Why do oscilloscopes have bandwidth limits?

[Anding]

“… because an oscilloscope can be considered as a low-pass filter and this is the -3db cutoff frequency”, yes but why should it be the case that an oscilloscope is a low pass filter?

“… because it’s good for the market if manufacturers can segment buyers according to need”, yes but that’s a way of taking advantage of bandwidth limits - not a reason for them

If my scope can sample at, say, 8GSa/S and I am happy to sample my waveforms with 8 sample points each, why do I not have a scope capable of measuring (some aspects of) a 1GHz digital signal regardless of the scope “bandwidth”?

[ataradov]

The fundamental reason is that designing a high bandwidth front end is not easy and there are physical limitations. There are further limits set by marketing, of course, but usually they are not too far away from the physical limits, especially for higher end models in the same range.

And of course it is possible to design 1 GHz front end, but that would put the whole device into much higher price bracket.

[Anding]

From experience how far can you use an oscilloscope beyond it's bandwidth limit?   Specifically, I'd like to compare the phase difference between in and out DDR2 data busses between an FPGA and a DRAM chip.  This is important because the input signals will need to be captured inside the FPGA using a clock with a phase delay to the output clock.  The slowest speed of the DRAM is 125MHz, which is starting to imply a 1GHz 'scope for many thousands of dollars.  Just for one measurement!  Will a 350MHz (soft upgraded) scope be any help to me at all?

[ataradov]

125 MHz square wave will look like a sine wave with 350 MHz bandwidth. The rest of it would be filtered out. Remember the bandwidth of the scope is stated for sines, and if you look at the spectrum of the square wave, only the first harmonic would fit.

And if you want to preserve the phase to that extent, you would need a way better equipment. This is not going to happen on consumer-level gear.

[jonpaul]

Not a simple question!

SR, Bandwidth, risetime, aberration are all different but interrelated specifications.
Digital scope vertical BW is a spec for the amplifiers.
Read about  the theory of ADC, SR, BW.

Sample rate is not directly related, and there is a limit to the BW for a particular max SR, the Nyquist rate. BW < 1/2 SR.
The filter shape for antialiasing is the determinant.

See the Tektronix Measurement and circuits concepts book on vertical amplifiers.

Finally the best analog scopes are  TEK


The most unreliable specs are on the Chinese knockoffs.

We use Yokogawa DL and DLM series digital scopes, and TEK 2465/7/B, 7000 analog.

Jon

[BeBuLamar]

“… because an oscilloscope can be considered as a low-pass filter and this is the -3db cutoff frequency”, yes but why should it be the case that an oscilloscope is a low pass filter?

“… because it’s good for the market if manufacturers can segment buyers according to need”, yes but that’s a way of taking advantage of bandwidth limits - not a reason for them

If my scope can sample at, say, 8GSa/S and I am happy to sample my waveforms with 8 sample points each, why do I not have a scope capable of measuring (some aspects of) a 1GHz digital signal regardless of the scope “bandwidth”?

Why does my tape measure has length limit. Mine is 20ft but why not unlimited?

[Fungus]

Why can't my car go faster than sound?

[Anding]

Why can't my car go faster than sound?

No reason why it shouldn't but that would require an engine sufficiently powerful to accelerate the car against the restive forces of friction and air resistance which increase with velocity (reference Newton II: F=ma), and sufficient fuel onboard (mass which itself needs to be accelerated) to complete the acceleration.  In the case of a car, it would require rolling gear that maintains it's mechanical integrity at the high RPM for travel above 330 m/s.

Can you put the equivalent physics for a 'scope in two sentences?

[Brumby]

Why can't my car go faster than sound?

No reason why it shouldn't but that would require an engine sufficiently powerful to accelerate the car against the restive forces of friction and air resistance which increase with velocity (reference Newton II: F=ma), and sufficient fuel onboard (mass which itself needs to be accelerated) to complete the acceleration.  In the case of a car, it would require rolling gear that maintains it's mechanical integrity at the high RPM for travel above 330 m/s.

Can you put the equivalent physics for a 'scope in two sentences?

There is more that can be said about the subject, but I'll focus on one....

"Do you want to pay for it?"

[Anding]

Why does my tape measure has length limit. Mine is 20ft but why not unlimited?

Because the length limit is set by the amount of material in the tape.  Whilst there is no theoretical upper limit, above say 20ft it becomes hard to layout the tape and keep it rigid and inconvenient to rewind it.  Above 20ft we can use surveyors tools or nowadays cheap laser meters to accomplish the measurement.

Can you explain the engineering constraints on the bandwidth limit of a scope in two sentences?

