Why the lack of open source scopes?

[GeoffS]

Can the code be stolen?

It's intended to be open source.

[Seg]

Hardware guys don't understand Open Source. But then again, most programmers don't either.

[SArepairman]

i mean why cant we steal the code from rigol or agilent

tools for the people!
oscilloscopes for everyone!

[dannyf]

why cant we steal the code from rigol or agilent

Why stop there? Why cannot we steal money from you? Why cannot we steal car from  you? Why cannot we steal house from you? ...

Really, people? Is there really no hope for humanity?

[poorchava]

Perhaps I'm being naive, but I don't think building a reasonable scope frontend is that difficult. Take the low-end of the Tek 7000 series for example, they had bandwidths of around 100-200MHz and were entirely built from through-hole components and double-sided/4-layer boards. Designing one, however, is probably going to be a bit more difficult, especially if OP hasn't done it before. A good start might be to get a vertical amplifier from an old mainframe-type scope (the 7000 series again comes to mind, although I'm sure there are cheaper alternatives) and get your digital back-end working well with that before trying to design your own frontend.

You won't get the ultra-low noise floor you'd expect from a modern commercial scope without serious black magic, but you should be able to get something that's Good Enough™.

As for why there are no open-source scopes, there are plenty. They're just extremely limited in sample rate and storage capacity (for digital), or require parts that aren't easy to get anymore (for analog). My guess as to the reason why there aren't many good open-source scopes is simple: For a very long time, high-speed ADCs and micros/FPGAs were expensive, hard to get, hard to use, and often covered by NDAs. It's changing now, but unlike, say, software defined radio, most of the people who want one either have one or can easily get one, so there's less desire to build one from scratch.

That's another major issue you are touching there: patent issues. The scope would have to be designed in such a way, as to not infringe on any patents held by Agilent, Tek, Hameg, Toshiba, Philips, Hitachi, Rigol, Atten, Siglent, LeCrap, Rhode&Schwarz etc.

I think that actually the analog part is the hardest. Digital hardware is pretty simple. Stuff like FPGA/DDR3/ADC schematic and layout can be copied from reference designs and application notes almost 1:1. The most critical part is the code inside FPGA or processor, which doesn't cost anything if you do something wrong - you just upload new version and test.

With analog design things are much harder, you can easily expect to have at least 3-4 revisions of the analog frontend before it works even remotely ok, and every revision costs a lot of money.

As for stealing code... First you have to get to that code and then defend yourself from omnidirectional legal attack. And that's assuming you live in a place with even half-sane patent law (in other words: not USA).

[Marco]

The design of this thing could be a massive flame war if you don't have strong leadership. Also, to complete with current generation scopes like the Rigol 2000 or Agilent DSO series you need some extremely high performance FPGA code written, which is very much non-trivial.

It's open source ... in the beginning it's good enough if it can do all the straightforward trigger modes at 100 Hz update rate. If it takes off someone will come along to improve it eventually (assuming the hardware picked has sufficient headroom).

I actually think you could do the front end with double sided PCBs and hot plate soldering ... as was said by Fanta Tek and others were doing 200 Hz scopes on double sided PCBs decades ago and a lot of their design effort went into what we would now use an integrated PGA for. Running through design iterations with double sided PCBs and hotplate soldering isn't going to drive cost to high heaven.

The FPGA/CPU need to be on a 4+ layered PCB, but as I said before you can simply buy the Parallela which is ridiculously good value. A huge advantage of the Parallela which I don't see any way to replicate with a processor of similar cost is that it can take the full 1 GB/s data stream. Normally this would require a processor with something like a multilane PCIE interface (or maybe you could hack something with a large number of GPIO and DMA but it would be quite a feat). So you could even just not do any processing inside the FPGA at all for the initial prototype just decode the LVDS and pass the data on to the CPU.

Maybe the ADC needs 4 layers too, but in the short run you could just use an ADC evaluation board with 50 Ohm inputs ... and then just copy paste that reference design later, that's what it's for.

[SArepairman]

why cant we steal the code from rigol or agilent

Why stop there? Why cannot we steal money from you? Why cannot we steal car from  you? Why cannot we steal house from you? ...

