-290 dB Distortion?

Status
Not open for further replies.
You can already make a headphone amplifier that achieves nearly unmeasurable distortion without your error correction circuit. That's been done by JCX and a few others with composite amplifiers. It will be a hard sell compared to the THX AAA based amplifiers like the Massdrop AAA 789 for only $349.

Well, you guys can say what you want but based on all of my testing I know I can spank the THX specs. Even if my EC somehow does have reduced performance if the main amplifier is already low distortion which I still have strong doubts about, the amount of overkill I can apply to the EC is ridiculous so unless it is completely useless after a certain mythical point then it still comes out ahead. I still find it ridiculous the suggestion that the EC can be rendered useless after a certain point when there has been zero evidence of this during any of my tests and no applicably stated theory to back it up other than the noise floor issue which applies to all things anyway. I'm hoping that guy that Scott recommended me to can help me formulate some proof in one direction or the other.

Does the performance gain over the THX matter practically? No. It also wont matter without a decent amount of marketing. But if I can maintain this performance with a 100W+ amp then I can at least have a provable advantage to show.
 
Last edited:
I still find it ridiculous the suggestion that the EC can be rendered useless after a certain point when there has been zero evidence of this during any of my tests and no applicably stated theory to back it up other than the noise floor issue which applies to all things anyway.

🙄 I still find it ridiculous the suggestion that you are not trolling
 
Last edited:
About passive component non-linearity: if the gain depends on a resistor ratio and the resistors have essentially equal currents flowing through them, like in an ordinary feedback amplifier, you could try to cancel the effect of the resistor's voltage coefficients and temperature variations. Just choose an integer gain and use a string of series-connected resistors of equal type and value, mounted in a similar way. So for example, don't use 100 ohm and 2 kohm, but use 100 ohm and 20 times 100 ohm in series.
 
You are not correcting the basic flaw in your argument, which is first you state something hard to believe if not physically impossible and then your only proof is your own statement that things work as you say.

Not sure about trolling but sad you don´t *see* the weakness of your arguments.
And that being polite.

Well, you guys can say what you want 😕 but based on all of my testing 😕 I know I can 😕 spank the THX specs.
there has been zero evidence of this during any of my tests and no applicably stated theory to back it up other than the noise floor issue which applies to all things anyway.

In a nutshell: your statements are extreme, only "sworn witness" is yourself and that is not enough in the Physics World.
 
If I put my QA401 on high resolution (FFT bins are 1.48 Hz IIRC - I can go lower but no point) and 50 averages, with a loopback test I get a noise floor reading of -145 to -150 dBV. It takes over a minute for the grass to settle (might be longer - I’d need to stop watch it).

But, as soon as you switch in the -20 dB attenuator, the noise floor comes up. I use a separate -40 dB attenuator (99 Ohms and a 1 ohm ) for power amp testing.

On a power amplifier I have, the non EC distortion at 200 W out into 8 Ohms is 30 ppm. When the EC is connected the distortion goes down to 7 ppm which is the noise floor. So the distortion is sitting below the 7 ppm threshold. The EC provides 20 dB distortion suppression across the audio band and all power levels.

Measuring these levels and getting consistent results is very difficult (consistent means I can repeat the experiment on different days and get the same result for 5 or 10 tries, so it’s not one- off).
 
Last edited:
About passive component non-linearity: if the gain depends on a resistor ratio and the resistors have essentially equal currents flowing through them, like in an ordinary feedback amplifier, you could try to cancel the effect of the resistor's voltage coefficients and temperature variations. Just choose an integer gain and use a string of series-connected resistors of equal type and value, mounted in a similar way. So for example, don't use 100 ohm and 2 kohm, but use 100 ohm and 20 times 100 ohm in series.

True, valid point.
In a very low distortion amp, the feedback network can become the main contributor to distortion when thinking that a resistor is just a resistor and that only the wattage has to be correct.
Took me days to find out when I used SMD components for the first time.

