Open-source digital oscilloscope "ScopeFun"

[prasimix]

The crowdfunding campaign just started: https://www.crowdsupply.com/scopefun/open-source-instrumentation

[james_s]

Yeah I think it's been said already that $200-$250 is about the right price point for this thing. That's cheap enough that it has something to offer over more polished commercial products.

[dejanpri]

I am reading this with interest and I just wanted to say, because you seem to be very focused on the price of ScopeFun...

$250 is about right for the "old" version of ScopeFun which was released few years ago. The "new" version of ScopeFun which is currently available on Crowd Supply is an improved design with higher specifications, so the price is also higher. If you check the ScopeFun BOM file on gitlab, you can see that there are more than 100 different types of components used and some of them are quite expensive.

And if you are comparing ScopeFun with other commercially available oscilloscopes, you should check all specifications, because a lot of times there are limitations which other manufacturers usually hide in small print.

But anyway, If you focus only on price, then you are probably missing the main idea of ScopeFun. It's not trying to directly compete with other commercially available oscilloscopes in terms of price, but more to build an open T&M platform which can be fully customized.

--
Dejan,
ScopeFun

[james_s]

Sure but in the end it all comes down to price for most people, price is king, and virtually all successful open source projects are successful because they are either cheaper than comparable commercial products or they do something unique that is not easily done by commercial products. I cannot think of one single successful open source project that costs as much or more than a COTS product of similar capability. Only a very, very tiny number of people have the skills and desire to modify and customize their test gear. Most people who have the skills to do that are going to prefer to spend their time using their test gear to accomplish other things rather than customizing their test gear.

I use quite a lot of open source stuff, both software and hardware, always because it's either cheaper than commercial equivalents or it's so niche that a commercial equivalent doesn't exist.

[OwO]

Well price is kind of important because >$500 is a major investment and most people simply can't afford/can't justify the spending...

[wilfred]

Why would I NOT expect open source to be cheap? If I can't save money then why on earth would I go with a half baked open source project vs a polished, completed, supported commercial product?

You'll have to tell me why you expect open source to be cheap (or free). Free has two meanings in the world of open source. Not all commercial products are polished, complete or supported very well or for very long. Eg smart TV's or really any device that requires firmware. Open source projects are not necessarily half-baked either.  I'm assuming by half baked you mean subject to continuing change. Which would be unfair because it is in essence the aspect of open source that appeals to those not strictly motivated by the free price tag.

[OwO]

I had a quick look through the BOM and it looks very un-optimized; why do you need both 1.25V and 1.8V voltage references? 1.8V is also an uncommon voltage (1.25V and 2.5V are the most common). What is that soldered through hole jumper for? No way to route a signal (this is a 6 layer board??) or use 0ohm smd jumpers? There are both 0402 and 0603 10nF caps, get rid of one. Why both 0603 and 0805 0ohm jumpers? You have 4 different voltage references and 4 different TVS diodes, surely this can be reduced? There is also what looks like 4 different op-amp types with similar specifications.

From the BOM it looks like someone designed this with no consideration of manufacturing or parts sourcing. The parts choices in general is also quite strange, niche and uncommon parts are used for very ordinary and common applications. I'm sure there is huge leeway in cost reduction without compromising quality. Even the open source 3GHz VNA designs have no more than 60 BOM lines.

[OwO]

...and the elephant in the room is that this is a 6 layer board when it really should be 4 layers. Every single open source VNA design I've seen are 4 layers...

I don't think the open-sourceness has anything to do with the price, it's all about design optimality and (trying to) do more for less, exactly the same thing in the closed source world.

[dejanpri]

OK, you are looking at BOM and you are starting to see why it costs so much...

Actually a lot of time and work was invested into optimizing BOM, and let me assure you that every component is there for a reason:
- 1.8V ref. is used for control DAC's and 1.25V ref. is for AWG DAC. We could use only one, but this would reduce precision for AWG and also increase noise
- there are no wire jumpers on PCB, you are probably looking at heatsink anchor
- 10 nF 0603 capacitors are used on digital inputs (more resistant to ESD), others are used for HF decoupling
- regulators+tvs diodes are located on different input blocks and for different reasons - protection requirements for analog front end are not the same as for ADC or digital inputs
Also, there are no niche and uncommon parts, all parts are available at mouser, except FPGA and heatsink assembly.

If you need to reduce the costs as much as possible (and are willing to make some compromises regarding reliability and performance), then your suggestions could make sense, but even then the difference in price would be small, because the most expensive components are FPGA, ADC, DAC and high speed op-amps.

