Class A vs AB vs D amplifiers

[SSS]

I have similar Tek scopes and Agilent digital scope. If there is a gross crossover distortion it may be seen, tiny ones not. Distortion may not be a spike or irregularity in the waveform like a slight edge. I mean just a deviation of the ideal sine wave form.

 

[SSS]

Would it be generally accurate to say that there's more difference between amps due to the rest of the implementation, rather than whether the core technology is B or D? I also suspect much of what we hear is beyond the range of what we have statistical instruments for, otherwise there's no sense in how amps sound different despite having equally good stats and being run within their power ranges.

Yes, much depends on the final design and implementation including component selection.

 

[Doodski]

Would it be generally accurate to say that there's more difference between amps due to the rest of the implementation, rather than whether the core technology is B or D? I also suspect much of what we hear is beyond the range of what we have statistical instruments for, otherwise there's no sense in how amps sound different despite having equally good stats and being run within their power ranges.

The frequency range of what you might hear of let's say 20Hz to 20kHz (A fantasy number but it doesn't matter here for my purpose.) That frequency range is very low frequency stuff and meters can easily meter that signal. You are trying to make excuses for this and that and your ideas simply do not fit reality.

 

[Doodski]

I have similar Tek scopes and Agilent digital scope. If there is a gross crossover distortion it may be seen, tiny ones not. Distortion may not be a spike or irregularity in the waveform like a slight edge. I mean just a deviation of the ideal sine wave form.

I've seen a lot of people claim to have Tek lab scope(s) and they in the end have those pedestrian cheap Tek models with no dual time base that have a very ugly trace and can't show what we discuss. Do you have a actual lab scope or the cheaper variety?

 

[oleg87]

NO. Since as I said there is no perfect amp available. Each class has its benefits but also drawbacks. But one idea, why there are at least to my knowledge no preamplifiers in Class-D technology if this type is called better than A or AB.

Well, because that's a silly idea. The inherent advantage of class D is efficiency. You can make it perform very well, as we see, but it requires more design effort and expertise to make it do so than linear topologies. Since a preamp does not need to deliver power, a class D implementation would be a pointless exercise.

 

[DonH56]

My statement was a more theoretical one. If the amplification factor is kept linear over level so class-A has no switchower effects nor any artefacts from switched transistors like at class-D or at class AB where there is a transient region. Since I have no access to high-priced class-A power amps I can't compare or judge them. There are technical books available covering the amplifier classes with their goods and bads. It is true on the other hand when power is needed at relatively low cost then a Hypex and a Purife module may serve well.

Class A, like any amplifier, will have other distortion products even if there is no crossover distortion.

Class AB amplifiers are class A at low levels, so no crossover distortion at low power levels, and as the signal moves to higher levels and thus class B operation, feedback reduces the distortion. Some of the best-measuring amplifiers utilize class AB cores (e.g. Benchmark). "Switchover" effects are grossly misrepresented in marketing literature IME, just like the influence of cables and other issues.

Class D is basically a PWM signal so crossover distortion in the conventional sense does not occur; the signal is constantly switching and then filtered to produce a continuous output. Same as happens with a delta-sigma DAC. Modern designs using self-oscillating designs incorporate feedback well above most conventional A or AB designs, resulting in much lower distortion.

NO. Since as I said there is no perfect amp available. Each class has its benefits but also drawbacks. But one idea, why there are at least to my knowledge no preamplifiers in Class-D technology if this type is called better than A or AB.

The advantage of class D operation is high efficiency at high power, something not of concern for a preamplifier. Class A is also arguably the simplest way to design the preamplifier's circuits. Conventional design serves well and is what most manufacturers know. I also suspect there would be consumer pushback based upon (falsely) poor perception of class D performance, as you exhibit for power amplifiers. All discrete device preamplifiers I have seen operate with class A circuits, not because AB or D doesn't work, but because class A is generally a simpler design, works adequately, and at such low power levels nothing more is required. Note that many op amps, often used in preamplifiers and the low-level circuits in a power amplifier, include class AB output stages within the IC.

I admit that with a regular oscilloscope you may not have seen crossover signal degrading. But an oscilloscope is not able to display tiny distortions in the waveform what we talk about. Digital scopes have a resolution of 8 bit data witdh, very good ones 12 bit. There may today be now ones with more than 12 bit, I did not check Tektronix, (HP) Keysight and Rohde&Schwarz or LeCroy products for this.

