Switching Regulator for Audio Amplifier

[Fred]

Where are you Fred ?

In the UK Margaret Thatcher really kicked this off. We are now 70% dependent on
the service sector and wait 'til that starts getting outsourced too.

Manufacturing industry is but 16-17% of GDP and much of the rest is low-wage
agriculture.

And this country was once called 'the workshop of the world'. We built everything
and anything. Now we can't even make cruise liners for the (American owned) P&O
Line and Cunard. Instead they come from Germany, France and Italy.

Graham

I'm in the states. From what little I've read, I think the economic
situation at the moment is even worse there in the UK.

 

[Eeyore]

I'm in the states. From what little I've read, I think the economic
situation at the moment is even worse there in the UK.

The business with Lehmann Brothers doesn't sound too good for sure.

Oh, I forgot to say, according to informed sources the UK is now a 'post-industrial
economy'. I suppose the next step is the 'mud hut economy' ?

Graham

 

[Fred]

The business with Lehmann Brothers doesn't sound too good for sure.

Oh, I forgot to say, according to informed sources the UK is now a 'post-industrial
economy'. I suppose the next step is the 'mud hut economy' ?

Graham

Conspiracy theories abound. But most of them have the global economic
crisis going well... as planned.

Ten major banks have failed this year here. Now Lehman Brother. The
media is giving it all the attention of a fender bender.

What is it they say about things beyond you control again?

Cheers,

Fred

 

[[email protected]]

Generally, not. Look at the regulation vs frequency curves.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultanthttp://www.abvolt.com

It depends on the pass device, which unfortunately is bipolar for
most"canned" devices. If you have a mos pass device, either integrated
for low power or discrete for high power, it all comes down to a ratio
of capacitors, i.e. the feed-through capacitance of the pass device
and the filter capacitor.

It is actually ironic in a sense that poorer regulators, i.e those
with less transconductance in the pass device, generally have better
high frequency performance (less feed-through capacitance) due to the
pass device being smaller.

 

[Jan Panteltje]

So where are all the jobs now ?

Selling cheap Chinese products of course

 

[Vladimir Vassilevsky]

It depends on the pass device, which unfortunately is bipolar for
most"canned" devices.
If you have a mos pass device, either integrated
for low power or discrete for high power, it all comes down to a ratio
of capacitors, i.e. the feed-through capacitance of the pass device
and the filter capacitor.
It is actually ironic in a sense that poorer regulators, i.e those
with less transconductance in the pass device, generally have better
high frequency performance (less feed-through capacitance) due to the
pass device being smaller.

I'd say it is determined by the parameters of the feedback loop rather
by the output device itself. The loop has to be stable in the broad
range of the parameters and loads, so they make it slow and overdumped.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[MooseFET]

But not green ? That's only for CDs !

The dynamic thing does actually have some merit amongst similarly amps of continuous
power rating since the one with sloppy regulation will have higher off-load rail
voltages and will reproduce brief transients without clipping more easily. The design
principle goes back decades in fact.

Yes, I first saw that argument in a discussion of a tube circuit. It
seems that now we should have out grown this idea. We can make stiff
power supplies and very powerful output stages and circuits to monitor
the temperature.

 

[Eeyore]

So where are all the jobs now ?

Selling cheap Chinese products of course [/QUOTE]

And the money to buy then with ?

In the case of the USA also from the Chinese of course !

Graham

 

[Eeyore]

Yes, I first saw that argument in a discussion of a tube circuit. It
seems that now we should have out grown this idea. We can make stiff
power supplies and very powerful output stages and circuits to monitor
the temperature.

As used and designed by audio professionals of course.

Graham

 

[Jan Panteltje]

And the money to buy then with ?

In the case of the USA also from the Chinese of course !

You got it!

 

[Nico Coesel]

I have an audio amp based on the LM3886. In order to get 50W out of
it I need to use a transformer who voltage rises too high for the
LM3886 to tolerate during no or light load times.

I have modeled a linear regulator for the amp, but is is very
wasteful, it dissipates over a hundred watts a channel. I had to
parallel four regulators on each rail in order to keep their
individual dissipations manageable.

I have started considering a switched mode power supply, and have been
looking at chips such as the LM5116. Some time ago, I heard using
SMPSs in audio amplifers in problematic because of the difficulty in
filtering out the switching noise from the output.

Any suggestions would be appreciated.

Several years ago I designed a class-D amp with an off-line SMPS. I
used a common mode filter transformer at the output of the SMPS to get
rid of the noise (windings arranged so that the AC currents cancel
each other). It works great.

