I want to know what good for is that Bandwidth in a DSO

[_q12x_]

- I want to know what good for is that Bandwidth in a DSO !
With examples! What is BEST used for? Why is so important?
I mean... I know I heard some guys, on youtube videos and even here on the forum, who have old types osciloscopes, before these DSO models appeared, in 2000 I guess or so, they told me theirs worked at very low bandwidth. Like 10 or 5MHz.
Also I was reading a post of a romanian compatriot that bought the lowest bandwidth DSO from the list, for a bit more cheaper price than on the original website. That DS1054Z at 419 USD, he got it at 401 USD from here in romania as a 'cheap' offer. (Is still not cheap for me)
The actual page: 1000Z Mixed Signal Oscilloscopes RIGOL

 

[Julie.]

Well bandwidth is the point where your signal drops by 3dB .

So very over simplified if you measure a 10kHz signal as 10V p-p that same signal will only measure 7V at 50MHz on a 50MHz 'scope - even though the actual signal is 10V.

In practice it is much more than this though. Say you are measuring a 20MHz square wave, in order to represent this, that's a 20MHz sine wave plus harmonics - very quickly the harmonics will go beyond the 50MHz capacity of the 'scope!

This would mean that a perfect 20MHz square wave would show as a very distorted "square" wave on the 50MHz 'scope.

Typically you need 5x bandwidth to signal frequency, as a minimum, ideally >10×

So it depends what you are doing, audio is 20 - 20kHz - so you need >200kHz bandwidth, but for computer signals this may operate at 20MHz or more, so even a 100MHz 'scope is only 5x , you really want a 200MHz or more 'scope.

If you are using it as a logic analyser - you need to take the switching speed - convert to frequency, then choose the bandwidth.


EDIT

Should really add as it's a DSO, you really need the details of sampling rate, plus memory, which is also dependent upon the timebase chosen , so a 5G sample rate could easily drop to 20M depending upon the timebase - given a signal to sample rate of 10 or 20x the max bandwidth would become 1-2 MHz irrespective that the 'scope has a bandwidth of 100MHz

Digital 'scopes tend to hide the practical limitations of sample rate, storage, etc

BTW most 'scopes share this sample rate across channels, reducing it even further

 

[_q12x_]

Very cool answer ! Thank you.

 

[_q12x_]

Speacking from fv pov, I'm more an automation type of guy. Im less into radio or mobile phone/tablet. So going on your last idea, I made a quick search for crystals to check their MAX fv and they are at max 300MHz.
I go and look at the wi-fi and bluetooth as well and here is what I find. I only hunt for the maximum frequencies in this search.

Also like @nsaspook mentioned, "Spend a little more on a modern DSO with signal decoding and capabilities." But that will jump up the price with probably another 1000$. So thats no bueno. But it will be very nice to have these options, mmmm. Especially in automation.
- Beside >signal decoding<, what else is very important to have? Especially for automation line.
I then go into 7400-series frequency list and I couldnt find any concludent answer, and absolutely nothing from Wikipedia, which was shockingly odd. The same for TTL logic IC's. Nothing. Hmmm. It must be up to 200MHz as papabravo said. Im not sure if logic IC's are quartz dependent, because if they are, then it will jump them to 300MHz. I know MCU are quartz dependent. But CPU ?

These NXP are MCU and CPU. Then the Microchip:

Ok... haha.
In general CPU fv:

What is using a CPU for such (4.3GHz) high fv if not a quartz?
Im doing a bit of homework here live. Haha. I want to get a better idea of how things tick, literally. Haha.
So the conclusion is simple.
The optimum DSO for --ME-- is in accordance with the highest crystal fv, so a 300MHz. For the moment and as my first professional DSO.

 

[pc1966]

@Julie. has covered the key points, that typically you want about 5 times the fastest signal rate you are looking for, but the Devil is very much in the detail.

