The Fundamentals of FFT-Based Signal Analysis and Measurement in LabVIEW and LabWindows/CVI - National Instruments
Long FFT = small frequency bins = small amount of noise energy per bin. Obviously we're now integrating the noise over more bins so the total noise energy remains linear with time. No violation of physics required here.
If your signal/distortion spikes are stationary (stable oscillator/signal source), then they will sit in one bin, thus raising the per-bin SNR tremendously. A stable signal source isn't trivial, nor are the distortion products of the rest of the measurement chain going to be easy to get below -160 dB (at a bare minimum) to successfully measure a DUT to -150 dB errors, no matter how long your FFT is.
The fact that I have to explain this is *why* I made my initial comment in October, and why I followed up that we're dealing with someone who doesn't even know what they're actually asking for but wants others to do his/her work.
Long FFT = small frequency bins = small amount of noise energy per bin. Obviously we're now integrating the noise over more bins so the total noise energy remains linear with time. No violation of physics required here.
If your signal/distortion spikes are stationary (stable oscillator/signal source), then they will sit in one bin, thus raising the per-bin SNR tremendously. A stable signal source isn't trivial, nor are the distortion products of the rest of the measurement chain going to be easy to get below -160 dB (at a bare minimum) to successfully measure a DUT to -150 dB errors, no matter how long your FFT is.
The fact that I have to explain this is *why* I made my initial comment in October, and why I followed up that we're dealing with someone who doesn't even know what they're actually asking for but wants others to do his/her work.
I've made it clear my knowledge and learning has been specialized toward amplifier design only and my free time is limited. I can do without the snide remarks, thanks.
I think I will try the distortion magnifier before buying a sound card for use with third party software. Unless the QA401 can be used with third party software?
I think I will try the distortion magnifier before buying a sound card for use with third party software. Unless the QA401 can be used with third party software?
Hi hellokitty,
I have one of the RTX 6001 from a group buy last year, and it doesn't really show anything you could bank on below -130 dB. The QA401 sure won't. Have a look at analysers from companies like Audio Precision, Keysight or Stanford and you will see that their analysers won't either.
I had the same problem you are having and even used spectrum analysers connected to the monitor output of my HP 339A. That didn't work. Earlier I had tried some pretty good sound cards, but they also wouldn't reach that far down. Looking at this from the other side, how many products are honestly rated to -122 dB? A few only. How about -130dB? None.
Although it would be nice to think of your creation making fantastic numbers, it simply isn't realistic. As was suggested before, look at the thermal noise of a resistance and the first active device. These will limit how quiet your circuit could possibly be.
Best, Chris
I have one of the RTX 6001 from a group buy last year, and it doesn't really show anything you could bank on below -130 dB. The QA401 sure won't. Have a look at analysers from companies like Audio Precision, Keysight or Stanford and you will see that their analysers won't either.
I had the same problem you are having and even used spectrum analysers connected to the monitor output of my HP 339A. That didn't work. Earlier I had tried some pretty good sound cards, but they also wouldn't reach that far down. Looking at this from the other side, how many products are honestly rated to -122 dB? A few only. How about -130dB? None.
Although it would be nice to think of your creation making fantastic numbers, it simply isn't realistic. As was suggested before, look at the thermal noise of a resistance and the first active device. These will limit how quiet your circuit could possibly be.
Best, Chris
I'm honestly not trying to be snide, but before you ask a thousand questions and try to have everyone do your work for you, maybe step back, scale down your ambitions, and actually understand what you're trying to accomplish and the difficulties thereof?
Otherwise literally you are being exactly a help vampire. That's not going to help you get your work done, anyhow.
Yes I noticed that the major players had specs around -120db which I thought was weird. I mostly care less about noise, it's mostly harmonics I'm interested in.
But here of an example of such a measurement.
Sourced here Official THX Measurements of the Massdrop x THX AAA 789 | Audio Science Review (ASR) Forum So it must be doable.
Forgive me if I'm not a fountain of perfect knowledge. If my existence is such a burden to you then you have no need to respond to my posts. Get off your high horse please and thank you
But here of an example of such a measurement.
Sourced here Official THX Measurements of the Massdrop x THX AAA 789 | Audio Science Review (ASR) Forum So it must be doable.
Presumptuousness is the epitome of foolishness. I don't know what you think I'm "trying to accomplish" but your obviously way off if you think your reaction is warranted. All I'm doing is asking questions on a forum designed for asking questions in order to fill in the gaps in my knowledge for my specific questions which I have a very specific need to know in order to continue with my research and learning. I always spend at least 30 to 60 minutes and too often weeks/months researching answers to my questions before asking it here, not quite the help vampire. Also as I've said many times I have little free time to begin with.
Forgive me if I'm not a fountain of perfect knowledge. If my existence is such a burden to you then you have no need to respond to my posts. Get off your high horse please and thank you
Last edited:
-130 dBV noise floor is not too shabby. You should have no trouble measuring the THD of an amplifier as long as you can provide a clean signal. You'll either need to build a precision oscillator or have a look at Baxandall's summing circuit to extend the capability of your existing system.
I do.
Tom
Not true. My Audio Precision APx525 hits -158 dBV noise floor reliably on the analyzer section. Sadly, the analog signal source is considerably noisier.
Tom
I have one of Bob Cordell's distortion magnifier kits being mailed to me soon. CordellAudio.com - The Distortion Magnifier I'm considering one of the super duper low distortion oscillator kits offered by the Diyaudio member Vicnic but I don't pretend to have a full understanding of these types of measurements yet so I am not entirely sure how much I would benefit since the distortion magnifier subtracts the fundamental before feeding it into the analyzer.