Really, people? Is there really no hope for humanity?

i would not keep tools out of the hands of the people 

What the hell will agilent or rigol care if a few hundred people (maybe) decide to build some time consuming oscilloscope.

in the end having the technology in the open will probably bring a net benefit to society at the cost of rigol/agilent anyway.

[LoyalServant]

i would not keep tools out of the hands of the people 

What the hell will agilent or rigol care if a few hundred people (maybe) decide to build some time consuming oscilloscope.

in the end having the technology in the open will probably bring a net benefit to society at the cost of rigol/agilent anyway.

Speaking as a software developer that has intimate knowledge of open source vs. closed source.
I have worked in both.
The obvious benefits to open source is that the end users can alter it in a way that suits their needs or to fix bugs.
Close source... well we all know that means we are at the mercy of the manufacturer and quite often that means
you get one release, possibly two then the hardware is EOL and as a result whatever you got... your stuck with.

Since Rigol is being used as an example let's go with that....
Rigol is software disabling features on their kit that that means open source is right out.
This is the way things are heading so obviously for them it will not bring a benefit.

If a few hundred people took their software to put on their cloned hardware you can bet their lawyers will be after someone.
So no, you can't steal their software and get away with it.. they will not allow that.

[Marco]

I wouldn't be so sure. Open source software where it costs you nothing but time to download and modify is one thing, but to work on this scope you would need expensive hardware. If it only has basic features to start with and isn't super cheap I can't see many people buying it, which means the pool of people who are interested and knowledgeable enough to improve it is going to be extremely small, and quite possibly less than 1.

Okay maybe I was a bit optimistic, there is a bit of luck involved fining those few talented individuals with more time than money or sense (or too much money and boredom).

Look at the various open source logic analyzers.

Between Logic Sniffer and Sigrok it seems to be doing alright.

Look at the commercial ones with open APIs. Even though writing protocol decoders isn't too difficult hardly anyone bothers.

Actually I'm relatively sure a lot of people bother, it's just that most people don't want the headache which comes with sharing code so they just use it privately ... the subset of people using explicitly open source tools obviously will have more people among them willing to share.

Any way, improving the oscilloscope's FPGA programming wouldn't require expensive hardware ... a signal generator would be enough.

[homebrew]

Hey guys,

where is your open source spirit? Those projects are not about developing competitive products. They can't be!
Software is another story. Open source software communities can (and did) develop competitive alternatives (like the linux kernel for example) for existing products in the market (Unix-derivates of that time). Why? (1) Because those projects just relay on brain-power.  (2) Because it is very cheap to alter the existing product for an individual contributor.

Hardware on the other hand, might be very expensive to alter (think of throwing away a 6-layer board of the oscilloscope example), smd parts can't be reused easily etc...

To my opinion, OS hardware projects exist for two reasons: (1) Entering the market with a new and innovative product where the business-case relies on the openness (like Raspberry-PI at the time). (2) Because we can! Because we want to learn and understand! OS hardware artifacts are a perfect way of sharing design knowledge when documented properly. Others thereby can understand why certain design decisions are made etc.

Many hardware projects are like that. Take for example all the hobbyists CNC-machines. If those projects are of real value, they tend to be expensive. In fact so expensive that it really makes no sense from an economical perspective. Those hobbyists could send of their milling parts to fabs till the end of their lives for the price of their machines ... And there is generally no argument made that they can not reach professional CNC standards by any means... it's simply not about that.

So in my opinion there will never be competitive OS scopes (neither in price nor in performance). Get realistic, let's build a well engineered 100MS/s 20MHz bandwidth scope, 32M samples memory with a standalone display. That would be useful, affordable and most important understandable. And a lot of fun of course!

The people participating could practice all flavours of designing electronic products:
Digital design, analog design, embedded programming, fpga-design, power supply design, user interface design, case design, etc, etc … lots of opportunities!

Take 10-20 interested people, one year of their spare time and a budget of 300$ per prototype …

Would be a great endeavour. Wanna pull that off anyone? I'd definitely be interested …

Cheers,
Peter

[openSourceScope]

Thanks for the positivity Peter! :-)

[madshaman]

I'd also be interested in such a project, even of only to verify others' designs, testing etc. but could only dedicate so much time.