 

[Eeyore]

Several years ago I designed a class-D amp with an off-line SMPS. I
used a common mode filter transformer at the output of the SMPS to get
rid of the noise (windings arranged so that the AC currents cancel
each other). It works great.

Ummm..... Where in the circuit ? If the DC drain on the +/- supplies is
asymmetric (which audio is) the core will saturate.

Graham

 

[Fred]

Ummm..... Where in the circuit ? If the DC drain on the +/- supplies is
asymmetric (which audio is) the core will saturate.

Graham

I got to thinking. If we work backwards from the speaker to the
transformer some thing interesting shows up.

We need 20Vpk and 5Apk at the speaker to get 50Wrms into 4 Ohms.

So add 4 for the dropout voltage and 0.7 for the bridge, that gives us
a minimum of +/-24.7V at full load load.

Since the 5A figure is peak the rms is 5/sqrt(2)= ~3.6A

Thus, the transformer needs to supply a minimum of 24.7V @ 3.6A for
each rail. That is approximately 89VA for each rail.

Accordingly for the four rails, the transformer must supply a minimum
of 4 * 89VA = 356VA.

Regardless of how good the ripple rejection of the chip is, in order
for it to supply a certain amount of power to a load, it must first
receive at least that much power from the power supply. It simply
lacks the ability to create energy out of thin air.

Therefore, a 360VA rating is the minimum to get 50rms out of the the
chip. 500VA is the closest standard size I can find available. I
think the extra capacity should at least reduce the requirement for
the reserve caps.

Fred

 

[Phil Allison]

"Fred"

I got to thinking. If we work backwards from the speaker to the
transformer some thing interesting shows up.

We need 20Vpk and 5Apk at the speaker to get 50Wrms into 4 Ohms.

So add 4 for the dropout voltage and 0.7 for the bridge, that gives us
a minimum of +/-24.7V at full load load.

Since the 5A figure is peak the rms is 5/sqrt(2)= ~3.6A

** This is where you went 100% WRONG !!

Current is only drawn in HALF cycles from each DC rail - so with a peak of
5 amps, the *average DC value* is 5/pi = 1.6 amps per rail.

The actual current comes mostly from the FILTER CAPS - not the
transformer windings which merely top up the cap's lost voltage twice each
cycle of the AC supply.

Thus, the transformer needs to supply a minimum of 24.7V @ 3.6A for
each rail. That is approximately 89VA for each rail.

** BOLLOCKS.

The correct VA figure can only be approximated by calculation - the
formula for which is 1.6 times the DC power draw - which is 40 watts for
each rail ( ie 1.6 x 25).

Accordingly for the four rails, the transformer must supply a minimum
of 4 * 89VA = 356VA.

** The correct figure is 4 x 40 x 1.6 = 256 VA ( approx.)

Therefore, a 360VA rating is the minimum to get 50rms out of the the
chip.

** Nonsense.

You could easily use a 160 VA rated tranny with music signals.

That IS the size used in the majority of 100 watt class B audio
amplifiers.



...... Phil

 

[Eeyore]

I got to thinking. If we work backwards from the speaker to the
transformer some thing interesting shows up.

We need 20Vpk and 5Apk at the speaker to get 50Wrms into 4 Ohms.

So add 4 for the dropout voltage
OK

and 0.7 for the bridge,

More like 1V at rated load.

that gives us a minimum of +/-24.7V at full load load.

+/- 25V

Since the 5A figure is peak the rms is 5/sqrt(2)= ~3.6A

Thus, the transformer needs to supply a minimum of 24.7V @ 3.6A for
each rail. That is approximately 89VA for each rail.

No ! The current taken from the transformer is non-sinusoidal. Apply a typical
form factor and you need to multiply this figure by anwhere between 1.6 and 1.8

Allow 6.5A at 25V-0-25 pk = 17.8 V rms = 116VA/ch - it only draws power from
the + or the - not both at the same time except for a low quiescent current. !

Oh but you haven't allowed for ripple. Haha more fun. That depends on the size
of your reservoir caps.

Accordingly for the four rails, the transformer must supply a minimum
of 4 * 89VA = 356VA.

Regardless of how good the ripple rejection of the chip is, in order
for it to supply a certain amount of power to a load, it must first
receive at least that much power from the power supply. It simply
lacks the ability to create energy out of thin air.

Therefore, a 360VA rating is the minimum to get 50rms out of the the
chip. 500VA is the closest standard size I can find available. I
think the extra capacity should at least reduce the requirement for
the reserve caps.

Eh ? Something went askew there. Anyway, bear in mind that almost no-one sizes
their transformers for continuous sinewave load.

