I achieved -290 db distortion about a year ago using a design I came up with. The distortion doesn't change much with load until the parts are literally melting off the board.
How did I measure it? Well the design is a bit like an error correction system so I used an amplifier with a known -150db distortion and drove it out of spec until it reached to 0db distortion and then applied the EC. Using a QA401 and Cordell's distortion magnifier with quality oscillators I can get -150db measurement on the QA401 itself. Giving me a total measurement range of -300db. So I take the result of the post EC measurement and add it to the amplifiers original distortion number. If this is somehow an invalid testing method let me know.
Viewable results here
I got -290 dB Distortion
Edit: Hey mod, could you name the thread something a little less boasty and errogant? Kinda not my style. Especially since it's not confirmed yet. I kind of didn't want to advertise the fact either.
Last edited:
You are probably fooling yourself. Any pictures of the distortion spectra and noise floor that you are interpreting to be at this level?
Using Bob's distortion magnifier the noise floor is unlikely to be below 4 or 5nV so the equivalent BW of the measurement needed would be ~.0000006Hz or about 450 hours of data. This would also require some kind of synchronous averaging or auto-correlating instrument.
I didn't take any pictures because it was being fed through the distortion magnifier in an odd set-up so it didn't mean much to anyone but me. I was over driving the amp by pushing the output voltage swing very close to the rails so there was distortion on every harmonic, mostly on the odd harmonics if I recall. I tweaked the gain until the distortion peaks hit 0db. Then I applied the EC and measured again to measure the improvement of linearity above 0db that the EC provided. This effectively gave me a -300db distortion measurement range. Noise floor is -130 +/- 5db on the QA401 without the distortion magnifier. I didn't really care about the noise floor measurement since I couldn't see it on the scope with a 100x magnifier so pretty inconsequential I think.
Last edited:
What??? -290dB distortion means a total of 3 femto volts harmonics to discriminate from noise. That means you either kept the DUT close to 0 Kelvin, or you spent years in auto correlating the signal from noise.
Okay I'll try to explain it once more as clear as I can. I have a power amplifier that I know measures at about -150db distortion. I also have an EC system that I will apply to this amplifier at a later point.
So, let's call the -150db amplifier A1, and the EC circuit A2.
Step 1: I applied a load to A1 and increased the gain until the voltage swing at the output was close enough to the rails to put A1 into a non linear state, very very close to the point of clipping. I tweaked this gain until the harmonics on the distortion analyzer peaked at 0db.
So what we have right here now is an amplifier that originally had a distortion figure of -150db and has been overdriven to now have a 0db distortion figure.
Step 2: I applied A2 to A1 in this newly overdriven state and measured the distortion. Since A1 is now 0 db, any distortion I measure when A2 is attached to A1 is going to be the improvement that A2 provides to the entire system in this overdriven state.
In other words, If we measured an improvement of -140db and we now stop driving the amp out of spec, A1 will return to its original distortion figure of -150db and and A2 will provide an addition bonus performance of -140db making the total distortion performance equal to roughly -290db.
Hopefully that clears up any misunderstandings. I'd like to hear your opinions.
So, let's call the -150db amplifier A1, and the EC circuit A2.
Step 1: I applied a load to A1 and increased the gain until the voltage swing at the output was close enough to the rails to put A1 into a non linear state, very very close to the point of clipping. I tweaked this gain until the harmonics on the distortion analyzer peaked at 0db.
So what we have right here now is an amplifier that originally had a distortion figure of -150db and has been overdriven to now have a 0db distortion figure.
Step 2: I applied A2 to A1 in this newly overdriven state and measured the distortion. Since A1 is now 0 db, any distortion I measure when A2 is attached to A1 is going to be the improvement that A2 provides to the entire system in this overdriven state.
In other words, If we measured an improvement of -140db and we now stop driving the amp out of spec, A1 will return to its original distortion figure of -150db and and A2 will provide an addition bonus performance of -140db making the total distortion performance equal to roughly -290db.
Hopefully that clears up any misunderstandings. I'd like to hear your opinions.
It did.
- To start with, a power amplifier with -150dB distortions is a big question mark by itself.
- Secondly, what is an "EC System" and how did you switch it in/out?
