Balanced amplifier - lower distortion?

[barefoot]

I'm wondering if there is any inherent advantage with respect to distortion in driving a speaker (or any other load) with a balanced (bridged) circuit?

For a given voltage across the load each amplifier only needs to swing half way. However, the effective load on each amp is 1/2, so the current draw remains the same. Output stage crossover distortion depends on the current rather than the voltage, so there isn't much benefit there.

Maybe there's something in the fact that the two amps swing in opposite directions, resulting in cancellation of some type of common mode distortion?

So far the only major advantage I can find in this configuration is a +6dB gain in dynamic range for a given power supply voltage - assuming the amps can provide enough current.

Any thoughts on all this?

Thanks,
Thomas

 

[Kev]

I don't think the speaker driver is going to know how it is driven.

As for any common mode ... :roll:
I guess there will be some common mode rejection/cancallation but I don't see that it is any different to a normal Bridged amp config.

Yes this can give more volts at the output than the rail voltage and with each amp seeing half the load then there is another 3 so intotal 6db.

Both amps will now need to be able to deliver the extra current ... :roll: ... unless you are aiming at a higher imp driver.

as with all speaker amp combinations it seems to come down to a compromise of current and rail voltage and speaker efficiency.

more volts or more current
a never ending merry go round
I tend to lean to more volts as I think that current can take its toll.
and yet
a very good speaker builder friend of mine has just made a double voice coil subwoofer with two 2 ohm coils and in parrellel that's 1 ohm.
and at a little over 2000 watts thermal it is a beast.

This is all about car competition sub woofer stuff ... crazy. They have rules and i think the low ohms is becausde the battery volts is limited to 12 volts and so the available BIG amps have a way of doing things.


eh ??
it's all a compromise
:roll:
what was the question ... :shock:

 

[barefoot]

Maybe I should be more specific with my questions.

Lets say each amplifier has an identical Class AB output stage.

1. Does the fact that each amp only has to produce half the output voltage for a given current through the load result in any distortion reduction?

2. Does the push-pull symmetry of the output stages result in any distortion reduction?

 

[Kev]

I don't think I am seeing this as symmetrical as you are
:roll:

I am picturing a normal analog amp with complimentary pairs and AB operation.
One amp is in the positive while the other is in the negative ... and so different output units.
:roll:
two wrongs don't make a right and I don't think this is as common as it first looks.
BUT
I think I am beginning to see what you are driving at.

Let me turn it all around.
What distortions are you trying to cure ?

 

[NewYorkDave]

Thomas, read up on the Crown "grounded-bridge" amplifiers (Macrotech, Powerbase, etc.). Their main advantage was greater output swing for a given supply, but other advantages of the topology may be addressed in the literature. From a user's/repairman's standpoint, anyway, those were good solid amps.

Here's a layman's overview of grounded-bridge:
http://www.crownaudio.com/pdf/amps/grbgpapr.pdf
If you dig around on the web, you may find a more in-depth technical treatment.

 

[barefoot]

Maybe this stripped down situation will make things clearer.

Will this circuit in principle yield lower distortion than just a single AB stage with a grounded load?

 

[bcarso]

"2. Does the push-pull symmetry of the output stages result in any distortion reduction?"

It can if you are dominated by second. But usually you are already not, with a reasonably good class AB complementary-symmetry output.

Bridge-Tied-Load or "BTL" amps are de rigeur for economy and low supply rails---they have been the rule for low-cost automotive electronics for years and years now. They facilitate not only higher power for a given Z load and supply voltage, if the current capability is there, but eliminate an output coupling capacitor. Often the chip amps are configurable as BTL, or two channels of SE (single-ended) and cap-coupled, for each pair of internal amps.

Another advantage is that the power supply ripple frequency is doubled, which helps the filter caps to work better.

There is an advantage in power supply noise rejection if the chipmaker or discrete circuit designer has done a good job already with the input stage(s). But most often they have not, so again there is little advantage unless the design is perfectly symmetrical and matched---thus having the input stage contributions be equal and thus common-mode at the load. Most of the time these conditions are not satisfied.

For switchmode amps there is often a greater advantage, especially the "pure digital" types that run essentially open-loop. Feedback that is effective at high frequencies is difficult or impossible, so having the output switches see the same power supply noise and hum more or less means the load gets that noise in common-mode and thus rejects it. Note that this is at zero signal---these folks can cheat a lot by specifying S/N as the Max output ratioed to the noise level. If that noise level pumps a lot with high signals....your mileage may be lower.

quote: "They have rules and i think the low ohms is becausde the battery volts is limited to 12 volts and so the available BIG amps have a way of doing things."

