rise time on a hacked Rigol scope (Re: Oscilloscope training class)

[Stiege]

I've got a Rigol 1052E that I think I've hacked to 100Mhz by following the instructions off this site, and my Rigol function generator arrived last week and I was hoping to prove that I'd done the hack properly by measuring the rise time of a square wave of the function generator on the scope...

Unfortunately I must be doing it wrong, I just basically hook up the function generator to the scope, get all the probe settings correct, tried outputting a 1khz square wave and then went to measure the rise time - it's 17ns!

I'm pretty sure I've done the hack correctly but I'm not sure what I'm doing wrong with regard to measuring the rise time, is it something to do with impedance matching or could it just be that the function generator actually has an inherent rise time?

Just want to add, the waveform of the square wave has an under-damped appearance, hope this helps...

Any help appreciated.

Kind regards

Alex

[ejeffrey]

If you are connecting the generator to the scope with a standard BNC cable you need to put a 50 ohm terminator on the scope side using a BNC tee.

Otherwise, if you have a high enough bandwidth 10x probe you can probe the center of the BNC output of the function generator directly.

Your function generator may simply not be fast enough.  If it has a SYNC output that may have a faster rise time than the signal output.

[alm]

ejeffrey: for best results, the output of the generator should be terminated, whether you connect it to the input directly or use a probe. The only thing the probe accomplishes is reduce the size of the stub (and change the frequency response of the system).

Ideal would be 50 ohm termination in the scope. Since low-end scopes like the Rigol lack this, the next best thing is a feed-through terminator, this minimizes the size of the unterminated stub. But in this case I believe a BNC tee connected directly to the input with a terminator attached should be just as good.

You don't mention the source you use, but my guess is that it's indeed the limited edge rate of the generator, what's the specified rise time?

[Stiege]

Hey,

Thanks for the help, looking at the specs for the Rigol DG1022 I see that they state the rise time for the function generator is less than 20ns, it also states it has the 5% overshoot that I mentioned.

I didn't realise that there could be such a huge variation in rise time for function generators; not that it will matter for the applications I will use it for, but if I did want to create a waveform that had a rise time of less than 3.5ns so I can see this limit on my scope, what is the best way to do this? Do I have to go out and buy a really expensive function generator, I did a bit of browsing through specs for some various function generators and couldn't find many that went below 10ns for the rise times for their square wave function, so how do you guys do it when you want to create a pulse with a rise time which is below this 3.5ns limit?

Thanks again for your help.

Alex

[ejeffrey]

Function generators always have internal series termination at the source.  Between that and the short cable length, poking the probe directly into output will work perfectly fine at this frequency.

If you want to get a fast edge time, any old 74HC / 74AC logic gate will do.  Take a hex NOT gate and wire it up on a bit of perf board.  Add bypass capacitors, and tie all the unused inputs to ground.  Send a 0..5V square wave from your function generator (terminated with 50ohm to ground at the gate) to one of the inputs and look at the output with a 10x probe.  Most logic gates can't drive a 50 ohm line, so you need to use the probe.  You could also connect several gates in parallel to drive a 50 ohm load, but as long as you have a sufficiently high bandwidth passive probe that isn't necessary.

Obviously you have to take care of all the normal requirements of high speed logic: bypass capacitor across the supply pins and connect all unused inputs to ground.

[Richard W.]

[...] so how do you guys do it when you want to create a pulse with a rise time which is below this 3.5ns limit?

You need a fast avalanche pulse generator.
http://cds.linear.com/docs/Application%20Note/an47fa.pdf page 93 

[alm]

Function generators always have internal series termination at the source.  Between that and the short cable length, poking the probe directly into output will work perfectly fine at this frequency.

For a typical function generator, yes. For a generator that is actually fast enough to test a 100MHz scope, I would use termination. With a generator with a 1ns rise time, I can definitely see a difference between unterminated, terminated at the source and terminated at the scope.

[ejeffrey]

Just from the length of the connector inside the function generator?  I have never seen a problem from that, but then I haven't looked very carefully.  1 nanosecond is also pretty fast for most passive probes regardless of the signal source.

I didn't realise that there could be such a huge variation in rise time for function generators; not that it will matter for the applications I will use it for, but if I did want to create a waveform that had a rise time of less than 3.5ns so I can see this limit on my scope, what is the best way to do this?

I misread that.  High speed CMOS logic gates will switch considerably faster than your function generator, but I think the 74AC series is still on the order of a few nanoseconds. 

To go faster than 3.5 nanoseconds you have to move away from standard function generators.  They simply aren't designed for that.  There are some dedicated pulse generator instruments that can do it.  Avalanche pulsers are one option, as are step recovery diodes.  There are extremely fast logic families like ECL and newer standards like LVDS  that can have really high toggle frequencies and correspondingly fast edges.  ECL can have edge times <200 ps.

Needless to say, the faster you go, the harder everything gets.  The source may be blazing fast, but the connection to your scope can be as big of a performance limit as the scope itself.  At several GHz everything is a transmission line, and even a slight reflection can totally smear out your rise time.

[Stiege]

Cool, well I learned a lot today.

[saturation]

Depending on what type of FG you have and what it's made off, the rise time of the output is no faster than the particular technology's intrinsic delay times.



With most device being made of a variant of CMOS, you can't expect any less than the limits of what's pictured above.

If you have a 100 MHz scope, and you can see the theoretically limited rise time of near ~ 0.35/rise time in s, you are unlikely to get any faster.  You need to do the measurement cognizant that the cabling, connectors, and impedance matching will affect the result and won't give its best so you have to isolate each of these as a factor in performance.

Try taking a signal out of the sync port.  On the Instek 1003 I get as low as 4.5 ns.  On the Hantek 3x25, I think I got between 10-15 ns.  Both of these are only $150 each new, so the Rigol should be able to do similar or better if it was built well.

Hey,

Thanks for the help, looking at the specs for the Rigol DG1022 I see that they state the rise time for the function generator is less than 20ns, it also states it has the 5% overshoot that I mentioned.

I didn't realise that there could be such a huge variation in rise time for function generators; not that it will matter for the applications I will use it for, but if I did want to create a waveform that had a rise time of less than 3.5ns so I can see this limit on my scope, what is the best way to do this? Do I have to go out and buy a really expensive function generator, I did a bit of browsing through specs for some various function generators and couldn't find many that went below 10ns for the rise times for their square wave function, so how do you guys do it when you want to create a pulse with a rise time which is below this 3.5ns limit?

Thanks again for your help.

Alex