Advance in Class D amps?

[zenelectro]

All the numbers for ME amps exceed the abilities of the human ear to resolve.

Years ago I and others did tests on a SS amplifier adding H2 and H3 to see what levels were perceptible. Measurements done with HP spectrum analyser and AP

System1. Baselines for the HP and AP1 were around 0.003% and 0.0005% respectively.

We could hear very low distortion levels, depending on spectrum down close to

the baseline of the test sets.

This pretty much correlates with other work such as that of Rupert Neve who

found that once you get to higher orders (h7) extremely low levels are audible.

From memory he was stating around -120dB.

Howeve, there are a number of very critical figures that are almost never mentioned by manufacturers. The standard figures are:

Frequency response: DC ~ 150kHz +0 -1dB

THD (measured from 20Hz ~ 20kHz at any power level): <0.1%

IMD: <0.1%

Rise time: 1uSec

Damping factor (measured from DC ~ 50kHz): 200 (model dependent)

However, there are a couple of 'gotchas' in the figures. Note the damping factor. Not only is it specified from DC ~ 50kHz, but it is the open loop figure. No global NFB is required to drive the figure artificially low.

Briefly, specs - ME linear amp versus ncore class D

Distortion: - ncore superior by a huge margin on all fronts and has a flat distortion

versus frequency characteristic which was specifically intended for good sonic results.

Damping factor – ncore is >2000 up to 20kHz = extremely load invariant. At 30kHz it is

>400 which is a precedent for class D.

WRT ‘artificial’ low OP Z from feedback - white paper covers FB paranoia nicely .

http://www.hypex.nl/index.php?option=com_content&view=article&id=89

Also not stated is:

Square wave performance. The square wave is perfectly square (within rise time and slew rate limits), regardless of load impedance. The load may highly capacitive or highly inductive and the square wave remains almost perfect. No Class D amp can manage this.

Driving square waves into cap loads can be a tough call for any amp, class D Or linear. Looking at the ncores very flat OP Z it should drive reactive loads very

well but unfortunately no data on this that I know of ATM.

In the case of some of the Class D amplifiers mentioned, the frequency response reveals that critical flaw. A frequency response that is not PERFECTLY linear to beyond 20kHz indicates that there will be phase errors well down into the audible region. It is for this reason that ME (and most other truely high end amplifier manufacturers) ensure that the frequency response of their products goes to at least 100kHz or beyond. Class D amps are not even in the race in this area.

WRT Phase errors you mean group delay errors. From data sheet: it has -3dB at around 60kHz. For a single order LPF this is

about -0.4dB at 20kHz and a group delay error of less than 0.3 micro seconds. So a

20kHz signal will arrive at the ear 0.3 micro seconds after it should.

I suggest you measure the phase distortion of any multiway speaker to put things in

perspective.

T

[davewantsmoore]

Anybody, do you agree with ZB when he says;

"A frequency response that is not PERFECTLY linear to beyond 20kHz indicates that there will be phase errors well down into the audible region"

Absolutely. What is the conjecture?

Edited July 4, 2011 by davewantsmoore

[Orpheus]

Absolutely. What is the conjecture?

I don't know. I'm just trying to understand whether there is any doubt about it. Technically, this is all beyond me.

[davewantsmoore]

I don't know. I'm just trying to understand whether there is any doubt about it. Technically, this is all beyond me.

No doubt. Basic physics. Amplitude and phase are inextricably related, you could say actually they're just a different view of the same thing.

If you feed a signal through an amp .... and if a given frequency that comes out the other end is different in amplitude (to the input) .... then the signal will also be altered in timing somewhere (phase or 'group delay' in this context)

As ZB mentioned.... You might be tempted to wonder why is your amplifier designed to handle signals waaaay higher up than the audible band .... it's because a change in frequency amplitude response also causes changes in timing/phase at some frequencies.

