Class B operation

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Does anyone have any links to sites with the basic theory of class B amp operation, or perhaps someone can give me a brief description here. I would also be interested in some links to simple class B schematics on the net. I'm just curious how class B operation is different from class A.

Cheers

Dan
 
Dan,
In a nutshell, (assuming a push-pull circuit, which is necessary for all classes other than A) a class A circuit is like two guys on one of those old-fashioned crosscut saws, each putting his strength into both the 'push' stroke and the 'pull' stroke. In a class B circuit, one fellow is pulling the saw, but the other is just standing there with his arms folded, resting for a second. This allows him to save strength and cool down for a brief time which sounds like it might be a good thing, but the fly in the ointment comes when he lays hands on the saw to take his turn pulling the saw back his way--as he grabs hold, and as the other guy lets go--there's a moment there when the saw is getting wiggly and neither one of them has good control over what's going on. This, in electronic terms, is called crossover distortion; a brief spike of really obnoxious distortion (this has nothing to do with tubes vs. MOSFETs vs. bipolars, it's simply a question of how they're used in the circuit) that's difficult to measure (using standard measurement techniques--it's not that hard to see with other methods) because it's of short duration. The problem is that the ear isn't fooled.
There are in-between classes, too. The guys can, instead of each doing 50% of the work (strict class B), do something more like two-thirds, which will involve a certain amount of overlap where they both have their hands on the saw for a portion of the cycle, but their rest period gets shorter. This leads to class AB. Note that AB isn't one set percentage, it's a sliding scale all the way from 50/50 (class B) up to 99/99 (meaning each guy gets this tiny little rest period of 1% of the saw stroke--hardly worth the trouble, really). Once you get to 100/100, then you're back at class A, with both guys pumping 100% of the time.
There are other classes, but how useful they are for high fidelity is open to debate. Class C, for instance, is slap-full, rail-to-rail clipping...all the time. Now, what the hell good is that? Well, it turns out to be really cool if you're running a radio transmitter (or a PC, for that matter...t'ain't nuthin' but a square wave, and they're really nifty for timing purposes, but we ain't building computers, here). Class D chops the signal into a skillion little bits (digitizes it) and then more-or-less amplifies it as class C, then reassembles the pieces to make something like a sine wave again. Other classes exist, but they're based on sliding bias or other oddities that get pretty slippery when you want to implement them cleanly.

Grey
 
Douglas Self has a decent website which goes into a little more technical detail, and gives a very good overall view of class-B. If you read the section on power amplifier distortion, there's a bunch of stuff dealing with diff pairs and the VAS, and about half way down the article, you'll find the section dealing with class-b output stages.

Now, I have to warn you - Doug tends to take the opposite extreme position to Mr. Rollins, dismissing everything subjective without giving much consideration to the possiblity that his test instruments aren't showing all the effects experienced by the listener. Anyway, Mr. Self does present a good method for optimizing class B to remove as much distortion as possible. While at one time, class B distortion was poorly understood and handled (leading to terrible sounding amps), modern designs can effectivley implement class-B outputs with very little distortion. In fact, enough class-B distortion can be removed to (IMHO) render it nearly inaudible. I say <i>nearly</i> because there is a difference between how class-A and good class-B amps sound - which i think has something to do with non-monotonic distortion (distortion which decreases with amplitude, instead of increasing like it should). That said, I regularly listen to a class B amp which sounds positively wonderful.

Class-B is a worthwhile pursuit if you're willing to invest the extra time and attention to deal with it's ugly side. But, with proper design, construction, and testing, class-B will deliver a substantial gain in efficiency, bringing down the cost of power supply components and output devices/heatsinks, which together form the bulk of an amplifier's cost. So, if a high-power amp is your gig, then going class-B can save big bux. For low-powered amps, the difference is much less, and I'd recommend sticking with class-A.
 
Thanks for the info everyone.

I have been reading up on this quite a bit and believe I now understand the subject a lot better. One thing I have been thinking about also is the whole amp/speaker interface. This is what I have arrived at:

A speaker is a current controlled transducer which produces movement of the cone by F = Bil.

The speaker is a reactive load with increasing impedance with frequency.

All amplifiers that I have seen are voltage amplifiers with current handling capabilities up to (and often exceeding for class A) their maximum power rating.

Controlling the voltage across a reactive load to produce current as a by-product doesn't sound like a good solution to me. I would think that a low distortion current waveform should be supplied to the transducer and let the voltage do whatever over the frequency range of the system.

I would like to hear some comments about this. Am I missing the point or trying to over simplify things?

