Class D: Turns Out it Does Suck!

[March Audio]

32 minutes ago, Shadders said:

Hi,

The quoted damping of 2,000 as per the review by soundstage which is contrary to another source such as Stereophiles output impedance measurement, creates a discrepancy.

 

The fact that the Stereophile output impedance measurement of class D correlates with the damping factor of Elektor class D amplifier, and that Stereophile output impedance measurements of many class A/B amplifiers correlate with Hifi News output impedance measurements, indicate that the Stereophile measurements for class D are accurate, and real world.

 

I expect that Stereophiles measurement is accurate.

 

Regards,

Shadders.

You don't know that.  It contradicts hypex measurements and the bhk labs measurements which BTW correlate extremely well with Hypex if you account for connections and leads from the modules. 

 

You are making assertions that you have no ability to back up. 

 

[Shadders]

1 minute ago, opus101 said:

 

Fixed it for you - you do seem rather in love with AB when that has its own issues. Are you aware of any of them?

Hi,

When you indicate issues - can you provide which ones you are referring to ?. Class A/B may have output stage distortion, VAS loading issues, LTP imbalance causing higher THD issues, can have a reduced efficiency if the bias current is set too high.

 

For a well design class A/B amplifier - it will always exceed the best performance of a class D amplifier. A class D amplifier is just an approximation of the signal. It requires an output filter to integrate the signal into a smoother function.

 

Regards,

Shadders.

[psjug]

Here is soundstage measurements of the same Bel Canto amp that sterophile measured.  It looks like the difference is in the amplifiers

 

https://www.soundstagenetwork.com/index.php?option=com_content&view=article&id=1515:bhk-labs-measurements-bel-canto-design-e-one-ref600m-mono-amplifiers&catid=97:amplifier-measurements&Itemid=154

 

[opus101]

1 minute ago, Shadders said:

When you indicate issues - can you provide which ones you are referring to ?. Class A/B may have output stage distortion, VAS loading issues, LTP imbalance causing higher THD issues, can have a reduced efficiency if the bias current is set too high.

 

If I remember correctly Doug Self in his book on power amp design says that AB is going to have always higher distortion than correctly biassed classB due to 'gm doubling'. There's an issue with maintaining a static bias point for the output stage idle current as the output stage temperature varies. There's also the thorny implementation issue of output stage load currents which have sharp edges (meaning fairly high dI/dt around the crossover region) and ensuring those don't couple into any of the rest of the circuitry.

 

If 'approximating' the signal is a problem then digital audio is worse - it also requires an output filter. As I see no problem with correctly implemented digital I equally see no problem with classD done right. A classD when designed well will certainly exceed a classAB on efficiency, so which metrics are you going to choose as relevant?

[Shadders]

1 minute ago, opus101 said:

 

If I remember correctly Doug Self in his book on power amp design says that AB is going to have always higher distortion than correctly biassed classB due to 'gm doubling'. There's an issue with maintaining a static bias point for the output stage idle current as the output stage temperature varies. There's also the thorny implementation issue of output stage load currents which have sharp edges (meaning fairly high dI/dt around the crossover region) and ensuring those don't couple into any of the rest of the circuitry.

 

If 'approximating' the signal is a problem then digital audio is worse - it also requires an output filter. As I see no problem with correctly implemented digital I equally see no problem with classD done right. A classD when designed well will certainly exceed a classAB on efficiency, so which metrics are you going to choose as relevant?

Hi,

Isn't class A/B a biased class B, where true class B will have worse crossover distortion ??? As soon as you bias class B, it then becomes class A/B.

 

Is it not the purpose of the output stage bias/Vbe multiplier, whose transistor(s) are located on the same heatsink, is the regulation for thermal aspects, such that as the temperature rises, the Vbe voltage reduces, reducing the output bias of the output stage, thus reducing the temperature. A closed loop system ?

 

For the thorny issue - is it not correct that the bias currents in the output stage mean that the positive voltage transistors do not turn off when there is a negative voltage signal, and vice versa, such that there is a significantly reduced crossover distortion, which is further reduced to 0.0001% THD level due to negative feedback. Implementing extra output stage transistors reduce the levels significantly which then requires less feedback.

 

For the approximation aspect - class A/B amplifiers are continuous functions - so their ability to replicate the input signal will always exceed the capability of class D, especially at high frequencies.

 

For the power efficiency, agree - class D is superior, but i do not see this as something the listener can experience through their speakers.

 

Regards,

Shadders.

[opus101]

Looks to me like you have a few misconceptions about the more arcane aspects of classB and classAB amps that a read of Doug Self's APAD and Bob Cordell's 'Designing Audio Power Amplifiers' might help to clear up.

 

On the thermal bias issue there is indeed a negative feedback loop typically but it has a significant time delay given that the instantaneous temperature of the output devices can't directly be sensed if the sensor's on the heatsink. Hence the development of 'ThermalTrak' transistors where diodes are included on the die.

 

On power efficiency it translates directly to noise on the power supply rails to the output devices - more wasted power means higher ripple hence for the same PSRR a classD amp is going to win out subjectively. This is experienced by the listener as classD's legendary 'grip' on the bass.

