A quick read of the code would say yes, that is what it does. The bandwidth = the noise bandwidth. I’m pretty sure I wrote it that way because it would seem strange to measure Noise+Distortion where you take the noise from one frequency range but distortion from a wider frequency range. I view it as you have a brick-wall filter with that bandwidth applied to the signal.
If you’ve got a reference that says this is wrong I’m happy to take a look at it again.
You are correct that THD is done by adding up the first N harmonics (up to 7 harmonics are taken up to sampling rate limit). I think I’m more likely to reduce the harmonics in THD to the noise bandwidth than vice versa take harmonics past the noise bandwidth in THD+N, but I’m open to other specs.
it were ongoing stalls, so that finally the programm need to be shut down by taskmanager.
This happened for example when changing generator voltage under a running spectrum measurement.
It´s no major issue. When one knows one can work around …. and yes, I know … changing a parameter on a software whilst it is running can always bear risks of stalling, or strange behaviour.
I’m still not sure what an ongoing stall is. Maybe a picture of the output would help but it sounds to me like you raised the voltage high enough that it was overloading the inputs which causes a relay click every second or so and produces high levels of distortion. Just a guess.
When a test starts each sweep it reads the gui parameter changes and caches them then during the sweep it only uses those cached parameters (with some leeway for things like turning plot lines on and off). It’s pretty robust when it comes to changing stuff during the run.
By the way – if you check the video on this page–> Attenuation · MZachmann/QA40xPlot Wiki at around 2:25 it actually shows an example of overloading the inputs and you can watch the plot as I let it overload.
Using also QA40xPlot a question came up. Maybe the function is already implemented and I didn’t find it. What I would like to have is that the calculation THD and THD+Noise and NOISE only can be filtered to a definable bandwidth. This because I use still analog distortion measurement instruments and there are a hi-pass filter (400 Hz) and two low-pass filters (30 and 80 kHz) available in order to depress hum and possible hi-frequencies which are not created by the measurement specimen. If such a function would be included then best the filter cut-off values may be freely set within the measurement frequency limits. Thanks in advance. Of course if the original QuantAsylum analyzer could have it too, this would be great.
Currently QA40xPlot supports a noise bandwidth setting (so low pass filtering) but there’s no high pass option to remove hum. I think you could do this with the factory QA40x by defining a user-defined weighting function that drops off the low and high frequencies.
I’ll take this as a feature request. It’s not that hard to implement.
No problem. It was so easy and I was so unhappy with the constant hardcoded 20Hz throughout that I just took care of it. It will be in the next release. It’s a little clumsy for now - you have to set the min frequency and bandwidth in Settings - at which point they apply to all test pages.
Mark
And note this only applies to noise math. It does not affect THD calculations. It does affect THD+N math.
Back to doing some stuff. I am so glad we have your software!
I was trying to measure THD for an external Oscillator. And ran into problems with high readings. Just for the hack of it, I decided to look at the waveform in your Oscilloscope page. Much to my surprise the scope shows a very non-linear artifact to the top of each waveform as the frequency rises. No problem at 1K, but as you go towards 10K it gets really ugly. This is reflected in the THD readings on the Spectrum test as well, so iI do not suspect it is the oscillator only. When I look at the waveform on a real scope it looks perfect, regardless of frequency. I then ran a loopback THD test in the Quant native sw, and the results were as expected. So in my mind it seems to be isolated to your sw, but of course I could be wrong. Any thoughts?
This sounds like a slight input overload or too low a sample rate. If you’re using an external oscillator (not driven by the QA40x) the automatic attenuation math isn’t going to work at all since there is no correlation between the output amplitude and the input amplitude, so you’ll have to manually set the attenuation. As a test, try at 42 and see if the issue continues.
If it’s not that, could you post a screenshot so I can see your settings? There’s no way to post a binary file on this forum, unfortunately. Also, help me out. What external oscillator are you testing? What settings do you currently run at to test it? Maybe a screen shot of the distortion at 10KHz?
I had tried different Attn settings, with no change.
I was wrong about the Spectrum distortion test showing High THD at 10K. I cannot duplicate. Sorry.
So the problem appears (to me) to be just on the Oscilloscope page. Even at 192K SR. Low freqs fine, high freqs ugly. This is reading the Quants internal Generator, with a balanced loopback setup.
Screenshot:
On the subject of THD. When running the Automated test for THD (Freq and Amplitude sweeps) in the native software, the THD readings (with the same loopback setup) are in the -80 dB range. I wonder why the readings are not super low, as in your THD tests. Any thoughts as to why? Or maybe I am doing something wrong?
Warning: Feature request: On the Oscilloscope page (your software), the fit all shows a big blob of color as it attempts to fit the entire timespan, which no one wants to see. And mousing back and forth many, many times in both directions is a royal PITA. I find the Osc page to be very handy, so this is a pain point for me. Any way to maybe show only 5 cycles, use the arrow keys to change the view, or use some other method?
I took a look at the photo and it looks fine. It would be clearer to you if you turned on Data Points in the graph but you’re just seeing the effect of sampling at a low frequency and not harmonic of the 10KHz. Your oscilloscope is probably sampling at least 100 times faster than the QA403.
If I run the native application (this is loopback for the QA403) it shows this
which is pretty identical to the output from QA40xPlot. Even running at 192000 sps that’s only 20 data points per 360 degrees and they data points are sampled on each cycle from slightly different points on the sine wave so it looks all strange.
Here’s the display on QA40xPlot of two curves. Both show a 4800Hz sine wave. One is sampled at 48000sps and one at 192000sps.
hold down the Alt key and drag a rectangle to set zoom like the factory app
or
enter values into the X Axis range on the right (which is from/to)
Don’t forget Pin to keep the graph boundaries fixed and scroll the mouse wheel inside of the X axis at the bottom of the screen for fastest zooming in-out.
I have exciting news! QA40xPlot finally supports Open/Save document! This will be available today or tomorrow after a bit more testing. It has been on my TBD List for a year.
There is a one minute demo showing a file load and short explanation here (you may want to raise the volume):
