Crossover notch distortion

[Jim]

From the Understanding Amplifier Classes article, I noticed "Let me stress, however, that Class A/B amplifiers can sound really really good. That crossover notch was our one major problem, and we have entirely eliminated it."

Do I read that as Emotiva class A/B amplifiers don't suffer from crossover notch distortion? I thought that all A/B amplifiers did (I'm not basing that on anything concrete, that's just what I thought).

[audiofile]

I think that is correct simply because the Class A/B amplifiers, by reason of their design, eliminated the crossover notch distortion problem.

What KeithL may have meant is that "...and we have entirely eliminated it [by utilizing the Class A/B design for our amplifiers.]

[Jim]

Mar 18, 2013 14:25:12 GMT -5 audiofile said:

I think that is correct simply because the Class A/B amplifiers, by reason of their design, eliminated the crossover notch distortion problem.

What KeithL may have meant is that "...and we have entirely eliminated it [by utilizing the Class A/B design for our amplifiers.]

I don't think that's true, I think that you can have it in Class A/B --- but eliminate it with careful design.

Like here for example, it illustrates it with a class A/B Push/pull:

www.aikenamps.com/CrossoverDistortion.htm

And here:
en.wikipedia.org/wiki/Crossover_distortion

This is interesting (from wiki):
"In the case of a class B/AB amplifier, crossover distortion can be reduced by using a slight forward bias in the base circuit such that the transistors are idling at a small output current. The forward bias causes the circuit to operate in class-AB mode, so both transistors are slightly on during crossover. This can reduce or eliminate the characteristic kink of crossover distortion, although other types of crossover distortion will remain."

The only reason I thought to ask, was that the XPA-1L page lists "Zero crossover distortion", which is true to Class A, but it made me think that the other Emo Class A/B amps had it then (Not sure if that makes sense!)

[Chuck Elliot]

Mar 18, 2013 14:31:41 GMT -5 Jim said:

Mar 18, 2013 14:25:12 GMT -5 audiofile said:

I think that is correct simply because the Class A/B amplifiers, by reason of their design, eliminated the crossover notch distortion problem.

What KeithL may have meant is that "...and we have entirely eliminated it [by utilizing the Class A/B design for our amplifiers.]

I don't think that's true, I think that you can have it in Class A/B --- but eliminate it with careful design.

Like here for example, it illustrates it with a class A/B Push/pull:

www.aikenamps.com/CrossoverDistortion.htm

And here:
en.wikipedia.org/wiki/Crossover_distortion

This is interesting (from wiki):
"In the case of a class B/AB amplifier, crossover distortion can be reduced by using a slight forward bias in the base circuit such that the transistors are idling at a small output current. The forward bias causes the circuit to operate in class-AB mode, so both transistors are slightly on during crossover. This can reduce or eliminate the characteristic kink of crossover distortion, although other types of crossover distortion will remain."

The only reason I thought to ask, was that the XPA-1L page lists "Zero crossover distortion", which is true to Class A, but it made me think that the other Emo Class A/B amps had it then (Not sure if that makes sense!)

Some CND will still be there in a traditional Class A/B amp stage. It is partly the purpose of negative feedback to remove that remainder.

[audiofile]

And as stated "In the case of a class B/AB amplifier, crossover distortion can be reduced by using a slight forward bias in the base circuit such that the transistors are idling at a small output current." this is what the article goes on to discuss how that notch distortion is eliminated. Although, I believe a positive and negative bias is needed to truly achieve full elimination of the distortion.

As I read that my question was, "Is this distortion noticeable and if so is it only at lower power levels?"

[Chuck Elliot]

The issue is that a bi-polar transistor need .7 volts between the base and emitter just to reach the start of conduction. Now I'm no amp designer, but I'm sure ways exist to keep that B-E threshold voltage such that it doesn't need to be overcome.

Regardless, transitioning operation from one device to another is not going to be perfectly linear.

Feedback is not bias. Feedback applies a small amount of the output back to the input which negates the error. This feedback can be local to just one stage of amplification or global from the last stage to the first!