- Third, you need to understand that not all distortion types are correctable by a feedback loop. It all depends on the distortion type
- Fourth, you need to better understand the distortion reduction mechanisms. There is no LGM or daemon driving out distortions, it's all about negative feedback (in your case).
- Fifth, you need to better understand how distortions are added. It's not that simple as adding dB numbers
- Sixth, you need to better understand what an EC circuit does and what are it's limitations and cost (in terms of stability). Any EC circuit driving distortions down -150dB is highly suspicious.
I'll stop here...
It's pretty simple to make an amplifier have this kind of spec. You can buy ones with better specs (more or less). Especially a lower power one. Mine was only capable of about 10W into 8 ohms.
Error correction system. I switched it in by placing it in the loop.
I know, I'm just reading to you what the data and spice tells me.
I understand.
Normally I'd agree, but the circumstances I placed the test in were designed to show the combined performance of both systems. If I simply reduced the voltage swing at the output to stop being at near clipping the performance would remain the same but shifted to a lower point on the graph. The amplifier doesn't know any better. I've also tested it using a much worse version of the EC and main amplifier circuit where I didn't need to drive the main amplifier to 0db in order to see the distortion and the results of the improvement were consistent across a variety of circumstances. The main stated test was simply one of the more extreme tests that I did of that nature.
Well considering how I designed it I'm pretty aware of how it works. Stability is not really a concern as of yet. I haven't tried it in an very high power system yet. It's kind of expensive to test that since I don't own a high power system. It's also not really a true EC circuit, I'm not sure how I'd define it. It's sort of a mix of things. But I used it as an EC in the tests.
Last edited:
That is not correct. At the moment you apply the EC the distortion is 0dB, NOT -150dB. Then you apply the EC and get -150dB*, so now your amp has -150dB. That's all.
* I do question a) the original -150dB and b) the 150dB reduction by the EC. This is what the best minds on the planet have been working at for decades and not succeeding. You just did that on the kitchentable? (Sorry to exaggerate a bit for effect).
You really must provide more facts and figures, measurement protocols etc. Extraordinary claims require extraordinary proof.
Edit - this is LTspice, not actual hardware? Ideal parts, perfectly matched devices? What use is that?
Jan
One of us is misunderstanding. The -140db measurement after EC is applied is how the main amplifier behaves when pressed to 0db so if the main amplifier is returned to its normalized state then the distortions performances add. For example if I reduced the voltage swing on the output of the main amplifier to the point where its distortion is no longer 0db but -1db then the absolute performance will be -141db.
So in other words if the main amplifier is being overdriven to 0db and then I apply a -140db EC to it, the EC performance is independent of the main amplifier performance even though it is affecting the absolute performance of the system. As I've said I've done many tests with lower performing versions to confirm that the EC performance is consistent across all circumstances that I've tested such as current overload, voltage overload and even simply incorrectly connecting or powering the main amplifier in certain ways. The increase in distortion performance of the EC has always remained stable and predictable.
So if I get a result where the main amplifier is 0db distortion and the EC applies a -140db boost in performance to the 0 db state of the main amplifier then the absolute distortion figure of the full system will shift downward as I return the main amplifier to lower distortion state, IE in spec.
Therefore if the main amplifier has a normalized distortion figure of -150db, then the normalized distortion figure of the post corrected amplifier must be -290db because the -140db is always applied.
Well, I don't know why you think the -150db performance of the main amplifier is such a big deal. Various forms of high feedback or EC circuits can already do this and some are available for purchase on the open market. Benchmark and THX are an easy couple I can think of off the top of my head. There's multiple threads stretching back decades on this forum showing these things including this one.
I agree about extraordinary proof though. I've tested this from all of the angles that I can think of and I kind of don't want to post the schematic on the internet because I can use it to get out of my crappy life situation potentially. I'm used to getting screwed over by people so I don't want to waste what now amounts to years of effort. I've spent waay too much money R&Ding this thing and I'm close to having it in a salable state, as least as a lower powered system. High power system R&D is in the future if I can afford it. Although I'm not really sure how to get beyond this point without showing someone. I've kind of run out of expendable money. Not sure what to do.