To get the power now they are all using switchmode power supplies regardless of driver impedances, and often class D amps following the boosted supplies. Crown is offering a 6kW amp I believe.... It is now common to have multiple alternators and batteries/giant caps for the d.c. input power. Madness, but it keeps people out of the house ;-)

 

[barefoot]

Ok, now that I look at my last circuit there do seem to be a couple obvious things that could lead to lower distortion.

1. There is always both an NPN and a PNP transistor conducting during positive and negative voltage swings.

2. The current through each device is the same, but the voltage is cut in half, so the power dissipation in each device is also cut in half. In other words, each half of the circuit shares the power dissipation.

Though I'm not sure what kinds of distortion these might aid with. Odd order harmonic and intermodulation distortion of course are the most annoying, so my hope is that they would be reduced.

Those power supply rejection advantages sound pretty good too. How much rejection do you think one might expect?

 

[bcarso]

It doesn't help odd-order unfortunately. As far as PS rejection and even-order, it all depends on how well-matched the two halves are.

Note that for bipolars the principal failure mode at high power is due to "secondary breakdown"---localized hot spots on the die that hog more as they get hotter. The lower voltages in general make the given transistor operate in a region more conducive to avoiding this. See the power dissipation curves for biploar power transistors where there are lines with various breakpoints and ever steeper slopes as the voltage shown on the x-axis goes up. There are families of them for different pulse widths. Compare to the similar graphs for power MOSFETs, which often are practically rectangles for short enough low-duty-cycle pulses.

I believe there are some audiophile amps that are "single-ended push-pull"---you have lots of current on each side at zero signal, like a single-ended ~class A amp, but the inevitable even-order is cancelled at the load hung between to the extent that the two sides are matched. I don't know if these disappoint those who like to listen to very simple, nearly non-polyphonic material and think the second harmonic restores some missing stuff, because they won't get much of that effect.

 

[PRR]

> any inherent advantage with respect to distortion in driving a speaker (or any other load) with a balanced (bridged) circuit?

No. Many small things to consider, but in the context of a complete design, they all vanish in the haze.

As you figured, the current is unchanged, for specified impedance and load.

Push-pull stages almost always have a single-ended stage in front, which has 2nd harmonic trouble. Two of these in opposition change the 2nd to a lower 3rd. The THD number will be better, sound may not be. But once you commit to two single-ended stages, you have enough parts to make one single-ended stage that won't have any 2nd harmonic problem.

Most of the power supply rejection issues are, again, driven by parts cost. The digital amps may have poor PSRR, but (not counting battery operation) the only reason you would use such an absurd thing to make music is to reduce cost (or size/weight).

The ONLY for-sure reasons to bridge-tie are when you are stuck with a too-low supply voltage (in cars, 12V is cheap, anything else is costly) or low-voltage transistors (it is hard to make those 6KW at speaker impedance with 100V-200V transistors unless they are stacked somehow).

> To get the power now they are all using switchmode power supplies regardless of driver impedances

Essentially all car-audio external power amps make their own high-voltage rails.

Essentially all in-dash car radios seem to cling to bridge-tied 12VDC 4Ω amps. An ideal amp makes 20, 22W this way. Published numbers range from 18W to 25W, but the actual power below clipping will be the same within a dB for any of these.

There are lesser radios. My wife got a Miata without a radio, asked the dealer to put something in, and didn't say what. The idiot found a single-ended radio, good for 5 watts. In a drop-top convertible! To be far, when I replaced it with a powerful (22W/ch) in-dash, it didn't get a LOT louder, just the obvious 6dB which is not even twice as loud as "not loud at all". What the car needs is 100W/ch, and a place to fit a 100W amp.

 

[bcarso]

"The idiot found a single-ended radio, good for 5 watts."

Definitely not a "top-down" radio.

Wow. I was almost to the point of thinking they were hardly doing those anymore. Still in all, a bit less silicon, and 1000uF caps are about 5 cents in the Far East, at least before the recent decline in the dollar.

There are some 2 ohm load BTL dual amps from Phil*ps and T*shiba. On a good day with the alternator whizzing and the volts high they will do about 40W x 2 at ~10% THD (sort of the automotive standard for power rating, roughly a 32% advantage over barely-onset-of-clipping power, given constant rail voltage). H*rm*n was buying those for well under $2 each for a while, definitely a serious contender in the watts per dollar sweepstakes.

 

[pstamler]

There's an advantage if the amplifiers are class-A: the current drawn from the power supply changes very little. James Bongiorno, designer of the Ampzilla series, later designed an amp with balanced class-A outputs.

Peace,
Paul

 

[Kev]

Maybe this stripped down situation will make things clearer.
... and conducting during positive and negative voltage swings.

yep
as I said, I thought I was already getting in tune with the basic idea
... but still not convinced it will bring the benefits in lower distortion.
This would be implying that the pos and neg are not tracking well in the original design.

