How to build your own high frequency oscilloscope

[rppearso]

Does anyone know if there is a textbook on the design and construction of high frequency oscilloscopes (up to 10 GHz)? I am guessing they would be some form of ADC circuit with perhaps a mixer and could use a lap top as the "DSP" and display.

How would such an O-scope be calibrated/tuned?

Thank you

 

[f95toli]

Define "oscilloscope"
Are you talking about a real-time oscilloscope or a sampling scope?

Either way; it is not something you can build from scratch at home.
Moreover, the components you would need are seriously expensive so it is not like you would be able to save much money.

 

[Rive]

Up to 10 GHz

Well, in DIY class it would be already a big feat if you can make it up to a few hundred MHz.

 

[anorlunda]

Here is a recent thread with lots of discussion on that topic.
https://www.physicsforums.com/threa...d-cheap-usb-oscilloscope.936167/#post-5915650

 

[f95toli]

Here is a recent thread with lots of discussion on that topic.
https://www.physicsforums.com/threa...d-cheap-usb-oscilloscope.936167/#post-5915650

I am not sure that thread is very relevant. The technology behind microscopes that can operate at a few GHz is very different from what you would find in a normal oscilloscope. Making an oscilloscope/digitizer is "easy" if you can base it around an ADC that has a wide enough analog BW for whatever max frequency you are after. The problem here is that there is no such thing as a 10 GHz (well, 20 GSPS) ADC chip meaning you need to use various tricks to get to that BW (the fastest ADC you can buy is something like 6 GSPS). This is why fast oscilloscopes were traditionally sampling scopes that "build up" a (repetitive) waveform over time using a slow ADC.
Really fast oscilloscopes are of course available (we have a couple in my lab, both sampling and real-time) but they are seriously expensive even for microwave equipment ($70 000 or so) and very big and bulky. Also, the price of the computer HW is insignificant compared to the MW part of the instrument so there is no reason why you would try to save money by using a laptop to control everything. .

This is NOT something you can DIY. A few hundred MHz might be possible if you are able to (and have the tools) to solder SMD and are willing to pay for a company to make a multi-layer PCB.

 

[berkeman]

(up to 10 GHz)

As others are saying, that's too high for a DIY project (heck, even for me!). What is motivating you to want to build a 'scope up to that frequency range? Maybe you should think "spectrum analyzer" above a few 100MHz instead?

For reference, here is one of the latest 'scopes from LeCroy -- 12 bits up to 8GHz. Wow!

 

[diogenesNY]

Notwithstanding the difficulty of 'rolling your own', perhaps the following links might be fun or of interest:

https://archive.org/details/UnderstandingUsingTheOscilloscope

https://archive.org/details/HeathkitO12UOscilloscopemanual

and more generally:

https://archive.org/search.php?query=oscilloscope

diogenesNY

 

[rppearso]

Yea the 70k price tag was the motivation for wanting to design my own. It’s obviously been done so it’s not impossible and with the advent of tormach or Volteta cnc it’s possible to litterally build your own chips -

The voltera looks slick but is pretty steep for only circuit boards so I am wondering if the tormach could do the same thing with the right bit.

The rf circuits I want to build are in those frequencies so dealing with MHz would be non productive.

I am not sure that thread is very relevant. The technology behind microscopes that can operate at a few GHz is very different from what you would find in a normal oscilloscope. Making an oscilloscope/digitizer is "easy" if you can base it around an ADC that has a wide enough analog BW for whatever max frequency you are after. The problem here is that there is no such thing as a 10 GHz (well, 20 GSPS) ADC chip meaning you need to use various tricks to get to that BW (the fastest ADC you can buy is something like 6 GSPS). This is why fast oscilloscopes were traditionally sampling scopes that "build up" a (repetitive) waveform over time using a slow ADC.
Really fast oscilloscopes are of course available (we have a couple in my lab, both sampling and real-time) but they are seriously expensive even for microwave equipment ($70 000 or so) and very big and bulky. Also, the price of the computer HW is insignificant compared to the MW part of the instrument so there is no reason why you would try to save money by using a laptop to control everything. .

This is NOT something you can DIY. A few hundred MHz might be possible if you are able to (and have the tools) to solder SMD and are willing to pay for a company to make a multi-layer PCB.

 

[Stephen Tashi]

I am guessing they would be some form of ADC circuit with perhaps a mixer

Are there actually scopes that can capture digital data at a frequency of 10 GHZ? Can they store 10 seconds of the data? What digital memory technology do they use?

