Class A vs AB vs D amplifiers

[voodooless]

Multitone is a good test, I'd like to see what happens when you switch off any one of the carriers to see what is actually in it's place

What would you expect in its place?

 

[Doodski]

Multitone is a good test, I'd like to see what happens when you switch off any one of the carriers to see what is actually in it's place. Multitone is what is used to measure composite triple beat.

I guess I should be quite on this subject. Maybe ask questions instead.
But it certainly turned out to be a lively subject with lots of postings.

I have electronics training and I just watch or ask questions when I have a deficit in my knowledge and you know what.> There are peeps lined up assisting when people ask questions here. Foremost I respect the advanced experts here because they deserve that.

 

[Blenderguy]

@voodooless
Depending where the composite triple beat builds up. On broadband rf amplifiers it shows up worse at the higher frequencies, so I think if you turn off the last carrier you will probably see a distortion product. With systems that have 100 or more carriers a new distortion product shows up called composite second order, and it generally is in the middle of the band. It has been over 20 years since I was involved with it, so I don't remember the specifics.

 

[SIY]

@voodooless
Depending where the composite triple beat builds up. On broadband rf amplifiers it shows up worse at the higher frequencies, so I think if you turn off the last carrier you will probably see a distortion product. With systems that have 100 or more carriers a new distortion product shows up called composite second order, and it generally is in the middle of the band. It has been over 20 years since I was involved with it, so I don't remember the specifics.

Ten minutes spent with @pkane 's free software for arbitrary multitone generation will be time well spent if it disabuses you of some incorrect notions.

 

[voodooless]

@voodooless
Depending where the composite triple beat builds up. On broadband rf amplifiers it shows up worse at the higher frequencies, so I think if you turn off the last carrier you will probably see a distortion product. With systems that have 100 or more carriers a new distortion product shows up called composite second order, and it generally is in the middle of the band. It has been over 20 years since I was involved with it, so I don't remember the specifics.

We don’t care about RF here though… audio frequencies are our domain. So my question should be answered with that context in mind.

 

[Blenderguy]

What would you expect in its place?

The principles of distortion in devices is always the same. A transistor does not know whether it is amplifying RF or lower frequencies. Second order distortion follows the 1:1 ratio, for every 10db increase in level a well behaved amplifier will have a 10dB increase in distortion, third order distortion, which is called intermodulation, follows the 2:1 ratio. You do need two tones for intermodulation to appear. For every 10db increase the intermodulation rises by 20dB. So when you see components on the spectrum analyzer you can determine which is which by raising the level 10dB.

 

[Blenderguy]

@antcollinet​

here is an interesting comparison on how music was mixed then and now. Looking at your spectrum capture it look like the lows are significantly higher below 200Hz and the spectrum drops off like a cliff at 7 or 8kHz. Compare it to this spectrum. From what I see is that the original master had significant components past 20kHz as you can see by the steep drop off caused by the CD sampling filter. The CD is Dave Grusin Cinemagic, the song is Fratelli Chase. Many of the cuts on this CD are quite challenging for loudspeakers since the average level is quite low compared to today's compressed recordings. If you played this fairly loud at 80db SPL, then the peaks would exceed 100 dB. But it sounds great at that playback level. If you do a quick calculation, the highs are only 40dB down from average level, so if you were to remove those and did an AB test, people with extended high frequency hearing would notice them lacking. This is a good song for testing mp3 conversion because your ears are not flooded with high levels of bass distracting you from hearing the artifacts. . Back 30 years ago when my hearing was good, I searched out for recordings that specifically had significant high frequency content just for testing.

pare it to this recording, one of my favorites to test audio equipment with.

 

[antcollinet]

You have to double the inductance because a 12ft long speaker cable is actually 24ft long

I accounted for that - '12 ft of two core"

 

[Blenderguy]

I measured with the wire I bought at 5 uH each way. Have you measured any speaker wires recently?
Now I use my old Kimber Kable that is braided and 3uH each way at the expense of more capacitance which I don't care about. I use the DE-5000 LCR meter that people on the EEVblog recommend

 

[MaxwellsEq]

@antcollinet​

here is an interesting comparison on how music was mixed then and now. Looking at your spectrum capture it look like the lows are significantly higher below 200Hz and the spectrum drops off like a cliff at 7 or 8kHz. Compare it to this spectrum. From what I see is that the original master had significant components past 20kHz as you can see by the steep drop off caused by the CD sampling filter. The CD is Dave Grusin Cinemagic, the song is Fratelli Chase. Many of the cuts on this CD are quite challenging for loudspeakers since the average level is quite low compared to today's compressed recordings. If you played this fairly loud at 80db SPL, then the peaks would exceed 100 dB. But it sounds great at that playback level. If you do a quick calculation, the highs are only 40dB down from average level, so if you were to remove those and did an AB test, people with extended high frequency hearing would notice them lacking. This is a good song for testing mp3 conversion because your ears are not flooded with high levels of bass distracting you from hearing the artifacts. . Back 30 years ago when my hearing was good, I searched out for recordings that specifically had significant high frequency content just for testing.
View attachment 384023pare it to this recording, one of my favorites to test audio equipment with.

Noise and resonance

 

[Blenderguy]

Noise and resonance

?

 

[pma]

It is neither resonance, nor high frequency content. It is noise shaped dither.

 

[DonH56]

A transistor might not "know" the signal frequency, but the designer and user will. Low-frequency and high-frequency transistors tend to be very different beasts, with different characteristics, to suit their applications. A 300+ GHz HEMT will have a LF (1/f) noise corner well above the audio band, for example, and a large audio power transistor will suck exhibit poor performance amplifying a 10 GHz RF signal.

