extending oscilliscope bandwidth

[kripacharya]

My Oscilloscope has an input b/w of 100Mhz. Now I want to sometimes work with circuits in the range of 100-200Mhz. But shelling out 4x the money when i already have a decent 100Mhz, and that too for the occassional project, is hard to do.

My circuits are bandlimited to below 200Mhz, and its usually repetitive signals.

So my natural thought was -- why not downconvert everything by use of a GOOD downconverter/mixer , a good LO of (say) 100MHz/ 80MHz/ whatever, do a bit of filtering etc -- and view it on my 100Mhz'er ?

As long as I know the broad nature of my measured signals, surely this would be a much cheaper yet equally viable method ?
Or not ?

 

[barry]

Just because the scope has a bandwidth of 100MHz doesn't mean it can't display 200MHz signals-they'll just be attenuated. More important would be the SAMPLING frequency(assuming it's a digital scope) As far as the downconverter approach, there's nothing wrong with that, in theory. In PRACTICE, however, you need to consider the effects of the filter, etc. on your original signal

 

[chuckey]

It would work providing the bandwidth of the original signal can be encompassed by the CROs bandwidth. Also there will be some "dead" bands where the low pass filter is starting to have an effect also the sidebands will be reversed in frequency (no problem?).
Frank

 

[kripacharya]

Just because the scope has a bandwidth of 100MHz doesn't mean it can't display 200MHz signals-they'll just be attenuated. More important would be the SAMPLING frequency(assuming it's a digital scope) As far as the downconverter approach, there's nothing wrong with that, in theory. In PRACTICE, however, you need to consider the effects of the filter, etc. on your original signal

yes i agree. As of now i know i can see a (for example) a 120Mhz signal. My max sampling is 40Msa/s. But when i use lower sampling rates - say 25Msa/s or 37Msa/s (non submultiples) i can see a signature at 5Mhz (for the 25Msa/s rate for e.g.) and at 9Mhz (for the 37Msa/s rate). But these are wayy down almost into the noise floor.

Regarding effects of filter.. what effects do you mean ? Lets say I use a 100Mhz LO with some nice mixer from MiniCircuits followed by NO filter to measure this same 120Mhz.
Wouldn't i get a measureable (after doing some gain/loss adjustments) response at 20Mhz ? And the others above 100MHz are attenuated anyhow by the 'scope b/w limits?

- - - Updated - - -

It would work providing the bandwidth of the original signal can be encompassed by the CROs bandwidth. Also there will be some "dead" bands where the low pass filter is starting to have an effect also the sidebands will be reversed in frequency (no problem?).
Frank

What dead bands ? As long as i know the limits of this setup, i can always make allowances for that.
Reversed frequencies no probs... been doing that already ! Note : my 'scope is 100MHz b/w with 40Msa/s. So yes, i know how that works already.

 

[barry]

You are sampling a 100MHz signal with a 40Ms/sec sampler? THAT'S going to give you some problems i.e., aliasing.

If you are looking at continuous signal through a filter then it should probably be ok, but filters will affect pulse response, etc. if you were looking at single shot event, for example.

 

[FvM]

I presume you are using an oscilloscope because you want to see signals in time domain and e.g. distinguish between different waveforms? Otherwise you could use a level meter or a specrum analyser, depending on the purpose of your measurement.

So the question is if you expect to somehow reconstruct the original signal after downmixing it? If you want to reproduce the phase relation of signal components, you typically need a quadrature mixer (not in every case, depends also on the nature of the input signal).

The classical way to record periodical signals at a lower sampling frequency is time equivalent sampling. It's apparently also done to some extent by your oscilloscope to extend the time resolution when displaying periodical signals, otherwise the combination 100 MHz analog bandwidth with 40 MSPS rate won't make sense. A more extreme ratio is implemented by dedicated samling oscilloscopes that achieve several 10 GHz analog bandwidth with MHz sampling rates.

