, use a 1350W SMPS for a 100W into 8ohms amplifier and one ensures that the SMPS can supply the 15A pk currents that can be demanded by a single 8ohms severe reactance speaker.
If one knows that the decoupling and small smoothing allowed on the output of the chosen SMPS is say 4Apk, then one can downgrade the SMPS specification to speaker peak demand minus the short term capacitor capability. i.e. 11Apk @ +-45Vdc. This 990W smps would indicate a secure 15Apk into the amplifier if the 8ohms speaker demands that much.

Its generally agreed that so-called "high-current" amplifiers on the average fair better in sound quality, all else remaining equal. For example, it is nice to have an amplifier that can deliver 50A peaks. Put another way, I believe that, at minimum, an amplifier should be able to deliver a large fraction of its maximum output voltage into a 2-ohm load. Consider a 200W amplifier. Its maximum output into 8 ohms is about 60Vpk. That amounts to 30A into 2 ohms. The rails for that amplifier need to be at least 65V. If you consider only one rail at a time (e.g., 20Hz full voltage into 2 ohms), the SMPS must be rated to deliver about 2000W. If you want to put in a 2kW SMPS for a 200W monoblock, it may be OK. That is a serious SMPS. Never overlook the fact that SMPS transformers and inductors are subject to saturation, and things get very ugly very fast when the magnetics in an SMPS go into saturation.

But you may be better off not depending totally on the SMPS for peak current, no matter how fast the reservoir capacitors are replenished. This is where good old reservoir capacitors come into play. Some purists might argue that big reservoir capacitors help keep the SMPS out of the signal path.

It is true that not all SMPS are happy starting up into a huge capacitive load. But if designed with this in mind, it is not difficult. In fact, it is important in a PFC SMPS that there be a soft start and that the subsequent isolated converter have either a delayed start or a soft start of its own. These are not difficult things to implement.

Finally, one may not always want the audio amp SMPS be extremely stiff, since some compliance on the reservoir capacitors enables them to do their job. In such a case, the isolated SMPS can be designed so that its regulation control is derived from a separate small secondary on the line side for regulation. This also eliminates the need for an opto-coupler in the feedback control circuit. Then there will be no direct negative feedback regulation on the amplifier rails, so they will be a little less stiff due to transformer realities. This allows for more graceful degradation under transient loads, just as does the transformer and mains impedance in a conventional supply. We don't want the SMPS falling on its sword just to keep the rails within a tenth of a volt regulation.

Most SMPS are designed for relatively constant or very slowly changing loads, like those encountered in a PC or telecommunications gear, as opposed to the demands of an audio power amplifier where big peaks are occasionally encountered. In my view, the peak output current of a power amplifier should be limited only by the hard-earned SOA of its output stage, NOT by the power supply.

If an SMPS is designed with the needs of an audio power amplifier in mind, and decent reservoir capacitors are included, a good result can be obtained without extreme over-specification.

Cheers,
Bob
 
www.hifisonix.com
Joined 2003
Paid Member
I think I also said the same in that off the shelf SMPS units are probably not suitable.

However, I also think you need to consider the fact that SMPS' do transfer energy into the output caps far more effectively than a conventional PSU. So, if designed right, they can support the peak power requirements of Audio.

Where's Bruno when I need him. :D

:)
 
Last edited:
Member
Joined 2011
Paid Member
I've been toying with a soft-start-for-SMPS concept, that seems to work pretty well: between the SMPS and the 0.1 farad capacitor bank, insert a soft start circuit which delivers a linear current ramp to the capacitor bank upon startup. About 0.5 to 1.0 amperes per second seems to be a relatively sweet spot, protecting the SMPS while starting up the amp in reasonable time AND while minimizing (I x V) power dissipation in the current ramp generator power-FETs. Fun fact: parts cost and board area appear to be lower with a PIC microcontroller chip in the design, instead of an old-school, all discrete design on a hand-layed-out single layer phenolic thru-hole PCB.
 
IDK perhaps it's the view (inside to outside ) from the perspective of the high end audio PA world. e.g. driving power into unknown speaker loads. System design is viewing from the top down taking into account tradeoffs from the amplifier, power supply, and load. Consider a SMPS that controls power very well pulse by pulse. Such a SMPS could offer both output voltage and current control depending on the load. Flip a switch for 4 ohms and below e.g. the SMPS now has half the voltage double the current. Instant SOA control AND better value added, than simply offering (overspec'd) massive OP stage and bulk capacitors.
 