When using an oscilloscope for this you are looking at a very small signal, either a very low output level, or a large signal with the fundamental notched out so you can observe fine details of the waveform. Most high-speed 'scopes are 8 bits, but low-speed 'scopes are usually 12 to 16 bits, with a few incorporating 24-bit delta-sigma ADCs. But with such low signal levels that is usually more than enough. Not if you wanted to measure 100+ dB distortion on a full-scale waveform, but when the waveform is only a fraction of maximum, a 'scope can provide a very good look. With a signal of say 100 mVpp (about 0.16 mW into 8 ohms), an 8-bit converter can resolve about 400 uV (about 2.5 nW into 8 ohms), and a 12-bit converter can resolve about 24 uV (~9 pW). Just how low do you want to go?

 

[MAB]

Class AB amplifiers are class A at low levels, so no crossover distortion at low power levels

This is a great point, and should be considered by the the 'first-watt' crowd, who seem to find deficiencies in AB design that only manifest in the 'first-watt', despite the fact that the first watt or so is where class AB operate exactly the same as Class A. It's ironic.

 

[MAB]

Would it be generally accurate to say that there's more difference between amps due to the rest of the implementation, rather than whether the core technology is B or D?

Yes.

I also suspect much of what we hear is beyond the range of what we have statistical instruments for,

No, actually the opposite.

otherwise there's no sense in how amps sound different despite having equally good stats and being run within their power ranges.

Actually, this is a great old article, from the days when these topologies were actively discussed, and we understood the interaction of speaker impedance with amps, and the need to do blind tests.
See page 78.
https://americanradiohistory.com/Archive-HiFI-Stereo/80s/HiFi-Stereo-Review-1987-01.pdf
I keep posting this since I recall when this test was published. I was working at a nice shop that sold Levinson amps including the monoblocks in this test, and I had already been convinced that this business of 'veils' being lifted, night and day differences due to cables, etc. was bogus. Like a circus.

The rest of the magazine is really fun to read.

 

[wwenze]

Is there a quick and dirty method to see how long a class AB amp stays class A? Using bias current? Let's say the bias current is 20mA therefore as long as my output current doesn't go beyond that I am staying in class A? On a speaker amp that would put us at around 0.003W or 30-40dB SPL before it switches to class B.

 

[SSS]

I've seen a lot of people claim to have Tek lab scope(s) and they in the end have those pedestrian cheap Tek models with no dual time base that have a very ugly trace and can't show what we discuss. Do you have a actual lab scope or the cheaper variety?

I have Tek 7633 with various plug-ins one of is with differntial input and Tek 454A as well an Agilent DSO5014A (4 traces). Further older Hameg and HP scopes. No, I don't have a very actual hi-end scope model since they are really expensive and I have no real use for them. My maximum frequency to measure is 140 MHz.

 

[SSS]

Is there a quick and dirty method to see how long a class AB amp stays class A? Using bias current? Let's say the bias current is 20mA therefore as long as my output current doesn't go beyond that I am staying in class A? On a speaker amp that would put us at around 0.003W or 30-40dB SPL before it switches to class B.

Seems to be right. One of my power amps is the Yamaha M-85. It has a switch to activate class-A setting. With this the heatsinks get pretty warm which means the bias current is much more than the usual 20 to 40 mA. Yet I didn't measure the current in this mode.

 

[SSS]

Class A, like any amplifier, will have other distortion products even if there is no crossover distortion.

Class AB amplifiers are class A at low levels, so no crossover distortion at low power levels, and as the signal moves to higher levels and thus class B operation, feedback reduces the distortion. Some of the best-measuring amplifiers utilize class AB cores (e.g. Benchmark). "Switchover" effects are grossly misrepresented in marketing literature IME, just like the influence of cables and other issues.

Class D is basically a PWM signal so crossover distortion in the conventional sense does not occur; the signal is constantly switching and then filtered to produce a continuous output. Same as happens with a delta-sigma DAC. Modern designs using self-oscillating designs incorporate feedback well above most conventional A or AB designs, resulting in much lower distortion.


