Theoretical question on amplifier design

Hak Foo

Active Member
As I understand it, normally amplifiers have to be built with (typically) pairs of output transistors, because one device can handle the positive voltages and one the negative. You end up either wasting a lot of energy running them to do nothing half the time (class A) or introduce switching noise cutting them off (class B/AB)

What would happen if you injected some DC into the signal to amplify, then removed it after amplification?

Instead of, say, a signal centered on 0 and ranging from -2v to +2v, it could range from 0 to 4v, centered around 2v.

Then you'd only need a "positive side" transistor to amplify that to, say, 0-30v centered around 15v. To get it back to a normal zero centered signal, a simple option would be to use a +15v reference as "ground" for the speakers.

This seems like it has some merit-- 50% fewer expensive output transistors, no switching distortion... but on the other hand you've got to keep the two offsets precisely tuned and if it goes wrong, you're running a huge DC offset by design. It also requires transistors that behave in a super linear way.

Has anyone ever tried a design like that? Or am I completely lost on how AC signals work?
 
As I understand it, normally amplifiers have to be built with (typically) pairs of output transistors, because one device can handle the positive voltages and one the negative. You end up either wasting a lot of energy running them to do nothing half the time (class A) or introduce switching noise cutting them off (class B/AB)

What would happen if you injected some DC into the signal to amplify, then removed it after amplification?

Instead of, say, a signal centered on 0 and ranging from -2v to +2v, it could range from 0 to 4v, centered around 2v.

Then you'd only need a "positive side" transistor to amplify that to, say, 0-30v centered around 15v. To get it back to a normal zero centered signal, a simple option would be to use a +15v reference as "ground" for the speakers.

This seems like it has some merit-- 50% fewer expensive output transistors, no switching distortion... but on the other hand you've got to keep the two offsets precisely tuned and if it goes wrong, you're running a huge DC offset by design. It also requires transistors that behave in a super linear way.

Has anyone ever tried a design like that? Or am I completely lost on how AC signals work?
A simple way to remove the DC offset is to run the output through a capacitor. Indeed, that's what capacitor-coupled amplifier designs do.

If I understand what you're describing -- and I'm not sure I do -- you're essentially describing Class A operation.

Switching distortion is a characteristic of Class B Push-Pull operation.

The solution is Class AB, which is effectively a combination of both, and smooths the crossover point to reduce switching distortion to a negligible amount. In this case, "negligible" means the switching distortion is lower than, or the same as, noise inherently present in any electrical system.

See https://www.electronics-tutorials.ws/amplifier/class-ab-amplifier.html for an explanation.
 
Another way of thinking about your proposal is that indeed you've only got one output transistor on the offset output, but the 15V reference "ground" would have to both sink and source current, thus requiring two output transistors to accomplish this. So, you've moved from two output transistors of the typical design to a total of three for your suggested design. I don't see the savings. Plus, the sink/source "ground" would have to be able to switch between these states smoothly, or else you'll be inducing the exact same switching distortion you're trying to prevent. Double whammy!
 
As I understand it, normally amplifiers have to be built with (typically) pairs of output transistors, because one device can handle the positive voltages and one the negative. You end up either wasting a lot of energy running them to do nothing half the time (class A) or introduce switching noise cutting them off (class B/AB)
Why don't you ask Nelson Pass on diyaudio.com? Are you aware he produces amplifiers using a single output device in his SIT series?
 
The OP is describing the quasi-complementary push-pull (solid state) amplifier technology.
This was used in the early days of solid state audio circuits because (EDIT) robust, reliable complementary PNP/NPN pairs of (output) transistors weren't yet available. Most - if not all - such amplifiers were capacitor coupled, and (to some ears, at least) the "deficits" of capacitor coupling had euphonic side-effects. YMMV.

In terms of crossover distortion, biasing the devices (in a "normal" complementary amplifier) into class AB can essentially eliminate it (albeit with some loss of efficiency relative to pure class B operation) -- but push-pull amplifiers will still (always, always) have a very different spectrum of harmonic distortion products than a single-ended amplifier (for better or for worse), because push-pull amplifiers cancel out even order harmonic distortion products of the fundamental. Comes with the territory.

That can have audible consequences.

Here's a pretty succinct and cogent description/explanation. The resource from which this comes is focused on vacuum tube amplification, but the topic at hand depends only on amplifier topology (SE vs. PP) and not on whether the amplification devices are tubes or transistors.


upload_2018-9-21_20-59-4.png

source:
http://www.x3mhc.no/dokumenter/SE-v-PP-Part1.pdf

There used to be a great write-up on the HH Scott quasi-complementary amps and receivers (specifically focusing on the Scott 260 amplifier), but even archive.org cannot find a copy :(
As it happens, I have an example of the kit version of the 260 (LK-60) and it is actually a decent sounding solid state amplifier.