[KeithL]

Now we're getting into the advanced lessons.....

First, the article you referenced is about tube amps, which do act a bit differently. It is true, however, that running in Class A/B doesn't automatically eliminate all possibility of some sort of "crossover distortion" at the points where things shift from one output device conducting to both.

Even Class A doesn't eliminate all distortion - because not all distortion is symmetrical in a way that causes it to cancel out. (It is a common misconception that, because the devices are operated asymmetrically, all distortions will cancel out. This is an oversimplification and is simply not true. A lot of the distortions cancel out to a significant degree - but not all, and not completely.)

However, there is a specific type of crossover distortion known as crossover NOTCH distortion which is characteristic of Class B amps, and which is caused by the two devices actually having a discontinuity where one stops conducting before the other one starts. This specific distortion was quite common in early Class B amps (it was difficult to make sure that the two halves NEVER conducted at the same time, but nonetheless conducted right down to zero volts). This type is especially nasty because it happens right where the signal is crossing zero, is often quoted as "what's wrong with Class B", and is entirely eliminated in Class A/B.

I think it is fair to say that moving to Class A/B changed crossover distortion from "a major problem that was very difficult to solve effectively" to "a minor problem that is easy to deal with very effectively by proper design". Since, in Class A/B, neither device stops conducting entirely, you never get that "hard corner" or "gap". Also, since the point where each device basically reaches it's minimum amount of conduction is NOT at zero volts, and the points for the two devices are in different places, you have less non-linearity at those points, it is in different places for the two devices, and it isn't happening at the zero crossing point. All of this makes it much less of a problem, and much easier to deal with. As you say, it still takes good design to get it entirely right.

Mar 18, 2013 14:31:41 GMT -5 Jim said:

Mar 18, 2013 14:25:12 GMT -5 audiofile said:

I think that is correct simply because the Class A/B amplifiers, by reason of their design, eliminated the crossover notch distortion problem.

What KeithL may have meant is that "...and we have entirely eliminated it [by utilizing the Class A/B design for our amplifiers.]

I don't think that's true, I think that you can have it in Class A/B --- but eliminate it with careful design.

Like here for example, it illustrates it with a class A/B Push/pull:

www.aikenamps.com/CrossoverDistortion.htm

And here:
en.wikipedia.org/wiki/Crossover_distortion

This is interesting (from wiki):
"In the case of a class B/AB amplifier, crossover distortion can be reduced by using a slight forward bias in the base circuit such that the transistors are idling at a small output current. The forward bias causes the circuit to operate in class-AB mode, so both transistors are slightly on during crossover. This can reduce or eliminate the characteristic kink of crossover distortion, although other types of crossover distortion will remain."

The only reason I thought to ask, was that the XPA-1L page lists "Zero crossover distortion", which is true to Class A, but it made me think that the other Emo Class A/B amps had it then (Not sure if that makes sense!)

[KeithL]

Actually that 0.7 volts varies from transistor to transistor, and shifts - significantly - when a transistor warms up. This means that it is actually very difficult to maintain a transistor precisely at the point where it is "just ready to turn on". This was the original theory about how things should be done, and the crossover notch was the result of the fact that you couldn't actually get it just right and keep it that way. Also, while you are actually in that "null zone" where neither transistor is conducting, feedback doesn't work properly, because there is no signal to apply it to.... so you are actually in a small but quite nasty nonlinear zone of operation....

[With modern designs and parts it is possible to make a very good Class B amplifier, but it has no significant benefits over Class A/B, and it's much easier to get an equal or better result with Class A/B.]

Actually, your summary works quite well....

Transitioning from one device to another is always going to be somewhat nonlinear. In Class A, the operating range is in essence "the transition zone" - so it is spread out as much as possible and is the smoothest. In Class B, the transition zone is a single point, which is the worst case scenario. In Class A/B, the transition zone is spread out (not as much as with Class A, but still a huge improvement over Class B).

Mar 18, 2013 15:06:42 GMT -5 Chuck Elliot said:

The issue is that a bi-polar transistor need .7 volts between the base and emitter just to reach the start of conduction. Now I'm no amp designer, but I'm sure ways exist to keep that B-E threshold voltage such that it doesn't need to be overcome.