Don't most signal processing devices at such high frequencies (such as radar receivers) rely on heterodyning the high frequency signal with a lower frequency signal?

 

[rppearso]

Are there actually scopes that can capture digital data at a frequency of 10 GHZ? Can they store 10 seconds of the data? What digital memory technology do they use?

Don't most signal processing devices at such high frequencies (such as radar receivers) rely on heterodyning the high frequency signal with a lower frequency signal?

I screen captured this and am going to look up how radar recievers are designed.

 

[Borek]

Yea the 70k price tag was the motivation for wanting to design my own. It’s obviously been done so it’s not impossible

That's interesting, do you have any links?

with the advent of tormach or Volteta cnc it’s possible to litterally build your own chips

Chips? The video was was just about making a PCB. It is a bit like saying "I will build my own car, I already have a screwdriver to screw on the lights" ;)

 

[rppearso]

That number is what I have seen on youtube videos, 70k was not explicit but they were all some crazy number. Anything more than 10k or so for a piece of equipment is all the same to me. Whether its 70k or 200k, I am not going to be buying it lol.

I think there would be solder mounted chips that you would have to buy, but the ability to make a lot of your own PCB or even more complex PCB's could defray the costs quite a bit is the hope.

I am kind of looking at the "buy a burned out H1 hummer" and rebuild it. So buy the chips that you absolutly need that can't practically be made (micro chips that require complex micro electronics fabrication) and make the rest. Analog devices sells a WIDE array of ADC and DAC chips, its just a matter of cleverally using mixers, variable matching, etc.

I would only want to use the lap top as a display, I don't see why there is a need for an oscope display when iphones and lap tops have amazing display capabilities. My borescope uses my iphone and blue tooths in and it works just as good as a $1000 borescope. Of course I already paid the $1000 for the processor and the screen in my iphone, I just needed, esentially a camera lense on a slinky.

I don't see why these same principals can't be applied with an oscope.

I didnt think the oscopes would be cheap but i was not expecting them to cost more than an intermediate level (tormach) cnc machine either

That's interesting, do you have any links?
Chips? The video was was just about making a PCB. It is a bit like saying "I will build my own car, I already have a screwdriver to screw on the lights" ;)

 

[Rive]

Can they store 10 seconds of the data?

They don't even try. Some thousand samples around a trigger usually enough. The bandwidth requirement is not OK for external memory chips, but given the amount of data it is more or less OK for small and fast internal buffers.

 

[Baluncore]

The highest speed oscilloscopes are usually based on sampling a signal trapped in a circular storage transmission line. They use expensive non-linear transmission lines, NLTLs, fabricated on ICs from GaAs microwave diodes and wire links to sharpen the sampling pulses.

10GHz oscilloscopes are not easy to use because they are difficult to connect to the system being observed without frequency dependent phase shifts that distort the signal. A spectrum analyser can be more help above 1GHz.

Fast sampling oscilloscopes are needed for time domain reflectometry. The use of reflected chirp radar techniques makes it possible to convert to baseband and use an FFT to get a better picture than can be done with a sampling scope.

The modulation is usually more important than the carrier. If the carrier is distorted it will show up as EHF harmonics, well above the fundamental and modulation. The modulation can be converted to base band and studied using digital signal processing.

You might consider building an oscoilloscope from an RF receiver front end such as a digital TV receiver or a programable mobile phone processor. Consider how you might apply the LM15851 “Ultra-Wideband RF Sampling Subsystem” made by TI. There are a few dozen chips like that out there just waiting to be misapplied.

 

[berkeman]

The highest speed oscilloscopes are usually based on sampling a signal trapped in a circular storage transmission line.

I actually worked on one of the first ones of those at Tektronix at a summer job during my undergrad EE studies. It was on the old 7000 series of oscilloscpes. Pretty neat technology.

Fast sampling oscilloscopes are needed for time domain reflectometry.

Yup! That's one of the things I've used them for in my EE work, especially for EMI debugging of PCBAs that seemed to have standing wave problems that were generating radiated EMI compliance problems. Good memories! (because we were able to fix the problems)...

 

[rppearso]

The highest speed oscilloscopes are usually based on sampling a signal trapped in a circular storage transmission line. They use expensive non-linear transmission lines, NLTLs, fabricated on ICs from GaAs microwave diodes and wire links to sharpen the sampling pulses.