The advantage of class D operation is high efficiency at high power, something not of concern for a preamplifier. Class A is also arguably the simplest way to design the preamplifier's circuits. Conventional design serves well and is what most manufacturers know. I also suspect there would be consumer pushback based upon (falsely) poor perception of class D performance, as you exhibit for power amplifiers. All discrete device preamplifiers I have seen operate with class A circuits, not because AB or D doesn't work, but because class A is generally a simpler design, works adequately, and at such low power levels nothing more is required. Note that many op amps, often used in preamplifiers and the low-level circuits in a power amplifier, include class AB output stages within the IC.


When using an oscilloscope for this you are looking at a very small signal, either a very low output level, or a large signal with the fundamental notched out so you can observe fine details of the waveform. Most high-speed 'scopes are 8 bits, but low-speed 'scopes are usually 12 to 16 bits, with a few incorporating 24-bit delta-sigma ADCs. But with such low signal levels that is usually more than enough. Not if you wanted to measure 100+ dB distortion on a full-scale waveform, but when the waveform is only a fraction of maximum, a 'scope can provide a very good look. With a signal of say 100 mVpp (about 0.16 mW into 8 ohms), an 8-bit converter can resolve about 400 uV (about 2.5 nW into 8 ohms), and a 12-bit converter can resolve about 24 uV (~9 pW). Just how low do you want to go?

Your view is alright. With hi-resolution ADC scopes it comes to the behaviour of audio ADC which can be and is used as low frequency scope with a PC program. But limited to around 80 kHz. Regular lab scopes need to display much higher frequencies so the resolution is lower.

 

[Doodski]

I have Tek 7633 with various plug-ins one of is with differntial input and Tek 454A as well an Agilent DSO5014A (4 traces). Further older Hameg and HP scopes. No, I don't have a very actual hi-end scope model since they are really expensive and I have no real use for them. My maximum frequency to measure is 140 MHz.

I have handled many Tek lab scopes in 3 slot and 4 slot configurations but I never had a Tek analogue storage scope. I had a nice HP analogue storage scope instead. Your Tek 7633 is a 100MHz mainframe dependent on the vertical amp(s) used in it. Apparently something has been slowed down a bit with the time base (The 7633 has a reduced-scan mode added to boost stored writing speeds at very slow CRT trace rates.) to make the analogue storage work better. I think this scope with a dual time base 7B53A plug in will magnify a waveform so much that you'll love it. I had the Tek 7603 a 100MHz to 110 MHz mainframe with the same vertical amps as yours takes and the same 7B53 time base as your Tek 7633 uses. Mine was calibrated and the engineer said it calibrated to 150MHz without fault. So I'm thinking maybe yours can do that sort of bandwidth too. All in all the Tek 7000 series lab scopes are still in demand and as one can see ebay has lotsa business in Tek, HP etc. If I where you I would hang onto it and see if it needs a control cleaning (With 100% anhydrous isopropyl alcohol as the contacts are all gold in the 7000 series and the vertical amps and time base have cam switches with silicone rubber contact suspensions instead of wipers so they will last a long long time and you can spin the control as fast as you want and not damage the switch assembly.) Tek engineered the daylights out of the 7000 series and the prices reflected that accordingly. Another side note; It appears the Tek came with a very very fine metal mesh filter for the front of the CRT and that is to prevent bombarding you with harmful energy from the CRT when you sit in front of it and operate it in analogue storage mode. If you have that screen take good care of it because they are expensive and a critical part of the safe operation of the analogue storage function.

 

[Doodski]

Is there a quick and dirty method to see how long a class AB amp stays class A? Using bias current? Let's say the bias current is 20mA therefore as long as my output current doesn't go beyond that I am staying in class A? On a speaker amp that would put us at around 0.003W or 30-40dB SPL before it switches to class B.

If me I would set up the amp for a test, offset calibration and the bias calibration. Then having the oscilloscope connected to the output and I would also be metering the bias with a meter and I would adjust the bias till the crossover distortion becomes evident @ the scope and then set the bias to calibration level and see what that is for current. That is actually how I saw if bias made a difference on a class AB amp. I just turned the bias control and observed the reaction while monitoring the bias with a meter.

 

[pma]

Is there a quick and dirty method to see how long a class AB amp stays class A? Using bias current? Let's say the bias current is 20mA therefore as long as my output current doesn't go beyond that I am staying in class A? On a speaker amp that would put us at around 0.003W or 30-40dB SPL before it switches to class B.

P(A) = 2*R*Iq^2

R … load impedance
Iq … idle current

In your case, for 20mA and 4ohm, yes it is 3.2mW in class A.