[branadic]

Good luck to find enough active fellowers. I've tried to assure people for a similar project multiple times, but without success. But instead of an 8bit design I was more interested in a 10bit design (something between Rigol and LeCroy) with up to GSps sampling rate and analog bandwidth of up to 500MHz. There are several adcs available for such a task.

For to analog input stage you can find an open source solution here, the pickaback board we designed for the Welec scope to replace the original one:

http://sourceforge.net/apps/trac/welecw2000a/wiki/HardwareImprovement

Furthermore you find an open souce FPGA design inkl. a NIOSII softcore here:

http://sourceforge.net/apps/trac/welecw2000a/wiki/niosII

[Marco]

So in my opinion there will never be competitive OS scopes (neither in price nor in performance). Get realistic, let's build a well engineered 100MS/s 20MHz bandwidth scope, 32M samples memory with a standalone display. That would be useful, affordable and most important understandable. And a lot of fun of course!

100 MS/s is the wrong point for a lowball design, 80 MS/s is a sweetspot ...

Would be a great endeavour. Wanna pull that off anyone? I'd definitely be interested

That's not really how amateur open source projects tend to work, it's usually one person creating a usable product before a community adds value.

BTW, one guy already half-made the hardware for something with these specs ... although he seems to have given up after getting a PCB made.

[homebrew]

100 MS/s is the wrong point for a lowball design, 80 MS/s is a sweetspot ...

Could you give a pointer of why that is? Definitely no availability issue of the ADCs... Just curiosity!

That's not really how amateur open source projects tend to work, it's usually one person creating a usable product before a community adds value.

Exactly! And in my opinion that's why most of the scope projects are half done, or half of the quality reachable. Doing all the design flavours is just to much for one person ...

[hlavac]

Real, Open source scope would be a great thing. As we can see from the ridiculous software crippling escapades some scope vendors are pulling, scope hardware is really quite cheap.
It would make great sense to have the open source firmware you can duplicate freely, this is where open source helps mankind as a whole!
You would need a worthy hardware platform, but thats something a Kickstarter would solve...
As for the PC scope concept, yes it makes perfect sense, dont be put off by the knob loving ludite crowd
All we need is enough processing power on the hardware to do the actual processing of the samples, leave the display and controls to the PC...

[Marco]

Could you give a pointer of why that is? Definitely no availability issue of the ADCs... Just curiosity!

Simply price/performance, sure it's only a 3$ drop in price (which is 33% drop in price for 20% drop in performance) ... but if you are going to lowball any way do it proper.

Though I still think the HMCAD1511 is the better IC to do a project around ... slightly higher end (1 GS/s single channel, 250 MS/s quad channel) but still affordable at 64$.

[Rasz]

Chris Gammell our saviour and prophet spoke >year ago on TheAmpHour about a need for a standardized protocol that would let all lab equipment speak/display/use Tablet as a standard UI.

Tablets horsepower quadrupled since then. Today it is possible to do all of oscilloscope computing inside a tablet. 4x Arm A9 = >20Gflops, new tablets with OpenCL will hit 60Gflops soon (Mali-T604) or even >250Gflops (top tier mali).
Randomly googled Lecroy paper shows some old 18GHz acquisition system doing DSP on Intel P4 (10Gflops) and reaching 2 Mpoint/s.

When you get right down to business Digital oscilloscope is just a software defined radio. All you need is an analog part (attenuators/filters/amps), digitizer, and finally some computing bloc + display. Currently computing is done in FPGAs and is secret sauce. If we decouple computing away from the magic closed box on the desk and move it upstream we would for example end stupid licenses for SPI/can/lvds/whatever decoding scam.

[Marco]

FLOPs don't help if you don't have the bandwidth ... for something like variable intensity displays with very high update rates (Digital phosphor as Tek calls it) you need to be able to process at wire speed.

[alm]

Real, Open source scope would be a great thing. As we can see from the ridiculous software crippling escapades some scope vendors are pulling, scope hardware is really quite cheap.

