Has anyone attempted a high performance DIY oscilloscope?

[Boscoe]

I did some forum searches and a bit of googling but not much came up although it's a noisy search. Having looked at some USB oscilloscopes recently it seems to me that a high performance oscilloscope of aroung 5GSPS could be one of those projects that actually make financial sense - assuming not too much prototyping is needed. I see an ADC will run you around £500 and the rest would probably be another £500 for a USB version. Has anyone tried such a thing? I would love to see some existing designs and perhaps contribute.

[ftg]

A 1000gbp price of admission per prototype round is likely why it has not really happened.
And those that have usually resulted in commercial products.
Like that one equivalent time sampling scope with a long thread here.

[dmendesf]

Thunderscope is the nearest thing I know about:

https://www.crowdsupply.com/eevengers/thunderscope

[Doctorandus_P]

A quick search:
https://html.duckduckgo.com/html?q=open+source+oscilloscope

finds things like:
https://www.crowdsupply.com/scopefun/open-source-instrumentation
https://www.eevblog.com/forum/testgear/a-high-performance-open-source-oscilloscope-development-log-future-ideas/
https://www.hackster.io/news/thunderscope-the-open-source-software-defined-oscilloscope-is-coming-to-crowd-supply-220eb2e458fe

One of the problems here is that it takes quite a lot of effort before a product becomes usable, and before it's usable, you have the chicken and egg problem, that not many people will be interested. And when a few people are investing a year or more of their time to make a usable product, they are likely to want some sort of compensation for it.

An additional problem is economy of scale. The amount of hardware for such a project is considerable. Apart from the analog frontend, you also need some decent processing power and digital bandwidth (USB-3?) Doing that for small production runs is not going to be cheap, and that makes competing with existing products difficult.

One path that could be interesting is to match OpenHantek with Owon Hardware. Owon makes quite decent and affordable USB scopes, but they are useless to me because their software does not run on Linux. (I'm not considering wine or VM).

The first version of bitscope (with a pic16F84) was an open source project, and it won a prize. But quite quickly after that it also went commercial, and buying one was quite expensive, compared to it's very moderate hardware limits.

I am not sure if there is commercial hardware that has a very open and open source friendly binary interface. Maybe Picoscope has a well documented interface that can be used without reverse-engineering, but I never looked into details.

[pcprogrammer]

A high performance oscilloscope is not something one simply puts together by getting an expensive high end ADC. High speed digital logic is rather complex and needs a lot of attention in how the components are placed on the PCB and the routing takes a lot of skill to get it right.

The OP mentions 5GSa/s which even at 8 bits per sample is more then USB3 can handle. (This is 40Gb/s) So you would need a lot of data processing on the board before even sending data to the PC.

With reasonably cheap 12 bit 2GSa/s oscilloscopes on the market it does not make much sense to try and make something yourself. Sure as a learning project and out of pure interest it can be fun, but I would not set the bar so high if it is a first time of doing such a thing.

[nctnico]

A high performance oscilloscope is not something one simply puts together by getting an expensive high end ADC. High speed digital logic is rather complex and needs a lot of attention in how the components are placed on the PCB and the routing takes a lot of skill to get it right.

Disagree. The high speed logic part is the easiest thing do by far compared to the rest. The real challenges are in the analog front-end and getting the software going. The digital part is like 5% (or less) of the work. Look at how Rigol and Siglent struggle to get the software in their DSOs to work well even after being in the business for 10+ years.

[SiliconWizard]

Agreed. While the high speed aspect of PCB design requires some care, the real challenge here would be the analog front-end, including proper protection.
A say >= 100 MHz front-end, with variable gain, very low distortion and well protected is in itself a full (and involved) project.

The software side is indeed a lot more work than many expect, although, to be fair, if the goal is to design some kind of USB oscilloscope, considering the software that comes with most commercial ones, this is doable. Not that it says anything really good about them.

Another tricky thing to implement is a reasonably useful and advanced triggering feature.

The "DIY oscilloscope" idea is one of the popular ideas that keep coming up - the OP will find endless threads about it already.
If anything, this will be a rabbit hole and a very expensive hobby.

[pcprogrammer]

A high performance oscilloscope is not something one simply puts together by getting an expensive high end ADC. High speed digital logic is rather complex and needs a lot of attention in how the components are placed on the PCB and the routing takes a lot of skill to get it right.

Disagree. The high speed logic part is the easiest thing do by far compared to the rest. The real challenges are in the analog front-end and getting the software going. The digital part is like 5% (or less) of the work. Look at how Rigol and Siglent struggle to get the software in their DSOs to work well even after being in the business for 10+ years.

