One more ADC to compare to, the Beis AD24QS DIY-ADC (CS5381 based: AD24QS and DA24QS - Audio ADC and DAC, 24 Bit / 192 kHz).
The ADC is single ended. Use the input floating or shunted does not make much difference, especially there are no 50Hz-line artefacts.
The measurement is made at 44.1kHz sample rate for convenience (Mac Pro optical-in limitation). The difference of l-r and the 30Khz spike of the Beis ADC is only there since a not 100% successful attempt to repair it.
The ADC is single ended. Use the input floating or shunted does not make much difference, especially there are no 50Hz-line artefacts.
The measurement is made at 44.1kHz sample rate for convenience (Mac Pro optical-in limitation). The difference of l-r and the 30Khz spike of the Beis ADC is only there since a not 100% successful attempt to repair it.
Attachments
Have you made the cable like in figure 17?
Note that pin 1 of the input connector is only connected to the shield of the cable, nothing else. Pin 2 and 3 are the two signal wires, where pin 3 is connected to the outer conductor of the male coax (in this case BNC). Pin 2 is connected to the centre conductor.
I just made a measurement with loop-back from output to input on both channels. The outputs were set to 0 dBV and the inputs were set to -20 dBV. The only difference was the cables used.
On the left channel I used a BNC coax cable on the output BNC connector. On the input side the shield was connected to pin 1 and 2.
On the right channel I used a cable configured like figure 17.
The results are quite different. Swapping the cables also swapped the results.
On the right channel the highest peaks are at around -130 dBV.
But yes, cabling can be a challenge. And ground loops are easily created. Much easier with balanced connections of course.
Attachments
Just for interest ...
I've never heard of an FFT window setting of Kaiser7 before, and I was suspicious of the exceedingly low distortion numbers I was getting using it. My normal setting for an FFT analyser is the old Hanning window. That one I know returns accurate answers, or at least answers that agree with other methods of getting a number for harmonic distortion.
Here are a pair of screens showing both windowing selections on my HP 339A distortion analyser. Which of these do you believe is correct?
-Chris
I've never heard of an FFT window setting of Kaiser7 before, and I was suspicious of the exceedingly low distortion numbers I was getting using it. My normal setting for an FFT analyser is the old Hanning window. That one I know returns accurate answers, or at least answers that agree with other methods of getting a number for harmonic distortion.
Here are a pair of screens showing both windowing selections on my HP 339A distortion analyser. Which of these do you believe is correct?
-Chris
Attachments
Kaiser window - Wikipedia
Kaiser Window - MATLAB & Simulink
I use Kaiser 7 for most of my measurements.
I have also used Hanning and Blackman Harris..
Second part of the statement is true, they are just different ways of applying windowing. Measurements performed with different windowing may or may not be directly comparable.
Kaiser Window - MATLAB & Simulink
I use Kaiser 7 for most of my measurements.
I have also used Hanning and Blackman Harris..
Second part of the statement is true, they are just different ways of applying windowing. Measurements performed with different windowing may or may not be directly comparable.
Hi Kevin,
I'm going to use the Hanning window. It agrees with the industry and gives me a more realistic answer. I don't need or want "sunshine up my rear numbers". The cold hard facts will do nicely, thank you very much.
The Hanning windowing generates answers close to what my distortion test set returns. and that agrees with almost every distortion analyser I have used in the past. The Leader instrument was also overly optimistic, especially when looking at 10 or 20 KHz fundamentals. I found out later that it's response was diving at 30 KHz (filter out) and not the 100 KHz claimed. Of course it read low distortion at higher frequencies. I had used it for years without knowing this. I can't stand it when equipment lies to you. This is what is happening with the THD measurement using a Kaiser7 window.
The Kaiser7 window might be useful for tracking down problems and for the IMD test. I'll have to see if it agrees with known correct answers. That's as soon as I can find something to compare to. But when designing and exchanging information about what numbers any of my creations are operating at, I could do without the embarrassment. If anything, you would want to be a little pessimistic.
-Chris
I'm going to use the Hanning window. It agrees with the industry and gives me a more realistic answer. I don't need or want "sunshine up my rear numbers". The cold hard facts will do nicely, thank you very much.
The Hanning windowing generates answers close to what my distortion test set returns. and that agrees with almost every distortion analyser I have used in the past. The Leader instrument was also overly optimistic, especially when looking at 10 or 20 KHz fundamentals. I found out later that it's response was diving at 30 KHz (filter out) and not the 100 KHz claimed. Of course it read low distortion at higher frequencies. I had used it for years without knowing this. I can't stand it when equipment lies to you. This is what is happening with the THD measurement using a Kaiser7 window.
The Kaiser7 window might be useful for tracking down problems and for the IMD test. I'll have to see if it agrees with known correct answers. That's as soon as I can find something to compare to. But when designing and exchanging information about what numbers any of my creations are operating at, I could do without the embarrassment. If anything, you would want to be a little pessimistic.
-Chris
Last edited:
The second graph in post #1623 is very interesting - the THD+N figure is lower than the THD figure.
