It is quite normal for the person who does not understand the question he has just asked to come to the conclusion that people replying do not understand the question.hellokitty123 said:
Someone pops up and says that he has invented a special camera which enables him to see fairies at the bottom of his garden. He realises that most people will doubt him, but he wants to assemble some evidence to show to prospective investors so he asks us how to demonstrate that his camera works OK. Of course, he cannot tell us how the camera works. He is basically asking us how to test a black box with inputs and outputs which we regard as nonphysical. He has a special lens which can form a start-of-the-art image even when viewing through a dirty window; he assumes that it will do so much better and see the fairies so much more clearly when he cleans the window. He bats away our comments that seeing fairies at any level of resolution is extremely unlikely.
I'm not sure what being Russian (or thereabouts) has to do with anything. This forum is thankfully largely empty of international politics. People from one part of the world may sometimes find that people from another part of the world think differently about some technical issues but that is not surprising. It is unlikely that any trolls on here have political motives.
Something I have occasionally seen is someone who wants to ask a newbie-level question without revealing too much to those he is trying to have a higher-level discussion with may use an alternative login (forbidden by the forum rules?) to ask the newbie question.
I understand the difference, but there lies an inflection point in practice where sinad and distortion become indistinguishable, or at least inseparable in measure. Quite unlike spice where we can make so many things perfect for all eternity as to push noise down below the purported distortion level (in that bin). So really, it's hard to talk about validating linearity at -290 dB without invoking fantastical instruments.
So, yes, we can talk about -290 dB distortion, but it must remain at this time a purely academic discussion without a test in sight to prove it disprove.
Concepts and silly notions. Where would we be without them? When a company announced it would be sending rockets into space, everyone got out their stopwatches to see how fast they would go bankrupt.
Here is a proposition:
The error amplifier cannot be embedded in the amp that is driving the load and within the loop gain of the main amp. This limits the degree of distortion correction. The error amp must be independent. In order for the error amplifier to mobilize large amounts of distortion correction, it must 'know' what to do. In other words, it needs a reference. We can provide a reference amplifier that swings like the amplifier driving the load. It does not need an output stage and can be fast. As it is basically driving fresh air, feedback, phase and gain margins can be set to ridiculous values. The error amp is now a slave correction amp - SCA.
Taking things one stage further, how much gain can the SCA have? Can it be a comparator?
Here is a proposition:
The error amplifier cannot be embedded in the amp that is driving the load and within the loop gain of the main amp. This limits the degree of distortion correction. The error amp must be independent. In order for the error amplifier to mobilize large amounts of distortion correction, it must 'know' what to do. In other words, it needs a reference. We can provide a reference amplifier that swings like the amplifier driving the load. It does not need an output stage and can be fast. As it is basically driving fresh air, feedback, phase and gain margins can be set to ridiculous values. The error amp is now a slave correction amp - SCA.
Taking things one stage further, how much gain can the SCA have? Can it be a comparator?
My first problem with this thread was that alhough distortion is a "bad thing", it´s annoying and it should be avoided, it must first be perceived so as to become annoying.
So if anything is 140dB BELOW noise level, only way to detect it is (forget hearing, it is to use test instruments, computing power, analyze noise for HOURS/DAYS/MONTHS/YEARS .. what my (or any Human) ear does NOT do, then why is it even worth posting here?
Same as being worried because a Martian flea bit a Martian cat ... in Mars ... and obsessing about its influence on a Dire Straits album I am listening to ... on Earth.
That such un-event reaches >400 posts on DIYAudio only means a lot of people have a lot of spare time on their hands and nothing better to do.
Oh well.
So if anything is 140dB BELOW noise level, only way to detect it is (forget hearing, it is to use test instruments, computing power, analyze noise for HOURS/DAYS/MONTHS/YEARS .. what my (or any Human) ear does NOT do, then why is it even worth posting here?
Same as being worried because a Martian flea bit a Martian cat ... in Mars ... and obsessing about its influence on a Dire Straits album I am listening to ... on Earth.
