Sound Quality Vs. Measurements

[DF96]

To assert the presence of 1KHZ residual there should be peaks
comparable to the 18 and 21KHZ ones , that is the peaks have two clear bars
with a minimum of width between them.

Why do we need to have any 1kHz? That is second-order. The 18 and 21kHz are third-order. You can have one without the other. The apparent absence of 1kHz second-order IMD says precisely nothing about the presence of third-order IMD at 18 and 21kHz.

If there were any 1kHz, it would be a single peak - not two peaks with a gap between.

 

[wahab]

This is a linear scale so a peak at 1KHZ if there was one should
appear in the graph with the same shape as the mentionned third
order products.

That said i generaly have relatively high 1khz products in simulations
wich invalidate the claims that simulators are linearity friendly...

 

[ThorstenL]

Hi,

This is a linear scale so a peak at 1KHZ if there was one should appear in the graph with the same shape as the mentionned third
order products.

And it is there, compressed in width due to the use of linear scale.

So, please note:

1) There is a 1KHz at around -70dB (~0.03%) in the Plot
2) If 2nd hd was very much lower it would not show up at all

Ciao T

 

[DF96]

This is a linear scale so a peak at 1KHZ if there was one should
appear in the graph with the same shape as the mentionned third
order products.

Yes, one peak. You appeared to be asking for two peaks.

That said i generaly have relatively high 1khz products in simulations
wich invalidate the claims that simulators are linearity friendly...

Or could be that your designs have relatively high levels of second-order distortion, even with ideal perfectly-matched devices?

 

[ThorstenL]

Hi,

Moreover , if we look at the Goldmund s IMD graph , wich according to Thorson is a measurement at 20V PP ,that is 7V RMS , we can see that at 1KHZ there s nothing else than a notch and the close peaks
are not distinguished from noise.

First, my name is Thorsten. It is actually quite clear.

Second, there is a peak at 1KHz. It is above the noisefloor.

Anyway, the output level allow the amp to stay in a better linearity zone, albeit the 18/21khz products are high for a so called high end amplifier.

On what basis do you claim that the 3rd order products "are high for a so called high end amplifier".

Can you present evidence that suggests audible problems being introduced by around 0.03% 3rd HD/IMD at 7V output (that would be a peak SPL of around 92dB in room for a stereop pair of "average" Speakers per JA's research).

And do you have evidence for Loudspeakers that have less than 0.03% 3rd HD or 3rd order IMD when having 7V RMS applied?

Ciao T

 

[ThorstenL]

Hi,

That said i generaly have relatively high 1khz products in simulations wich invalidate the claims that simulators are linearity friendly...

A sufficiently ham-fisted design will also show high levels of non-linearity even in a simulator.

However, taking your previous simulator example in post 1286 that you attempted to pass as actual performance measurements:

I so no particular evidence of significant levels of 1KHz IMD products contrary to your claim.

And I note that at a claimed 34V Peak (that is around 10dB higher level or 10 times the power of the Stereophile measurement and incidentally 72W into 8 ohm) shows the 3rd order products at around -100dB (around 0.001% IMD), which is around 30dB better than real measured performance of a suitably similar amplifier at 10dB lower level as shown by stereophile.

So, in this case I would suggest that the performance from the simulator will be hard to replicate using real devices, but I will be convinced otherwise if you can actually show real measurements.

Ciao T

 

[wahab]

Yes, one peak. You appeared to be asking for two peaks.


Or could be that your designs have relatively high levels of second-order distortion, even with ideal perfectly-matched devices?

One peak that has width as the visible ones at high frequency.

It s when simulating any amp , not only the one i design myself.

Generaly i noticed that the first stage linearity and gain
are instrumental in reducing IMD and a current mirror
as the input differentials load is mandatory to have
top of the barrel IMD performances.

 

[wahab]

Hi,

First, my name is Thorsten. It is actually quite clear.

Sorry for mispelling your name, it was unintentional...

Second, there is a peak at 1KHz. It is above the noisefloor.
On what basis do you claim that the 3rd order products "are high for a so called high end amplifier".

Can you present evidence that suggests audible problems being introduced by around 0.03% 3rd HD/IMD at 7V output (that would be a peak SPL of around 92dB in room for a stereop pair of "average" Speakers per JA's research).

And do you have evidence for Loudspeakers that have less than 0.03% 3rd HD or 3rd order IMD when having 7V RMS applied?

Ciao T

The 1KHZ peak is quite non significative , i think you ll be agree
on this point.