4 layers??? I'm not saying it's impossible, but it would be extremely difficult to route high speed signals between FX3 <> FPGA <> DDR3 only on 4 layers and almost certainly not without blind/burried vias which costs the same as adding 2 more layers. High speed signals require a solid plane reference and more layers also improves EMI which is important for low noise.

[OwO]

4 layers??? I'm not saying it's impossible, but it would be extremely difficult to route high speed signals between FX3 <> FPGA <> DDR3 only on 4 layers and almost certainly not without blind/burried vias which costs the same as adding 2 more layers. High speed signals require a solid plane reference and more layers also improves EMI which is important for low noise.

Look at my example design of 32 bit DDR3 + Zynq on 4 layers: https://github.com/gabriel-tenma-white/sdr5

I'll take a more detailed look at the schematic rather than just the BOM and see if I can find any more low hanging fruit to fix.

[free_electron]

I had a quick look through the BOM and it looks very un-optimized; why do you need both 1.25V and 1.8V voltage references? 1.8V is also an uncommon voltage (1.25V and 2.5V are the most common). What is that soldered through hole jumper for? No way to route a signal (this is a 6 layer board??) or use 0ohm smd jumpers? There are both 0402 and 0603 10nF caps, get rid of one. Why both 0603 and 0805 0ohm jumpers? You have 4 different voltage references and 4 different TVS diodes, surely this can be reduced? There is also what looks like 4 different op-amp types with similar specifications.

From the BOM it looks like someone designed this with no consideration of manufacturing or parts sourcing. The parts choices in general is also quite strange, niche and uncommon parts are used for very ordinary and common applications. I'm sure there is huge leeway in cost reduction without compromising quality. Even the open source 3GHz VNA designs have no more than 60 BOM lines.

copy paste of appnote designs ....

[james_s]

Why would I NOT expect open source to be cheap? If I can't save money then why on earth would I go with a half baked open source project vs a polished, completed, supported commercial product?

You'll have to tell me why you expect open source to be cheap (or free). Free has two meanings in the world of open source. Not all commercial products are polished, complete or supported very well or for very long. Eg smart TV's or really any device that requires firmware. Open source projects are not necessarily half-baked either.  I'm assuming by half baked you mean subject to continuing change. Which would be unfair because it is in essence the aspect of open source that appeals to those not strictly motivated by the free price tag.

I already did tell you, in fact it has been mentioned several times. I expect open source to be cheap because it virtually always is. I expect it to be cheap because it usually is not quite finished, not quite as polished as a commercial product. I have to do some of the work myself, I have to do my own digging for information, I cannot simply call up the manufacture and get support.

Can you point me to even one open source project that is successful despite costing as much or more than a comparable closed source commercial product? I can't think of one. If you still cannot understand why most people expect open source to be cheaper then you are being willfully ignorant.

[thm_w]

I had a quick look through the BOM and it looks very un-optimized; why do you need both 1.25V and 1.8V voltage references? 1.8V is also an uncommon voltage (1.25V and 2.5V are the most common). What is that soldered through hole jumper for? No way to route a signal (this is a 6 layer board??) or use 0ohm smd jumpers? There are both 0402 and 0603 10nF caps, get rid of one. Why both 0603 and 0805 0ohm jumpers? You have 4 different voltage references and 4 different TVS diodes, surely this can be reduced? There is also what looks like 4 different op-amp types with similar specifications.

From the BOM it looks like someone designed this with no consideration of manufacturing or parts sourcing. The parts choices in general is also quite strange, niche and uncommon parts are used for very ordinary and common applications. I'm sure there is huge leeway in cost reduction without compromising quality. Even the open source 3GHz VNA designs have no more than 60 BOM lines.

Most of your points are valid, but, multiple packages or 0 ohm shunts? We are talking about shaving cents off of the design at that point. Its not worth the time.
Going off jlcpcb prices, 4-layer pcb is ~$2.40 each and a 6-layer is $3.50. So saving $1.10 per board is somehow a big savings, considering the hours of work you'd put into it?

Their goal is to sell ~100pc, so mouser prices are:
KAD5510P-25Q48 $44
XC7A35T-FTG256 $40 (don't see it on mouser)
LTC6957HMS-2#PBF $5.49
ADA4817 $5.00 x 2
AD8337 $4.40 x 2
MAX15053 $2.20 x 5
LTC4360-1 $1.80
OPA2830 $2.20

If you can figure out how to save on any of these parts its going to make much more difference. LCSC has the FPGA but not enough pieces sadly.