[Anding]

I'll do it in one....

"Do you want to pay for it?"

That is a financial constraint.  What is the necessary physics that one needs to pay for?

[Brumby]

There is a reason electronics in the RF region is often referred to a voodoo.

I know enough about the subject to know I understand extremely little - just a few principles and almost and next to no practical experience.  I've tuned 27MHz mobile antennae and not much else.

[Someone]

Why can't my car go faster than sound?

No reason why it shouldn't but that would require an engine sufficiently powerful to accelerate the car against the restive forces of friction and air resistance which increase with velocity (reference Newton II: F=ma), and sufficient fuel onboard (mass which itself needs to be accelerated) to complete the acceleration.  In the case of a car, it would require rolling gear that maintains it's mechanical integrity at the high RPM for travel above 330 m/s.

Can you put the equivalent physics for a 'scope in two sentences?

Same as the car needing more power and expensive fancy wheels/tires, fast amplifiers need more power and expensive components. Jamming in more power needs all the supporting expense or it blows up.

[Anding]

There is a reason electronics in the RF region is often referred to a voodoo.

I know enough about the subject to know I understand extremely little - just a few principles and almost and next to no practical experience.  I've tuned 27MHz mobile antennae and not much else.

Yes, this is bringing back memories of simulating RF circuits with a package called "Super Compact" many years ago (I was just running the scripts and printing out the logs ;-) .  Answering my own question here but

The lumped circuit abstraction model starts to break down at RF frequencies and we can no longer treat components as discreet.  The circuit board itself has to be modelled.  At the same time as an oscilloscope is a measurement tool, a specification has to be given over which it maintains linearity of response and for which calibration is accurate.  A convenient way to do that is to _impose_ (I'm guessing here) a low bandpass filter on the front end to exclude the region which the maker has not designed for (price point of the design considered)

I'm curious, with ingenuity what use can be gotten "above the bandwidth limit".  This can be experimented using a simple FPGA to generate fast (200MHz) clocks and waveforms.  Ok there's a project!

[Anding]

fast amplifiers need more power and expensive components

That is very helpful thank you

[alm]

Read the Tektronix Circuit Concepts book on vertical amplifiers that Jon referenced, at least chapter one. Obviously no one makes vertical amplifiers with discrete transistors or tubes anymore, but the description about the relationship between bandwidth and transient response is still relevant.

The bandwidth is limited by the design of the amplifier. In the case of an op-amp, it would be related to the gain-bandwidth product and the gain it is configured for in the circuit (very simplified). There might be additional limitations, like the configurable bandwidth limit (e.g. 20 MHz), or bandwidth limits added so they can offer lower bandwidth models without much additional engineering. That's why some scopes can be 'hacked' to a higher bandwidth, but this is generally limited to the highest bandwidth model in the same series, like 50 MHz Rigol DS1052E to 100 MHz (DS1102E equivalent), but not higher.

One important distinction is if you want to look at the bus to decode what is being sent, or if you are verifying the signal integrity of the bus where you want to look at details like meeting setup-and-hold timing requirements. There's a rule-of-thumb chart posted here with the necessary bandwidth for decoding and checking signal integrity for common buses.

[jonpaul]

à moment please

analog and digital scopes are completely different in the archetecure design and components

All blocks in the signals path affect the overall observed bandwidth



For example an analog scope CRT with a simple deflection plate has perhaps 10..50 MHz BW

At 100..300 MHz a dome mesh PDA

For 400 MHz to 1 ghz distributed deflection with controlled delay and Zo

Tektronix 2467/B and 7104 use that and microchannel intensifier plate

see the Tektronix CRT concepts book

Bon courage

Jon

[robert.rozee]

i am quite certain that i will regret asking this, but...

if you're just looking at the rise-time of digital (0-5v) signals, why do you need an amplifier in front of the ADC? most oscilloscopes are designed to handle input signal levels ranging from 10's of mV up to 10's of volts, but in the specific case outlined in this thread it seems that this sort of flexibility is not required.

might you just need perhaps an impedance buffer, and possibly a differential driver (in the case of the ADC having a differential input). and i'm wondering if you even need more than ONE channel? yes, this might end up being a very specialized instrument, one that can barely be called an oscilloscope.


cheers,
rob   :-)

[TheUnnamedNewbie]

I'll risk going over board here, but as my dayjob is working against these limits, I'll give a brief insight in what is going on.
For scopes, I think there are a multitude of factors: Cost, interconnections, and technology.

Cost is simple: faster needs more design expertise, fancier technologies, more hardware, more storage, and at the end of the day things are more expensive.