As for the 0.03% ratio i agree that it s surely below audibility
but since IMD is nastier than THD and that current technology
allow for a magnitude better numbers , why not ?...

High distorsion of speakers should not be an incentive
to neglect amplifiers parameters that seems to largely
surpass the speaker s possibilities.

Hi,
A sufficiently ham-fisted design will also show high levels of non-linearity even in a simulator.

However, taking your previous simulator example in post 1286 that you attempted to pass as actual performance measurements:

I so no particular evidence of significant levels of 1KHz IMD products contrary to your claim.

And I note that at a claimed 34V Peak (that is around 10dB higher level or 10 times the power of the Stereophile measurement and incidentally 72W into 8 ohm) shows the 3rd order products at around -100dB (around 0.001% IMD), which is around 30dB better than real measured performance of a suitably similar amplifier at 10dB lower level as shown by stereophile.

So, in this case I would suggest that the performance from the simulator will be hard to replicate using real devices, but I will be convinced otherwise if you can actually show real measurements.

Ciao T

This is an unadequate exemple since this design is among
the best i managed to bring from simulation to building while
keeping a good stability and reproducibility despite a 16 transistors
front end , but generaly numbers are not as good when using
simpler designs , wich i prefer , though.....

Could you provide me with the link of the preview you did
find in the net?.



.

 

[ThorstenL]

Hi,

The 1KHZ peak is quite non significative , i think you ll be agree on this point.

I agree the 1KHz peak measured for the Goldmund Mimesis 8 Amplifier is low in level, but it is at the level I would expect from the THD Measurements.

As for the 0.03% ratio i agree that it s surely below audibility but since IMD is nastier than THD and that current technology
allow for a magnitude better numbers , why not ?...

But why? I fail to see any merit to doing so.

Sure, we can make it better than this, but what do we expect to gain for making it better? And what price do we need to pay to attain such an improvement?

As for IMD being nastier, IMD is simply a different angle on HD, if the underlying non linearity is simple and consistent. If the non-linearity causing the HD and IMD is inaudible then it is inaudible, no matter if you measure HD or IMD.

High distorsion of speakers should not be an incentive to neglect amplifiers parameters that seems to largely
surpass the speaker s possibilities.

But what is the benefit?

The Speaker will limit the system performance, an amplifier with much lower 3rd HD than the speakers will not benefit the system.

This is an unadequate exemple since this design is among the best i managed to bring from simulation to building while
keeping a good stability and reproducibility

Yet in post #1286 you presented the graph like this:

IMD 1920 products seems quite high unless it is close to clipping zone.

An exemple of an average 100W amplifier using lateral fets :

So what you claimed to be an "average 100W amplifier" has suddenly become "this design is among the best i managed"?

So, what you showed not only was a simulation and not actual measurement when you criticised another amplifier designs actual measured performance which was worse than your simulation.

In addition, when you claimed to show an "average" design's results you actually did not show an "average" design, but one the of best you did.

Interesting.

Could you provide me with the link of the preview you did find in the net?.

Sorry, I do not understand, what preview?

Ciao T

 

[wahab]

Hi,

But why? I fail to see any merit to doing so.

Sure, we can make it better than this, but what do we expect to gain for making it better? And what price do we need to pay to attain such an improvement?

As for IMD being nastier, IMD is simply a different angle on HD, if the underlying non linearity is simple and consistent. If the non-linearity causing the HD and IMD is inaudible then it is inaudible, no matter if you measure HD or IMD.

But what is the benefit?

Improving an amp linearity will improve all parameters , IMD among others.

So what you claimed to be an "average 100W amplifier" has suddenly become "this design is among the best i managed"?

In addition, when you claimed to show an "average" design's results you actually did not show an "average" design, but one the of best you did.

There s no contradiction , it s one of my best but it is average
compared to many famed and highly regarded amps.

Sorry, I do not understand, what preview?

Ciao T

This one :

Goldmund Mimesis 8 power amplifier Measurements | Stereophile.com

When you did display their IMD measurement i was surprised
by the difference with simulations , but i had a doubt about
the sample that was sent to stereophile so i made some sims
and approached accurately the measurement results by increasing
the two 4.7pF Cdom to 47pF.

Then reading stereophile review of the Mimesis , i noticed this sentence :

the risetime, judging from the 10kHz squarewave shown in fig.7, was significantly higher than expected at 4.4µs vs 0.7µs.

Well , Goldmund did send a sample that was way more compensated,
surely to tame down the worst thing that can be said of an amplifier
in a review ,far ahead of other parameters as THD/IMD/DF ect
, i.e, instability...

