Can a scope test it's own bandwidth?

[injb]

I just bought this scope, a Picoscope 2405A, and my first real oscilloscope. I've been playing around with it to get familiar with the features. It has a signal generator built in, and I wanted to see how well the scope works at high frequencies. The scope is supposed to have 25 Mhz bandwidth. But the signal generator only goes to 1 Mhz, and I don't have any way of generating a higher frequency with a known waveform.

So I hooked up the scope to the 1 Mhz square wave, and I was a little disappointed to see that it was pretty distorted. In fact the highest frequency that looks anything like a square is a few hundred Khz. But some rounding off is to be expected, after all, it can't show all the harmonics since it's limited to 25 Mhz anyway right? I know the theory of this, but I'm short on experience, so my problem was, I didn't know how to tell *how* distorted a 1 Mhz square wave should be on a scope with 25 Mhz bandwidth. Then I had an idea - the software has a spectrum analyzer feature. Could that tell me exactly how much harmonics I'm picking up, and thereby show me the true limit of the scope's bandwidth? I thought this would be a valid test, since the spectrum analyzer is just a software feature and can only display information that the scope has successfully picked up.

Here's what I got when I tried it out (see attached pic).

The 25 Mhz peak is actually fairly distinct, and in fact the 27 Mhz one is visible too. After that it's just noise.

So I take that to mean, my funny looking signal *is* what a square wave should look like with bandwidth limited to 25 (or maybe 27) Mhz. Is that right? (Not that I had any reason to doubt the specs, but I just like to see things for myself...that's why I have a scope!)

[ataradov]

This looks more like under-compensated probe. How do you connect your generator to the scope?

[injb]

This looks more like under-compensated probe. How do you connect your generator to the scope?

That's what I thought at first! But when I tried adjusting the compensation, the amplitude changed, but the shape did not. The probes are supposed to be compensated already, and the instructions say to use a 1 Khz square wave and adjust until it looks square - at 1 Khz it was already spot on out of the box. I'm using the probe on the x10 setting.

The generator is built into the scope - it's just another BNC connector beside the channels. So I just put the probe tip directly into it. It's all grounded together inside the scope anyway. I did try attaching the ground lead to the shell of the signal output BNC, but it made no difference.

[ataradov]

Here is how 1 MHz square wave would look like after filtering through a filter with 25 MHz bandwidth.

Your result is definitely just some side effects of the way signal generator interacts with the scope.

Or siggen is just at it limit and can't output a good wave.

[xrunner]

I just bought this scope, a Picoscope 2405A, and my first real oscilloscope. I've been playing around with it to get familiar with the features. It has a signal generator built in, and I wanted to see how well the scope works at high frequencies. The scope is supposed to have 25 Mhz bandwidth. But the signal generator only goes to 1 Mhz, and I don't have any way of generating a higher frequency with a known waveform.

No you can't use the 1 MHz generator sine wave to test a 25 MHz bandwidth. At 25 MHz the signal amplitude the scope measures would be 3 dB down from the actual value coming from a known calibrated source - but you are a far cry from getting up to that frequency.

But the good news is you now have an excuse to buy more test equipment. 

[Circlotron]

Here is how 1 MHz square wave would look like after filtering through a filter with 25 MHz bandwidth.

After going through a DSP filter maybe, but this situation is more like an RC filter I expect.

[ataradov]

After going through a DSP filter maybe, but this situation is more like an RC filter I expect.

Good point.

In any case, more equipment required to prove either instrument.

[injb]

I just bought this scope, a Picoscope 2405A, and my first real oscilloscope. I've been playing around with it to get familiar with the features. It has a signal generator built in, and I wanted to see how well the scope works at high frequencies. The scope is supposed to have 25 Mhz bandwidth. But the signal generator only goes to 1 Mhz, and I don't have any way of generating a higher frequency with a known waveform.

No you can't use the 1 MHz generator sine wave to test a 25 MHz bandwidth. At 25 MHz the signal amplitude the scope measures would be 3 dB down from the actual value coming from a known calibrated source - but you are a far cry from getting up to that frequency.

But the good news is you now have an excuse to buy more test equipment. 

I was using the square wave, not the sine wave. My reasoning was that the square wave has higher frequency components than 1 Mhz.

It sounds like I need more stuff anyway! I do have an old function generator that currently doesn't work. I can fix it, but it's max frequency is also 1 Mhz. Maybe it'll produce a better square wave than this though.