Regardless, transitioning operation from one device to another is not going to be perfectly linear.

Feedback is not bias. Feedback applies a small amount of the output back to the input which negates the error. This feedback can be local to just one stage of amplification or global from the last stage to the first!

[Chuck Elliot]

Thanks for the reply Keith.

[KeithL]

We need to be very careful about terminology here if we want to generalize.

ALL active devices (tubes, transistors) produce come distortion.

With transistors, the lowest amount of distortion tends to be when they are biased rather heavily on. This is the operating point normally used for Class A amplifiers.

In a Class B amplifier, each transistor shuts entirely off at the crossover point. This happens to be about the worst place for distortion. You also have the possibility that each one won't turn on at exactly the point where the other turns off, leaving an actual gap or step in the waveform (this is the "crossover notch".)

In a Class A/B amplifier, each transistor is biased such that it is operating well out of the range where it will distort the most, and safely away from the possibility of shutting off before its partner is ready to take over.

Transistors each only conduct in ONE DIRECTION, so "positive and negative bias" is a meaningless term. What we are doing is biasing both the positive and negative transistor to be slightly on all the time. (With bipolar transistors you do this by allowing current to flow into or out of the base; with FETs you use a voltage; with tubes it is a voltage.) Whether the actual current or voltage will be negative or positive depends on the type of device we are using, but it will always be in the same direction for each.

As for your question:

Distortion occurs at all signal levels (some distortions vary with signal levels and others do not) Since crossover notch distortion happens at the zero crossing point, and is mostly independent of signal level, it is especially noticeable when there is very little signal. (When there is no signal, there is nothing to distort, but at very low signal levels, the notch could make up 50% of the signal, at which point the THD would be 50%.)

Most amplifiers show increasing THD+N at very low power levels, but most of that is noise (background hiss). An amplifier with bad notch distortion will actually have rising distortion levels at low power levels - enough that it will sound quite nasty when playing quietly. This was a problem with many very early Class B designs (it was often described as "getting gritty sounding when you played them quietly"). There are many other distortion mechanisms, and some get worse at high signal levels while others do not. For most well-designed Class A/B amplifiers, none of the forms of distortion that remain are especially worse at low or high power levels..... if you look at one of our AP reports, you will see that distortion remains very low over a wide range, rising slowly and steadily until you reach full power. At that point it rises sharply when the amplifier actually starts to clip. (The rise you see at the bottom of the THD+N graph at VERY low power levels is more noise than distortion.)

Mar 18, 2013 14:46:59 GMT -5 audiofile said:

And as stated "In the case of a class B/AB amplifier, crossover distortion can be reduced by using a slight forward bias in the base circuit such that the transistors are idling at a small output current." this is what the article goes on to discuss how that notch distortion is eliminated. Although, I believe a positive and negative bias is needed to truly achieve full elimination of the distortion.

As I read that my question was, "Is this distortion noticeable and if so is it only at lower power levels?"

[Jim]

Thanks Keith, that really helps!

I didn't realize I was mixing up "Crossover distortion" with "crossover notch distortion"

Fascinating stuff, I appreciate you taking the time to write out such a clear, detailed explanation!

[mgbpuff]

O.K. Here's a question - how does differential circuitry as in the XPS-1 differ from Class B? I think, but I'm not sure, that the answer is that differential is technically Class B, but unlike a power amp, a preamp does not couple to an inductive load with EMF kickback which is the real reason for crossover notch distortion. Right or wrong?

[KeithL]

You can go HERE to read about amplifier classes....

emotivalounge.proboards.com/index.cgi?board=magazine&action=display&thread=28322

In summary, in a Class B design the positive rail transistors switch off at "exactly" the same time as the negative rail transistors switch on, and vice versa. In theory this is not an issue but, unless they transition perfectly, there is a "notch" in the waveform where they don't quite line up. That is "crossover notch distortion". In Class A/B, the conduction zones "overlap" a little, which prevents this notch from occurring. In a Class A design, both transistors are on all the time (complete overlap), which is even more linear.