As somebody else mentioned (sorry, work is calling).... Does a small amount of amplitude distortion (say 0.5db down at 20khz) cause phase shifts which are a massive deal breaker in audio amplifiers??? Maybe not, especially when compared with speakers and rooms, etc ..... but, in any other type of amplifier (non-audio) this would be certainly be considered totally unacceptable. Not every amplifier is connected to something as crude as a speaker ;-)

The more I learn about audio, the more it appears that our ears are somewhat forgiving towards amplitude distortion (eg. it is difficult to hear a sharp narrow notch in the frequency response, most of the time) ..... but extremely sensitive to timing/phase distortion.

I think ZB (eeek, I'd hate to paraphrase a boffin, and get it wrong) ... is less coming from the angle of saying that a certain amplitude or phase distortion is horribly audible (even if true), and deal breaker for the technology (cos there's always horses for courses) ...... and more from the POV that at a technical level it is 'unnecessary' to have an amp with such a compromise .... of course, that's not saying that people should abandon furthering the technology behind class D amps, or stop buying class D amps (as noted there is big potential) .... it's just putting things in perspective.

Edited July 5, 2011 by davewantsmoore

[Zaphod Beeblebrox]

Years ago I and others did tests on a SS amplifier adding H2 and H3 to see what levels were perceptible. Measurements done with HP spectrum analyser and AP

System1. Baselines for the HP and AP1 were around 0.003% and 0.0005% respectively.

We could hear very low distortion levels, depending on spectrum down close to

the baseline of the test sets.

How low? Bear in mind that 0.1% corresponds to -60dB. Even a quiet room adds around 40dB to that figure. The ability to hear distortion levels below 0.1%, reliably and using even very high quality loudspeakers (ESLs would be the choice here) is quite rare.

This pretty much correlates with other work such as that of Rupert Neve who

found that once you get to higher orders (h7) extremely low levels are audible.

From memory he was stating around -120dB.

I find that very hard to believe. I do agree that high order harmonic distortions are to be avoided, but the ability to hear distortion levels that are at -120dB is difficult to accept.

Briefly, specs - ME linear amp versus ncore class D

Distortion: - ncore superior by a huge margin on all fronts and has a flat distortion

versus frequency characteristic which was specifically intended for good sonic results.

No argument from me. Many amplifiers exhibit vastly lower levels of THD than ME amps do. In general, however, those distortion figures are:

* Largely academic.

* Obtained at the expense of other, potentially more audible, figures.

Damping factor – ncore is >2000 up to 20kHz = extremely load invariant. At 30kHz it is

>400 which is a precedent for class D.

Indeed it is. It is superior to most linear amps too.

WRT ‘artificial’ low OP Z from feedback - white paper covers FB paranoia nicely .

I'll take your word for it.

http://www.hypex.nl/index.php?option=com_content&view=article&id=89

Driving square waves into cap loads can be a tough call for any amp, class D Or linear. Looking at the ncores very flat OP Z it should drive reactive loads very

well but unfortunately no data on this that I know of ATM.

Get back to me when you can show me that data. Square wave performance can reveal some fatal flaws in otherwise excellent amplifiers.

WRT Phase errors you mean group delay errors. From data sheet: it has -3dB at around 60kHz. For a single order LPF this is

about -0.4dB at 20kHz and a group delay error of less than 0.3 micro seconds. So a

20kHz signal will arrive at the ear 0.3 micro seconds after it should.

Which may well be audible to listeners who listen critically, using high quality loudspeakers. If the cited amplifier exhibited a frequency response that was perfectly linear to 20kHz, we would not be having this discussion. A -0.4dB fall at 20kHz is not acceptable in a high quality system. I am, frankly, surprised that you continue to argue this point.

I suggest you measure the phase distortion of any multiway speaker to put things in

perspective.

T

I suggest you examine the phase response of a good quality ESL to put things into perspective. You'll get no argument from me WRT the following:

* That many speakers do not provide an accurate phase response.

* That Class D amplifiers are generally acceptable for a large number of speakers and listeners.

* That I will reserve judgement on the cited amplifier until I see some independent test data.

