I was playing with SDS814X FFT math (Hanning window) comparing results with my SSA3021X (Gaussian filter), in attachment some screens of 50MHz 0dBm CW sine wave (from SDG2042X) with different SPAN / RBW.
I would lik to know your opinion about SDS814x spectrum images @ 100Hz RBW, there is something wrong.
Am I doing something wrong or am I pushing the measure system over some limit ?
The first thing I noted was the seemingly random usage of window functions. Hanning and Gaussian are not comparable and for a meaningful test we should stick with the Flattop window on both instruments.
Then I wonder how the signal can actually be 50 MHz / 0 dBm. If so, then the SSA shows a level error of -0.91 dB, which might be attributed to the Gaussian window. But even the rather unsuitable (for spectrum analysis) Hanning window cannot explain a +0.97 dB amplitude error on the SDS800X HD.
Let’s compare the first test with 30 kHz RBW now:
SDS824X HD_FFT_50MHz_0dBm_RBW28kHz
Now compare this with the SSA screenshot. The SDS level measurement isn’t too far off at -0.288 dBm, the filter shape factor is as expected for a spectrum analysis and the noise floor is <-90 dBm. At 51 MHz that is some 20 dB better than the SSA – and the difference would be even greater if we measure closer to the carrier.
Next is 1 kHz RBW:
SDS824X HD_FFT_50MHz_0dBm_RBW900Hz
Yes, we see some modulation sidebands below -72 dBm now, and again we cannot decide if they would be visible on the SSA as well, as the noise floor there completely masks everything that low. By contrast, the SDS800X HD shows a noise floor of -100 dBm.
But we can try the same scenario with a higher class DSO, like the SDS2000X HD:
SDS2504X HD_FFT_50MHz_0dBm_RBW700Hz
This proves that the signal source is clean and the spurious modulation happens within the SDS800X HD somehow.
Finally 100 Hz RBW. This just isn’t possible at 50 MHz – the SDS800X HD is a DSO, hence a baseband instrument and not a spectrum analyzer. The FFT would need to compute at least 4 Mpts to handle this.
Yet we can try to come as close as possible: using the full 2 Mpts at 250 MSa/s we can get about 440 Hz RBW with the Flattop window. In a pinch we could use the Blackman window, which has still acceptable amplitude accuracy and a great side-lobe suppression. This way we can get ~200 Hz RBW:
SDS824X HD_FFT_50MHz_0dBm_RBW200Hz
Yes, there are those spurious modulation sideband signals again, and these are not visible on an e.g. SDS2504X HD:
SDS2504X HD_FFT_50MHz_0dBm_RBW165Hz
So yes, the SDS800X HD has a problem there. Since the FFT should be the same across the platform, this is more likely to be a hardware issue.
In attachment 2 additional spectrums of 50MHz carrier AM modulated with 1KHz sine wave, 100%.
Once again, 1 kHz is a tough request. It is just about manageable:
SDS824X HD_FFT_50MHz_0dBm_RBW200Hz_Mod1kHz
Yet a closer look reveals a 10 kHz modulation frequency for the provided examples anyway, so here we go:
SDS824X HD_FFT_50MHz_0dBm_RBW200Hz_Mod10kHz
Yes, there are various spurious signals again. On the SDS2504X HD we get the third order intermodulation products only:
SDS2504X HD_FFT_50MHz_0dBm_RBW165Hz_Mod10kHz
These are not unexpected though. A wideband amplifier can only have so much linearity, and I’ve covered this topic in more detail here in this thread (reply #6, section “Distortion Measurements”):
https://www.eevblog.com/forum/testgear/sds800x-hd-review-demonstration-thread/msg5293759/#msg5293759The main source of distortion in a modern DSO frontend would be the PGA, where third order intermodulation distortion is specified as -65 dBc (typically) at 250 MHz. Harmonic distortion is even worse at -50 dBc (typical) at 100 MHz. Both SDS meet the expectations at ~-67 dBc.
https://www.ti.com/product/LMH6518BTW, in the SDS800X HD datasheet, the spurious-free dynamic range SFDR is specified as only ≥35 dBc.
Same topic, comparison with DHO1000 that does not show unwanted spuries :
Yes, that’s quite amazing. As stated above, such distortion figures cannot be expected with industry standard parts like the LMH6518.