Last edited:
Here is another interesting aspect of PFC and SMPS. SMPS behave like constant-power devices. They will do whatever they need to do to deliver the power needed by the load. If input voltage falls, they will draw more current in order to supply the same needed power. This is usually a good thing, but comes with an interesting twist. They in many ways look like a negative resistance to the mains because of this behavior. This is true even in the case of a PFC SMPS.

We tend glibly to say that PFC achieves near-unity power factor simply by emulating a resistive load. Not quite true, however. It is looking more like a negative resistive load, unlike, say, an incandescent lamp. Ultimately, this behavior of an SMPS'd load might conceivably crash a mains that has too much impedance in it.

The world is quickly moving to more SMPS. The introduction of widespread LED lighting will accelerate this. I wonder if any of this stuff has preventive circuitry to mitigate the possible negative resistance effect. Think about utility brown-outs, especially when they want to shed some load. They expect that when they turn the voltage down, the current will go down. True in general today, but our loads may be moving away from that paradigm.

There already may be localized examples of heavy-hitter loads that behave this way. Consider these huge data-centers that gobble huge amounts of power to feed their zillions of SMPS-powered computing engines. Assume for the moment that they do not use huge amounts of "conventional" power for their cooling. Assume further that they use all LED illumination to make things a bit worse. Problem for load stability? I don't know, but I wonder if some kind of mitigation strategy has been put into place.

Of course, if we choose, we can design design an SMPS that acts to draw less power when the source voltage goes down, but then where does that leave the load that needs that power? I'm sure a PC would not be happy with that. For SMPS illumination loads that of course would be an option.

Just some interesting musings; maybe I've got it all wrong.

Cheers,
Bob
 
Administrator
Joined 2007
Paid Member
Slightly OT but maybe a sign of the future.

Our regional electricity supplier has invited customers (and I'm one of them) to take part in an 18 month survey as they try something akin to load management. What it comes down to is voltage reduction within legal limits.

I wonder how that will actually be done though... its gonna need a big pot :D

My thought at the time was "do they realise that as voltage goes down current demand will rise" when thinking of all the SMPS units out there. I believe the UK network is based on only around 1.5kWh usage per household at any one time.

CLASS - Customer Load Active System Services
 
I find the term "negative resistance" a poor one, it implies the load is supplying current to the source - better, to my mind is "negative dynamic resistance" ...

I disagree. Negative resistance is a well-understood term in the industry, and it has nothing to do with the load supplying current to the source. If you want, you can say that a negative resistance is a load whose current draw decreases if the voltage is increased, or, a load whose current draw increases when the voltage is decreased.

Cheers,
Bob
 
Bob in #2243 & Infinia in #2248 make some very good points.

The issue is what do we want the PSU of an ideal but practical amplifier to do. Here is my list.
  1. Constant voltage (ie no ripple or HF hash) until the maximum power rating of the amp is reached at some LoZ eg 2R but maybe even 4R if the behaviour specified in 3 is good. This is particularly important in Class A amps cos the huge standing current means huge sawtooth ripple which is superimposed on the signal on overload.
  2. At lower Z loads, the PSU reverts to supplying constant power. This is almost trivial to design in as most SMPS do this when free-running. The rail voltage will drop but current supply is 'effectively unlimited'.
  3. When the rails drop, they STILL need to be free of ripple & HF hash. The rate of drop is of course determined by the zillion uF caps on the supply. The rate of recovery is ultimately determined by the constant power limit but the PSU needs to intelligent enough not to have discharge / recharge cycles lead to ripple. Perhaps lower the regulated 'constant voltage' if the 'constant power' mode is activated more than once or twice.
  4. Infinia's switch is useful if the load is known to be LoZ to match the amp's initial constant voltage rails to it so the constant power stuff happens rarely.
  5. PFC as Bob describes to minimize hash generated and for immunity to hash on the mains.
I discussed the design of such a PSU for a 1000W amp with a SMPS designer circa 1980 but the project didn't go ahead.

The important features come out of DBLTs on PSUs for power amps.

I suspect Bruno Putzey's semi-regulated supplies for his Hypex modules may be similar in behaviour.
 