[georgehifi]

ZB, Anybody, do you agree with ZB when he says;

"A frequency response that is not PERFECTLY linear to beyond 20kHz indicates that there will be phase errors well down into the audible region. It is for this reason that ME (and most other truely high end amplifier manufacturers) ensure that the frequency response of their products goes to at least 100kHz or beyond."

Yes absolutely, and the faster it rolls off after 20khz the worse it is and there is even more phase shifts down into the audible range and so on. This is why brick wall filters were so nasty in early cdp's as they have the worst phase shifts and ringing down to 5k or even lower.

And same goes for Class D it has phase shifts because they have to try get rid of all the switching noise after 20khz.

Cheers George

Edited July 5, 2011 by georgehifi
spelling

[zenelectro]

ZB, which model are these measurements taken from?

Zen Electric, what are your comments on these figures, and what other measurements would you like to know?

Anybody, do you agree with ZB when he says;

"A frequency response that is not PERFECTLY linear to beyond 20kHz indicates that there will be phase errors well down into the audible region. It is for this reason that ME (and most other truely high end amplifier manufacturers) ensure that the frequency response of their products goes to at least 100kHz or beyond."

Absolutely. What is the conjecture?

I don't know. I'm just trying to understand whether there is any doubt about it. Technically, this is all beyond me.

OK, I will explain further. Phase shift does not necessarily cause audible errors.

An audio signal made up of frequencies from 20Hz to 20kHz can go through a filter

(or amplifier) and have phase shift but as long as it comes out the other end looking the

same as it went in there has been no phase distortion.

Group delay = the time it takes a signal to pass through the device (or filter). If all

frequencies take the same time to pass through the device, it doesn't matter how

long that time is, the complex signal will be unchanged at the other end.

Analog low pass filters will result in phase shift, as ZB correctly states, but if all audio

frequencies (20Hz - 20kHz) pass through that filter with the same group (time) delay

there has been -no- phase distortion.

Take a simple 1st order RC low pass filter that is down -3dB at 80kHz. (ref Pic)

Yellow = amplitude - at 20kHz it is down about 0.25dB

Green = phase shift - at 20kHz it is ~ 14 degrees. This seems a lot but more important

is the change in group delay.

Pink = Group Delay (the time it takes the signal to pass thru filter)

- It is constant at 2uS till about 5k.

- At 10k it is about 1.96uS

- At 20k it is 1.88uS. *Change* in group delay is ~ 0.12uS (120 nano sec).

The 20kHz sig is arriving out of the filter 120 nano sec time shifted from where it should.

[davewantsmoore]

The 20kHz sig is arriving out of the filter 120 nano sec time shifted from where it should.

Yes, inaudible (!?) .... but I'd imagine this makes NFB problematic (understatement?!)

EDIT: and now I'm reading about feedback in class D amps .... (and yes, it seems group delay makes traditional NFB impossible) .... my boss hates you people distracting me ;-)

[zenelectro]

Yes absolutely, and the faster it rolls off after 20khz the worse it is and there is even more phase shifts down into the audible range and so on.

George, this is not necessarily so.

A single order filter has more phase distortion than a linear phase 3rd order (or higher)

Bessell or Gaussian rolloff.

I think you are maybe referring to digital brickwall filters (yes?) which are in fact

perfectly linear phase in the audio band but the very high order (100's dB/octave)

causes other problems.

Ref to pic below. Same -3dB at 80kHz, this time it is a 3rd order Gaussian roll off.

The group delay error at 20 kHz is now 0.03uS at 20kHz or 30 nano seconds!

And same goes for Class D it has phase shifts because they have to try get rid of all the switching noise after 20khz.

Cheers George

The OP noise of this class D amp is pretty low. Looking at the HF IMD graphs, noise

floor and RMS noise specs referred to 1 watt it's hard to fault it.

*************

For me - I think this discussion has pretty much gone the distance.

I've tried pretty hard to reply in an even handed manner and as technically correct as I

possibly can within the limitations of my knowledge.