[DaJMasta]

Easy way to check if it's rise time from the sig gen: get a TTL buffer chip (hex inverter, schmitt trigger, etc.) with reasonable bandwidth (almost any will do, but something like HCT is a good choice) and pipe the signal into one side and monitor the output with the scope.  A relatively fast TTL family logic chip is usually rated to at least 30MHz or so, so you'll get nice clean edges on a 1MHz signal, and even with the parasitics from a solderless breadboard, they shouldn't take the edge off things too much at that frequency.  Just make sure you bias the input to keep it positive (AC coupling cap and then a resistor divider will do the trick).  If you're into electronics and have a scope, you've likely got all the parts you need for it, and if you take a look at the datasheets for the chip you select, you could easily get one capable of use at 60+MHz, which would certainly give you an edge hard enough to check 25MHz bandwidth (though, you may want to stick it in some protoboard to keep the capacitive parasitics down).

That said, it's worth mentioning that everything important to edge times and calculating bandwidth works on the falling edge too, which looks faster.  Calculate the speed of the falling edge (1 / fall time) and multiply by a coefficient for your appropriate front end filtering (I'd start with 0.45) to figure your bandwidth based on edge rate.  Otherwise, find yourself a sig gen that can go to 25MHz+ and see where your -3dB point is.

[vk6zgo]

I just bought this scope, a Picoscope 2405A, and my first real oscilloscope. I've been playing around with it to get familiar with the features. It has a signal generator built in, and I wanted to see how well the scope works at high frequencies. The scope is supposed to have 25 Mhz bandwidth. But the signal generator only goes to 1 Mhz, and I don't have any way of generating a higher frequency with a known waveform.

So I hooked up the scope to the 1 Mhz square wave, and I was a little disappointed to see that it was pretty distorted. In fact the highest frequency that looks anything like a square is a few hundred Khz. But some rounding off is to be expected, after all, it can't show all the harmonics since it's limited to 25 Mhz anyway right? I know the theory of this, but I'm short on experience, so my problem was, I didn't know how to tell *how* distorted a 1 Mhz square wave should be on a scope with 25 Mhz bandwidth. Then I had an idea - the software has a spectrum analyzer feature. Could that tell me exactly how much harmonics I'm picking up, and thereby show me the true limit of the scope's bandwidth? I thought this would be a valid test, since the spectrum analyzer is just a software feature and can only display information that the scope has successfully picked up.

Here's what I got when I tried it out (see attached pic).

The 25 Mhz peak is actually fairly distinct, and in fact the 27 Mhz one is visible too. After that it's just noise.

So I take that to mean, my funny looking signal *is* what a square wave should look like with bandwidth limited to 25 (or maybe 27) Mhz. Is that right? (Not that I had any reason to doubt the specs, but I just like to see things for myself...that's why I have a scope!)

A 1MHz square wave should look like, surprise, surprise, a 1MHz square wave on a 25MHz Oscilloscope!
You only need response out to about the 7th harmonic of the square wave fundamental frequency for it to look OK, so a 10MHz 'scope would do a pretty good job.

It looks as if the "signal generator" is a pretty basic function generator, which " runs out of grunt" at 1MHz.

Obviously, there are three things you can do.
(1)Make up or borrow a source of good 1MHz square waves,
(2)Borrow a different 'scope & look at your existing square wave,
(3)As someone else has suggested, borrow a good RF signal generator, & test the 'scope out to 25MHz with sine waves.

[ADT123]

The PicoScope 2205A signal generator / AWG output has a bandwidth of 1MHz that is why the edges of the square wave are rounded so all look normal.  See specifications here https://www.picotech.com/oscilloscope/2000/picoscope-2000-specifications

As others have said look at the edge from a fast logic signal to see higher bandwidths / rise times.

[Berni]

Yep you need a faster generator that can produce a square wave with more than 25MHz of bandwidth.

One of the forum members has created such a generator for this very purpose of testing scopes ( https://www.eevblog.com/forum/projects/yet-another-fast-edge-pulse-generator/ ). The incredibly fast edges it generates are faster than even a 4GHz oscilloscope can see. So this means that when you look at the square wave of the scope you are seeing how much the scope rounds off the edges rather than showing how round the signal of the generator is.

You can probably build one yourself out of some fast 74xx logic and a oscillator. It most certainly wont push a 4GHz scope to the limit but it will certainly be faster than your 25MHz scope without a problem.(It can generate about 1ns rise times)

[tggzzz]

Yep you need a faster generator that can produce a square wave with more than 25MHz of bandwidth.

One of the forum members has created such a generator for this very purpose of testing scopes ( https://www.eevblog.com/forum/projects/yet-another-fast-edge-pulse-generator/ ). The incredibly fast edges it generates are faster than even a 4GHz oscilloscope can see. So this means that when you look at the square wave of the scope you are seeing how much the scope rounds off the edges rather than showing how round the signal of the generator is.