To the likes of Rigol and Atten. Not to the hobbyist in one-offs (or hundred-offs). Why not just write an open-source bitstream and firmware for the Rigol DS2000, or the Hantek scope that already runs Linux? Forum user tinhead may have already done some of the reverse engineering for the latter.

As for the PC scope concept, yes it makes perfect sense, dont be put off by the knob loving ludite crowd

The usability of a PC-based GUI just sucks compared to a scope with dedicated buttons and knobs. Ask anyone who's used both. I think I've met one person who after using both a stand-alone scope and PC-based scope preferred a PC-based scope. And that was mostly because bench space was at a premium and he found the worse UI and analog performance to be acceptable for his purpose. PC-based is easier for a DIY design if you're building a scope from scratch. But it will again make the product less competitive with the 'real' commercial scopes.

All we need is enough processing power on the hardware to do the actual processing of the samples, leave the display and controls to the PC...

I would be more worried about how to get the data in the PC at a decent rate, if you don't want to be one of those PC-based scopes with way too slow update rates. USB 2.0 is ubiquitous, but slow. PCIe is fast enough, ubiquitous in desktop PCs, but not that simple to interface. USB 3.0 might be fast enough, but is not that ubiquitous and most USB 3.0 interface ICs are not available in small quantities.

Tablets horsepower quadrupled since then. Today it is possible to do all of oscilloscope computing inside a tablet. 4x Arm A9 = >20Gflops, new tablets with OpenCL will hit 60Gflops soon (Mali-T604) or even >250Gflops (top tier mali).

Except that the fastest interface available on many tablets is USB 2.0 high speed, which might be able to transfer 24 MS/s or so on a good day.

When you get right down to business Digital oscilloscope is just a software defined radio.

Just like a software defined radio, which only has maybe 20 MHz bandwidth at best. And is not DC coupled. Or has to receive signals from millivolts to tens of volts. Or needs a flat response across its frequency range. And a well defined 1 Mohm input impedance.

[AndyC_772]

Any time I'm using my scope, my PC is usually running a compiler, or serial terminal, or viewing a PDF data sheet.

It's simply not available for use as a scope UI.

[Marco]

And a well defined 1 Mohm input impedance.

1 MOhm input scopes are just a nasty bit of history which should have been ditched long ago, the front end belongs on the probe IMO. It's only kept around at this point so they can upsell the sane way to do it.

[alm]

From a practical point of view: probes designed for a 1 Mohm input impedance are available cheaply with a bandwidth of about 100 MHz and suitable for voltages from a few mV to a few hundred volts. The cheapest commercial active probe designed for 50 ohm input impedance will probably cost a few hundred bucks. And it will probably only have a fixed range of about +/- 10 V. Useless for low level audio signals or higher level power supply rails.

Putting the full attenuation, controls and compensation network in the probe head, which has to be small by necessity, will be tricky. I'm not aware of any commercial product doing this. Designing custom probes, in addition to a scope, does not help in making the project feasible. And what's the advantage? It allows for a lower input capacitance, but that's not exactly your biggest issue when designing a < 100 MHz scope.

I guess you could design an external attenuator / amplifier that connects to the 50 ohm scope input, but why? The less boxes on my bench, the better. The only commercial equivalents that come to mind are the buffer amplifiers that allow high-end scopes that are 50 ohm only to interface with high-impedance probes like the Tek TCA-1MEG, but these are not exactly cheap (read: cost in excess of $1k).

[Rasz]

All we need is enough processing power on the hardware to do the actual processing of the samples, leave the display and controls to the PC...

I would be more worried about how to get the data in the PC at a decent rate, if you don't want to be one of those PC-based scopes with way too slow update rates. USB 2.0 is ubiquitous, but slow. PCIe is fast enough, ubiquitous in desktop PCs, but not that simple to interface. USB 3.0 might be fast enough, but is not that ubiquitous and most USB 3.0 interface ICs are not available in small quantities.

USB 3.0 is available in small quantities, FX3 does 250MB/s

Tablets horsepower quadrupled since then. Today it is possible to do all of oscilloscope computing inside a tablet. 4x Arm A9 = >20Gflops, new tablets with OpenCL will hit 60Gflops soon (Mali-T604) or even >250Gflops (top tier mali).

