Is this standard tolerance for square wave generator output?

[pwnell]

I have a Rigol MSO5104 scope and tested its built in AWG.  I have read that it is only a basic waveform generator and not very good, so I am wondering if the square wave output I measured is within spec.  I get that a square wave has lots of extra harmonics and that these get attenuated more the higher the frequency of the harmonic, hence distorting the wave.  However the max output of the AWG is 15MHz for a square wave, the scope is 100MHz and should deal with a 100/5 = 20 MHz square wave just fine.  And as I understand it, due to the oversampling (8GSa/s) it should even be fine for higher frequency square wve signals...

So I tested 1kHz - looks good to me.

Then 10kHz - still good.

At 1Mhz I can see some overshoot at the rise of the wave, but it still looks decent.

At 10MHz it is not really a square wave any longer.

At 15MHz it looks more like a distorted sine wave.

That said - is this to be expected?

Oh - I forgot to mention, I am using the standard 10x passive probe connected to a 50ohm BNC cable from the function generator to try and eliminate wave reflections.

[Doctorandus_P]

Your squarewave looks quite normal for the output of a function generator. These things have filtering inside and also opamps which are optimized for analog use and have a limited slewrate.

[rstofer]

It looks about right.  Getting to the 5th harmonic doesn't produce a very good square wave.  Maybe the 9th?  In which case, a 100 MHz scope will have a decent display of a 10 MHz square wave.

https://www.khanacademy.org/science/electrical-engineering/ee-signals/ee-fourier-series/v/ee-visualize-fourier-series-square-wave

There are many animations on the Internet

http://www.sigcon.com/Pubs/edn/TaketheFifth.htm

There's a bunch of math behind this.

[pwnell]

Hmm but my 10MHz square wave is not very good, so does that mean it is the AWG or the scope input that is causing this? Unfortunately I do not have another scope to test with.

[rstofer]

If it shows rounding, it is very probably the bandwidth limitation of the scope.  The way to test the scope is with a very fast rise time square wave generator.  BandWidth (GHz) = 0.35 / RiseTime (ns) for whatever the scope shows for 10%-90% risetime.  A 100 MHz scope should have 3.5 ns risetime.  There are some fast rise time circuits around this forum.  Probably search for 'risetime'

[MarkF]

Here is a fast (2ns) rise time pulse generator:
I only used four inverters with 200Ω resistors to get exactly a 50Ω output.



Here is what I built with a switchable input of the oscillator or an external source (via a jumper):

 

[srb1954]

Oh - I forgot to mention, I am using the standard 10x passive probe connected to a 50ohm BNC cable from the function generator to try and eliminate wave reflections.

Did you terminate the end of the BNC cable with a 50 load? You will need this to get proper pulse fidelity as a 10x passive probe doesn't present the correct termination load to the cable and you may get ringing on the waveform.

A better option for maximum pulse fidelity is to run a 50 cable direct to the scope's BNC input, through a feed-through 50 BNC to BNC termination.

[pwnell]

Like this? Bnc Adapter, P57 1 Stks 50ohm Feed Through Terminator Bnc Seat 50ky Device Q9 Adapter https://www.amazon.ca/dp/B07SD3TC51/ref=cm_sw_r_cp_api_fabc_PaT1FbZ9MH67M

[srb1954]

Yes

[pwnell]

Ordered one, will test. Might also try and build that circuit if I can find the components. Thanks for the help.

[S. Petrukhin]

Ordered one, will test. Might also try and build that circuit if I can find the components. Thanks for the help.

Сan't the scope itself connect a 50\$\Omega\$ resistor? In the generator output settings, there is an output type.

[pwnell]

Yes - that is the output impedance of the AWG that is 50 ohm.  But the input terminal of the scope is 1M (HighZ) and cannot be adjusted.

[S. Petrukhin]

Yes - that is the output impedance of the AWG that is 50 ohm.  But the input terminal of the scope is 1M (HighZ) and cannot be adjusted.

It seems to me that one resistor on any side is enough. 