No, but I do use spice to brainstorm and optimize. In spice I usually use an inverting amplifier and an external resistor divider between the input and output to see the resulting distortion in volts because the distortion is always below the noise floor when running a transient analysis. I've found the resistor divider method to be better anyway because it will show you things such as crossover distortion that won't show on the transient analysis and it's easier to optimize and compare designs.
Last edited:
Well you need to find a way to convince people that matter. And a picture is worth a 1000 words, as they say.
A start would be to show that you have the tech capability to measure -150dB (r? V?) distortion.
Jan
Well as I said I have a QA401 distortion analyzer plus Bob Cordell's distortion magnifier and also Victor's oscillators in 1khz, 10khz, 11khz, and 20khz as well as the power supply board that is offered here on the forums here somewhere.
When used together I get a maximum reading of a little more than -150db in the 100x setting. I've switched back and forth between the 1x and 100x setting to make sure I'm getting the right numbers and it lines up. It's too easy to break the -150db point with low powered amplifiers so the only way I could really measure the 'ultimate' version of the circuit was to artificially drag down the main amplifier so I can see what the EC performance is.
My confidence is pretty high because of the large variety of tests that I've done and the fact that results are always the same. But of course I need an outside opinion at some point regardless of how many tests I've done. I'm not sure what sort of 'picture' would be sufficient for anything useful other than a flat out schematic.
When used together I get a maximum reading of a little more than -150db in the 100x setting. I've switched back and forth between the 1x and 100x setting to make sure I'm getting the right numbers and it lines up. It's too easy to break the -150db point with low powered amplifiers so the only way I could really measure the 'ultimate' version of the circuit was to artificially drag down the main amplifier so I can see what the EC performance is.
My confidence is pretty high because of the large variety of tests that I've done and the fact that results are always the same. But of course I need an outside opinion at some point regardless of how many tests I've done. I'm not sure what sort of 'picture' would be sufficient for anything useful other than a flat out schematic.
OK so you put in a sine wave and drive it to the rails until it clips enough to make "0dB" distortion. To correct this would require the output to go beyond the rails, clearly this is not what is happening.
Primarily odd distortion is compressive, to correct it you need an expansive non-linearity. In general these things behave very poorly when put in a feedback loop. Another point is that non-linear behavior does not obey superposition so the 300dB would be just a number 150 + 150 with no basis in reality. If you are measuring a "lot" of off the shelf amplifiers at -150dB you are probably misinterpreting something along the way (the Benchmark does not claim that performance BTW).
Measurements here Review and Measurements of Benchmark AHB2 Amp | Audio Science Review (ASR) Forum
Last edited:
No, I didn't make it clip, I put it very near clipping. Just close enough to make it distort. So in other words it was close enough to the rails to cause the output stage to function in a very non linear range of operation to the degree where the feedback of the amplifier could not correct it.
I used the scope to make sure I didn't actually chop off any of the top of the signal. But as I said I've also tested current overdrive and incorrectly connecting and configuring the circuit in certain ways and the results are the same in terms of improvement.
Generally speaking it seems that if it is capable of correcting it will seemingly always correct to the same degree in DB. At least as far as my tests have gone.
I used the scope to make sure I didn't actually chop off any of the top of the signal. But as I said I've also tested current overdrive and incorrectly connecting and configuring the circuit in certain ways and the results are the same in terms of improvement.
Generally speaking it seems that if it is capable of correcting it will seemingly always correct to the same degree in DB. At least as far as my tests have gone.
Last edited:
Since a square wave does not have 0dB harmonic distortion you are clearly using another definition of THD.
What is your dB reference? dBr? dBV? referred to what?
And you really need to spend some time to write something like a test report with all details, test protocol, result graphs, clear explanation step by step. As Scott mentioned, measuring -150dB referred to say 1V is not something that's done routinely, and it is definitely NOT true that there are lots of power amps that do that, not even the Benchmark.
Maybe review the paper from the man who lead his name to this thread, he managed -180dB. Try to duplicate that as a start. All that can be done without disclosing your circuit.
A 4-sentence post on diyaudio won't make you rich. Not trying to discourage you, but getting famous is hard work ;-)
Jan
Last edited:
- Status
- Not open for further replies.