With measurements you will see rise of high order harmonics when transition from class A to AB.

 

[antcollinet]

I also suspect much of what we hear is beyond the range of what we have statistical instruments for,

Your suspicion is seriously incorrect. Our measurement instruments are orders of magnitude more sensitive than our hearing, and can measure frequencies vastly outside the range we can hear also.

 

[antcollinet]

In your case, for 20mA and 4ohm, yes it is 3.2mW in class A.

Don't you get a higher output current in class A than the bias current?

 

[Doodski]

Would it be generally accurate to say that there's more difference between amps due to the rest of the implementation, rather than whether the core technology is B or D?

Can you please rephrase your question and make it a little less general. I am trying to avoid mistaking what you want to know and telling you what I think you meant...LoL.

I also suspect much of what we hear is beyond the range of what we have statistical instruments for

OK. I will put things into perspective for you.

When I was a bench tech I used:
- a 100kHz bandwidth differential input HP analogue storage scope.
- a 500MHz rack mount 7000 series Tektronix lab scope.
- a 100MHz 7000 series Tektronix lab scope.
- a cheap Kenwood 40 MHz scope.

The metrology shop where I bought my Tektronix scope and HP scope gear was selling into the GHz bandwidth oscilloscopes.

Now... If your hearing is very very good you might hear from 20Hz to 20kHz

When I am using a 500,000,000 Hz bandwidth oscilloscope for 20,000 Hz maximum frequency signal the test gear is barely working at all and is nowhere near it maximum capability.

Then there are GHz scopes. It looks like this.
1,000,000,000 Hz capable scopes being used to measure 20,000 Hz.

Do you see now how your idea about the test equipment not being able to measure all of what we can hear is not an accurate belief?

 

[SSS]

I have handled many Tek lab scopes in 3 slot and 4 slot configurations but I never had a Tek analogue storage scope. I had a nice HP analogue storage scope instead. Your Tek 7633 is a 100MHz mainframe dependent on the vertical amp(s) used in it. Apparently something has been slowed down a bit with the time base (The 7633 has a reduced-scan mode added to boost stored writing speeds at very slow CRT trace rates.) to make the analogue storage work better. I think this scope with a dual time base 7B53A plug in will magnify a waveform so much that you'll love it. I had the Tek 7603 a 100MHz to 110 MHz mainframe with the same vertical amps as your takes and the same 7B53 time base as your Tek 7633 uses. Mine was calibrated and the engineer said it calibrated to 150MHz without fault. So I'm thinking maybe yours can do that sort of bandwidth too. All in all the Tek 7000 series lab scopes are still in demand and as one can see ebay has lotsa business in Tek, HP etc. If I where you I would hang onto it and see if it needs a control cleaning (With 100% anhydrous isopropyl alcohol as the contacts are all gold in the 7000 series and the vertical amps and time base have cam switches with silicone rubber contact suspensions instead of wipers so they will last a long long time and you can spin the control as fast as you want and not damage the switch assembly.) Tek engineered the daylights out of the 7000 series and the prices reflected that accordingly. Another side note; It appears the Tek came with a very very fine metal mesh filter for the front of the CRT and that is to prevent bombarding you with harmful energy from the CRT when you sit in front of it and operate it in analogue storage mode. If you have that screen take good care of it because they are expensive and a critical part of the safe operation of the analogue storage function.

Thank You for the tips. You seem to have much more experience with Tek than I. I worked for HP, so I knew the hp scopes like the 1740A and storage version 1741A. Now I am long time retired and thus have no actual grip to the Keysight products.

 

[Doodski]

Thank You for the tips. You seem to have much more experience with Tek than I. I worked for HP, so I knew the hp scopes like the 1740A and storage version 1741A. Now I am long time retired and thus have no actual grip to the Keysight products.

I am familiar with those HP models as well. When I was a tech specializing in mechatronics I worked part time at a metrology lab that also bought, refurbished and calibrated test equipment for resale. The owner would go to large corporations' auctions in the USA and buy used operational top name brand test gear and after he got it calibrated and ready for sale he was averaging about 500%+ markup. Very lucrative operation. The HPs scopes that you have use a fine trace so the resolution can be pretty detailed as compared to a fat bright blurry trace found on inferior scopes. Those HP scopes are fine pieces of gear.