What distortions are we trying to cure ?

... but in the context of a complete design, they all vanish in the haze.

thanks PRR ... definitely the diminishing returns type of situation, I think.

The ONLY for-sure reasons to bridge-tie are when you are stuck with a too-low supply voltage ...

yep
or to just bring a different feature to the amp to make it more versatile.

are we looking to a powered monitor ??
and
are you looking to a chip amp ... cos a chip amp may require a technique like this to get the powers you want into a single driver
:roll:
I'd still be heading back to a more simple idea when looking to the higher quality product.
KISS ... keep it simple ... etc

what are you up to Thomas ?

 

[Kev]

[quote author="pstamler"]There's an advantage if the amplifiers are class-A: [/quote]
we posted near the same time
yep
and that I do get as there have been a few HiFi people to go this way ... with both tubes and SS.

The broadcast tech in me say not a good idea BUT the Esoteric HiFi side of me say
Yes !!
with the right presentation and the marketing to go with it
:green: you could put together a powered monitor with a very expensive price tag ... that would help to heat your cold control room deep into the winter.

 

[bcarso]

"James Bongiorno, designer of the Ampzilla "

Saw him just the other night at that Whitlock AES presentation, in his inimitable pink hat. I was happy to see he has apparently recovered after his health crisis.

 

[Gus]

For some reason this thread reminded me of the Cotter amp(?) from the late 70's that I read about in the audio critic IIRC. I think both channels where "standard" ab out: however one channel was inverted buffer before it and the loudspeaker was connected inverted phase.

The idea was for more of a constant current draw between the split power supply rails IIRC.

 

[bcarso]

Yeah, I've done that a lot with budget powered speakers. I later noticed that Ph*lips was configuring some of their chip amps to facilitate this (one channel inverting, one not). When it's a single-supply amp this means you can use a single common coupling capacitor to common, providing that the bass is mixed to center as it is most of the time---hence the cap isn't "there" for pure differential output signals.

I believe the field of power amp design has yet to do a methodical approach to power supply drain vs. signal, and active device dissipation vs. signal (emphasis on methodical there---plenty of analyses of various common configurations exist). The assumption is usually that the signal self-heating effects are rendered negligible by global feedback---and yes, they typically are. Challenged by some highend folk to eliminate global feedback (concerns I view with healthy skepticism btw) I have done some exploring of just how one could preserve the measured performance without it.

For the photodiode array spectrometer I designed years ago, which required a complex set of clock drive signals that needed to be both fast and exquisitely stable, I found ways to make most of the signal generating circuits be both constant power supply drain and constant active-device dissipation. It was relatively easy for pulse circuits with essentially only two states to do the latter. For the former, rather than using shunt regulators I produced compensatory switched current loads local to each section, which was a tad faster than a local regulator. Among the benefits of this methodology was that a large copper sheet ground plane could be used which worked well at high frequencies and low, and avoided the issues with the inductances associated with star grounding conductor runs.

 

[Kev]

[quote author="bcarso"]For the photodiode array spectrometer I designed years ago, which required a complex set of clock drive signals that needed to be both fast and exquisitely stable, ... [/quote]
interesting stuff
just how fast are we talking about ?

 

[bcarso]

The output amp that steered the currents was about a 2ns rise/fall (10-90%).

The 50mA current source had a trifle higher output C than it might have. I used FD700 diodes steering in and out of Vishay 100 ohm R's, then ~100 ohm coax cables took the signals into the cold chamber where the array was. Since it was a mostly capacitative load I series-terminated with 100 ohms to have the reflection double the risetime at the array pins without overshoot. At the array pins the signals were 15ns rise/fall.

The trick of the whole system: you had to sit with everything off for up to an hour at the telescope, then switch on and do your readout. If anything much shifted, especially the cooled preamps' input offset voltages, this appeared as a error in readout. It was necessary to stabilize the input FET temp to about +/- 10 millidegrees K or better, in order that the offset error be on the order of the readout noise. Since there was no feedback resistor there was no easy way to inspect the preamp input voltage. Feedback C reset was done with selected rf bipolars (<100fA Icb at room temp, lower with cooling) and a very elaborate adaptive controller.

The paper describing the preamps and clock drivers appeared in an SPIE Instrumentation in Astronomy 1982 symposium publication. Anyone interested that can't find it in a library can PM me and I will snail mail a copy.

 

[clintrubber]

from Brad:

There are some 2 ohm load BTL dual amps from Phil*ps and T*shiba.

FWIW, we've about finished a thing here that does 1 Ohm drivers as well (giving 150W max.power typ. downhill, 14.4V & other fancy ways of calculating - as this game is played by everybody unfort.).