Below are the sims of IMD and rise time of the Mimesis 8.
We clearly see that the rise time given in the review correlate
with the increased compensation....

 

[a.wayne]

Hi,



Amplifier, details of it's design and how it fared in subjective evaluation, the whole Stereophile article is here:

Goldmund Mimesis 8 power amplifier | Stereophile.com

Having heard an Amplifier from this generation (not sure if it was a Mimesis 8 or Mimesis 9 though - I suspect a 9 though) I have to say I liked it by far better than most solid state amplifiers I had the chance to hear. It is one of the few SS Amplifiers I'd be willing to live with.

It way bettered big ML Monoblocks we had for comparison in all areas but it played second fiddle to an open-loop tube amplifier (845 Single Ended), so personally I'd probably take a Kondo Ongaku or Baransu if I could have ANY Amp I liked but only one.

Speakers during this evaluation where quite high efficiency I might add (15" PHL Woofer & 2" Compression Driver/Horn giving around 98dB/W/m), which does play a role.

Ciao T

It appears avg at best on the test bench and sonically. I would suggest the weak sloppy bass was due to the small PSU and miniscule output stage, well vs the rest ...

 

[tvrgeek]

The HCA1200 was built in Taiwan, very different from mainland China, the factory looks like one might look in Silicon Valley. Don't worry too much, you will just break something.

I shall be respectful.

Yes, the island is very different. I stand corrected.

 

[john curl]

Your Rotel was probably made in the same factory. '-)

 

[mr_push_pull]

this is probably my preferred subject but I wonder why I imagined that it will turn into anything but few people fighting over personal obsessions.

we're fighting over SS vs tubes, FB vs NFB, PP vs SET but we somehow forget that our auditory system does not care about these but only on the final result.

any, and I repeat ANY discussion about audio is absolutely irrelevant if psychoacoustics are ignored. I wish there were more serious studies on this (correlation of perception with measurements) but my personal research showed that real scientists seem to be uninterested in it.

I'm somehow tired of reading assertions from self-proclaimed gurus when could use science.

 

[mr_push_pull]

PS: I would add that the few designers which are also interested in the way our ear/brain system works seem to consistently deliver good, widely recognized products

 

[Wavebourn]

The problem is, the final result depends on topology. That means, for different topologies different weak points need to be considered, so proper measurement of them would lead to better final result.

 

[mr_push_pull]

The problem is, the final result depends on topology. That means, for different topologies different weak points need to be considered, so proper measurement of them would lead to better final result.

see? that's exactly what I was talking about.

IMO, this a random assertion. and unless it's backed up by some research it stays that way.

what needs to be studied (and for some reason isn't) is what makes a specific topology sound better.

 

[a.wayne]

this is probably my preferred subject but I wonder why I imagined that it will turn into anything but few people fighting over personal obsessions.

we're fighting over SS vs tubes, FB vs NFB, PP vs SET but we somehow forget that our auditory system does not care about these but only on the final result.

any, and I repeat ANY discussion about audio is absolutely irrelevant if psychoacoustics are ignored. I wish there were more serious studies on this (correlation of perception with measurements) but my personal research showed that real scientists seem to be uninterested in it.

I'm somehow tired of reading assertions from self-proclaimed gurus when could use science.

You do realize we are talking about an subjective evaluation of the science, problem is most scientist can't hear . so they discount what they do not understand.

Getting a group of inexperienced listeners together who are easily fooled, rigging the results is not science ...

Just saying ...................

 

[a.wayne]

The problem is, the final result depends on topology. That means, for different topologies different weak points need to be considered, so proper measurement of them would lead to better final result.

Ask before , so i will sneak it in again , why no Hybrids, seems it would be the best compromise between all topologies, so why not more Hybrids...?

 

[mr_push_pull]

You do realize we are talking about an subjective evaluation of the science, problem is most scientist can't hear . so they discount what they do not understand.

Getting a group inexperienced listeners together who are easily fooled, rigging the results is not science ...

Just saying ...................

I totally agree with this.

I've read many papers on various aspects of psychoacoustics. most of the times, nothing is mentioned about: the subjects used (are they trained listeners? are they average listeners? are they young, old, proven to have good hearing etc?), used gear, listening room etc. also, short sound samples are often used (isolated piano, sax you get the idea), while things can largely change with real music. also, many aspects of perception seem to be deliberately ignored (transient portion of notes is ignored, only the steady-state portion is analyzed).
it kinda makes sense, why would a scientist having all the meas at hand (most likely working for a corporation) be interested in something that appeals to a small percentage of the public? well... it's sad.