Another power amp idea

[PigletsDad]

I am pretty sure that the cost of a power amp is dominated by the cost of the power supply parts. Designs that use multiple transformers are even more expensive.

For active systems, or I guess home theater, large numbers of power amp channels are needed, and the total cost can be quite high. An active home theater system needs LOTS - 5.1 with 3 way everywhere uses 16 channels!

In this type of case, it is worth thinking through if there are design tricks that can maintain the quality level, while allowing a shared supply to run many channels.

A design with high PSRR obviously helps, but there are real issues with managing the high return currents into a shared star earth.

An alternative strategy is to use a balanced circuit using Nelson Pass' "super-symmetry" idea. This is slightly more complex in practice than a single ended design - the main expense is the need for two pairs of output devices, rather than one. However, the ability share power supplies over many channels will more than make up for this.

I have some simulations that suggest this can really help, as you start off with very high intrinsic PSRR just from the balanced layout, even before the normal circuit measures come into play.

You get the added benefit that the SOA curves are more favourable because the load is shared across more output devices.

 

[john.luckins]

Hi PD,

Are you talking about a bridged arrangement where there is no net current in the 0 volt line? If so the doubling of the output devices means a lower intrinsic damping factor. I'm not sure how well Naim have overcome this with the NAP500 or whether this is a real world shortcoming of such an arrangement.

John

 

[flashgo]

I'm glad you're thinking along these lines, PD. I would be interested in an add-on surround system – i.e., service for one middle and 2 rear speakers. For 2-ways, that would be 6 channels in one box. There are other applications as well....

 

[mudlark]

on a personal basis I would like you to have a look at the active filter. I use an ESPaudio P09 filter which is great but seems to have no competition.

Personally I can't see me going active on my 5.1 system as the speakers are the expensive bit.

I may be interested in trying a 4 channel for my stereo system. 5.1 is eleven channels!!

 

[PigletsDad]

Hi PD,

Are you talking about a bridged arrangement where there is no net current in the 0 volt line? If so the doubling of the output devices means a lower intrinsic damping factor. I'm not sure how well Naim have overcome this with the NAP500 or whether this is a real world shortcoming of such an arrangement.

John

Yes, I mean a design with no net current in the earth line.

Good point, but damping factor is genuinely not a big deal; once it say 10 or so further increases make no real difference and the output impedance in my simulations looks to be very small indeed; I think damping factors of several hundred will be on hand.

 

[martin clark]

the main expense is the need for two pairs of output devices, rather than one.

That might need to be four pairs depending on power targets. Each half 'sees' half the nominal speaker impedance, perhaps requiring a load capability well below 2ohms per half of the bridge.

Still cheaper than more transformers, agreed!

 

[PigletsDad]

That might need to be four pairs depending on power targets. Each half 'sees' half the nominal speaker impedance, perhaps requiring a load capability well below 2ohms per half of the bridge.

Still cheaper than more transformers, agreed!

Yes, but the voltage on the transistors is halved, so if you look at the SOA curves, the allowable current goes up a lot. I initially dismissed balnced operation for exactly the arguments you make, but when I looked at the real data sheets you can safely deliver much more current than you think - especially into reactive loads.

 

[martin clark]

Aha... <penny drops>

 

[john.luckins]

This is starting to sound like a 16 channel version of the NAP500! For each channel could you not connect two NAP circuits like the input amps of an instrumentation amplifier. There would be no need for a 0v reference at the input, the feedback paths could be commoned and disconnected from the reference earth witha single bipolar/non polar cap and there need only be one Zobel network.

If you wanted to be really clever you could try earth regeneration on a single rail ouput stage PSU to reference it to the 0v rather than the direct connection of a split rail supply. I think Jiim would have to take the credit for that. It worked on the rectifier/switching noise on the class A preamp that was the Starfish, so why not on a class AB power amp where there would appear to be more to be gained?

John

 

[PigletsDad]

Actually, if you look at the Nelson Pass patent, you can see a much simpler circuit than an instrumentation amp layout; it really only involves just a few more parts than the NAP layout!

Schematic to follow in a day or two.