1Gs/s Digital scope project. Anyone interested?

digital scope projects

Hi all,

I would love to have a digital oscilloscope capable of at least 1Gs/s. Does anyone think it would be possible to make a homebrew one? If anyone is interested, I'm sure we could work this out. This is roughly the plan:

Use 10 ADCs capable of 100MS/s each, and interleave the sampling. They would all be controlled by an FPGA.

I look forward to hearing from anyone.

thanks
Hugo

circuit digital scope

A design like this requires *lots* of development resources: digital and analog circuit design, FPGA design, PCB design, PCB manufacture (no perf boards or even quality 2-sided PCBs for this !) Is this a work project ? If so, you had better already be in the digital scope business. If not, this is definitely *not* a home-brew project.

Am I that this is to be a USB/PC device ? If so, take a look at the Link Instruments PC Based USB 60MHz Model MSO-19 ($250). It is *only* good for 60 MHz. Think how many man-hours it would take to develop just a properly working prototype !

usb oscilloscope 1gs/s

Hi Pascor,

Yes, this is definitely a hard project. But I figure that the kind of people who might need this (like myself) might also be those kind of people who would be able to help.

I hope you don't think I was imagining this soldered on perf board! It will obviously require properly designed PCBs. I've just purchased a seat of Altium Designer, so I've got access to fairly good tools.

I know it's hard, but if we manage to come up with a design, it would be such a great resource for the community.

Hugo

good will digital scope

take a look at this - slow (100Ms/s) but interesting

https://www.fpga4fun.com/digitalscope.html

make digital scope circuit

I too thought of designing one of these and did a bit of work to get there. At these frequencies everything becomes hard! PCB tracks and even the PCB material have to be chosen for the right properties. All tracks are transmission lines and very high speed logic techniques must be employed. I'm not suggesting you don't try it - its a fun challenge. But if you want a 1GS/s scope - do what I did and peruse ebay. With a little patience you can pick one up for ~$400. It will cost you more than this to start to build one and the design equipment required even more.
However, the path I would wander down if I wanted to build one would be to look at the evaluation boards of the fast ADCs out there (usually $1000 to buy). All these systems use very large Xilinx FPGAs as the interface and RAM for the ADCs. Wideband amplifiers have another swag of complexities to consider along with the input attenuators. Of course all of these would have to be SMDs. Also read the HP journals for highspeed scope design articles. Although old, all the basics are described very well and these still apply nowadays.
Good luck!

swingbyte

OK, here are some ideas, anyone want to comment?

There are some fairly low cost 250Ms ADCs, like the AD9480: https://www.analog.com/static/imported-files/Data_Sheets/AD9480.pdf
They have input bandwidths up to about 700MHz.

Four of these together would give 1Gs. The data then just need to be clocked directly into four memory ICs.

The datasheets give example schematics and layouts, which don't look overly hard.

We use Altium Designer at work, and regularly have multi-layer PCBs made. Altium also has powerful features for designing FPGA systems. I have lots of experience designing electronic systems with SMD, but have never tried anything this fast.

Surely there's enough brain power on this forum to get something like this working?

Hugo

Adding some useful links:

100Ms/s FPGA scope: http://www.fpga4fun.com/digitalscope.html
How probes work: http://www.signalintegrity.com/Pubs/straight/probes.htm
Bitscope: http://www.bitscope.com

I am looking for a new challenge. Something that involves some more complex PCB design and, if possible, uses an FPGA. One of my ideas was to build a spectrum analyzer using some high speed ADCs. I did some research a while back. Funny that I found this thread and I came up with something similar. I looked at the AD9481 and also the ADC08200.

I have designed FPGA, ARM9 and AVR32 boards of up to 8 layers. I have also done quite some VHDL design and C programming. The new thing for me would be higher speeds both on analog and digital sides.

So when do we start? Maybe it, or part of it, can be put on opencores? In my case I would even consider using FFT stuff from opencores.

Ok, 1GS sampling is unbelievably difficult to achieve, getting 4 250MHz clock perfectly shifted to achieve this is not an easy task. Sure, it can be done, but you need some really expensive equipment to test/verify this.