But the input stage of a sampling oscilloscope has an extremly short aperture time, so the sequence of sampled data points can be easily displayed as time domain signal (with some special tricks). I don't think that it can be done similarly easy with your downmixed signal.

 

[kripacharya]

You are sampling a 100MHz signal with a 40Ms/sec sampler? THAT'S going to give you some problems i.e., aliasing.

If you are looking at continuous signal through a filter then it should probably be ok, but filters will affect pulse response, etc. if you were looking at single shot event, for example.

LoL !! yes it cerainly gives me problems, and I typically need to cross-check with different sampling rates to work out what the actual freq is versus an aliased one. Painful process, and not entirely error free. Obviously I cannot measure single-shots or non-continous signals above a fairly lowish freq limit.

However even with these constraints, I manage to measure signals - using the Spectrum Analyser part of the DSO - close upto the 100Mhz limits.

- - - Updated - - -

I presume you are using an oscilloscope because you want to see signals in time domain and e.g. distinguish between different waveforms? Otherwise you could use a level meter or a specrum analyser, depending on the purpose of your measurement.

So the question is if you expect to somehow reconstruct the original signal after downmixing it? If you want to reproduce the phase relation of signal components, you typically need a quadrature mixer (not in every case, depends also on the nature of the input signal).

The classical way to record periodical signals at a lower sampling frequency is time equivalent sampling. It's apparently also done to some extent by your oscilloscope to extend the time resolution when displaying periodical signals, otherwise the combination 100 MHz analog bandwidth with 40 MSPS rate won't make sense. A more extreme ratio is implemented by dedicated samling oscilloscopes that achieve several 10 GHz analog bandwidth with MHz sampling rates.

But the input stage of a sampling oscilloscope has an extremly short aperture time, so the sequence of sampled data points can be easily displayed as time domain signal (with some special tricks). I don't think that it can be done similarly easy with your downmixed signal.

My bad.
I should have mentioned that i use the Spectrum Analyser part of the DSO. My current purpose is to simply be able to detect the freq components and have a reasonable confidence in their amplitude's for continous signals only beyond the DSO's analog sections 100Mhz 3db limit. I do not know how the signal is attenuated beyond this limit by the DSO. Hence my idea to downconvert.

btw - I am referring to BitScope 325

- - - Updated - - -

So coming full circle - any suggestions / circuits / components etc for such a downconverter design ?

 

[Borber]

What about this: **broken link removed**

 

[FvM]

A spectrum analyzer can be extended to effectively any frequency range by using harmonic mixers, as long as you don't mind aliasing and an ambiguous spectrum display. Old style SAs don't have selection filters, instead you identify the harmonic order of a line by tuning the centre frequency and watch how the line shifts.

I think that Mini-Circuits mixer, oscillator and filter components are basically O.K. to make a down mixer. You should be aware of possible cross products of oscilloscope sampling frequency and external oscillator some filtering beween mixer and oscilloscope will be most likely required. This is essentially different from a dedicated spectrum analyzer which usually generates the carrier for an external mixer internally and optionally bypasses the first internal mixer in external mixer mode.

 

[kripacharya]

Actually I would love to not have to deal with aliasing etc, but with my current constraints I don't think thats possible.
But why specifically a harmonic mixer ?

 

[FvM]

But why specifically a harmonic mixer ?

I was referring to a classical spectrum analysator design that achieves a wide frequency range at the cost of ambiguousity. If you want a limited frequency range, you'll go for the opposite way, filter possible alias signals before entering the mixer.

 

[kripacharya]

I was referring to a classical spectrum analysator design that achieves a wide frequency range at the cost of ambiguousity. If you want a limited frequency range, you'll go for the opposite way, filter possible alias signals before entering the mixer.

Oh right then. I was planning to put in a 100-200Mhz bpf. That won't eliminate my aliasing probs totally - since it's a 40msps system - but it'll help.

Of course the challenge THEN is figuring out how-to align the filter in the first place !?