scope question

[p.larner]

i have a 50 mhz dso with a 1gs/s sampling rate,i see an owon 200mhz scope with a 2ghz/s sample rate,how can it not need double the sample rate for 200mhz analog bw?.

[Psi]

i have a 50 mhz dso with a 1gs/s sampling rate,i see an owon 200mhz scope with a 2ghz/s sample rate,how can it not need double the sample rate for 200mhz analog bw?.

Your 50mhz DSO does not actually need 1gs/s, it only has it because the manufacturer wants to keep their products the same between all their different 50/100/150/200 mhz product versions.  There is high RnD cost designing the main chipset/ASIC inside the scope, so they only want to design one of them and use it for all their scopes in that class. Lower spec versions get the 1gs/s by using the same core chip as the 250mhz version, not because they need it.

If you go by Nyquest rate, to display a 50mhz waveform you need at absolute minimum 2x the sample rate, so for 50mhz that would be 0.1gs/s.
That would capture the signal but it would look like shit. So 4x is probably the minimum you'd want. That would be 0.2gs/s.
So you can see that 1gs/s is far more than you need.   10x would be the height of luxury, and 1gs/s for 50mhz bandwidth is 20x.

At 1gs/s you could probably handle 250mhz bandwidth (4x) without it looking like shit.
Your scope only supports 50mhz bandwidth because of a combination of one or both of these reasons.
A) The software is artificially limiting the bandwidth to 50mhz
B) The analog front-end/input circuit is likely different between different bandwidth models, and you only have the cheap/basic 50mhz front end.

This is why some scopes are hackable to get more bandwidth, sometimes with just a firmware update, and other times with a simple hardware mod

[CosteC]

In most of entry level and mid-level oscilloscopes sampling is divided between channels.
Rigol DS1054Z has 1 GSample/s if ONE channel is used, 0.5 GSample/s if two, and 0.25 GSample if three or four.
This way it is sensible sampling for ALL channels: 250 Msamples for each 50 MHz channel.

Not all cheap scopes are so generous

[TimFox]

Again, one of my pet peeves:  the frequency unit "mHz" = 10^-3 Hz does exist, and is relevant to the bandwidth of a DC measurement on a DVM, but a 'scope should have bandwidths measured in MHz.

[tggzzz]

Again, one of my pet peeves:  the frequency unit "mHz" = 10^-3 Hz does exist, and is relevant to the bandwidth of a DC measurement on a DVM, but a 'scope should have bandwidths measured in MHz.

Yeah

My pet peeve is measuring time in nS. My Fluke 25 DVM does have a fairly well-hidden nS range - as an extension of the resistance ranges.  Some datasheets have both s and S, e.g. FETs.

For the ignorant, s is the SI unit seconds whereas S is the SI unit Siemens.

Then we can move on to the loonies that measure power station output in mW, and confuse miles and miles per hour (actually they confuse MW and MWh, but the point is the same).

[TimFox]

I miss the pre-SI conductance unit  "mho", spelled as an upside-down  , which is hard to find on today's software.

[CosteC]

My pet peeve is measuring time in nS.

Unit of time, second is abbreviated to "s", lower case. Siemens, unit of conductance is "S" - upper case.
1 nS = 1 G

Corrected attribution.

[TimFox]

1 nS = 1/ G\$\Omega\$

Once again, what is quoted above as from me was not posted by me;  it was posted by tggzzz, who was replying to my post.

[tggzzz]

My pet peeve is measuring time in nS.

Unit of time, second is abbreviated to "s", lower case. Siemens, unit of conductance is "S" - upper case.
1 nS = 1 G

Wrong attribution.

Can I suggest you click the "quote" button above the top right of any post, and start from that.

[JeremyC]

i have a 50 mhz dso with a 1gs/s sampling rate,i see an owon 200mhz scope with a 2ghz/s sample rate,how can it not need double the sample rate for 200mhz analog bw?.

I suggest you check more about “Nyquist Theorem”. Check the links below:
- Read the “Electrical Measurement” section in https://www.sciencedirect.com/topics/engineering/nyquist-theorem
- Wikipedia https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem

[tggzzz]

i have a 50 mhz dso with a 1gs/s sampling rate,i see an owon 200mhz scope with a 2ghz/s sample rate,how can it not need double the sample rate for 200mhz analog bw?.

I have a 140ps/4GHz Tek 1502 with 25kS/s.

I have a 200ps/1.7GHz Philips PM3400 where the signal path's transistors are f_T=200MHz BC107s

When considering the Nyquist limit you have to pay very careful attention to the definition of "signal bandwidth".