Maybe we should all sit tight and wait till the modules are available. See how they

sound / measure in reality. I am certainly considering them for a few applications.

Whether they stack up to the very best linear amps - I don't know for sure but based

on what we know so far they should perform very well indeed.

We will see.

I've got a ton of work to get through and products to design / test so can't really

justify spending much more time on this thread. Typing / explaining tech issues quickly

is not a strength for me so hopefully there has been some benefit to all this discussion.

cheers

T

[Nigel]

This thread has convinced me to go for a Class D for bass duties. < 200Hz.

Can anyone recommend one about 350W(guess) into 4 ohms with variable phase adjustment ?

[davewantsmoore]

been some benefit to all this discussion

Thankyou. I do crave the objective discussions.

[georgehifi]

This thread has convinced me to go for a Class D for bass duties. < 200Hz.

Can anyone recommend one about 350W(guess) into 4 ohms with variable phase adjustment ?

Don't know about ClassD but this has had good reports, switchable to 230v 300w discrete high current, with our exchange rate $420 for two of them and Paypal excepted

https://www.madisound.com/store/product_info.php?products_id=371

And this class D http://cgi.ebay.com/500-Watt-RMS-Home-Theater-Sub-Woofer-Plate-Amplifier-/390327573446?pt=Speakers_Subwoofers&hash=item5ae1561bc6

Cheers George

Edited July 5, 2011 by georgehifi

[tktran]

A -0.4dB fall at 20kHz is not acceptable in a high quality system

At the end of the day we're all listening to systems, not just amplifiers connected directly to our ears.

Now a typical speaker, dynamic or electrostatic, has a far poor high frequency response than amplifiers or sources.

At 20KHz, a typical ESL is 2-10dB down on-axis. And typically 10-20dB down off axis, which explains why

there's such a narrow sweet spot.

At 30 degrees off axis, a good 1" dome tweeter is 2-3dB down at 10KHz and 10-15dB down at 20KHz.

Most high quality Class A or A/B solid state amps also have a gradual roll-off, and are usually down 0.25-0.5dB by 20KHz, which is magnitudes better.

So I'm not sure how 0.4dB down at 20KHz qualifies as unsuitable for a high end system.

I can also tell you that my 32 year old ears are at least 10dB down at 20KHz, many men in their mid 50-60's would be lucky to hear anything at all above 15KHz.

Edited July 5, 2011 by tktran

[Orpheus]

At the end of the day we're all listening to systems, not just amplifiers connected directly to our ears.

Now a typical speaker, dynamic or electrostatic, has a far poor high frequency response than amplifiers or sources.

At 20KHz, typical ESL is 2-10dB down on-axis, and typically 10-20dB down off axis.

At 30 degrees off axis, a good 1" dome tweeter is 2-3dB down at just 10KHz and 10-15dB down at 20KHz.

Most high solid state amps also have a gradual roll-off, and are usually down 0.25-0.5dB by 20KHz, which is magnitudes better.

So I'm not sure how 0.4dB down at 20KHz qualifies as unsuitable for a high end system.

I can also tell you that my 32 year old ears are at least 10dB down at 20KHz, many men in their mid 50-60's would be lucky

to hear anything at all above 15KHz.

In that case, you haven't been paying attention. It appears generally accepted that phase coherence is effected at audible frequency levels by frequency response issues above 20Khz.

[tktran]

It appears generally accepted that phase coherence is effected at audible frequency levels by frequency response issues above 20Khz.

No it's not generally accepted that a fraction of a dB down at 20Khz, and 3dB down at 60KHz makes a difference to phase coherence.

But yeah, if you've got all kinds of high frequency hash and wild eg. +/-10dB swings going on above 20Khz, then yeah, then people might be bothered worrying.

[davewantsmoore]

No it's not generally accepted that a fraction of a dB down at 20Khz, and 3dB down at 60KHz makes a difference to phase coherence

Can you explain to me?!