Sure there is a lot to get a complete high performance oscilloscope up and running, but I think you are underestimating the percentage of the digital part a bit. For someone with no experience it won't be easy to get any of the parts done, and you would need to be specialized in multiple fields to get it done. Analog, Digital and software, where on the software end there are also lots of layers you need to know about.

My point was in reference to the original post where it is opted that getting a 500 GBP costing ADC and some additional components also for around 500 GBP would get you a long way. It just does not.

If I recall correctly, one of the things with tom66 his scope, was that the front end still needed to be done and was not easy to get it right.

Another aspect of designing something like this requires knowledge in working with FPGA's or have the ability to design a digital ASIC. Building something like this without one of those will be almost impossible.

My conclusion is that it is not worth doing it as some project to eventually make money of. (Other than landing you a good job that is)

Agreed. While the high speed aspect of PCB design requires some care, the real challenge here would be the analog front-end, including proper protection.
A say >= 100 MHz front-end, with variable gain, very low distortion and well protected is in itself a full (and involved) project.

I agree on that and it takes knowledge, skill and a lot of, or access to good testing equipment to do this. Without the testing equipment you will never know if your design is actually any good.

The software side is indeed a lot more work than many expect, although, to be fair, if the goal is to design some kind of USB oscilloscope, considering the software that comes with most commercial ones, this is doable. Not that it says anything really good about them.

Yes the software takes time too. Getting out an initial version maybe not that long, but there will always be new features to think of and other improvements to make before it is really mature.

Another tricky thing to implement is a reasonably useful and advanced triggering feature.

This is also an important issue. Getting good trigger functionality requires a lot of thought. Like where to implement it. In the analog section, the digital section or a combination of both.

The "DIY oscilloscope" idea is one of the popular ideas that keep coming up - the OP will find endless threads about it already.
If anything, this will be a rabbit hole and a very expensive hobby.

It is, and it can be fun, but best to keep the bar set low with like a 1GSa/s setup or much lower but with a 14 bit ADC to get better resolution.

[BennoG]

you know you can buy a picoscope 2406B for GBP 500 that can do 1Gsps and has 32Msamples of memory on board.
The frontend is only 50Mhz, if you want higher bandwidth frontend the const will go up to 100fold of this  for example the 3Ghz frontend  and 10Gsps will cost GBP 15000.

[2N3055]

Look at how Rigol and Siglent struggle to get the software in their DSOs to work well even after being in the business for 10+ years.

You should stop your generalizations..They are not the same in that regard and you know it...

[nctnico]

A high performance oscilloscope is not something one simply puts together by getting an expensive high end ADC. High speed digital logic is rather complex and needs a lot of attention in how the components are placed on the PCB and the routing takes a lot of skill to get it right.

Disagree. The high speed logic part is the easiest thing do by far compared to the rest. The real challenges are in the analog front-end and getting the software going. The digital part is like 5% (or less) of the work. Look at how Rigol and Siglent struggle to get the software in their DSOs to work well even after being in the business for 10+ years.

Sure there is a lot to get a complete high performance oscilloscope up and running, but I think you are underestimating the percentage of the digital part a bit. For someone with no experience it won't be easy to get any of the parts done, and you would need to be specialized in multiple fields to get it done. Analog, Digital and software, where on the software end there are also lots of layers you need to know about.

My point was in reference to the original post where it is opted that getting a 500 GBP costing ADC and some additional components also for around 500 GBP would get you a long way. It just does not.

If I recall correctly, one of the things with tom66 his scope, was that the front end still needed to be done and was not easy to get it right.

Another aspect of designing something like this requires knowledge in working with FPGA's or have the ability to design a digital ASIC. Building something like this without one of those will be almost impossible.

My conclusion is that it is not worth doing it as some project to eventually make money of. (Other than landing you a good job that is)

For sure such a project is a good one to grind your teeth on and learn a ton of useful stuff. I went on a similar adventure over 20 years ago and wished somebody told me to look at the software side first as hardware is much less problematic in comparison. Nowadays there are interesting software projects like ngoscopeclient which could serve as a backend for a digitizer (= acquisition unit like the Thunderscope mentioned earlier).

[Postal2]

The desire to make some unnecessary crap is usually very strong, I barely overcame this desire to make an oscilloscope, having 7 ready-made ones. I still have unclaimed ADCs for this project.

[vk6zgo]

A high performance oscilloscope is not something one simply puts together by getting an expensive high end ADC. High speed digital logic is rather complex and needs a lot of attention in how the components are placed on the PCB and the routing takes a lot of skill to get it right.

