[SOLVED] Oscilloscope parameters

[mrinalmani]

I am looking forward to purchase an oscilloscope but I am confused with the specifications.
A typical example of specification is:
Bandwidth: 25MHz
min time step: 5ns
Sampling rate: 100Ms/S
Rise time: 14ns

But I dont understand somethings.
1. If sampling rate is 100Ms/S then the min time step should be limited to 10ns. How is 5ns possible?
2. Bandwidth should be 100MHz, why lower?
3. what is rise time?

I need an oscilloscope to measure frequency response of high frequency transformers (100kHz to 250kHz) and also to analyze switching delays (in order of 10ns to 20ns).
4. What minimum ratings should I go for?
5. How about PC based oscilloscope modules... any idea?

 

[FyrewurXedge]

With your specs, I think your oscilloscope can read/measure on what you stated that you need.
By the bandwidth 25MHz, it is lower maybe by design so it not as good as oscilloscope that has 100MHz.

The min time step means you can view your wave/signal into 5ns so if you want to measure anything lower that 5ns you can't view it.

Rise time is the amount of time your signal goes from minimum to maximum.

 

[mrinalmani]

Thanks. But if the sampling rate is 100MHz, it means 1 sample every 10ns. How then, can the oscilloscope claim a minimum time step of 5ns?

 

[expert_vision]

At 100MSPS a 25MHz signal will only have 4 samples per period. Trying to display signals higher than that would be just unreadable. The maximum theoretical bandwidth is 1/2 sampling rate (Nyquist frequency).

By minimum time step they mean the minimum time per x division on your screen. At 5ns time step you'll have a sample every 2 x divisions on your screen.

 

[mrinalmani]

I still dont understand the meaning of rise time in context of oscilloscope specification.
When the sampling period is 10ns, regardless the rate of change of signal it must be sampled at 10ns and the results be displayed on the screen. Then why a limitation of 17ns in rise time?
Also I dont suppose Nyquist criteria holds here. It is intended for reconstruction of wave and not for displaying sampled points!!

 

[expert_vision]

Well, the oscilloscope is capable to sample every 10ns, but it won't be able to detect a signal with a full swing in less than 17ns. Effectively, you won't be able to measure rise times smaller than 17ns, even though the sampling period is 10ns. This limitation is due to step response of the analog input circuit of the oscilloscope.
The 10ns sampling rate is a limitation of the digital input circuit (and other things).

The Nyquist frequency limit is used for signal reconstruction, but if you would try to display a signal sampled at a rate smaller than 2*f, you would deal with aliasing issues. If you don't have at least a few samples per period, you won't be able to make any reliable measurements.

The oscilloscope bandwidth is 25Mhz instead of 50Mhz(Nyquist frequency) because of the analog input circuit of the oscilloscope. Signals with frequencies higher than 25Mhz will be attenuated more than 70% (-3dB).

 

[FvM]

A point that hasn't been considered yet in the discussion is that periodical signals can be displayed with more points per time division using equivalent time sampling. The bandwidth and respective risetime limitations will be still in effect of course.

 

[mrinalmani]

All right I get it. But what is "equivalent time sampling"? Is it somehow related to a few manufacturers quoting stuff like "dynamic step resolution", which is much higher than the actual sampling resolution. Kindly explain...
Thanks

 

[FvM]

It means that the display is generated from multiple triggers. There are different names are used for the feature, I'm under the impression that most manufacturers are using it but I'm not quite sure.

 

[crutschow]

I believe "equivalent time sampling" is used for display of a repetitive waveform and consists of varying the trigger point slightly between multiple samples of the waveform period and then combining (averaging) them to give a smoother displayed waveform. It would reduce the coarseness of the samples.

 

[expert_vision]

I believe there are multiple techniques to achieve that, but in principle, high sampling rate of a period is obtained by acquiring samples from multiple periods, which then are interlaced.

 

[mrinalmani]

I still quite don't understand "equivalent" sampling rate (which by looking into specification sheets ,seems like a warehouse if the "actual" sampling rate were a retail outlet!!)

 

[D.A.(Tony)Stewart]

Old Analog Sampling oscilloscopes work on the basis of the resolution of the time delay resolution and not just the sampling rate. That is how they achieved ns resolution repetitive waveforms using very fast sample and hold methods.

The better DSO's do the same thing to improve bandwidth resolution for repetitive waveforms. This of course doesn't work for 1 shot waveforms.

 

[mrinalmani]

Ok, I got it somewhat. But is the plot obtained by equivalent sampling an accurate picture of what is going on?
Consider the following....
Suppose the "actual" sampling interval is 100ns.
Now if there is a local oscillation or spike that lasts less than, say 30ns. Then will the scope show this oscillation/spike, assuming the "equivalent" sampling interval to be as low as 5 ns

 

[FvM]

Modern real time oscilloscopes have reasonable balance of sampling rate and analog bandwidth. In so far your example is somehow unrealistic. But generally, signals faster than the sampling interval can be recorded/reproduced if they are periodical and pass the analog bandwidth.

 

[expert_vision]

You can't use equivalent time sampling to measure spikes. Equivalent time sampling works only with repetitive/periodic signal (a sum of sines).

The way equivalent time sampling works is by shifting the sampling in time after after a trigger.
Let's say you have a 100MHz sample rate and you want to display a 100MHz signal. The trigger is set to say 0V and a set of samples are being made. Being that the sample rate = signal freq you will get a sample in each period. That's pretty useless. Next, another set of samples are being made, but with a small delay after the trigger hits. This results in another set of samples, one per period, but with a known time offset from the initial ones. The process is repeated until the delay reaches the sampling period. The final result is the reconstructed waveform obtained from multiple repetitive waveforms.

The second picture in **broken link removed** depicts this process. Each sampling cycle is done at a different time, that's why your signal must be strictly repetitive.

 

[mrinalmani]

Now I get it... Thanks a lot!