Taking your statement literally ... you are contradicting the basic laws of physics ..... So do you mean that the phase difference is not audible? (I agree, it probably isn't) .... or are you saying something else?

EDIT: Just FTR, in case anyone is following along.

http://www.ugrad.math.ubc.ca/coursedoc/math100/notes/trig/mathgifs/phase_3.gif

Here is a formula showing t=constant, A = amplitude, w = frequency, and Phi (O with a vertical line) = phase .... If frequency and t remain constant ... and you adjust the amplitude ... then phase will also change.

Edited July 5, 2011 by davewantsmoore
add some maths

[tktran]

All filters, bar none, will introduce a phase shift. By definition. Correct.

What i'm talking about, is the audibility of a low (or high) pass filter whose rolloff is very shallow, as is demonstrated by virtually all modern amplifiers. If you look at their measurements, virtually all are down a fraction of a dB by 20Khz, and approximately down 3dB by 40-80KHz.

I won't even get started about the importance (or lack thereof) of square wave reproduction...

[davewantsmoore]

Gotcha. Thanks.

[davewantsmoore]

Gotya. Thanks.

[biologist]

I've not been able to follow much of what has been discussed, my technical ceiling is pretty low, but I have some class D amps, namely the Firestone Audio MASS Pre-Amplifier and Firestone Audio BigJoe IV Mono Power Amplifiers, specs below.

Firestone Audio MASS Pre-Amplifier

Circuit Structure: Infrared Control Pre Amplifier

Power Structure: Switching Power Supply / Soft Start Circuit

Control Mode: Stand By / Power ON / Volume Up / Volume Down / Input Select / Mute

Gain Range: -95dB to 31.5 dB in .5 dB increments

Circuit Protection: Output Short / Over Current

Volume Control Chip: TI - PGA2310

Audio Performance: (1 kHz sine wave, 2Vrms output, 24 Bit/48kHz)

Noise Level: (1 kHz, A-Weighted) -107.5 dB

Dynamic Range: (1 kHz, A-Weighted) 107.5 dB

THD: 0.0018%

Stereo Crosstalk: -100dB

Connectivity

Inputs: 2 sets of Gold Plated RCA

Output: 1 set of Gold Plated RCA

Firestone Audio BigJoe IV Mono Power Amplifier.

Amplifier Structure: Coupling Capacitor, Mono-Mono Class D Amplifier

Power Structure: Soft Start Circuit, Single-supply

Power Output: 8ohm @ 44W (0.3 THD) 4ohm @ 88W (0.3 THD)

Circuit Protection: Output short / Over Current / Over and Under Voltage / Temperature Protection

Power Stage: Texas Instruments - TAS5121

PWM Controller: Texas Instruments - TAS5010

ADC Chip: PCM1803A

I've run these amps through several sets of speakers so far including Quad 12L, Krix Acoustix and some new red Usher S 520's and have been very happy with the performance. I've pushed the Quads and Ushers pretty hard on a few occasions but this system is right next to my dining table and is mainly used to listen to music while eating or talking so its rarely loud. Thats left for times when I'm home alone.

I've also substituted the MASS pre amp with a Carot One Ernestolo Class T amp using only its tube Pre Amp and I did notice a distinctly warmer sound with this combination especially with vocals and acoustic recordings. The Ernestolo was not bad by itself with acoustic music but only really did well at low listening volumes. Hence I resorted to using the tiny tube pre amp and the Firestone monoblocks. The specs for the Ernestolo make for interesting reading...