I’ve never seen this before, and I fail to understand how it can happen.
I’ve never seen this before, and I fail to understand how it can happen.
I don't think it's lying. I think it's just using a Kaiser window, which is different to a Hanning window. There is no "right" or "wrong", it's just different. Like apples and oranges.
Uhm, well, maybe that software is lying. 😀
Well, I know that the answer depends on the window being used, but I still think there is a right and wrong. The Hann window is appropriate for single tone THD tests whereas the Kaiser7 window is better for the two tone IMD spaced closely.
What I think is important is that the answers you get and figures you quote should agree with the world's accepted standards. It is entirely pointless to quote numbers using a system intended for a different measurement. So in that sense, using Kaiser7 for THD tests would be incorrect, it's the wrong window type if you read up on the various window types. As a crosscheck, the Hann window generates numbers that can be verified using other test instruments.
I can't understand why anyone would wish for overly optimistic figures in the first place. The answer just is, and looking at numbers generated by using the wrong windowing is a complete waste of anyone's time and energy. I don't believe there is a conversion that can be used between the two THD numbers either, making the wrong windowed numbers utterly useless.
It would seem that either members using this equipment bone up on windowing and what each type is good for, or they need a short course on the subject.
-Chris
What I think is important is that the answers you get and figures you quote should agree with the world's accepted standards. It is entirely pointless to quote numbers using a system intended for a different measurement. So in that sense, using Kaiser7 for THD tests would be incorrect, it's the wrong window type if you read up on the various window types. As a crosscheck, the Hann window generates numbers that can be verified using other test instruments.
I can't understand why anyone would wish for overly optimistic figures in the first place. The answer just is, and looking at numbers generated by using the wrong windowing is a complete waste of anyone's time and energy. I don't believe there is a conversion that can be used between the two THD numbers either, making the wrong windowed numbers utterly useless.
It would seem that either members using this equipment bone up on windowing and what each type is good for, or they need a short course on the subject.
-Chris
Chris,
Check your last post with the plots. The THD and THD+N readings for both don't make any sense at all. The "THD+N" should be more with the Hanning window not less. Put the cursor on the 2nd harmonic and try showing the peak of the fundamental.
Check your last post with the plots. The THD and THD+N readings for both don't make any sense at all. The "THD+N" should be more with the Hanning window not less. Put the cursor on the 2nd harmonic and try showing the peak of the fundamental.
There is something very wrong with the Hann measurement you've shown in post #1623, and it's pretty meaningless to compare Hann with other windows based on that measurement.
I use Hann 99% of the time also (the only other 1% is when comparing results with other people, like in here, so using the same windows others have chosen). I just compared Hann and Kaiser 7 in ARTA using the same setting you've shown, the THD difference is within 15% but not 20 times.
The measurement of the THD+N that is bigger than THD is not right!
From the ARTA user manual, page 32:
When applying the signal window, it is recommended to follow these rules:
for continuous nonperiodic signals (noise) use the Hanning window
for measuring harmonic and intermodulation distortions use the Kaiser5 or the Blackman4 window, but to get 24-bit resolution use Kaiser7 window
for calibration with a sine signal use the Flat-top window
for periodic noise, multitones and other signals that are periodic within the acquisition window use the Uniform window.
From the ARTA user manual, page 32:
When applying the signal window, it is recommended to follow these rules:
for continuous nonperiodic signals (noise) use the Hanning window
for measuring harmonic and intermodulation distortions use the Kaiser5 or the Blackman4 window, but to get 24-bit resolution use Kaiser7 window
for calibration with a sine signal use the Flat-top window
for periodic noise, multitones and other signals that are periodic within the acquisition window use the Uniform window.
When you know that the frequencies involved are a synchronous
subharmonic of the sample rate. Then you get absolutely no spill
into the neighbour bins and maximum energy where it belongs.
And short time series are enough. Just at least one full cycle of
every frequency involved. And only full cycles.
That is essential on a wafer tester. Every millisecond counts
and costs money.
There is no Hanning window, really. The guy was Justus von Hann,
an Austrian meteorologist.
The manual of my Agilent 89441A analyzer recommends the
Hann window for noise measurements.
cheers, Gerhard
Last edited:
Hi ej25awd,
Anyway, there is nothing that I can do to make ARTA read that way. It does it's own thing. The Hann window gives me numbers that correlate to what my analog distortion measurement set reads, an HP 339A. The Kaiser7 window is completely unrealistic.
I just tried the same measurement again just now. All I had to do was turn the equipment on and plug the oscillator output from the 339A into the RTX-6001. I still get similar numbers with the THD+Noise reading much lower than the THD only reading. THD= 0.0016% THD+Noise=0.00068%. So that's a day later with the equipment having been turned on just now. So it's impossible that the program was in an odd memory type problem, a total, hard reboot of software and hardware.
This isn't a problem with the RTX-6001. It is an issue with this expensive software program. It isn't anything that I am doing as this is beyond my control.