Interconnections: You can't just send 10 GHz signals over cheap connectors. The connector starts to have a big impact, as do the cables. Everything starts becoming part of the circuit and that requires more finicky interconnections. I wouldn't want to use our 110 GHz UXR scope to measure an arduino output, because it would load it with 50 ohm which the arduino probably wouldn't be very happy with. High speed scopes start using connectors like 3.5 mm, 2.4 mm, etc, which are very fragile, require care in connecting-disconnecting, and are very expensive - a 2.4 mm to 3.5 mm metrology-grade adapter might cost several hundred euros a piece. I don't want that on a scope I want to use to measure some SPI busses with, so there is little point making those scopes that fast, in a right-tool-for-the-job kind of way.

At the upper limit is technology. There are two metrics with regards to the 'speed' of an active device,  the 'transit frequency', or ft, and the 'maximum power gain frequency', or fmax. This is the point where, due to the fundamental physical limits (things like the mobility of carriers in semiconductors) make it so a device no longer produces current gain (for ft) or power gain (for fmax). At this frequency, you need more current into a device than you get out, so you no longer have the ability to make signals larger - let alone do more complex processing of this signal. This puts a first-order boundary on a scope. (actually above-fmax circuits are a thing but lets not open that can of worms)

For discrete active electronics this is going to be in the few GHz (usuall) at best. It goes higher when you use integrated circuits - small CMOS goes to about 300 GHz (depends on the technology, but most people agree that it peaks at around 40nm CMOS nodes). SOI can go further, to 400 GHz, SiGe can do 500 GHz, InP has technologies that go up to 1 THz (but that technology is trash for anything but a very simple amplifier).

On top of that, a significant additional factor is that designing these high speed 100 GHz circuits is by no means trivial. It is by many considered a black art, and us millimeter-wave and RF designers are sometimes treated as voodoo magicians, wrestling waves out of devices. Sparameters, matching, smith charts...

An on top of that, to get to these points, you need to start canceling out poles with zeros, or as an RF designer would see it, tune out capacitance with inductance. This introduces a bandwidth limit - you only tune out the capacitance at a certain frequency. Make it more wideband? That requires more complex interconnections of inductors and capacitors, which in turn leads to lower gain, and at some point you lose all the benefits you were trying to get. Even though you can design a 230 GHz to 250 GHz amplifier in 45 nm CMOS, you cannot design a DC-200 GHz amplifier in the same technology

[Brumby]

I wouldn't want to use our 110 GHz UXR scope to measure an arduino output...

Thank you for that.  That was a brilliant mental picture!   

[pdenisowski]

Here's my answer

[pdenisowski]

From experience how far can you use an oscilloscope beyond it's bandwidth limit?   

It depends on the input waveform.  If you're looking at a pure sinusoid, you can often use a scope well past it's 3dB bandwidth without the waveform being distorted, although the amplitude of the displayed sinusoid will be reduced.

On the other hand, if you're looking at a signal with high frequency components (e.g. a square wave), then these higher frequency components will be attenuated more than the lower frequency components and waveform distortion will result.  This is why a square wave (fundamental and odd harmonics) ends up looking line a sine wave when you have insufficient bandwidth:  the higher frequency components have all been significantly attenuated.  (ataradov mentioned this above as well )

[joeqsmith]

I attempted to make a virtual oscilloscope to demonstrate LeCroy's Digital Bandwidth Interleaving.  The basic idea was to combine two channels to double the bandwidth rather than the time resolution.   Rodger Delbue chimed in with some additional comments on the subject:



A few of us were attempting to build a simple oscillator on a breadboard for the fun of it.  Far from a scope or anything useful, but it my give you some insight to the problems.   

https://www.eevblog.com/forum/projects/challenge-thread-the-fastest-breadboard-oscillator-on-the-mudball/msg3081669/#msg3081669

In this video I tried to measure some pulses that I sent down different media, from copper to optics.   

[RoGeorge]

Why there has to be a limit?

For oscilloscope it's easy to say, you have a given budget (price, power, thermal, etc.) than there is the sampling theorem, and for your 8GSa/s ADC the most you can reconstruct (for an unknown non-repetitive signal is 4GHz).

The question is much deeper, as in "why there is no observable infinite in the physical world"?

Nobody knows, but so far infinite only exists as a concept, i.e. in math.
In the physical world (as far as we can observe) nothing is infinite. 

[Fungus]

Can you put the equivalent physics for a 'scope in two sentences?

I can do it in two words: "Inductance" and "Capacitance"