For others not the expert Blenderguy is, here is a short article discussing how harmonic and intermodulation distortion are related and where IMD products land: https://www.audiosciencereview.com/...armonic-and-intermodulation-distortion.25436/ The short answer is that even-order IMD products tend to land at very low and very high frequencies (around DC and 2x the signal frequencies), whilst odd-order products land near the original signal frequencies.

Speaker cable, or any wire, inductance depends upon configuration. Many folk have measured cables but, unless you have an outlier for amp and speaker impedance and/or a very long cable run, all that matters is DCR. Some 12 AWG coaxial cable I have around measures ~0.5 uH/m (both shield and center conductors considered); 12 AWG zip cord I have runs a little higher, around 0.7 uH/m. That is at audio frequencies, basically "DC" to our network analyzer. You can braid or weave insulated conductors to reduce inductance; the trade is higher capacitance.

 

[Theta]

Short answer - Yes.

Some of the more economical class-D "chip amps" can be load sensitive, and get a bit messy at the high end on differing loads - but the better designs don't have this issue.

The quality manufacturers/designs classA/AB/D all sound the same in double blind tests.

So at an engineering level, there is no reason to use Class A anymore (just wasting power as heat) - and AB vs D, are usually much of a muchness in terms of idle/standby power consumption, and pretty closely match in real life use too.

There are state of the art designs in each category, that achieve measurable results far beyond any audibility threshold... those are typically far cheaper now than they were 30 years ago, but still not exactly chump change.

A $100 ClassD chipamp - may well have issues (as described above!) - where a similarly priced AB, will suffer from lack of power, but won't misbehave otherwise.

The other thing to look out for, is traditional AB designs have substantial headroom above and beyond their rated loads... Typical Class D designs don't. - Hence a 50W AB may be able to put out 75W or even 100W for brief periods of time, where a class D design might be rated at 75W... and peak at either 75W or marginally more (like 80W).

Often these limitations are due to the use of regulated SMPS power supplies for ClassD and unregulated transformer based power supplies in AB. An AB with a SMPS might also end up with its rated and max power being quite close to each other - also the amount of capacitor reserve in the design will determine the short term peaks possible...

Look at Benchmark for SOTA class AB, and Hypex/Purifi/ICE for Class D modules (and many manufacturers who then use the modules in their amps)

Benchmark is class H, Class AB don't all sound the same, DC class AB and capacitor copled AB are diferent, Also Mosfet output amps sound diferent compared to bipolars. Of course it all depents on the load.

 

[dlaloum]

Benchmark is class H, Class AB don't all sound the same, DC class AB and capacitor copled AB are diferent, Also Mosfet output amps sound diferent compared to bipolars. Of course it all depents on the load.

That is the point I have made ... within their design performance envelope (including the load they are designed for) - all halfway decent amps sound identical.

Once you push an amp outside of its performance envelope, then it's sound will start to change - and different designs will misbehave differently - hence they sound different.

So, in most cases, with easy to drive 8ohm relatively efficient speakers - all amps will sound the same.

In my setup, with my 4 ohm nominal speakers, dropping to 1.6ohm minimum - all amps do NOT sound the same...

But all amps that can handle 1 ohm speakers... do sound the same.

 

[pogo]

If it were that simple. Listen to a bridged and unbridged NAD 1ET400A. A difference is perceptible at any volume and load.

 

[pma]

If it were that simple. Listen to a bridged and unbridged NAD 1ET400A. A difference is perceptible at any volume and load.

Any proof?

 

[antcollinet]

I measured with the wire I bought at 5 uH each way. Have you measured any speaker wires recently?
Now I use my old Kimber Kable that is braided and 3uH each way at the expense of more capacitance which I don't care about. I use the DE-5000 LCR meter that people on the EEVblog recommend

I've not measured recently - since retirement I don't have access to the kit so far.

Are you measuring each conductor separately? You should be measuring by connecting the speaker cable two conductors together at one end and measuring the inductance between the two wires at the other end. This gives a different result from measuring the inductance of a singe wire end to end.

This is because when in use as a speaker cable with both wires in operation simultaneously, the send and return currents interact, changing the "system" inductance compared with a single wire conducting in free space.

 

[ShadowFiend]

Right, but that's a lowpass at 20khz or 40khz, either way the noise is not really likely to actually reach your tweeters in any case.

In midrange or bass, where low frequency eclectrical low pass filter is applied (in passive speaker) and the natural roll off of driver occurs in fairly low frequency, your statement is true.
But why the noise is not reaching the tweeter? When you connect a tweeter in active or passive speaker to amplifier, there will be no electrical low pass filter in between, so noise from amplifier will transmit directly to tweeter. Meanwhile, many tweeters like Diamond tweeter is flat on axis at 40kHz. Even the normal 1" fabric ones have flat on axis until 30kHz.

 

[dlaloum]

If it were that simple. Listen to a bridged and unbridged NAD 1ET400A. A difference is perceptible at any volume and load.

As soon as you bridge an amp, you effectively halve the impedance of the load...

So a 4ohm speaker is then "seen" by the now bridged amp, as a 2ohm load...

The likelihood of an amp sounding "different" once bridged, increases massively.

The surprising amps are those like the Quad 606, 707, 909, or the Crown XLS series.... they can all handle 1 ohm speakers without a problem...
I've run the Crowns into my 1.6 ohm gallo speakers... bridged - so they were effectively driving a 0.8ohm load.... they sounded the same as they did in stereo mode (and the same as the Quads too...) - the Quads are rated "stable into any load", the Crowns are rated down to 1ohm load (and demonstrably handled a 0.8ohm load in my case)

These all sounded identical.... whereas my Integra DRX3.4 sounded "different" - and not in a good way.