But a 250VA is going to be ample (regulation issues aside which I hope have
already been covered in depth).

You should see the transformers I spec for commercial pro-audio amps. You'd
think they're tiny for the kW amp ratings, But they're specially wound.

Graham

 

[Fred]

"Fred"







** This is where you went 100% WRONG !!

Current is only drawn in HALF cycles from each DC rail - so with a peak of
5 amps, the *average DC value* is 5/pi = 1.6 amps per rail.

The actual current comes mostly from the FILTER CAPS - not the
transformer windings which merely top up the cap's lost voltage twice each
cycle of the AC supply.


** BOLLOCKS.

The correct VA figure can only be approximated by calculation - the
formula for which is 1.6 times the DC power draw - which is 40 watts for
each rail ( ie 1.6 x 25).


** The correct figure is 4 x 40 x 1.6 = 256 VA ( approx.)


** Nonsense.

You could easily use a 160 VA rated tranny with music signals.

That IS the size used in the majority of 100 watt class B audio
amplifiers.

..... Phil

I work up i the middle of the night and realized that.

I got there just slightly differently. I used 0.3535 to determine RMS
voltage for each rail, since only half wave rectified. So,

20Vpk = 14.14Vrms + 4 + 0.7 = 18.84Vrms and 5Apk = 3.54A

18.84 * 3.54 = 66.7VA --> four rail total 267VA.

The regulation I think I got. Plitron lists their 2x25V 300VA at 6%
regulation. That would work, but they now have a $250 minimum order.
And Hammond omits any mention of regulation. I guess with with them
its buy and see...

Power rating of the music signal. There's another thought. I suppose
I could write some code to analyze PCM files of the music I listen and
find what the averages are.

It bothers me to say it, but it would be a lot cheaper just to buy six
regulators and regulate the transformers that destroyed the first
chips I used.

 

[Jan Panteltje]

I got to thinking.

We need 20Vpk and 5Apk at the speaker to get 50Wrms into 4 Ohms.

So add 4 for the dropout voltage and 0.7 for the bridge, that gives us
a minimum of +/-24.7V at full load load.

Since the 5A figure is peak the rms is 5/sqrt(2)= ~3.6A

Thus, the transformer needs to supply a minimum of 24.7V @ 3.6A for
each rail. That is approximately 89VA for each rail.

****NO**** THE POSITIVE AND NEGATIVE PEAK DO NOT HAPPEN AT THE SAME TIME.

+ peak

 

[MooseFET]

I got to thinking.  If we work backwards from the speaker to the
transformer some thing interesting shows up.

We need 20Vpk and 5Apk at the speaker to get 50Wrms into 4 Ohms.

20 * 5 / sqrt(2) = 70W

You have done something wrong in your math.

So add 4 for the dropout voltage and 0.7 for the bridge, that gives us
a minimum of +/-24.7V at full load load.

I'll bet that at full load, the bridge is losing more than 0.7.
During the peak of the sine wave, there is a very large current in the
diodes.

Since the 5A figure is peak the rms is 5/sqrt(2)= ~3.6A

Thus, the transformer needs to supply a minimum of 24.7V @ 3.6A for
each rail.  That is approximately 89VA for each rail.

When converting Watts in the load to VA in the transformer, you need
to multiply by 1.5. No matter which sort of rectifier you pick, the
current drawn on the secondary is not a sine wave so there is extra
demand on the transformer.

The audio amplifier will have an efficiency of about 60% or 70% at the
best so taking your 50W I say:

50W * 1.5 / 0.60 = 125VA per amplifier

This is about where you should end up when when the design is done.

Accordingly for the four rails, the transformer must supply a minimum
of 4 * 89VA = 356VA.

You have assumed that to make the current on the output, both rails
must produce the current at the same time. Since this is not a full
bridge design, this is not right. The current flows on one rail at a
time.

 

[MooseFET]

[....]

But for audio the average power is much lower, at least for music,
so 100VA would be just fine.

Transformers aren't all that expensive. Going a bit over designed on
the transformer will make it run cool.


There is another issue that has been ignored all the way through these
arguments. The actual mains voltage can be 10% either way of the
nominal. The design needs to be done at the two extremes if you want
it to produce the rated power at the low line case and survive the
normal high line case. A thermal shutdown is a very good idea here
too.

 

[Phil Allison]

"MooseFET"

We need 20Vpk and 5Apk at the speaker to get 50Wrms into 4 Ohms.

20 * 5 / sqrt(2) = 70W


** Huh ????????

Why the root 2 ??

Peak power = double "watts rms" - you dumb asshole.



...... Phil