 

As regards your other claims against classD I suggest doing more reading - try some of the links/posts contributed by DIYA member Eva, she's a very knowledgeable classD amplifier designer. There's one relevant paper attached to this post : https://www.diyaudio.com/forums/class-d/90404-signal-modeling-self-oscillating-amplifiers-post3915499.html

[Shadders]

8 minutes ago, opus101 said:

Looks to me like you have a few misconceptions about the more arcane aspects of classB and classAB amps that a read of Doug Self's APAD and Bob Cordell's 'Designing Audio Power Amplifiers' might help to clear up.

 

On the thermal bias issue there is indeed a negative feedback loop typically but it has a significant time delay given that the instantaneous temperature of the output devices can't directly be sensed if the sensor's on the heatsink. Hence the development of 'ThermalTrak' transistors where diodes are included on the die.

 

On power efficiency it translates directly to noise on the power supply rails to the output devices - more wasted power means higher ripple hence for the same PSRR a classD amp is going to win out subjectively. This is experienced by the listener as classD's legendary 'grip' on the bass.

 

As regards your other claims against classD I suggest doing more reading - try some of the links/posts contributed by DIYA member Eva, she's a very knowledgeable classD amplifier designer.

Hi.

Cordell text is as per my statement of class B (page 97, 98 details), and Doug Self does state class B is slightly biassed. As per Cordell, it is semantics - just because some people later state class B has a bias current, is neither there nor there. Class A/B exceeds the performance of class D.

 

Thermal - it is only an issue if you have not adequately designed the heatsink correctly for the conditions required.

 

An adequately designed power supply for class A/B is no worse than class D power supply. The fact that many class D amplifiers use a switched mode power supply means that the power supply may have less ripple. You can use a "normal" power supply (transformer, rectifier, reservoir capacitors) for class D. You can use a SMPS for class A/B - and some designs do this - i have seen many reviews in Hifi News of class A/B amplifiers using a SMPS. Hmmm, win out subjectively ??? - maybe. Will a class A/B with SMPS be better than a class D with a normal power supply ?? Again it is subjective ??

 

Whatever the processing in class D, it will always be an approximation that has to be filtered to interpolate, yet still has a significant high frequency carrier on its output. Class A/B is no such approximation function, nor does it have significant high frequency noise continually on its output.

 

Regards,

Shadders.

[opus101]

3 minutes ago, Shadders said:

An adequately designed power supply for class A/B is no worse than class D power supply.

 

You've lost me completely. I shall bow out here.

[Shadders]

2 minutes ago, opus101 said:

 

You've lost me completely. I shall bow out here.

Hi,

Should have stated "class D amplifier power supply" and not "class D power supply".

Regards,

Shadders.

[Jud]

2 hours ago, mansr said:

The recording microphones would pick up any IM products that our ears would.

 

I've read this response before and certainly understand it (particularly as an older male whose hearing extended to maybe 16kHz at the top end the last time I tried it a few years ago).  I do still continue to wonder whether anything in the subconscious experience of listening live, even down to bone conduction of very low levels of ultrasonics, renders the sensation of audible IM products live at all different than hearing them reproduced by a system that lacks IM distortion.  In different words, what is it about IM distortion that leads many people to describe the sound of NOS DACs, for example, as "natural"?

 

Anyway, this is getting off topic.

[Shadders]

3 hours ago, Shadders said:

As an aside, the Hifi News November 2012 review of the MBL Corona Line C51 measures the output impedance to be 0.016ohms to 0.051ohms

Hi,

I need to correct this - the MBL Corona is a class D amplifier - so lower output impedances are possible. It is their own design, so it would be good to know why it measures this low and others measure much higher. Damping is 500 to 157.

 

Regards,

Shadders.

[esldude]

14 minutes ago, Jud said:

 

I've read this response before and certainly understand it (particularly as an older male whose hearing extended to maybe 16kHz at the top end the last time I tried it a few years ago).  I do still continue to wonder whether anything in the subconscious experience of listening live, even down to bone conduction of very low levels of ultrasonics, renders the sensation of audible IM products live at all different than hearing them reproduced by a system that lacks IM distortion.  In different words, what is it about IM distortion that leads many people to describe the sound of NOS DACs, for example, as "natural"?

 

Anyway, this is getting off topic.

Maybe similar reasons some like low level distortion in LP , tubes or tape? Sounds more richly detailed.

[psjug]

7 minutes ago, Shadders said:

Hi,

I need to correct this - the MBL Corona is a class D amplifier - so lower output impedances are possible. It is their own design, so it would be good to know why it measures this low and others measure much higher. Damping is 500 to 157.

 

Regards,

Shadders.

Just looking at the ncore products - the measurements of output impedence or damping factor vary quite a bit.  It seems to me that it shouldn't be too surprising that tens of milliohms of output impedence could be added to the ncore by the OEM.

[Shadders]

3 minutes ago, psjug said:

Just looking at the ncore products - the measurements of output impedence or damping factor vary quite a bit.  It seems to me that it shouldn't be too surprising that tens of milliohms of output impedence could be added to the ncore by the OEM.