10GHz oscilloscopes are not easy to use because they are difficult to connect to the system being observed without frequency dependent phase shifts that distort the signal. A spectrum analyser can be more help above 1GHz.

Fast sampling oscilloscopes are needed for time domain reflectometry. The use of reflected chirp radar techniques makes it possible to convert to baseband and use an FFT to get a better picture than can be done with a sampling scope.

The modulation is usually more important than the carrier. If the carrier is distorted it will show up as EHF harmonics, well above the fundamental and modulation. The modulation can be converted to base band and studied using digital signal processing.

You might consider building an oscoilloscope from an RF receiver front end such as a digital TV receiver or a programable mobile phone processor. Consider how you might apply the LM15851 “Ultra-Wideband RF Sampling Subsystem” made by TI. There are a few dozen chips like that out there just waiting to be misapplied.

Very cool, I am going to add TI to my list, I did not even think of ordering chips from TI. The circuit I am building will be concerned with phase shifts of high frequency signals, so I will need to be able to see how the circuit responds to an incoming signal and then how the DSP creates a shifted signal using a VCO and retransmits it.

I was looking at doing it all analog but then the reciever antenna and transmitter antenna would be fighting with each other.

 

[Baluncore]

I was looking at doing it all analog but then the reciever antenna and transmitter antenna would be fighting with each other.

Analogue is good. The distortion or phase shift of a carrier is best analysed by the amplitude and phase of the carrier harmonics. Modulation is best analysed by synchronous down converting the input and output with the same oscillator, in the same way that a network analyser extracts the transfer function of an RF block. If the up and down conversion local oscillators are phase locked, the phase shift of the modulation will also be locked. There are many low-cost quadrature transceiver chips designed with IQ image-rejection mixers on the Tx and Rx channels. There are modulators and demodulators available for every mobile phone or data link band in common use. The name Maxim comes first to mind;
https://www.maximintegrated.com/en/pl_list.cfm?filter=wr

All singing, all dancing, frequency agile chips like the AD9361 from AD are expensive, but they can make life easier if you need a dedicated network analyser.
http://www.analog.com/en/products/r...eceivers/wideband-transceivers-ic/ad9361.html
The chips cost about $100 each in small quantities. There are often AD designed, Chinese mass-produced evaluation boards available at good prices. Search eBay and AliExpress.com for commonly used evaluation boards and chips. For example, complete AD9361 evaluation boards can now cost anywhere between $500 and $3500.

I do not believe you need a really fast oscilloscope, you actually need a heterodyne based network analyser. We need a better idea of the carrier frequency range and the modulation characteristics of the signals you are employing, before we can identify a quickest to implement, low cost solution.

 

[davenn]

I think there would be solder mounted chips that you would have to buy, but the ability to make a lot of your own PCB or even more complex PCB's could defray the costs quite a bit is the hope.

I am kind of looking at the "buy a burned out H1 hummer" and rebuild it. So buy the chips that you absolutly need that can't practically be made (micro chips that require complex micro electronics fabrication) and make the rest. Analog devices sells a WIDE array of ADC and DAC chips, its just a matter of cleverally using mixers, variable matching, etc.

Unless you have good experience in designing and building RF circuits over 1GHz, I suspect you don't really realize what is involved
in building multi-GHz RF gear. You cannot "just throw a few chips onto a circuit board and hope it works" when it comes to those frequencies.
board layout becomes very critical, track lengths over a 10mm start becoming RF radiators.
Every stage must be in screened boxes to avoid RF from one stage getting into another

Take it from some one who has built gear going up to 24 GHz

Dave

 

[rppearso]

Unless you have good experience in designing and building RF circuits over 1GHz, I suspect you don't really realize what is involved
in building multi-GHz RF gear. You cannot "just throw a few chips onto a circuit board and hope it works" when it comes to those frequencies.
board layout becomes very critical, track lengths over a 10mm start becoming RF radiators.
Every stage must be in screened boxes to avoid RF from one stage getting into another

Take it from some one who has built gear going up to 24 GHz

Dave

Hello Dave,

That is a good point, perhaps I will table this endeavor to revisit another time. Plus with the loss of my oil and gas job I pretty much have to focus full time on just investing what I have saved so it can grow and putting away every additional cent I have at my pedestrian job until I can generate 2k a month in purely passive income so I can essentially buy my own time back. This set back may cost me a decade or more if I can't duplicate or come close to duplicating my previous income.