The capture rate of DSO has improved massively over the years, but at the very top end they are still "sampling scopes" that work on repetitive signals, so first time round they capture at say 100MHz so every 10ns, then next time they shift by 0.1ns and repeat, and 3rd time by 0.2ns and repeat, etc, so if the signal is stable and repetitive you get 10GHz sample rate.

Unfortunately many real-world applications are for glitches and other non-repetitive phenomenon so to capture them you need to spend a small fortune on a DSO that genuinely can sample in the multi-GHz region.

 

[pc1966]

What is using a CPU for such (4.3GHz) high fv if not a quartz?

Most fast semiconductors have internal clock "multipliers" that are typically PLL systems that lock to a modest frequency crystal in the 10-50MHz region and then generate one or more stable clock signal in the hundreds of MHz to GHz region.

Im doing a bit of homework here live. Haha. I want to get a better idea of how things tick, literally. Haha.
So the conclusion is simple.
The optimum DSO for --ME-- is in accordance with the highest crystal fv, so a 300MHz. For the moment and as my first professional DSO.

Generally speaking crystals above 20-30MHz are overtone mode. They still have the fundamental resonance in the few/tens of MHz region, but they have been cut so the 3rd, 5th, etc, overtone is on the wanted frequency. These are known as "overtone" and not harmonic as they are not exact multiples of the fundamental.

To make an oscillator that works on overtone mode crystals you need a frequency-selective amplifier as the crystal always wants to run in fundamental mode. So typically you see an LC tank (inductor & capacitor) in the circuit tuned to provide sufficient gain only at the overtone. Some very high performance mid frequency oscillators (typically 5MHz or 10MHz frequency standards) also use 3rd overtone cut crystals as it allows the circuit impedances to be sane and still not significantly degrade the loaded Q-factor of the very best crystals.

But these days if you want a high frequency you would look as a frequency synthesiser chip that takes, say, 10MHz and allows you to programs the desired output (within reason). Quite a few of such chips have built-in VCO (voltage controlled oscillator) so they really are complete "systems on a chip" solutions, while others for higher signal purity still make use of an external VCO and then the design and specification of that aspect can be critical to signal purity.

Which matters not one bit for a CPU clock! In fact they often deliberately jitter the clocking (spread-spectrum) to reduce radio interference!

But if you are design a radio or radar then phase noise / spectral purity is a very important parameter.

 

[pc1966]

Here is an example of a sampling scope:

DCA-X Sampling Oscilloscopes

Keysight’s DCA-X sampling oscilloscope is part of our Digital Communication Analyzer family. See available mainframes and modules with superior precision.

www.keysight.com

Capture rate is 256k sample/sec but bandwidth a whopping 110GHz. Starting price is from $90k

Edit: Just to add that they also do one with real-time bandwidth that is comparable as 256G sample/sec capture. Apparently price is around $1.3M!

 

[_q12x_]

Im not into analyzing crystals to be clear. Im only referencing and guiding after the top speed of one.
The electronics that are using crystals or lower speed than a crystal, to be easily read, especially in my automation direction.
I like your explanations mister @pc1966 .

 

[Julie.]

Exactly!

With an analogue storage 'scope or standard 'scope the bandwidth dictates the display.

However when it's a digital, the sampling rate actually dictates the storage.

There's a whole host of sampling theories, most notably Shannon - Nyquist, which dictates the sample frequency has to be twice the measured frequency in order to measure the frequency, although the display will show a clear triangular waveform rather than the actual sine wave, if you need true representation then you need 10x.

So let's say we have a sample rate of 4G, but you are using 4 channels, this gives only 1G samples per channel per second.

If you have 100k of memory points, the memory would be full in 1/10000 (100us) seconds

Now your timebase reduces this, at 10ms/div over 10 div , if you have 100k of memory, you would fill this with a sample rate of 1000 samples per second.

So you don't have the 1G sample rate at all, if you were looking at a 50/60 Hz signal looking for a 100kHz superimposed distortion (say from a smps) - the storage 'scope wouldn't be able to see this signal - your conclusion being there is no superimposed signal from the smps on the supply side, which would be completely wrong.