Hi,

Yes - i am thinking that the specific modifications each amplifier manufacturer makes is on the output filter (series inductor part). This will have a significant effect on the output impedance, albeit the impedances are low.

Regards,

Shadders.

[diecaster]

2 hours ago, opus101 said:

 

Fixed it for you - you do seem rather in love with AB when that has its own issues. Are you aware of any of them?

 

My Class A/B tube amp sounds better than my solid state Class A amp. Go figure....

[fas42]

6 hours ago, Jud said:

 

In different words, what is it about IM distortion that leads many people to describe the sound of NOS DACs, for example, as "natural"?

 

Anyway, this is getting off topic.

 

What makes audio reproduction "live", "natural", "have the oomph of the real thing" is lack of distortion - not additives. Why NOS DACs may appeal is that their topology limits other, more irritating artifacts - as almost always, it's not the obvious, 'measurable' distortion that matters, it's the stuff that nearly everyone hasn't got a handle on.

 

Once one has achieved competent sound in a rig, it is then very educational to see how easily this quality can be disrupted - have the slightest whiff of an unpleasant artifact intrude, and the sound literally becomes "unlistenable to".  Generic distortion is normally miles from being the cause of this, so discussions of one number being slightly better than another, for some reason, bear close to zero relevance to the matter.

[sandyk]

8 hours ago, psjug said:

Here is soundstage measurements of the same Bel Canto amp that sterophile measured.  It looks like the difference is in the amplifiers

 

https://www.soundstagenetwork.com/index.php?option=com_content&view=article&id=1515:bhk-labs-measurements-bel-canto-design-e-one-ref600m-mono-amplifiers&catid=97:amplifier-measurements&Itemid=154

 

Another one from Bel Canto

[fas42]

6 hours ago, esldude said:

Maybe similar reasons some like low level distortion in LP , tubes or tape? Sounds more richly detailed.

 

Ummm, try the other way around, perhaps ... . "More richly detailed" is precisely what well sorted solid state digital delivers - when one has experienced what is possible, then LPs and tape are just a variation of tonality, which may appeal if one's digital is not up to scratch ...

[sandyk]

8 hours ago, opus101 said:

There's an issue with maintaining a static bias point for the output stage idle current as the output stage temperature varies.

Quote

Is it not the purpose of the output stage bias/Vbe multiplier, whose transistor(s) are located on the same heatsink, is the regulation for thermal aspects, such that as the temperature rises, the Vbe voltage reduces, reducing the output bias of the output stage, thus reducing the temperature. A closed loop system ? -Shadders

Some designs use On Semiconductor "ThermalTrak" transistors , which have integral diodes.
 The literature for these devices claims that they eliminate the need for  quiescent current adjustment,
 as well as providing much better thermal tracking than a traditional VBE multiplier circuit.-Silicon Chip magazine

[Shadders]

6 minutes ago, sandyk said:

Some designs use On Semiconductor "ThermalTrak" transistors , which have integral diodes.
 The literature for these devices claims that they eliminate the need for  quiescent current adjustment,
 as well as providing much better thermal tracking than a traditional VBE multiplier circuit.-Silicon Chip magazine

Hi,

Yes - there are alternative methods, but the Vbe multiplier circuit works. The variation in bias is negligible once the designed stable bias has been reached.

 

A well designed amplifier, is immune to performance changes if there are small changes in bias current. It really is not an issue, else every amplifier would be using ThermalTrak. The best output devices (On Semi, Sanken) do not have ThermalTrak after 20 years of availability. ThermalTrak is not really needed as existing methods are ok.

 

The device selection is critical - output devices (and drivers) do need to have low beta droop to ensure performance.

 

Regards,

Shadders.

[March Audio]

9 hours ago, psjug said:

Here is soundstage measurements of the same Bel Canto amp that sterophile measured.  It looks like the difference is in the amplifiers

 

https://www.soundstagenetwork.com/index.php?option=com_content&view=article&id=1515:bhk-labs-measurements-bel-canto-design-e-one-ref600m-mono-amplifiers&catid=97:amplifier-measurements&Itemid=154

 

 

Helpful info Implementations of course have the possibility of causing negative effects. The point to take away is that it's not a fundamental problem of class d that has been stated by another poster. 

[sandyk]

14 minutes ago, Shadders said:

The device selection is critical - output devices (and drivers) do need to have low beta droop to ensure performance.

 

 Especially into the now more common 4 ohm loads.

 

 I agree with what you are saying about existing methods, and that variation in bias is negligible once the designed stable bias has been reached.

 

[GeneZ]

On 1/6/2019 at 5:31 PM, fas42 said:

Listening to non-optimised audio rigs sucks - it only takes listening to most of the expensive rigs at an audio show to show how far we still have to go on that one ...

  

With 'trickle down' we can all wear a crown. 

 

😏 Patience...

[GeneZ]

A little off topic...

 

Class D: Turns Out it Does Suck!

 

 

Anyone ever wonder?   Why did the word "suck" end up meaning something with a negative connotation?