A low FFT “noise floor” is not the equal of low real time noise.
Hi Chris,
I think that you and I are pretty much in the same place regarding the noise.
We have wonderful tool available to us. With big enough sample size and many times over averaging, random noise will pretty much disappear from a FFT plot. With small FFT bin sizes the apparent “noise floor” falls away. Ed Simon shows -170dB “noise floors” in his resistor distortion plots.
A low FFT “noise floor” is not the equal of low real time noise.
Real time noise will swamp the low level distortion products that may be visible in a FFT plot.
DT
Hi Chris,
I think that you and I are pretty much in the same place regarding the noise.
We have wonderful tool available to us. With big enough sample size and many times over averaging, random noise will pretty much disappear from a FFT plot. With small FFT bin sizes the apparent “noise floor” falls away. Ed Simon shows -170dB “noise floors” in his resistor distortion plots.
A low FFT “noise floor” is not the equal of low real time noise.
Real time noise will swamp the low level distortion products that may be visible in a FFT plot.
DT
Good to remind us again of the real power of an FFT - to eek out the harmonic components from the noise floor.
Jan
As long as the sampling frequency remain stable, yes, that's a huge advantage of long-FFT's. (Time averaging, baby!)
I can't not wonder if there is something that might render the results of very long averaging unreliable. I agree that long averaging is the best to drag out harmonics and mains interference, providing a valuable tool for researching a problem.
I have to really work to accept those numbers as any statement of performance. Getting it to a level that it is repeatable in different labs would go a long way towards creating a metric that could be useful. Imagine the field day the advertisers would have with this!
-Chris
I have to really work to accept those numbers as any statement of performance. Getting it to a level that it is repeatable in different labs would go a long way towards creating a metric that could be useful. Imagine the field day the advertisers would have with this!
-Chris
Long averaging requires real stability in the sources. For noise you are really averaging the noise at a given frequency which is good. But for any stationary signals that are not that stationary they will be misrepresented.
Coherent averaging will actually remove any random signals and reduce the noise unmasking stationary signals (e.g. melt noise on the QA101) but its not an indication of noise energy.
High resolution FFTs also get very low noise in each bin. Useful for the same type of unmasking. But the actual noise is still the same, you are only getting tiny slices of it. Again you need really good stability for that analysis. Analog RC oscillators drift enough that you can see the harmonics moving from bin to bin and getting smaller than their actual energy.
Coherent averaging will actually remove any random signals and reduce the noise unmasking stationary signals (e.g. melt noise on the QA101) but its not an indication of noise energy.
High resolution FFTs also get very low noise in each bin. Useful for the same type of unmasking. But the actual noise is still the same, you are only getting tiny slices of it. Again you need really good stability for that analysis. Analog RC oscillators drift enough that you can see the harmonics moving from bin to bin and getting smaller than their actual energy.
Hi Demian,
Those are the types of things I was considering actually. I would think that your stimulus signal and sample clock should be locked and come from the same piece of equipment. You might need to assess the distortion in your stimulus signal and subtract that from the resulting output.
-Chris
Those are the types of things I was considering actually. I would think that your stimulus signal and sample clock should be locked and come from the same piece of equipment. You might need to assess the distortion in your stimulus signal and subtract that from the resulting output.
-Chris
Using the same clock can hide issues. Stabilty and isolated clocks plus the right frequencies will show problems that can hide otherwise.
I try to measure 44.1 systems with 48 clocks.
Sent from my LG-H811 using Tapatalk
I try to measure 44.1 systems with 48 clocks.
Sent from my LG-H811 using Tapatalk
Patiently waiting for release. Don't expect a lot of chatter until that time, or there is some kind of development.
-Chris
-Chris
We have no idea how much unpaid overtime he is having to do to complete this project inside the cost he screwed out of his employer !
Hi Andrew,
Why, of course or I wouldn't have posted that in the first place.
Your comments are specific and normally wouldn't be known outside of the parties involved directly. So:
If you know directly from Jens, I doubt he would want the information out in the public.
-Chris
Why, of course or I wouldn't have posted that in the first place.
Your comments are specific and normally wouldn't be known outside of the parties involved directly. So:
- How do you know this
- How do you know the specifics
If you know directly from Jens, I doubt he would want the information out in the public.
-Chris
I started with "we have no idea"
That tells you that I am guessing why he has little time to keep us informed of progress on an hourly basis.
Knowing that he gave us a price that he estimated was about half of the expected retail cost, I further guess that his employer is making little or no profit in the batch production for us.
I know from experience that costs can easily run away.
Somewhat tongue in cheek, I suggested that maybe he was so busy trying to keep the project going without incurring any extra costs, that he was doing some unpaid overtime.
But that is all fiction and should not be taken as anything else.
That tells you that I am guessing why he has little time to keep us informed of progress on an hourly basis.
Knowing that he gave us a price that he estimated was about half of the expected retail cost, I further guess that his employer is making little or no profit in the batch production for us.
I know from experience that costs can easily run away.
Somewhat tongue in cheek, I suggested that maybe he was so busy trying to keep the project going without incurring any extra costs, that he was doing some unpaid overtime.
But that is all fiction and should not be taken as anything else.
I guess we are all hoping it is still progressing. Certainly appreciate his effort.
Sent from my iPhone using Tapatalk
Sent from my iPhone using Tapatalk
It is not quite that bad 🙂
The GB does represent a huge discount though, which would not be possible on normal commercial conditions. I do put in some unpaid overtime (voluntarily) on some parts of the projects.
The preparations for production are still in progress. Expected to be ready from manufacturing late August or early September. So slightly delayed, mainly due to component sourcing issues.
The GB does represent a huge discount though, which would not be possible on normal commercial conditions. I do put in some unpaid overtime (voluntarily) on some parts of the projects.
The preparations for production are still in progress. Expected to be ready from manufacturing late August or early September. So slightly delayed, mainly due to component sourcing issues.
Damn awesome, just in time for the winter buckle-in with my fav hobby! atm, its kindof, mehh, sun is out, supposedly, and we just hit a new all-time record of how many june days have constant rain.. 26 days ! yeah, 26 damn days, and now I just had a shower from above, so were counting 27 days atm.. a few more and whole june was raining.
Glad the project is progressing.. cant wait.
Just wondering, Ive followed this thread for the last almost 2 years, and now that were waiting for the actual production, how would we expect the DAC part to perform as I can read the analogue part here is really really well designed ?
I mean compared to eg. a typically high end DAC ?
I mean compared to eg. a typically high end DAC ?