Cheers

Dan
 
Dan

Rather than repeating information available elsewhere, I will just draw your attention to the following articles at the ESP Audio Pages which cover current drive as opposed to voltage drive:

http://sound.westhost.com/z-effects.htm

http://sound.westhost.com/project56.htm

A thing to bear in mind is that all commercial speakers have been designed to work from normal commercial amplifiers and are therefore balanced for voltage drive.

Geoff
 
Dan,
Geoff's final point is quite correct. I tend to phrase it differently, however: If it ain't broke, don't fix it. Speakers work, hence the amps that drive them don't need fixin'. (That's not to say they couldn't use substantial fine tuning, though...)
Look at it this way.
At any given moment, IR=E *will* be satisfied. Period. No ifs, ands, or buts. If we substitute reactance for resistance, there's no change, really, we just need to start asking what frequency we're talking about. But, given that Ohm's law *must* be satisfied (even if we have to get picky about frequency), it doesn't matter whether you choose to regard it as a voltage that generates a current, or a current that generates a voltage. It's only semantics.
As a related point, people often try to claim that bipolar transistors are "current" devices, whereas tubes and MOS/FETs are "voltage" driven devices. It's a spurious distinction. The grid (tubes) or gate (MOS/FETs) is an inherently high impedance entry point compared to a bipolar base. Rather broadly speaking, the amount of voltage or current required to drive any of the three device types will be determined by Ohm's law in much the same way. (Don't nobody get fussy about the fact that we're talking about active devices rather than passive. I'm speaking in rough terms, here.)
There have been various commercial amps over the years that claimed to be "current" amplifiers as opposed to "voltage" amplifiers. Their ad copy usually made exactly the point that you made. I don't recall that any of them actually *sounded* significantly different, at least any more than you'd expect one amp to differ from another.

Grey
 
Grey/Geoff,

would I be correct in assuming that using impedance correction circuitry between the amp and speaker (i.e. notch filter for resonance peak and zobel network for rising impedance with frequency) to flatten the impedance/frequency curve of the driver would not produce any benefits/better sound quality. I am not interested in this circuitry as part of a crossover network but merely as a way of making a single driver load more amp friendly in the hope of adding to overall sound quality.

Cheers

Dan
 
Dan,
Life being what it is, there aren't too many free rides. The output network on a solid state amp is there for the stability of the circuit. To the extent that you might be able to get away without one then the circuit will be simpler and hopefully sound better, not to mention a dollar cheaper to build.
Note that the circuit is not a carved-in-stone sort of thing. Tube designs (even transformerless designs) do not have anything like it--they simply don't need it. There are also examples, such as Nelson's Alephs, that, though solid state, do not need them. (His A-75 is an example of one that does...)
Making a speaker easy to drive is another thing, entirely. I prefer a whip and chair, such as you see lion tamers use. Others recommend flame throwers, but I don't hold with such extreme measures.
Zobel networks are for the benefit of the crossover. A passive crossover will have a hard time of it if the driver impedance gets too far out of whack. On the other hand, if you biamp the amplifier will be quite happy to drive the speaker directly. Here we are assuming that the amplifier has sufficient current in reserve to deal with dips (peaks are no problem--the amp simply delivers less power). But given that passive crossovers always involve a certain amount of wasted power (called insertion loss), you'll end up ahead power-wise if you biamp.
Yet another approach--one that I'm fond of--is to use drivers that are purely resistive whenever possible.

Grey
 
Dan,
Planar drivers (I use Bohlender-Graebner RD-75s and the woofer panels from my old Magneplanar Tympani IVs) are good examples of purely resistive drivers. Ribbon tweeters are purely resistive as long as the resistance is not so low that you need a transformer to drive them. I use the ribbons out of my T-IVs, which are direct-drive (about a 3 ohm load).
If you're curious about the drivers, try these sites:
http://www.bgcorp.com
http://www.magneplanar.com
There are others, as well. Unfortunately, Strathern (Geoff, there's another company I miss...) is no longer with us, but there's a company called Orca (?) that makes a ribbon driver or two. Others will come to me, I'm sure, as soon as I've logged off. Carver had a ribbon driver, for instance.
If you're wondering about the quad-amped insanity I've cobbled together for speakers, poke around in one of these recent threads. Bryan asked, so I stuck references in there to the threads where I described what I'm using.
Beware...a number of companies call planar drivers ribbons as a marketing tool, but that doesn't make it so. A ribbon driver is free-floating on the sides, driven by magnets on the sides. A planar is bound on the sides (much like an electrostatic membrane), driven by magnets in front and back.

Grey
 
Dan,

I am a firm believer in active crossover networks, but I still use a Zobel network across the driver because it gives a more predictable result. Anytime you use a Zobel network with low order passive networks it usually helps. Sometimes Zobels in high order networks can have unusual results.
Grey may be able to shed some light on use of Zobels in high order passive networks. I for the most part use low order networks.

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