That such un-event reaches >400 posts on DIYAudio only means a lot of people have a lot of spare time on their hands and nothing better to do.
Oh well.
Last edited:
Correct, theoretically, but irrelevant from any practical perspective.
Assume you have an audio power amplifier with a SNR of 100dB at the room temperature of 300k. A 1M FFT in 20Hz-20KHz will render a noise floor of -100dB - 10*Log(1,000,000/2)=-157dB and now you cool it down to 4K (liquid helium). The noise floor of the same system will now be -176dB.
To further lower the noise floor you need to synchronous average the FFT. Set aside where are you getting the trigger source (at those level it is unlikely you can use the signal itself), to lower the floor noise by 80dB you need 100,000,000 averaging (since the noise floor goes with the square root of number of averaging).
Assuming that each 1M FFT takes 1 second (which is rather optimistic, I would say) you will need 3.17 years of 24/7 synchronous averages every second to get to -256dB. Meantime, you have to keep the system @ 4K above zero.
Last edited:
It's likely that was a display of the readout, (monitor output), of one of his distortion analyzers. Or he place a notch filter before the input of the QA. It's also possible he notched the data using his Minidsp or other. He does this to show the principle from time to time.
dBr is relative to some level, could be anything. dBV is referenced to 1Vrms.
My first experience of error correction was as a teenager when Quad introduced current dumping.I started designing and building my own valve amps aged 10.Now I`m 63 yrs old and can still see the attraction of the quest for distortion free amplification and so it seems can many others on this forum (heated posts).I do continue to maintain -290dB is not on.
The thing to do is file a patent app this will give priority then PUBLISH the circuit and method on this forum there is a world of expertese here and critical peer revue is no bad thing.There may just be some merit in the topology.
The thing to do is file a patent app this will give priority then PUBLISH the circuit and method on this forum there is a world of expertese here and critical peer revue is no bad thing.There may just be some merit in the topology.
See post 266. I think that my remarks about second-order effects that will limit the performance are practically very relevant.
Ah, ok, I missed that. Of course, second order effects are limiting well before three years
.No need to prefix "K" with "Degree". Kelvin (K) is the unit for temperature, defined as one of the seven SI base units.
SI base unit - Wikipedia
So the absolute zero temperature is written as 0 K.
SI base unit - Wikipedia
So the absolute zero temperature is written as 0 K.
The PCB image on the Cherry paper has a rather low quality.
That's why I have drawn in red the supposedly missing traces.
Here he connects resp. 1) the cold Input, 2) the cold LS wiring and 3) the Power supply Gnd all "somewhere" to the chassis.
This has to be handled with much greater care, not to nullify with ease all taken precautions to reduce mutual induction.
So what you do sounds a lot better.
Hans
Attachments
I see primarily as minimizing the loop areas - which would of course minimize the inductances Cherry mentions as well. There are clearly a number of ways to achieve this as there are a few designs on the forum at ppm levels that don't use this specific layout, but the designers have simply found a way to overcome the issue through good layout practice.
I like the way he has laid the emitter degen resistors next to the on board filter caps but balk at the speaker and input connections being next to each other.
(Thanks for clarifying the drawing BTW)
I like the way he has laid the emitter degen resistors next to the on board filter caps but balk at the speaker and input connections being next to each other.
(Thanks for clarifying the drawing BTW)
I have no idea what error correction was invented.
I made several attempts at inventing an EC.
It turned out, every time it was a variant of negative feed back.
So I gave up and now think there is nothing else than NF.
I suspect that most ECs are just variants of negative feed back, whatever smart and looking different.
I even think, this can be proved from equations that can be implemented in different ways.
I made several attempts at inventing an EC.
It turned out, every time it was a variant of negative feed back.
So I gave up and now think there is nothing else than NF.
I suspect that most ECs are just variants of negative feed back, whatever smart and looking different.
I even think, this can be proved from equations that can be implemented in different ways.
- Status
- Not open for further replies.