[TimFox]

With only a 50 ohm resistor at the source, your cable is “series terminated”, and the voltage waveform at the high impedance load should be correct.  It will have twice the amplitude of the “double terminated” case with a 50 ohm load added.  However, the double termination will be less sensitive to deviation of either resistance or cable impedance from nominal value.

[alsetalokin4017]

Something is wrong somewhere in the OP's setup. Or, if not exactly wrong, not exactly right either.  I'd like to see the test redone with a 50 ohm thru terminator at the scope input.

Here's my (unlocked) DS1054Z looking at a fast-edge oscillator circuit similar to MarkF's above, except using 74HC14n (didn't have any AC versions on hand).

My FEO isn't even built on a PC board with BNC connectors, it's just soldered together on vectorboard and using JST type 2-pin connection to a 50 ohm generic cable to a 50 ohm feedthru terminator at the scope end.

[pwnell]

I am going to pick up a Schmitt trigger and try this...

[srb1954]

With only a 50 ohm resistor at the source, your cable is “series terminated”, and the voltage waveform at the high impedance load should be correct.  It will have twice the amplitude of the “double terminated” case with a 50 ohm load added.  However, the double termination will be less sensitive to deviation of either resistance or cable impedance from nominal value.

Scope amplifier and probe bandwidths are generally measured/specified with a double terminated signal source to maintain best signal fidelity and to reduce the effect of the input capacitance of the scope input or probe degrading the test signal risetime.

[MarkF]

Looking at the DS1000Z's AWG square wave output specs, the bandwidth is 0.1 Hz to 15 MHz.
I'm not surprised that the edges roll off near the bandwidth limit.  Just what I would expect to see.
I would be interested to see the square wave output from another DS1000Z's built-in Arbitrary Function Generator.

I believe most Function Generators specify their bandwidth as the sine wave.
However, their square wave bandwidth falls far short of that of the sine wave bandwidth.

As far as doing the measurement:
  I would NOT use a combination of a 50Ω coax and the 10X probe.
  Just make the 1X connection with the 50Ω coax cable. 
  Then, add the 50Ω in-line termination, at the scope input, when you get one.

[David Hess]

That looks about right for a higher frequency function generator.  The waveform fidelity is limited by the linear output amplifier.  Pulse generators can produce much faster square waves because their output amplifier does not need to be linear.

Linear amplifiers which can drive 10 or even 20 volts peak-to-peak at 10s of MHz are not trivial so it is expected that higher frequency square waves will have considerable distortion.

[alsetalokin4017]

Output of an antique DataPulse DP-101 pulse generator, topped out at just over 10 MHz:

DP101 > 6 feet of RG58/U > 50 ohm thru terminator > scope

[pwnell]

@MarkF so I built your circuit as close I could.  I used a 74AC14 and the same components as you.  Hooked up VCC to +5V from my bench supply, and gnd to -.  I used the 10x probe and looked at the signal.  Here is the circuit and the waveforms.  Thoughts?  I have a 3.75ns rise - is that fast enough?  I looked at the datasheets for an 74AC14 and the only time I could find that might be rise time is tPLH which is 1.5ns to 11ns @ 5V which seems much higher than the 2.5 you got.

[alsetalokin4017]

Not too bad for a breadboard. It's actually kind of amazing that you got that good a result from your breadboard and standard probe. It will be even better when you connect directly with coax and a 50 ohm thru terminator at the scope, instead of using your regular 10x probe.

[pwnell]

Ps I mistyped my original post. I have a Rigol MSO5104 scope (100MHz)

[pwnell]

So I just connected the output of the circuit directly to the coax t splitter attached to my oscilloscope with one end of the t attached to a 50 ohm coax terminator. It made no difference except to attenuate the signal. Rise time stays fixed at 3.6ns

[MarkF]

Here is output for comparison:

I seem to remember that the actual measured bandwidth of the 100MHz DS1000Z is around 160MHz.
So, 3.2ns is about all the scope will be able to measure.