Hans
 
hellokitty123 said:
I still find it ridiculous the suggestion that the EC can be rendered useless after a certain point when there has been zero evidence of this during any of my tests and no applicably stated theory to back it up other than the noise floor issue which applies to all things anyway.
You still don't get it, do you? Are you expecting us to produce a theorem which says this can't be done; is educated common sense not sufficient for you geniuses?

If one day you reveal exactly what you have done then someone will be able to tell you in detail why it doesn't work as well as you imagine; you may or may not understand their critique. In the meantime you will just have to accept our assurance that it doesn't work.
 
You still don't get it, do you? Are you expecting us to produce a theorem which says this can't be done; is educated common sense not sufficient for you geniuses?

If one day you reveal exactly what you have done then someone will be able to tell you in detail why it doesn't work as well as you imagine; you may or may not understand their critique. In the meantime you will just have to accept our assurance that it doesn't work.

Exactly. The likelihood that anyone discovers something truly novel in analog design with conventional discrete devices today seems low. Every time I think I see something new it’s already been described in writing well before I was born. Especially by someone who has gaps in their knowledge of the discipline.


Well, you guys can say what you want but based on all of my testing I know I can spank the THX specs.

It’s always harder than it seems.

If you really really want to prove it, build an amp and pot the whole thing in epoxy or whatever you need to feel secure, and then see if you can get someone with a brand new AP and experience using it to measure it for you.
 
Last edited:
How do you propose to correct something that is below the real noise floor? What does the EC 'correct' if it cannot see the signal? Its already been explained that if you want to see 'deeper' into the noise floor, you have to wait a long time (see mu earlier post - -145dB takes > 1 minute) and that is completely incompatible with an audio amplifiers BW requirements - but ok of course for instrumentation where you can live with very narrow bandwidths of <<1 Hz.


In a real world EC system the noise floor limits the performance - doesn't matter what the simulation says.


To put this in perspective, imagine you stand at the edge of the shore line at low tide. There are sharp rocks (harmonics) beneath the water level and you cannot see them to avoid them because of the waves, foam etc (noise). The only way to fix this is to drain the ocean.
 
If I put my QA401 on high resolution (FFT bins are 1.48 Hz IIRC - I can go lower but no point) and 50 averages, with a loopback test I get a noise floor reading of -145 to -150 dBV. It takes over a minute for the grass to settle (might be longer - I’d need to stop watch it).

The screen grab said .182Hz (262K FFT). That's why the noise floor of -120dB on the QA bothered me, if the DUT is that noisy I don't see how the DM can have a -150dB noise floor. And as I said before a simulator has a -300dB numerical noise floor (I don't know why anyone would compile SPICE with 128Bit FP math) capacitor and resistor distortion as well as any EM or induced magnetic effects are non-existent there are no thermo-couples or any non-ideality in passives, the answers are fantasy.

From Bob's comments on the DM
Avoids noise issues in source and in THD analyzer
Does not avoid noise introduced by DUT
DUT noise rejection is the job of the spectrum analyzer
 
Last edited:
How do you propose to correct something that is below the real noise floor? What does the EC 'correct' if it cannot see the signal? Its already been explained that if you want to see 'deeper' into the noise floor, you have to wait a long time (see mu earlier post - -145dB takes > 1 minute) and that is completely incompatible with an audio amplifiers BW requirements - but ok of course for instrumentation where you can live with very narrow bandwidths of <<1 Hz.


In a real world EC system the noise floor limits the performance - doesn't matter what the simulation says.


To put this in perspective, imagine you stand at the edge of the shore line at low tide. There are sharp rocks (harmonics) beneath the water level and you cannot see them to avoid them because of the waves, foam etc (noise). The only way to fix this is to drain the ocean.

You can keep on repeating the same nonsense, but it remains nonsense. As hellokitty123 made no claims about SINAD but just about distortion, he can measure the harmonics in as narrow a bandwidth as he likes. Hence, at least in theory, there is no such thing as distortion below the noise floor, as you can reduce the noise floor as much as you like by reducing the measuring bandwidth. In practice things like the linewidth of the signal generator and the measurement time may limit how low you can go.
 
Status
Not open for further replies.