A fully differential amplifier means that you pretty much have two entirely separate amplifiers amplifying equal and out-of-phase signals, which cancels out certain forms of noise and distortion. An amplifier can be entirely differential, or only certain parts of it can be differential (like a differential input) - which still gives you certain benefits. A differential amplifier can be Class A, Class B, or Class A/B, and any class of amplifier can be differential - or not.

Most preamps are actually Class A, since they operate at low power levels and so it's simple to run them that way. An inductive load is more difficult to drive, and so may aggravate certain types of distortion in certain amp designs, but it is not the cause.

Apr 30, 2013 10:48:36 GMT -5 mgbpuff said:

O.K. Here's a question - how does differential circuitry as in the XPS-1 differ from Class B? I think, but I'm not sure, that the answer is that differential is technically Class B, but unlike a power amp, a preamp does not couple to an inductive load with EMF kickback which is the real reason for crossover notch distortion. Right or wrong?

[mgbpuff]

Thanks - guess I was confusing differential with Class B. Do the differential amplifier paths in the XPS-1 operate Class A?

[Deleted]

I sure as hell hope there is not going to be an exam on this! ;D

Keith is an absolute wealth of information and then some. Thanks!

[AudioHTIT]

Apr 30, 2013 11:16:35 GMT -5 @chuckienut said:

I sure as hell hope there is not going to be an exam on this! ;D

There is Chuckie, and you want to get an A in the Class, or maybe an A/B, but not a B.

[garbulky]

0Keith while we have your attention. Does a fully balanced amp from a fully balanced source yada yada basically mean that all distortion (analog wise) is cancelled out or at least to a large degree?

[KeithL]

The details are actually rather complicated....

With a fully balanced amplifier, a significant portion of the distortion generated BY THAT AMPLIFIER (by the two identical "halves" of the amplifier) will cancel out. The exact amount that cancels out will depend on the details of the distortion itself and of the circuitry involved. (Symmetrical low-order nonlinearities tend to cancel very well; random noise and higher-order nonlinearities not so well; random noise not well at all.)

[Note that, if the amplifier is already very low in distortion, there may not be anything significant, and of the right types, left to cancel out. Also note that it isn't going to magically eliminate distortion produced by something upstream in the signal chain.]

A balanced connection and a differential input and output together produce an almost entirely different (but also good) effect. What you get there is an almost total immunity to most of the noise and interference picked up by the interconnect. (Any noise that is picked up equally by both lines in the balanced interconnect is cancelled out. Balanced cables are usually twisted tightly together because this only works if both signal lines pick up exactly the same noise so it can cancel out properly. This works well to eliminate things like hum and other "interference", but doesn't do much for thinks like plain old random hiss on the input or output. You also sometimes gain a slight lowering of the noise floor - depending on the design and the other circuitry involved - but this is only 4-5 dB at best.

Each of these effects works by itself, and they complement each other to produce a low distortion and low noise signal path.

NOTHING will ensure that ALL distortion or noise is eliminated, and neither of these is 100% effective at delivering its benefits.... but both can help to a significant degree (but, like I said, the exact amount will depend on the details). A fully differential amplifier will usually be at least somewhat lower in distortion, but the difference may not be audible, and a balanced connection will usually help IF you have a noise problem but, if you don't have a noise problem, then there's nothing for it to help with...

Apr 30, 2013 12:04:55 GMT -5 garbulky said:

0Keith while we have your attention. Does a fully balanced amp from a fully balanced source yada yada basically mean that all distortion (analog wise) is cancelled out or at least to a large degree?

[garbulky]

Very informative Keith. Thank you for your time.

[KeithL]

Actually, in the XSP-1, we chose to go with Class A/B for most of the signal path.... it just worked out that way
(and we're quite pleased with the results)

Apr 30, 2013 11:15:49 GMT -5 mgbpuff said:

Thanks - guess I was confusing differential with Class B. Do the differential amplifier paths in the XPS-1 operate Class A?