Making a 1Gs oscilloscope is not difficult, making it good is the problem

Sorry, I can not agree; than what you basically need is a better delayed timebase oscilloscope and a good pulse generator_but your gen. you can build self too...
K.

Hi,
I found yet for what I in subject "high-speed generator" referenced:
Appl. Note 98, LT, 2004.
Check pls pp5--9, or maybe full
K.

I am also interested in this topic and although I am a novice in this area I recognize the difficulties involved. TI does provide high performance A/D and signal processing for this range of frequencies as discussed at the following link and in a single chip.
https://www.ti.com/ww/en/analog/ads5400/index.shtml?DCMP=hpa_dc_ads5400&HQS=Other+BA+ads5400-bs

This is a very costly chip but it does show that this is plausible. I wonder about approaching the design issues from another perspective. By starting at a lower sampling frequency and building upward. Many of us do not require the higher speeds. In fact I feel that the majority of users would be content at far lower frequencies. Building upward would mean that a stable design could be achieved for lower frequencies first at lower cost and at a lower investment of time. Then as the project continues lessons learned at lower frequencies would be applied. At various thresholds the design issues will become increasing difficult and require greater and greater expertise in the "art of PCB" layout. It is my impression that much of the difficulty will arise in this area as speeds increase. A greater portion of time will be devoted to isolating PCB trace related issues. In my opinion, challenging though this project may be it is worthy of pursuit and if we begin with the marginally achievable and build upward we can satisfy the requirements of many more users.

Welcome Imedia,
You are right, but I (personaly) can not achive in all points...
You see; its possible to have lot of problems with absolutly routine circuits too_you can read over their much topics at Eda too...
In my opinion its so; if you select i.e. an ADC of 100 Msps; your real bandwidth will be at 10Mhz, this is todays not a very interesting bandwidth and if you are investigated lot of time & many_I mean, it must have enough "profit", than a 100Msps digiscope is for good money to buy, their are no more expensive...
I belief, that the analog part of the circuit is relative not so much complicated_OK, trigger circuit is surly a problem, but Amps & Attenuatros + ADCs less.
OK, I have "some years" experiance in high-speed designs_up to over GHz logic and analog pulse techniques, with PCB routings too.
In all cases I can learn some new too...
I have in my home lab analog scope for i.e. ca. 400MHz bandwidth_its enough "high" for me, but so a project will me interest
I think its to drive with two versions; even
1, at 100Msps system,
2, at 1Gsps system.
So designed the project; are the most interested persons serviced.
Most important is in my opinion in all cases the signal handling software:-(, but these isnt my profession...
K.

The problem with a design like this is data management, at 1GS/second, you have 1gxnumber of bits data to take care of, say you have an 8 bits ADC, then your data glow will be 1Gbytes/second, you maybe able to get to this kind of speed with modern FPGAs by immediately expanding the data from 8bits to 32 bits and decrese your required clock for data flow to 250MHz which is manageable.

But, that is only part of the problem, the PCB design of such board will be really tough and expensive, for a hubby project, this can be an expensive experience, but duable.

/Farhad Abdolian

I've also thought about building such a beast in the past... However, I have not yet started but I will definitely participate.

As far as I understood the problematic part is the input jack to the ADC. What about building the project in different parts?

Complex analog part + FPGA part + µP part? + PC Software part

I think the ad9480 will be perfectly suited for this app...

I would love to write the Software part of the scope.. I think of a software that could be run on an embedded system or alternatively on a PC for hardware implementation without display as well.

Best Regards,
Manuel

the SW and the analog part are the most difficult ones. You can easily use a spartan6A development board for the digital part of the design, it has everything necessary for such project.

Hi,
I have some complexer high-speed mixed boards, inclusive low level broadband sensing, in my experience_maybe can see for the analog part "up to outputs" of ADC?
K.

I have the low frequency design in VHDL already, I am using my Spartan 3e board with a 60MSPS ADC and some analog parts to use as a low level oscilloscope, what I don't have is any kind of SW to connect it to the PC and control it, right now, it is all controlled via UART. If you want we can work together on this.

Do you know pls_or some body_ is spartan6 from PCAD 2006 supported?
Tnx in advance!
K.

I don't understand what you mean? You can create your own symbols in a day or so.