[Marco]

Trying to earn back the invested time takes enough margin, that worrying about a couple bucks worth of components/PCB isn't really worth it.

[Marco]

I cannot simply call up the manufacture and get support.

For manufacturers where doing so won't just aggravate you and help very little, the prices for this level of functionality will be more expensive.

[AndreZheng]

 ;DI can help give you a quote for this chip. The XC7A35t-2FTG256C FPGA costs around 15$, but this is the price from China local vendor.

[OwO]

Reducing BOM lines isn't saving cents, it can save up to $10 per unit on assembly based on various quotes I have gotten. Non-optimal parts choices add far more cost, because I see a large portion of the cost of the design is in the little things. I see ~$20 in ESD protection alone, there must be a more effective way of doing input protection. The choice of main system components may not be optimal either, for example I have seen far cheaper ADCs with similar specs, and if you are buying FPGAs from digikey you are getting ripped off (it's not too bad for prototyping but you don't want to do that for production).

[OwO]

MAX15053 $2.20 x 5

Replace with AP3429 ($0.168 on mouser @100pcs). Same specs, 1/10 the price. Also saves you from the PCBA cost associated with a stupid package.

[dejanpri]

4 layers??? I'm not saying it's impossible, but it would be extremely difficult to route high speed signals between FX3 <> FPGA <> DDR3 only on 4 layers and almost certainly not without blind/burried vias which costs the same as adding 2 more layers. High speed signals require a solid plane reference and more layers also improves EMI which is important for low noise.

Look at my example design of 32 bit DDR3 + Zynq on 4 layers: https://github.com/gabriel-tenma-white/sdr5

I'll take a more detailed look at the schematic rather than just the BOM and see if I can find any more low hanging fruit to fix.

I see you are using a larger footprint Zinq in your DDR3 design. Now try to add 2x10 bit ADC, 12-bit DAC, 12-bit digital GPIO, 32-bit GPIF FX3, etc... and 4 layers aren't enough anymore. But as I said, even if you could do it, I would be concerned about EMI.

[dejanpri]

MAX15053 $2.20 x 5

Replace with AP3429 ($0.168 on mouser @100pcs). Same specs, 1/10 the price. Also saves you from the PCBA cost associated with a stupid package.

Nice find, this could probably work

[dejanpri]

@james_s, but why do you think that ScopeFun costs more then a comparable closed source commercial product? I think you should check again...

[james_s]

@james_s, but why do you think that ScopeFun costs more then a comparable closed source commercial product? I think you should check again...

I didn't say it did, at $750 I suspect it does, at a lower price perhaps not, I haven't really checked. I was simply responding to the absurd "why would you expect open source to cost less?" question.

[nAyPDJ]

@dejanpri

One thing I', curious about is why write your own scope software from scratch? I've taken a bit of a look at Sigrok, and it seems to me like extending that would have save some work, especially with how constrained your project is in manpower.

[thm_w]

Reducing BOM lines isn't saving cents, it can save up to $10 per unit on assembly based on various quotes I have gotten.

So you are telling me adding a single BOM line item, increased the assembly quote by $1+ per board?

Changing this quote from 100 unique parts per board to 110 increased the cost by 40c per board, or 4c per component: https://www.pcbway.com/quotesmt.aspx
This page went up 16c per board or 1.6c per component: https://www.7pcb.com/PCB-Assembly-Quote.php

Non-optimal parts choices add far more cost, because I see a large portion of the cost of the design is in the little things. I see ~$20 in ESD protection alone, there must be a more effective way of doing input protection. The choice of main system components may not be optimal either, for example I have seen far cheaper ADCs with similar specs, and if you are buying FPGAs from digikey you are getting ripped off (it's not too bad for prototyping but you don't want to do that for production).

$20 ESD ICs is valid place where money could be saved. You might risk additional returns, hard to say without looking at the design.
Yes ADC/FPGA is where the big money is, as we've established.

[dejanpri]

@dejanpri

One thing I', curious about is why write your own scope software from scratch? I've taken a bit of a look at Sigrok, and it seems to me like extending that would have save some work, especially with how constrained your project is in manpower.

Probably the main reason is to have more flexibility with adding new features and also to achieve better performance. Pulseview is OK for basic stuff, but with using our software we can have features like dedicated threads, OpenGL rendering (3D view, virtual persistence) and other hardware control features like sever mode, Python API.