Hi cwtim01,
So what you are saying is that ARTA is broken software I guess. All I am bringing attention to is how unrealistically improbable the Kaiser7 window is. People were suggesting it's use only because it gave them the nicest numbers, and that is what I was getting at. The other issue of the THD+Noise numbers being impossibly low is another thing completely.
Anyway, the comparison is of ARTA and the two windows people are using. It is a perfectly valid comparison given the constraints. I used the HP 339A because it is well known and others here can duplicate my test with the information given. I was kind of hoping someone would do this to see what they get.
Hi lemon,
Yup, that's what it says all right. And the numbers you are seeing are not right for sure. See, we are in complete agreement. However, I come from an HP world with real spectrum analysers and network and signal analysers. I'm well aware how to apply the test equipment and setup.
My point was that I was told to use the Kaiser7 window because my loopback tests showed higher distortion for the RTX. I'm trying to say that you don't use a particular window just because you like the answer it gives you better. The Hann window is the proper one to use for single sine harmonic distortion tests, not the Kaiser7. The Kaiser7 looks like it would be suited to the IMD tests using more than one tone, but not a bunch of tones.
Hi Gerhard,
Yes. I just read off a source I found on the internet while trying to find out about the Kaiser7 window type. Before opening ARTA, I had never seen any reference to that window type before and I was extremely suspicious that it was the wrong window type to use for THD tests. That is a correct statement according to the internet sources I found.
It looks like many folks were prepared to use the Kaiser7 windowing to obtain THD results and report on that. I was hoping to bring attention to the fact that the KAiser7 window is not the correct one for THD tests. I think that you, me and Keysight can agree on that. I'll bet that even Tektronix would agree. 🙂
Best, Chris
Yup, I know. I copied the screen and saved into a file. Check the picture carefully and you will see the settings are as I indicated. ARTA seems to make gross errors sometimes. Please note that the connections and levels were not adjusted between screen captures. The only thing that changed was a new window type and acquisition. The average was linear at 100 samples as was recommended earlier by someone.
Anyway, there is nothing that I can do to make ARTA read that way. It does it's own thing. The Hann window gives me numbers that correlate to what my analog distortion measurement set reads, an HP 339A. The Kaiser7 window is completely unrealistic.
I just tried the same measurement again just now. All I had to do was turn the equipment on and plug the oscillator output from the 339A into the RTX-6001. I still get similar numbers with the THD+Noise reading much lower than the THD only reading. THD= 0.0016% THD+Noise=0.00068%. So that's a day later with the equipment having been turned on just now. So it's impossible that the program was in an odd memory type problem, a total, hard reboot of software and hardware.
This isn't a problem with the RTX-6001. It is an issue with this expensive software program. It isn't anything that I am doing as this is beyond my control.
Hi cwtim01,
I was comparing the THD numbers only since I can recognise that the THD+Noise numbers are out to lunch completely.
So what you are saying is that ARTA is broken software I guess. All I am bringing attention to is how unrealistically improbable the Kaiser7 window is. People were suggesting it's use only because it gave them the nicest numbers, and that is what I was getting at. The other issue of the THD+Noise numbers being impossibly low is another thing completely.
Anyway, the comparison is of ARTA and the two windows people are using. It is a perfectly valid comparison given the constraints. I used the HP 339A because it is well known and others here can duplicate my test with the information given. I was kind of hoping someone would do this to see what they get.
Hi lemon,
Yup, that's what it says all right. And the numbers you are seeing are not right for sure. See, we are in complete agreement. However, I come from an HP world with real spectrum analysers and network and signal analysers. I'm well aware how to apply the test equipment and setup.
My point was that I was told to use the Kaiser7 window because my loopback tests showed higher distortion for the RTX. I'm trying to say that you don't use a particular window just because you like the answer it gives you better. The Hann window is the proper one to use for single sine harmonic distortion tests, not the Kaiser7. The Kaiser7 looks like it would be suited to the IMD tests using more than one tone, but not a bunch of tones.
Hi Gerhard,
Yes. I just read off a source I found on the internet while trying to find out about the Kaiser7 window type. Before opening ARTA, I had never seen any reference to that window type before and I was extremely suspicious that it was the wrong window type to use for THD tests. That is a correct statement according to the internet sources I found.
It looks like many folks were prepared to use the Kaiser7 windowing to obtain THD results and report on that. I was hoping to bring attention to the fact that the KAiser7 window is not the correct one for THD tests. I think that you, me and Keysight can agree on that. I'll bet that even Tektronix would agree. 🙂
Best, Chris
Ooohps, Julius von Hann, not Justus.
I have no opinion on Kaiser windows. Most windows only make a difference
below -50 dB or so. If you add the power in the center bin +- 3 bins you
should be pretty safe from artefacts.
I have no opinion on Kaiser windows. Most windows only make a difference
below -50 dB or so. If you add the power in the center bin +- 3 bins you
should be pretty safe from artefacts.
Thanks everyone for all discussion on window types and application. There does seem to be a common trend on when to apply them although the theoretical explanation might be too mathematical to understand. Perhaps someone would like to sort it out into the wiki page? I could not find it because I don’t remember which directory it is in.