Introduction
This is a continuation of the verification tests on the waveform generators and comparing the AFG31052 results to other instruments.
Sine wave distortion
This is another measurement in the verification steps from Tektronix and is further supplemented by a total harmonics distortion and a spurious signal measurement. All measurements were made with the spectrum analyser built into the MDO34. There did seem to be a slight mismatch with the test recommendations. The table for the AFG310x had four settings up to 1 MHz but with only having a 50 MHz unit, I could not go that far. The option was of course to replace the 100 MHz setting with the 50 MHz rating of the unit, I elected to go for the previous table that offered settings from 100 kHz up to 50 MHz, which is also a bit more suitable for the DG1022. As the 3404A waveform generator is limited to 1 MHz output, it was omitted from this verification test.
Looking at the table, a dBc pass limit is advised for top to the 5th harmonic. These limits changed slightly for the lower frequencies I adopted. I also applied the limits from the respective manufacturer to their units as well as applying the limits for the AFG31052 across all of the units tested. The results are tabulated for each channel in the picture gallery below.
| {gallery} Harmonic distortion test results |
|---|
AFG31052 Channel 1 distortion measurements |
AFG31052 Channel 2 distortion measurements |
DG1022 Channel 1 distortion measurements |
DG1022 Channel 2 distortion measurements |
UTG962 Channel 1 distortion measurements |
UTG962 Channel 2 distortion measurements |
MDO34 distortion measurements |
The DG1022 was the only instrument that passed its own criteria for distortion, this levels for this are a little more forgiving than the requirements set out for the AFG31052. At 100 kHz, the AFG31052 failed at the second harmonic. However, I am not sure if I should have made this as part of the test. In my defence, I would have expected a higher specified unit to be able to pass the test limits for one of their lower specified units. The MDO34 could also not meet the specification for 100 kHz output on the 2nd harmonic. The UTG962 failed at two frequencies for the second harmonic, 100 kHz and 1 MHz.
With the majority of the units, seemingly having issues at the second fundamental level, I did start to wonder if there was an issue with the test set-up rather than the waveform generators. Not really being able to think what I could do to further investigate this, I decided to park the issue and move on to the next test.
I also saved some screenshots from the MDO34 to show what the actual harmonic looked like from the AFG31052 for each frequency and then the highest frequency for each of the other units.
| {gallery} Harmonic distortion screenshots |
|---|
AFG31052 - Spectrum analyser screen at 100 kHz, 1 Vpp sine wave |
AFG31052 - Spectrum analyser screen at 1 MHz, 1 Vpp sine wave |
AFG31052 - Spectrum analyser screen at 5 MHz, 1 Vpp sine wave |
AFG31052 - Spectrum analyser screen at 25 MHz, 1 Vpp sine wave |
AFG31052 - Spectrum analyser screen at 50 MHz, 1 Vpp sine wave |
MDO34 - Spectrum analyser screen at 50 MHz, 1 Vpp sine wave |
UTG962 - Spectrum analyser screen at 50 MHz, 1 Vpp sine wave |
DG1022 - Spectrum analyser screen at 20 MHz, 1 Vpp sine wave |
Total harmonic distortion
The total harmonic distortion, usually referred to as THD, is a test again specified by Tektronix. This is measured at one frequency of 20 kHz, with a 1 Vpp output. This time the measurement is taken to the 7th harmonic and as the MDO34 used for the measurement, does not have a THD calculation built into it, a manual calculation is made based upon the measurements obtained. This calculation is defined within the Tektronix manual along with a working example. The THD limit is specified as 0.2 % for the AFG3000 series.
However, the formula detailed in the Tektronix guide is not what I am used to for THD calculation and I could not actually make their example work either. I therefore opted to work to formula that I am more used to, which I could get to work with their example.
The table above details the results of the harmonic measurements made for the THD calculation. The Tektronix manual states that if the levels at each harmonic are all better than -62 dBm, then there is no need to carry out the calculation, as it will be better than 0.2 % THD. Immediately there is some concern as the second harmonic is clearly no where near -62 dBm. The fifth harmonic is just below -62 dBm, but this is likely to be offset by the other harmonics that are quite a bit better than -62 dBm. The THD calculation for each instrument follows.
| {gallery} THD Calculation Tables |
|---|
AFG31052 THD Calculation |
DG1022 THD Calculation |
UTG962 THD Calculation |
MDO34 and 3404A THD Calculation |
AFG31052 THD vs Test Frequency |
From the calculation results, it can be that none of the instruments reach the specified 0.2 %, the closest actually being the 3404A that reached 0.24 %. It can be seen that the predominant difference between the 3404A and the other instruments is the reading at the second harmonic. To further test the AFG31052 I took further readings and made THD calculations, as the frequency was increased. The plot shows that between 1 MHz and 1.5 MHz, the THD drops to below 0.2 %. The screenshot below shows the output of the AFG31052 at 20 kHz.
The next screenshot shows the output of the AFG31052 at 1.5 MHz. A clear drop in the second harmonic can be seen.
Reading through some specifications of the 3404A, I came to notice that the Picoscope software had a THD measurement built into its spectrogram, so I retested all units utilising the 3404A instead of the MDO34. There was some struggle with this test setup. At a 200kHz span, a lot of noise was picked up on the Picoscope and a full set of readings could not be obtained, as seen below.
Set to 500 kHz bandwidth and the noise disappears and the THD value drops down to 0.23 %.
The table below compares the THD calculations with the readings from the Picoscope. Only the AFG31052, DG1022 and UTG962 were tested on the Picoscope.