HH Scott LK60 DSC_0789 by Mark Hardy, on Flickr

EDIT: Here's a little info about the quasi-complementary PP topology:
http://db.audioasylum.com/mhtml/m.html?forum=vintage&n=216510&highlight=quasi-complementary&r=&search_url=/default.mpl?searchtext=&b=AND&topic=&topics_only=N&author=&date1=&date2=&slowmessage=quasi-complementary&ip=&sort=score&sortOrder=DESC&sortRank=Forum&forum=ALL
 
Last edited:
Heh... notice that Scott themselves in a contemporary ad* refer to the [EDIT] 260 as direct coupled. It is not, at least in the sense of the term as we understand it today, but it is also not transformer coupled, in either the driver or output stages -- in those days, interstage transformers were nearly ubiquitous in solid state amplifiers.

As the ad indicates, the quasi-complementary output stage has its DC removed by fairly large (ca. 3000 uF, if memory serves) output capacitors.

upload_2018-9-21_21-6-42.png

___________
*
source: https://www.americanradiohistory.com/Archive-Audio/60s/Audio-1965-03.pdf
 
Last edited:
Speaking of one output device per channel -- I am listening to a Class A, single ended solid state amplifier even as I type this (and FWIW) -- the Nelson Pass designed "Amp Camp Amp" (albeit a slightly tarted up variant of one).

:)

http://www.firstwatt.com/pdf/art_amp_camp_1.pdf

upload_2018-9-21_21-23-33.png

Q1 in the schematic is the output transistor -- there are a total of four transistors per channel in the whole circuit, though! :) Read the PDF for details.
Note also that the output is capacitor-coupled just like the 1960s push-pull quasi-complementary HH Scott amplifier, and for the same reason (to remove DC from the amplified output). A 3300 uF capacitor is used in the "ACA".

DSC_2871 (2) by Mark Hardy, on Flickr

DSC_6923 (2) by Mark Hardy, on Flickr
 
you'd only need a "positive side" transistor to amplify that to, say, 0-30v centered around 15v. To get it back to a normal zero centered signal, a simple option would be to use a +15v reference as "ground" for the speakers.
What you are describing sounds like normal transistor Bias. To eliminate the "crossover distortion" caused by the time it takes a transistor's Base-Emitter junction to conduct, a DC voltage is applied to the output transistors to just get them to conduct slightly all the time. This eliminates the time delay and thus the distortion.
 
As I understand it, normally amplifiers have to be built with (typically) pairs of output transistors, because one device can handle the positive voltages and one the negative. You end up either wasting a lot of energy running them to do nothing half the time (class A) or introduce switching noise cutting them off (class B/AB)

What would happen if you injected some DC into the signal to amplify, then removed it after amplification?

Instead of, say, a signal centered on 0 and ranging from -2v to +2v, it could range from 0 to 4v, centered around 2v.

Then you'd only need a "positive side" transistor to amplify that to, say, 0-30v centered around 15v. To get it back to a normal zero centered signal, a simple option would be to use a +15v reference as "ground" for the speakers.

This seems like it has some merit-- 50% fewer expensive output transistors, no switching distortion... but on the other hand you've got to keep the two offsets precisely tuned and if it goes wrong, you're running a huge DC offset by design. It also requires transistors that behave in a super linear way.

Has anyone ever tried a design like that? Or am I completely lost on how AC signals work?


What you are suggesting does not change the situation. You still need pullup and pulldown whether its to ground and B+ or +/- supplies of equal or unequal voltages. In reference to other posts here the first practical power amps used ground and B+ only. The output point was set to 1/2 B+ and a capacitor was used to couple out the signal and block the DC. Later and present amplifiers tend to use dual +/- supplies to eliminate the large coupling capacitor.

One might consider using a high frequency bias like we do on a tape recorder to get past the dead zone and then filter it out. But DC bias is so easy why bother. I dont think there is any way around it, though you aren't the only person thinking about it.
 
What you are describing sounds like normal transistor Bias. To eliminate the "crossover distortion" caused by the time it takes a transistor's Base-Emitter junction to conduct, a DC voltage is applied to the output transistors to just get them to conduct slightly all the time. This eliminates the time delay and thus the distortion.

You are correct that the output transistors are conducting all the time, bias current is typically 20-50 mA. It does reduce crossover distortion. However this has nothing to do with time or time delay.
 
Tim de Parivicini, the UK "Valve God", launched a single-ended one-transistor amp about 20 years ago (cover of HFN) to capitalize on the SET trend. Reviews were insanely great. So was the price. Unfortunately I don't recall the design details.
 
Back
Top Bottom