If only you were closer
Last edited:
I haven't been able to get my amps to produce harmonics above -130db until right at the edge of clipping and I have several variations of the circuit in question which I need to measure in order to optimize the cost and performance. I designed it to be extremely low in noise as far as parts choice and component values go so I'm not worried on that front but I can't really do anything further until I can see the harmonics. Spice suggests that the harmonics are well down into unicorn land but I won't believe it until I see it.
I suspect that your simulated THD is far below the one you would measure in reality. So I suggest you should do your first measurements with something like one off the shelf soundcard with maybe 110dB dynamic range. I am sure you will find your harmonics earlier than expected.
Would you pls elaborate on the signal generator you used for this measurement.
Well yeah, simulations are almost always far better than reality. However I designed the circuit to push the limits on practical performance and even the simulation isn't able to display the harmonics because they are down so far. Although I think there is a way to remove the noise floor in LTspice right? It's sort of irrelevant regardless.
You may have wanted to read the earlier posts
I have a QA401, my problem is the performance is holding up to promise but I need better equipment to see the harmonics and optimize it.
The one installed inside the QA401.
Last edited:
You want to measure extremely low distortion levels and no subtractive technique will perfectly subtract the input signal, so I expect you still need a quite good oscillator.
If it's a limited number of measurements you want to do, you could try using makeshift LC filters (with air-cored coils and NP0, polystyrene or polypropylene caps) to suppress harmonics of the oscillator and the fundamental from the amplifier - although even a good-quality LC filter probably distorts more than -150 dB.
Hi Tom,
Your residual noise floor isn't real. Wayne said just about everything that could be said.
We can get down to -150 dB in RF land because it's all a 50R impedance, and the measurement band is way up there. No 1/f to worry about.
Anyway, your gear is lying to you and you're letting it. I get averaged noise down in the -140 dB land, sometimes lower. I don't believe it for a second though!
-Chris
Your residual noise floor isn't real. Wayne said just about everything that could be said.
We can get down to -150 dB in RF land because it's all a 50R impedance, and the measurement band is way up there. No 1/f to worry about.
Anyway, your gear is lying to you and you're letting it. I get averaged noise down in the -140 dB land, sometimes lower. I don't believe it for a second though!
-Chris
Its real enough, just its not a 'floor'. Rather its the contents of a single FFT bin whose bandwidth is likely very much less than 1Hz.
Hi abraxalito,
That may be, but I wouldn't call that situation a real or useful result. It is, however, something averaged that math gave us. Useful for seeing into the grass and locate signals we increase or decrease depending on what we are doing.
My big problem with the acceptance of Tom's floor is that there is then the urge to quote that figure to others. Lay people who aren't familiar with how test equipment obtains its results would assume a figure that they might be given from using a metered instrument. Something like an HP 339A or any other like instrument. Even the A/D convertor is probably only spec'd to -128 dB at best. That alone should be a huge hint.
-Chris
That may be, but I wouldn't call that situation a real or useful result. It is, however, something averaged that math gave us. Useful for seeing into the grass and locate signals we increase or decrease depending on what we are doing.
My big problem with the acceptance of Tom's floor is that there is then the urge to quote that figure to others. Lay people who aren't familiar with how test equipment obtains its results would assume a figure that they might be given from using a metered instrument. Something like an HP 339A or any other like instrument. Even the A/D convertor is probably only spec'd to -128 dB at best. That alone should be a huge hint.
-Chris
Okay so if averaging allows us to peer into the grass then could someone explain to me why Quantasylum decided to limit the averaging on the QA401 to only a few seconds?
Unless the stuttered measurement that the qa401 makes over time is equivalent to the long term averaging I was recommended earlier in which case the QA401 is capped around -135db.
Unless the stuttered measurement that the qa401 makes over time is equivalent to the long term averaging I was recommended earlier in which case the QA401 is capped around -135db.
Last edited:
Hello hellokitty123,
I think that somebody you want to explain this works as an employee for the manufacturer. It certainly isn't up to anyone here to second guess them.
I'll take a guess. It might have something to do with the size of the dataset long term averaging can generate. I think that the QA400 and QA401 are excellent instruments judging from how many people are happy using them.
I have a test and measurement background, so I might be looking at this differently than you might. Too bad that Audio Precision and Keysight units are so expensive. But when you see the lengths they have to go to for the performance you can understand at least some of the cost. Look at the size of those instruments and compare it to the size of yours, or even my RTX. Just the shielding alone costs a fortune! Never mind the power supplies and possible signal contamination from within the instrument. My spec-ans run from 60 ~ 80 pounds each. They are not quoting -150 dB numbers, nothing even close. I wish I could afford a U8903B!
-Chris
I think that somebody you want to explain this works as an employee for the manufacturer. It certainly isn't up to anyone here to second guess them.
I'll take a guess. It might have something to do with the size of the dataset long term averaging can generate. I think that the QA400 and QA401 are excellent instruments judging from how many people are happy using them.
I have a test and measurement background, so I might be looking at this differently than you might. Too bad that Audio Precision and Keysight units are so expensive. But when you see the lengths they have to go to for the performance you can understand at least some of the cost. Look at the size of those instruments and compare it to the size of yours, or even my RTX. Just the shielding alone costs a fortune! Never mind the power supplies and possible signal contamination from within the instrument. My spec-ans run from 60 ~ 80 pounds each. They are not quoting -150 dB numbers, nothing even close. I wish I could afford a U8903B!
-Chris
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.