Carot One ERNESTOLO

Specific

T-Amp IC Tripath TA2024

Output Power

2x15W @4ohm

2x6W @8ohm(RMS)

Signal-to-Noise Ratio (SNR) 98dB

Dynamic Range 98dB

IHF IM Distortion 0.10% @ 1W, 4ohm

THD+Noise Audiophile Quality 0.03% @ 9W 4ohm

0.1% @ 11W 4ohm

0,1% @ 9W 6 ohm

0.1% @ 6W 8ohm

High Power 10% @ 15W 4ohm

10% @ 10W 4ohm

Power Efficiency 81% @ 15W 4ohm

90% @ 10W 8ohm

Specification of the Headphone Amplifier Unit

Signal-to-Noise Ratio (SNR) 92dB

Power Output 3.0 Watts per channel on 33 Ohm

Specification of the Pre Amplifier Unit

Vacuum Tube 6DJ8/6922 series or 12AU7 series

(exchangeable)

Signal-to-Noise Ratio (SNR) 92dB

THD + Noise 0.05% @ 10k ohm 0.15% @ 33 ohm

Input Audio CD IN RCA (Left/Right) x 1 [3Vpp max.]

For such tiny little boxes these units are pretty impressive, or maybe I'm easily impressed. I suppose I should try connecting the Firestones and the Ernestolo up to my Orpheus Auroras and see what happens.

[firedog]

Came across this thread and skimmed a few pages, decided to jump in. Excuse me if I'm a little off the discussion flow, but a response to some of what I read.

Just bought an assembled power amp from http://classdaudio.com ; they also sell kits.

250W X 2 into 8 ohms; 500 X 2 into 4 ohms

Price: $590 US plus shipping. Kits are significantly less. Kits are a simple build for any slightly experienced DIYer.

Theory about how class d can or can't sound is nice. Listen before you make up your mind.

This amp sounds significantly better than my Musical Fidelity X-150 amp. In pretty much every parameter. (I'm still using the same MF amp as the preamp, so the only change to the system was changing the power amp to the Class D.) Highs are not harsh. Bass, mids, transients, dynamics, mids, sound stage are all noticeably better.

I haven't done direct A/B testing, but I've heard other amps in my system. This one is at least as good or better than some traditional amps costing many times its price. It is also competitive with other Class D amps costing several times it's price.

I understand you personally might not like the sound, but don't dismiss the whole category out of hand like some on this thread have done.

BTW, I also saw the Devialet mentioned in this thread. It actually isn't a class D amp, if you read the specs closely. It's something else.

[georgehifi]

Just bought an assembled power amp from http://classdaudio.com ; they also sell kits.

250W X 2 into 8 ohms; 500 X 2 into 4 ohms

Price: $590 US plus shipping. Kits are significantly less. Kits are a simple build for any slightly experienced DIYer.

.

At least they use a linear power supply with that kit, like Channel Islands did with their D-100 that gets rid of half of the nasties, now just to get rid of the horror saw tooth noise across the top of the top of the square wave as they would still have this, because the technology is not ready yet for them to get rid of this Class D switching frequency noise, maybe in a couple more years the powers to be can develop it to be a much higher switching frequency and then not be so destructive to the audio band.

http://www.stereophile.com/content/channel-islands-audio-d-100-monoblock-power-amplifier-measurements

Cheers George

Edited August 14, 2011 by georgehifi

[eras]

George, do all Class D amplifiers have the sawtooth pattern/noise? The Halcro amplifiers seem to reproduce a very nice square wave.

http://www.stereophile.com/content/halcro-dm88-reference-monoblock-power-amplifier-measurements

[georgehifi]

I have never seen the circuit but Halcro's from what I was told are not digital amp's but push pull Mosfets (bi-polar are my preference for output stages) though the same source said they do have very very sophisticated smp (switch mode) power supplies, something so exotic you may never see in digital amps.

In digital amps to get rid of that nasty sawtooth noise you see like on the CI D-100 amps, the noise filtering would have to cut severely into the HF/upper mids audio band, as it is they come down enough to make it bearable, but the trouble is the amount of phase shifts the filters introduce down into the HF and upper mids area of the audio spectrum is not what you would call state of the art, as to me it makes the upper mids and treble sound opaque or lifeless, and if not filtered down enough it will make them sound hard and edgy, because it would let too much of the sawtooth crap through.

Cheers George

[Guest Bodhisattva]

D is for Duck....my 2 cents.

Bodhi