Siglent SDS3000X HD and upgraded SDS1000X HD

[Martin72]

Then he doesn't need to compare anything if the basic conditions are different.
Strange.
Maybe I'll be lucky and he'll answer my question.

[Bad_Driver]

I did the same with my little HD2000x HD, may be of interest for you.

https://www.eevblog.com/forum/testgear/siglent-sds2000x-hd-12bit-(published-for-chinese-domestic-market-only)/msg5632163/#msg5632163

[egonotto]

Hello,

I think we can all agree that the result Dave showed doesn't show the advantage of 14 bit, right?

Best regards
egonotto

[Martin72]

With reference to his statement in the video “And this is the difference between 14 and 12 bit ADC”, we have confirmed that this is not the case.

[kripton2035]

I recall he compared the output of the 14bits keysight and the 12bits R&S, with the small 50MHz FFT pulse signal.
but we see now that with a 12bits siglent HD we can see this small signal in the same conditions...
so where is the 14bits advantage ?

[Performa01]

At high sensitivities like 1 mV/div, the dynamic range is determined solely by the frontend noise. Even a totally noise-free 50 ohms system (resistive noise only) cannot provide even 7 bits of dynamic at 1 GHz bandwidth. At 500 MHz it is still less than 8 bits. In practice, even the lowest noise frontends exhibit at least twice as much noise as in the previous calculation, thus reducing the dynamic by another bit. Therefore 14, 12, 10 or even 8 bits don’t make the slightest difference at 1 mV/div.

The FFT reduces the bandwidth, hence lowers the noise, so we can get more dynamic range for individual frequencies. But this is due to the process gain of the math operation and does not benefit from more ADC-bits either.

The real difference can only be seen at the lowest sensitivity without attenuator, like 100 mV/div on the Siglents. Once again, with a noise-free Frontend, we could get 13 bits of dynamic at best. With a real low noise frontend and 2 nV/√Hz the dynamic is limited to 12 bits at 1 GHz bandwidth.

Higher bit counts may be beneficial at lower bandwidths, down to 10 MHz at best, but not lower, because the noise floor rises significantly at lower frequencies. At 100 mv/div and 20MHz bandwidth, we could make use of 15 bits if we have a very low noise frontend with only 2 nV/√Hz at 20 MHz. But even at such low bandwidths, it is less than 9 bits at 1 mV/div.

Long story short: for 1 GHz bandwidth, 12 bits is the absolute maximum sensible resolution, fully exploitable only in three vertical gain settings: 100 mV, 1 and 10 V/div. More than 12 bits make sense only for low bandwidths and low sensitivities of 10 mV/div and above. Thus 14 bits is of very limited value on a general purpose DSO.

[hpw]

At high sensitivities like 1 mV/div, the dynamic range is determined solely by the frontend noise.....

One important thing to consider as to use sample "AVERAGING" !!

Using my LeCroy WM or SDA, I have to use at least of an minimal 4 times averaging, otherwise the noise gets to high.

Just to measure the rtHz FFT figures, a high amount of averaging is a must use. As you did already.

[Bad_Driver]

I went trough the HP/KS HD3 datasheet and tried to do some measurements to compare my SDS2000 HD with the published datas. Some results will follow.

But I‘m wondering if it is possible to get with FFT and Math channels rtHz Power density values from the 2000/3000 HD?   
I found the following LeCroy paper:
https://www.teledynelecroy.com/doc/noise-measurements-using-your-lecroy-oscilloscope
but I‘m a little bit lost after figure 6 or 7…..

Can one of our well known expert explain if this is possible and if „yes“ how to do that?

PS: I saw the yt video from the IMSAI guy on zone triggering with the HD3. I‘m pretty happy to have this with the SDS2000/3000 HD series and before with my SDS2000X + for some time


Thanks in advance!

[Martin72]

At high sensitivities like 1 mV/div, the dynamic range is determined solely by the frontend noise. Even a totally noise-free 50 ohms system (resistive noise only) cannot provide even 7 bits of dynamic at 1 GHz bandwidth. At 500 MHz it is still less than 8 bits. In practice, even the lowest noise frontends exhibit at least twice as much noise as in the previous calculation, thus reducing the dynamic by another bit. Therefore 14, 12, 10 or even 8 bits don’t make the slightest difference at 1 mV/div.

The FFT reduces the bandwidth, hence lowers the noise, so we can get more dynamic range for individual frequencies. But this is due to the process gain of the math operation and does not benefit from more ADC-bits either.

The real difference can only be seen at the lowest sensitivity without attenuator, like 100 mV/div on the Siglents. Once again, with a noise-free Frontend, we could get 13 bits of dynamic at best. With a real low noise frontend and 2 nV/√Hz the dynamic is limited to 12 bits at 1 GHz bandwidth.

Higher bit counts may be beneficial at lower bandwidths, down to 10 MHz at best, but not lower, because the noise floor rises significantly at lower frequencies. At 100 mv/div and 20MHz bandwidth, we could make use of 15 bits if we have a very low noise frontend with only 2 nV/√Hz at 20 MHz. But even at such low bandwidths, it is less than 9 bits at 1 mV/div.

Long story short: for 1 GHz bandwidth, 12 bits is the absolute maximum sensible resolution, fully exploitable only in three vertical gain settings: 100 mV, 1 and 10 V/div. More than 12 bits make sense only for low bandwidths and low sensitivities of 10 mV/div and above. Thus 14 bits is of very limited value on a general purpose DSO.

I think this should also be posted in the Keysight HD3 thread.

[Martin72]

Just for fun, SDS1104X-E....
4µV/50Mhz.
The view should be enjoyed with a little caution, as the SDS1104X-E already displays various interference spectra in this range, even if no generator is active.

[Someone]

Just for fun, SDS1104X-E....
4µV/50Mhz.
The view should be enjoyed with a little caution, as the SDS1104X-E already displays various interference spectra in this range, even if no generator is active.

With a Δf of 7-8kHz many scopes can get that (not)special -105dBm including the "noisy" predecessor from HPAK. This is why the settings/configuration are so important to make the same in a fair comparison.

Sample rate, channel bandwidth, FFT length, averaging (or not), hi-res (or not), and how all of those change the bandwidth of the final measurement.

[Martin72]

Like I said, just for fun...
We won't find out the right settings like in the video any time soon, because he has forgotten them...