[radiolistener]

i have a 50 mhz dso with a 1gs/s sampling rate,i see an owon 200mhz scope with a 2ghz/s sample rate,how can it not need double the sample rate for 200mhz analog bw?.

It depends on oscilloscope specification.

There are two bandwidth for digital oscilloscope:
1) Digital bandwidth, which is a half of sample rate
2) Analog bandwidth which should be much smaller than digital bandwidth (at least 5-10 times smaller for budget oscilloscope and more than 20 times smaller for professional oscilloscope).

Larger gap between analog bandwidth and digital bandwidth allows to get better rejection of digital conversion aliases and images with analog filter on oscilloscope input frontend. Better rejection for aliases and images means that oscilloscope will show you less signal distortions. But it requires to use more expensive ADC converters and more fast and powerful FPGA for data processing.

Some oscilloscopes using more cheap ADC/FPGA with smaller digital bandwidth in order to provide you with more cheaper device. Yes, it can show you more distortions for signal than more expensive oscilloscope with more fast and powerful ADC/FPGA, but you get more budget solution. And for many cases such solution is pretty acceptable.

In short, if difference between analog bandwidth and sample rate/2 is higher it means that such oscilloscope will show you more true waveform with less distortions...

These distortions happening when measurement signal contains frequency components higher than digital bandwidth of oscilloscope. For 1 GS scope digital bandwidth is 500 MHz, so it will suffers from signals above 500 MHz on it's input, which usually have low enough level in home and diy applications, so you can use more cheaper oscilloscope with 1 GS for home needs.

PS: there is also exists specific type of oscilloscope - stroboscopic, it can have analog bandwidth more than digital one, but it has it's own limitation, because such oscilloscopes are intended to work with periodic signals, but this is different story.

[TimFox]

Your "digital bandwidth" = (1/2) sample rate is called the Nyquist Limit for the system.
That is a very important parameter for any discretely sampled system.
If an analog signal presented to the sampler has a periodic component above that frequency, it will appear in the sampled data as an "alias" that cannot be removed by filtering.

[tggzzz]

Your "digital bandwidth" = (1/2) sample rate is called the Nyquist Limit for the system.
That is a very important parameter for any discretely sampled system.
If an analog signal presented to the sampler has a periodic component above that frequency, it will appear in the sampled data as an "alias" that cannot be removed by filtering.

Nicely put.

For clarity, I have emphasised the point that is often not understood.

[TimFox]

Life gets more interesting in non-oscilloscope applications to determine the sample rate required to fully sample a bandpass-limited function, where there are no frequency components below a lower level and above a higher level.
This requires a band-pass filter (with appropriately steep slopes) ahead of the sample head.
The sample head must have a narrow sample-time aperture, but the required sampling rate is not so high as twice the high end of the passband.
See  https://personal.utdallas.edu/~raja1/EE4361%20Spring%2014/Lecture%20Notes/Bandpass%20Sampling.pdf

[wasedadoc]

Sub-Nyquist sampling has been successfully used on video signals. Most of the signal energy is concentrated into frequencies with spacing at video line (not AC supply) frequency. By appropriate sampling frequency the alias components appear in the gaps and can be sufficiently suppressed by comb filtering to achieve acceptable subjective performance.

eg https://rdweb1.lh.bbc.co.uk/rd/publications/rdreport_1975_04

[TimFox]

Yes, if you know where the alias frequencies in your system will appear, they can be suppressed by appropriate filtering after the sampling.
In the general situation, where all you know is the bandwidth of the system before sampling, the aliases are at frequencies not known a priori.

[tggzzz]

Life gets more interesting in non-oscilloscope applications to determine the sample rate required to fully sample a bandpass-limited function, where there are no frequency components below a lower level and above a higher level.
This requires a band-pass filter (with appropriately steep slopes) ahead of the sample head.
The sample head must have a narrow sample-time aperture, but the required sampling rate is not so high as twice the high end of the passband.
See  https://personal.utdallas.edu/~raja1/EE4361%20Spring%2014/Lecture%20Notes/Bandpass%20Sampling.pdf

See N-path, a.k.a. commutating capacitor filters, and the closely related Tayloe mixers. You don't need a narrow sample time aperture

I've just rebuilt - after a gap of 43years - an analogue bandpass filter with a Q>1000, rolloff of >30000dB/decade out of 10% components and eight FET analogue switches.

Since it is also a mixer, it is fun to use a standard scope's V-vs-t display to watch demodulation of an AM signal on the "carrier" at the centre of the bandpass, and a scope's X-Y display to watch FM signals on the quadrature filter outputs.

Lovely example of what Nyquist doesn't equire, and the power and simplicity of a sampled-time continuous-amplitude system.