The OP mentions 5GSa/s which even at 8 bits per sample is more then USB3 can handle. (This is 40Gb/s) So you would need a lot of data processing on the board before even sending data to the PC.

With reasonably cheap 12 bit 2GSa/s oscilloscopes on the market it does not make much sense to try and make something yourself. Sure as a learning project and out of pure interest it can be fun, but I would not set the bar so high if it is a first time of doing such a thing.

I'm sure artbyrobot could slap an "eiite" one together in his spare time, bur Treez would probably complain about the power supply!

[David Hess]

It is, and it can be fun, but best to keep the bar set low with like a 1GSa/s setup or much lower but with a 14 bit ADC to get better resolution.

Even in the 1980s, 100 MS/s DSOs used custom ASICs.  Unless the design is much slower, an FPGA is almost a requirement for reasonable sample rates.

[Boscoe]

Hi everyone, thanks for the replies. I know there are projects out there (through google searching) but I couldn't find anything with a sampling rate above 1GSPS which _is_ pointless to do DIY IMHO.

I think the digital design and PCB design would be failry straight forward, I've done more complex stuff in one revision before. The AFE would be challenging for me, however. I think I project for me would include a low voltage AFE to make things easier - I don't measure anything above 50Vpk, really. Perhaps this part can be modular for iterative development and future upgrades.

I think £1000 for a prototype is perfectly doable with parts sourced via JLCPCB.

[tszaboo]

I think the issue is more fundamental than that. You cannot really make a high speed oscilloscope without having a high speed oscilloscope to debug it. And if you already have one, than you are making it for other people, and for that, it needs to be competitive. Which is very hard to do when the far east decided to make scopes with "cost plus" pricing. My ultimate point is that the design would need to do something unique to stand out and gather interest.

[Boscoe]

Good point. I do have multi GHz bandwidith oscilloscopes and spectrum analysers at work that I can use...

[2N3055]

Just to make sure, you are doing this for fun and as a learning experience?

[David Hess]

Hi everyone, thanks for the replies. I know there are projects out there (through google searching) but I couldn't find anything with a sampling rate above 1GSPS which _is_ pointless to do DIY IMHO.

I do not see anything which makes higher sample rates unduly difficult.  This will depend mostly on the FPGA used.  Many have fast interfaces for ADCs, but of course it will cost.

I think a heroic discrete logic design could get to 100 MS/s or even 200 MS/s, but I do not recommend it unless your interest is pushing discrete logic performance.

I think the issue is more fundamental than that. You cannot really make a high speed oscilloscope without having a high speed oscilloscope to debug it.

I use a sampling oscilloscope to calibrate other oscilloscopes, and it is possible to design and build a sampling oscilloscope which is self calibrating.

I would prefer to design a sampling oscilloscope anyway because currently available sampling oscilloscopes are lacking.  My dream project is an DSO with modular front ends including a sampling front end.  DSOs and sampling oscilloscopes share most of the needed digital hardware.

[Jan Steinman]

Here's mine. "High performance" is relative, no? This one has the highest performance of any that I've built (sample size: 1).

[tggzzz]

I think the issue is more fundamental than that. You cannot really make a high speed oscilloscope without having a high speed oscilloscope to debug it.

In that case, how do companies make the world's fastest scope?

[xrunner]

I think the issue is more fundamental than that. You cannot really make a high speed oscilloscope without having a high speed oscilloscope to debug it.

In that case, how do companies make the world's fastest scope?

It's a Catch-22 

[Jan Steinman]

I worked at Tektronix, although not on scopes.

I do know that they used tunnel diode sampling to boost the maximum frequency of a scope, for periodic waveforms. That way, they could actually scope a scope with a scope that didn't have enough bandwidth on its own to scope the scope.

All clear now? :-)

[tggzzz]

I worked at Tektronix, although not on scopes.

I do know that they used tunnel diode sampling to boost the maximum frequency of a scope, for periodic waveforms. That way, they could actually scope a scope with a scope that didn't have enough bandwidth on its own to scope the scope.

All clear now? :-)

My Tek 485 has a 350MHz bandwidth. It triggers stably on a >1GHz input signal. The frequency limitation appears to be that the signal's amplitude is reduced to <0.2 divisions, i.e. the TD trigger is not the limitation.

Solid triggering is an underappreciated parameter affecting a scope's usability.

[NiHaoMike]

I say go straight for the capability that very easily beats most scopes on the market, and that's memory depth. A used server can be fitted with a few hundred GB of RAM for much cheaper than you think.