Except that the fastest interface available on many tablets is USB 2.0 high speed, which might be able to transfer 24 MS/s or so on a good day.

Every tablet soc these days has extremely fast camera interface. For example i.MX6 has csi-2 at ~3Gbps and cpi 20bit at 180MHz. That would be enough for 2 channel 20MHz, good start I think.

When you get right down to business Digital oscilloscope is just a software defined radio.

Just like a software defined radio, which only has maybe 20 MHz bandwidth at best. And is not DC coupled. Or has to receive signals from millivolts to tens of volts. Or needs a flat response across its frequency range. And a well defined 1 Mohm input impedance.

All of that is front end. Community build a family of frontends, starting with cheap ass 2x20MHz and see what software people can do with it.

Any time I'm using my scope, my PC is usually running a compiler, or serial terminal, or viewing a PDF data sheet.
It's simply not available for use as a scope UI.

So you are limited to only 1 computer per household?
You would swap magic black box Oscilloscope for a PC dedicated to being Oscilloscope. Its not about being cheaper, its about opening up that black box and letting you do whatever YOU want with data. No more stupid UIs, no more  $1000 licenses to decode serial protocols, no more being stuck with one color palette, limited to shallow sample memory, or shitty slow math functions. Did I mention no more 3 frames per second Owon GUI?

Every time Chinese come up with new scope (or ham radio or whatever doohickie) its ALWAYS the same story - decent hardware (for the price) married with total shit for software. Even western companies can frack up the simplest things, like FLIR E4 Mike tested recently (4 buttons that do NOTHING and 3 levels deep lagging UI to change one of the most often changed parameters). All of it can go away if we decouple presentation and computation layer from the frontend.

[alm]

Every tablet soc these days has extremely fast camera interface. For example i.MX6 has csi-2 at ~3Gbps and cpi 20bit at 180MHz. That would be enough for 2 channel 20MHz, good start I think.

Sure, the SoC has fast interfaces. How do you access this without hacking the tablet? Hacking the tablet is going to be particular to one type of tablets, and cheap tablets tend to come and go, so it's hard to standardize on one type unless you pick an expensive type like a Samsung/Apple product. Are you also proposing to build your own tablet?

All of that is front end. Community build a family of frontends, starting with cheap ass 2x20MHz and see what software people can do with it.

Many SDRs don't even have 20 MHz bandwidth, let alone dual channels. Most SDRs use USB 2.0, which can't sustain 2x20 MHz on a single port. So what components would scopes and SDRs have in common?

You would swap magic black box Oscilloscope for a PC dedicated to being Oscilloscope. Its not about being cheaper, its about opening up that black box and letting you do whatever YOU want with data. No more stupid UIs, no more  $1000 licenses to decode serial protocols, no more being stuck with one color palette, limited to shallow sample memory, or shitty slow math functions. Did I mention no more 3 frames per second Owon GUI?

You already can. Any DSO sold today can interface with a computer over USB and/or LAN. Often with a well-documented SCPI protocol if you buy a decent brand. Few people use this for general purpose use (as opposed to logging or automated setups) because it's a pain compared to the stand-alone scope and software availability is poor. Is either of that going to change when an open-source scope comes along? Look at the Sigrok project: none of their GUI's offer a decent interface for a logic analyzer yet. None of the GUI's for the open-source SUMP/OLS logic analyzer is that great either. And the GUI for a LA is much simpler than for a scope.

Every time Chinese come up with new scope (or ham radio or whatever doohickie) its ALWAYS the same story - decent hardware (for the price) married with total shit for software.

So how about taking that decent hardware that you're not going to exceed in either performance or price in an open-source design and replace that shitty software?

In general it seems that you guys are looking to increase the number of challenges, as if reaching your core target is not hard enough. Summarizing your and Marco's post together it seems that you guys want to build an analog front-end, an active probe, the ADC + digital hardware, build/hack a tablet and write software better than that by Owon/Hantek/Atten/Rigol. Each of which alone could easily take an experienced engineer a few months of work, depending on how high you're aiming. If you want to accomplish something, then you should identify the core problem you're trying to solve and focus on that.