You can probably build one yourself out of some fast 74xx logic and a oscillator. It most certainly wont push a 4GHz scope to the limit but it will certainly be faster than your 25MHz scope without a problem.(It can generate about 1ns rise times)

You can get pleasingly close; I built one with a ~350ps risetime. Key bit is 3 sets of (74lvc1g14 + 143ohms) in parallel, with decent decoupling.
https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg1902941/#msg1902941

[Mechatrommer]

One of the forum members has created such a generator for this very purpose of testing scopes ( https://www.eevblog.com/forum/projects/yet-another-fast-edge-pulse-generator/ ). The incredibly fast edges it generates are faster than even a 4GHz oscilloscope can see.

the earlier writing using such SiGe comparator is this http://www.starlino.com/build-a-really-fast-pulse-generator-50ps-rise-time-using-an-ultra-fast-sige-comparator.html and you can get the chip from ebay at a quarter the price. fwiw..

[MrW0lf]

You can get pretty good idea by just probing Arduino in this manner:



Or directly IC pins.

Here's reference using Arduino on my old 2205:



+-10V range, _A=average(A-2.5), x1.5 vertical zoom.

0.35 / 10ns = 35MHz

Result with x10 probe is better than tests here with Leo's pulser & 50Ω passthru directly on BNC. But this is different range. So it is either range or capacitive loading playing tricks. Also +5V signal takes only small portion of full ADC range (since there is no DC offset on scope and +-5V will just overrange). Have no time to look into it further now.

[med6753]

Here's a 1MHz square wave on a B&K 2120 scope with a 20MHz bandwidth. Sweep speed 2us with X10 engaged.

The signal source is a Heathkit IG-4244 Oscilloscope Calibrator with a rise time of approx 1ns and fed to the scope with a special 50 ohm terminated coax to minimize ringing and reflections.

Your picoscope should be able to produce a similar clean waveform.

 

[HKJ]

Remember to use the probes in x10 mode, the bandwidth of the probes are fairly limited in x1 mode.

[David Hess]

There are two ways (1) but they both require a calibrated signal source.

A reference level pulse generator can be used to measure the transition time of the oscilloscope which can yield the bandwidth.  Up to 200 MHz it is possible to construct a suitable source from some fast discrete logic.  The 75 ohm sync outputs from a VGA port work well for this also.

A leveled signal generator can be used to directly find the -3dB bandwidth.

(1) Sampling oscilloscopes, real ones, have other ways to measure their input bandwidth without a calibrated signal source which is just as well because calibrated signal sources at 1 GHz and higher are problematical.  At lower bandwidths, low being up to at least 10 GHz,  any signal generator can be used to find the first null of their non-linear sin(x)/x frequency response from which their 3dB bandwidth can be calculated.  At even higher frequencies, three sampling inputs can be connected to each other to measure their own kickout from which the 3dB bandwidth can be calculated.

[tggzzz]

Up to 200 MHz it is possible to construct a suitable source from some fast discrete logic.

With care you can push it higher than 200MHz, e.g. here's a ~250ps risetime. https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg1902941/#msg1902941

Not calibrated, but surprisingly good.

[macboy]

There are two ways (1) but they both require a calibrated signal source.

A reference level pulse generator can be used to measure the transition time of the oscilloscope which can yield the bandwidth.  Up to 200 MHz it is possible to construct a suitable source from some fast discrete logic.  The 75 ohm sync outputs from a VGA port work well for this also.

A leveled signal generator can be used to directly find the -3dB bandwidth.

(1) Sampling oscilloscopes, real ones, have other ways to measure their input bandwidth without a calibrated signal source which is just as well because calibrated signal sources at 1 GHz and higher are problematical.  At lower bandwidths, low being up to at least 10 GHz,  any signal generator can be used to find the first null of their non-linear sin(x)/x frequency response from which their 3dB bandwidth can be calculated.  At even higher frequencies, three sampling inputs can be connected to each other to measure their own kickout from which the 3dB bandwidth can be calculated.

I'll add another way to measure not only the bandwidth, but the actual frequency response curve.  This takes advantage of the fact that an impulse signal has equal energy at all frequencies. We can't practically generate an impulse, nor could we meaure it if we could (would you send a 1 gigavolt, 1 ns wide pulse into your scope? no thanks). We can however generate and measure very fast step (square wave) waveforms, and an impulse is the derivative of a step. We can show mathematically that the "derivative dv/dt of the response of a system to a step" is the same as the "response of a system to the derivative dv/dt of a step". Since the impulse contains all frequencies at equal amplitude, we can determine the frequency response of the system by looking at the frequency components in the derivative of its response to a step waveform. In other words, apply a very fast rise time step (at least 5x faster than the scope rise time), take the derivative of that using the scope's math functions, then do an FFT on that.  The result is something like this:



Refer to this post also: https://www.eevblog.com/forum/projects/yet-another-fast-edge-pulse-generator/msg1323191/#msg1323191
The latest version of the pulse generator used in this test has a rise time ~30 ps, and was designed and built and is being sold by a forum member (not me!) at his website for a very reasonable price. Refer the thread above.