With dsos you have to know exactly what the signal is, in order to see if the dso is representing it or not.

 

[_q12x_]

- Beside >signal decoding<, what else is very important to have? Especially for automation line.
or this is the most important function next to a common sampling of a waveform?

 

[pc1966]

Not sure what the top speed of commercially available crystals are. They used to top out around the 300MHz region but I suspect that and higher are now using SAW resonators instead.

 

[pc1966]

- Beside >signal decoding<, what else is very important to have?

Sadly I would have to now say "A sane user interface"

Especially for automation line.
or this is the most important function next to a common sampling of a waveform?

It really depends on what you are trying to do. If you are analysing digital signals then logic analyse features like triggering on a pattern are useful, for serial comms the ability to recognise and decode common formats, etc.

If working on audio and low frequency stuff then good bandwidth-limiting options, and also CAT-III probes if working on AC power circuits!

Size of display, battery/portable, etc are other factors. For many a PC adaptor and "virtual scope" is better so data can be captured and analysed, but then you have issues of long-term support as operating systems change, etc.

 

[pc1966]

Back you the OP's first question, then I would probably take the cheapest one as it is 200MHz and fastest.

4 channels is potentially very useful for logic, but remember many 2ch scopes can at least have an external trigger so you can look at 2 independent-in-time signals as well. However, prices is very much more. More importantly they seem to allow you to add ($200) a digital capture 16-channel thing for digital logic analysis.

It is easier to chose for a very specific job, as otherwise you can't tell if you would need faster in the future or more channels.

Ultimately you pay your money and take your choice!

 

[pc1966]

Just to add - if you do get one of the logic analyser sort of add-ons be very careful to only use it on well-behaved low voltage logic. Typically they have very limited protection against over-voltage!

 

[_q12x_]

Well, I already have a DSO138 for experimenting so far. I have it for 3 years or so.

What I dont like on it: It has a very unclean waveform, full of jag and stairs and noise. Not at all clean.
Then, I recently find out it is not suitable to measure logic IC's, so under its minimum limit of 10us.
I dont use it THAT much so a better osciloscope will most probably catch dust for a good percentage of time.
But it will be cool to have one though.
Very good explanations and I really enjoy reading them, very nice lecture to read.
I still have no idea what to choose. Haha.
I guess, the bottom line is... choose one, and live with it.

 

[Julie.]

Well, I already have a DSO138 for experimenting so far. I have it for 3 years or so.
View attachment 102652

What I dont like on it: It has a very unclean waveform, full of jag and stairs and noise. Not at all clean.

...

That's a result of the sample rate.

Misunderstood effects of this do lead to missing key information.

I have a similar oscilloscope to the ones you are looking at, actually a siglent sds1000, it is usable - but you do have to be aware of limitations, when looking for a signal, don't just leave it on the same timebase, apply the signal then sweep through. You get one picture on one timebase and a totally different one on another!

 

[pc1966]

At times i have made simple RC or op-amp filters just so I can keep the scope display to the low frequency signal as I knew there was a lot of HF that would just upset things.

With the older analogue scopes you got a more complete/less strange display under those conditions.

 

[pc1966]

I have a Keysight EDUX1002A which is OK, though now obsolete/replaced by something in darker plastic. Not so expensive I worry about using it.

I also have a Rhode & Schwartz RTB2004 which is far more capable, but also going to be painful if I trash it.

 

[_q12x_]

Does exist in this big world second hand oscilloscopes at lower prices than the new ones?
I never find them, maybe you can direct me to some websites that you may know. Doesnt matter what country they are. Just to be cheaper than the new ones.

 

[pc1966]

I don't know, you could try the usual market places, etc.

Sometimes universities will have old stuff they get rid of when a budget allows replacement but often they are really old and sometimes they just get put in cupboards "just in case". But it might be worth asking.