Disabled Account
Joined 2012
Its generally agreed that so-called "high-current" amplifiers on the average fair better in sound quality, all else remaining equal.

But you may be better off not depending totally on the SMPS for peak current, no matter how fast the reservoir capacitors are replenished. This is where good old reservoir capacitors come into play.

Most SMPS are designed for relatively constant or very slowly changing loads, like those encountered in a PC or telecommunications gear, as opposed to the demands of an audio power amplifier where big peaks are occasionally encountered. In my view, the peak output current of a power amplifier should be limited only by the hard-earned SOA of its output stage, NOT by the power supply.

If an SMPS is designed with the needs of an audio power amplifier in mind, and decent reservoir capacitors are included, a good result can be obtained without extreme over-specification.

Cheers,
Bob

:cool::)
 
3. When the rails drop, they STILL need to be free of ripple & HF hash. The rate of drop is of course determined by the zillion uF caps on the supply. The rate of recovery is ultimately determined by the constant power limit but the PSU needs to intelligent enough not to have discharge / recharge cycles lead to ripple. Perhaps lower the regulated 'constant voltage' if the 'constant power' mode is activated more than once or twice.
And the reality is very far from this. Practical measuring, and sim's point this out - real voltage rails turn into a complete mess once the pressure's on, when decent current spikes are demanded by the speaker. Only by the grace of reasonable PSRR does the amp manage to maintain its dignity ... ;)
 
Disabled Account
Joined 2012
[*]PFC as Bob describes to minimize hash generated and for immunity to hash on the mains.
[/LIST]

There are several $$ test equipment to measure the hash levels. But an inexpensive portable one made by Peter Madnick and Demian Martin (ENTECH Wideband Powerline & EMI analyzer) gives an audible output via small speaker so you can tell if there are audio freqs on the ac line. It also gives a metered display of relative level of DM and CM noise as well.
You can use it to do a before and after line filtering check for effectiveness of a line filter without the big $ expense.

-RNM
 
Last edited:
well by objective measures like word count, dedicated forums, ads, shelf space... the "pure subjectivists" are doing a bit more slinging - and by boycotting the Scientific Method, refusing Blind Tests, Controls, claiming literally incredible memory abilities they aren't likely to progress audio reproduction any more than they haven't in the last 1/2 century
It is true...but...
As always, it is a question of balance. "pure objectivism' is as stupid as 'pure subjectivism'. For a simple reason, if we know a lot about what's happens in the components we use, we still don't know everything.
Laws of physics are reproducible and demonstrated. Everithing witch goes against is snake oil.
Music reproduction is a make believe game, and our brain compute the signals after our ears...... It seems obvious we HAVE to listen carefully and lie *too* on our 'feelings' to design hifi gears. With care to not fool ourselves with auto suggestions.
Of course, the pure mathematic is less demanding, more reassuring, faster... And brings good results...till a certain point.
When two solutions measure the same and don't 'sound' the same, what is the 'pure objective' method to chose the best ?
 
There are several $$ test equipment to measure the hash levels. But an inexpensive portable one made by Peter Madnick and Demian Martin (ENTECH Wideband Powerline & EMI analyzer) gives an audible output via small speaker so you can tell if there are audio freqs on the ac line. It also gives a metered display of relative level of DM and CM noise as well.
You can use it to do a before and after line filtering check for effectiveness of a line filter without the big $ expense.

-RNM

I like that idea. I may look into it further.

Cheers,
Bob
 
It is true...but...
As always, it is a question of balance. "pure objectivism' is as stupid as 'pure subjectivism'. For a simple reason, if we know a lot about what's happens in the components we use, we still don't know everything.
Laws of physics are reproducible and demonstrated. Everithing witch goes against is snake oil.
Music reproduction is a make believe game, and our brain compute the signals after our ears...... It seems obvious we HAVE to listen carefully and lie *too* on our 'feelings' to design hifi gears. With care to not fool ourselves with auto suggestions.
Of course, the pure mathematic is less demanding, more reassuring, faster... And brings good results...till a certain point.
When two solutions measure the same and don't 'sound' the same, what is the 'pure objective' method to chose the best ?

I agree; a healthy combination of subjectivism and objectivisom is what is best. Indeed, sometimes things picked up on the subjective side lead to advances in the objective side when people look more closely for a cause of the subjective thing they heard.

Cheers,
Bob
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.