In my case, I see some ringing and overshoot on the step response. Looking at the frequency response, the scope front end clearly has a very sharp roll off above 2 GHz (I guess they wanted to avoid aliasing when using all four channels, giving 4 GS/s). The sharp filter completely explains the step response anomalies.

[David Hess]

I'll add another way to measure not only the bandwidth, but the actual frequency response curve.

That works also with a suitable calibrated source but very few oscilloscopes include the needed differentiation and FFT functions.  I do not know of any DSOs with bandwidths below 500MHz which do.

If the FFT returns phase results, then aligning the step response with the FFT will also allow the phase response of the oscilloscope to be measured revealing even more about the frequency response.  This can also be used for real time VNA measurements albeit with limited dynamic range.

[David Hess]

Up to 200 MHz it is possible to construct a suitable source from some fast discrete logic.

With care you can push it higher than 200MHz, e.g. here's a ~250ps risetime. https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg1902941/#msg1902941

Not calibrated, but surprisingly good.

Calibration is the problem.

That is actually a pretty horrid response with 5% aberrations; it is more than twice as large as the maximum aberrations I would expect from an oscilloscope but probably acceptable for measuring bandwdith.  Adding diode switching or a transistor cascode between the logic output and 50 ohm termination would improve it considerably without sacrificing speed.

[tggzzz]

Up to 200 MHz it is possible to construct a suitable source from some fast discrete logic.

With care you can push it higher than 200MHz, e.g. here's a ~250ps risetime. https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg1902941/#msg1902941

Not calibrated, but surprisingly good.

Calibration is the problem.

That is actually a pretty horrid response with 5% aberrations; it is more than twice as large as the maximum aberrations I would expect from an oscilloscope but probably acceptable for measuring bandwdith.  Adding diode switching or a transistor cascode between the logic output and 50 ohm termination would improve it considerably without sacrificing speed.

Calibration is, as usual, "an issue"!

It was only intended for <400MHz scopes, and it is sufficient for that. I would hope and expect that a differential output, e.g. any of the ECL variants, would have less "bounce" than a single-ended CMOS outputs.

At those frequencies it is quite likely that the parasitics associated with discrete components would do more harm than good. Even if a suitable scope was available to me, probing would be non-trivial.

[Mr Nutts]

I'll add another way to measure not only the bandwidth, but the actual frequency response curve.

That works also with a suitable calibrated source but very few oscilloscopes include the needed differentiation and FFT functions.  I do not know of any DSOs with bandwidths below 500MHz which do.

My 350MHz Lecroy LT264 can do the same, I think all of these Waveruner scopes can do that (maybe need a software option) 

I wouldn't be surprised if the older scopes (the ones with monochrome screen) could do it as well

[David Hess]

Calibration is the problem.

That is actually a pretty horrid response with 5% aberrations; it is more than twice as large as the maximum aberrations I would expect from an oscilloscope but probably acceptable for measuring bandwdith.  Adding diode switching or a transistor cascode between the logic output and 50 ohm termination would improve it considerably without sacrificing speed.

Calibration is, as usual, "an issue"!

My solution to the calibration problem was to get an old sampling oscilloscope.  So the sampling oscilloscope calibrates the reference level pulse generator which calibrates the non-sampling oscilloscope.

It was only intended for <400MHz scopes, and it is sufficient for that. I would hope and expect that a differential output, e.g. any of the ECL variants, would have less "bounce" than a single-ended CMOS outputs.

At those frequencies it is quite likely that the parasitics associated with discrete components would do more harm than good. Even if a suitable scope was available to me, probing would be non-trivial.

The idea is to "disconnect" the output from the logic's voltage output using a diode or common base transistor and parallel terminate the output leaving only the capacitance across the diode or transistor switch.  Now the only parts which matter are the RF termination and a single discrete diode or transistor.  The transistor version turns the low impedance logic output into a high impedance current output like CML (current mode logic) would use.  There are various tricky way to compensate the switch capacitance if necessary.

The parasitics from the gate version include all of those series resistors and even worse, the ground and power lead inductances.

An alternative circuit I have considered is to use a 74LVC125 as an open drain output to drive either a parallel termination or a diode or transistor switch driving the parallel termination.  I suspect it would deliver better pulse fidelity than the totem-pole output of standard logic.