I do not know why there is a high second harmonic on the readings from the MDO34 that is not replicated by the readings on the 3404A. The readings at a 200 kHz span are clear on the MDO34, yet on the 3404A there is an abundance of noise. I do not have access to any other apparatus to measure the THD or harmonics at such a high frequency, so my only option would be to repeat the tests at another location, but that will probably be unlikely for a while.
Spurious output measurement
This test, also defined by Tektronix, looks at the spectrum analyser readings at three frequency points and is a basic test looking for any spike that are above a defined dBc limit. Five frequency points are utilised with a 1 Vpp sine wave. The spectrum analyser is set with a much higher bandwidth than the previous tests and whilst there is a test specification for the spikes, it is also interesting to look at the various screenshots from the spectrum analyser.
The table below shows the highest spike observed at each of the test frequencies. It looks like the majority of the instruments cannot perform to the higher specification of the AFG31052. There are several points, especially at the 100 kHz test that are not up to specification. Overall, the UTG962 seems to have the noisiest output.
Here are some screenshots of some of the tests.
| {gallery} Spurious output spike screenshots |
|---|
AFG31052 spikes seen at 50 MHz |
DG1022 spikes seen at 5 MHz |
UTG962 spikes seen at 50 MHz |
MDO34 Spikes seen at 50 MHz |
Visually, the DG1022 shows a lot of spikes across the test span, but the spikes are generally small. The UTG962 shows a bunch of spikes close to the fundamental and then a further bunch of spikes around 300 MHz. This behaviour is not seen on the other instruments. Both the AFG31052 and the MDO34 show much cleaner outputs, presumably this is another benefit of the higher cost, more well developed instruments.
Triangular wave linearity
AFG31052 triangular wave output at 100 kHz, 2 Vpp and 50% duty.
This test is not specified by Tektronix. I used a test output of 100 kHz with a 2 Vpp level and 50% duty. The output was observed on the oscilloscope and saved to a csv file for analysis in Excel. In Excel, the ramp up line was isolated and compared to a perfect line from -1 volt to +1 volt. The highest deviation point between the actual and perfect line was then used to calculate the linearity. After the initial test, I measured the output of Channel 1 of the AFG31052 at 25%, 75% and then 90% duty cycle.
All of the instruments had a linearity between 3% and 4% except for Channel 2 of the UTG962 that slipped over to 4.2%.
The AFG31052 showed a reasonably linear response over the duty cycle range as seen in the table above.
Output symmetry
This is not a recommended test from Tektronix, but something I decided to look at. The test is conducted in a similar manner to how I take an air gap search coil measurement using the math function of the 3404A to compare both halves of the waveform by overlapping them and subtracting them from one another. I decided to carry out these tests at 800 kHz as this is the maximum frequency output for the ramp / triangle waveform. An 800 kHz signal has a period of 1.25 us. For the comparison formula, I need to divide this value by 2.
I then use the above equation, as a math channel within the Picoscope software. The time adjustment function is within the square brackets in the formula, [0.000000625]. This function is applied to Channel A, after it has been inverted, (-A[0.000000625]). The final element is subtracting the bracketed section from the original A Channel to create the complete formula, A-(-A[0.000000625]). If the waveform symmetry is matched, the output will appear as a straight line. I usually offset the math channel to give more clarity to the display.
The following screen captures are the output from Channel 1 of the AFG31052. Channel 2 was an exact copy, so I have elected not to include this.
| {gallery} AFG31052 Channel A Symmetry |
|---|
AFG31052 Sine wave symmetry |
AFG31052 Square wave symmetry |
AFG31052 Triangle wave symmetry |
AFG31052 50% Duty cycle pulse wave symmetry |
AFG31052 48% Duty cycle pulse wave symmetry |
The captures show excellent symmetry of each of the waveforms. As an example of the formula in operation, I have also included the screenshot of a pulse waveform at 48% duty cycle to show what is seen when the symmetry is not matched.
All the other units tested showed a very similar story. An exception was seen on the DG1022 where a slight discrepancy is seen at the corners of the square wave circle in the screenshot below.
This was also seen on the UTG962, but was a smaller spike than on the DG1022 and is harder to see.
The worst performer was the 3404A. In fairness, the waveform generator on this is not a high specification and the maximum output frequency is actually 1 MHz, so an 800 kHz signal is approaching its limits. However, it is another good example of the calculation displaying a symmetry error.
Conclusions
This was certainly an interesting set of measurements to perform across the different waveform generators and the AFG31052 performed exceptionally well in all the tests. In general the other units can match the AFG31052 in some of the basic tests. Where the AFG31052 started to excel, is in the more finer points of the tests and performance of the output.
The AFG31052 had a much better fall and rise time for a square wave and was only really matched by the MDO34, also from Tektronix. Spectrum analysis of the output waveforms was cleaner on the AFG31052 than the other instruments.
The AC flatness test turned out to be a bit of a mess and I hope to get the time to go back to this and get some meaningful results. I am convinced that this is a problem with the RF power meter and not any of the waveform generators. Likewise the THD measurement did not produce the expected results, but this was for all of the units, and again points towards an issue with the test set up rather than the waveform generators failing to meet their specification. I am not so sure what to about this aspect though, other than trying a different location.
The biggest surprise to this is how well the little Uni-T UT692 waveform generator performed. It doesn't have all of the capabilities of the AFG31052, but for a basic 60 MHz waveform generator, it would be hard to beat for its price.