Recommendations for an oscilloscope

User avatar
jerkstore

05 Oct 2019, 19:39

Hi Guys,
I have been meaning to buy myself an oscilloscope for a while now. Right now I have a 4-channel Saleae logic analyzer (1-channel of analog) which I got secondhand for a very reasonable price. This has served be well but I'm not sure it's fast enough / high enough resolution to get accurate measurements in some cases.

These days I have been playing a lot with Cortron / ITW Magnetic Valve switches, trying to write a keyboard firmware. Through talking with users (thanks @listofoptions) and by messing around with CommonSense (thanks @DMA), I think I have a basic understanding of how the switches should work. However, I'm not able to pinpoint the appropriate voltage thresholds you would need to see reproducible results.

I'd like to use the scope for this project, but also for other projects in the future as well.

I'm not really sure what I should be looking for in a scope. I don't have an electronics background so my shopping so far has been pretty uninformed.

Are there any nice options that would fit my use case well (i.e. something fast/accurate enough for keyboard stuff but not wildly overkill either)? Any recommendations would be really appreciated!

DMA

05 Oct 2019, 20:57

1054Z hacked to 1104Z is barely adequate for this.
If I would've known - I would probably go for 5074, although $1000 is A LOT for the tool you won't be using to earn money with.

I can publish KitProg-based cortron MCU replacement which drove the 74xx logic if you want to (I thought it didn't work reliably because 74 series officially doesn't work below 4.75V - but it was because of contacts oxidation).
It doesn't use any of the programmable hardware and just bit-bangs everything, so should be easy to convert to your favorite MCU.

If your favorite MCU doesn't have good (<50ns trigger time) comparators, fast VDACs _and_ an internal analog MUX - you'll be much better off just replacing the original MCU, leaving all driving and sensing logic alone. You don't need a scope for this.

User avatar
jerkstore

05 Oct 2019, 21:29

DMA wrote:
05 Oct 2019, 20:57
1054Z hacked to 1104Z is barely adequate for this.
If I would've known - I would probably go for 5074, although $1000 is A LOT for the tool you won't be using to earn money with.
Thanks for the insight on scopes. That perspective is really helpful and maybe my expectations of a cheaper unit are unrealistic. If nothing else, it gives me a starting place to shop around if I do want to pick something up.
DMA wrote:
05 Oct 2019, 20:57
If your favorite MCU doesn't have good (<50ns trigger time) comparators, fast VDACs _and_ an internal analog MUX - you'll be much better off just replacing the original MCU, leaving all driving and sensing logic alone. You don't need a scope for this.
This is something I probably haven't given enough consideration. I think because I'm barely a software guy and not at all a hardware guy, it seems a bit intimidating. I pulled the row driving chips and the column sensing chip off the board but I'd like to put them back and study it more. Maybe it will not be such a difficult task for me after all. Thanks again.

DMA

06 Oct 2019, 02:27

Sorry for leading you into expensive(-ish. Expensive ones are $15k+, frequently +++) scopes land.
Handheld Hantek is not significantly worse than Rigol and costs much less. There are USB scopes - they are not horrible either for signals with wavelengths longer than your room.
EEVblog is a good place to look at current state of oscilloscope-building.

It's just this particular task involves things like this:
Image

1 horizontal divider is 100 _nano_seconds. It's, like, SHORT. Light travels 30 meters in that time.

Yellow is the trigger (and totally uninteresting), blue line is comparator output (there's latch downstream so you don't have to read it exactly at that moment - I just want to show you how short it is), and violet is positive comparator input as sensed by 1054 - and I have a huge hunch those nice and tidy sinusoidal oscillations are a quantization/postprocessing artifacts - so I found a guy with a better scope and will confirm/deny that next week.

I'll push the "brain replacement" code this weekend - probably even today. Writing the docs for the inductive CS.

May I ask why you intend to walk the path of your own firmware? It's not a walk in the park, especially when 20ns pulses are involved.

What is it that's missing from QMK which led you to this path, being barely a software and not at all hardware guy?

I have my own reasons to go there - xwhatsit's cost(ironically I could have been bought 10 beamsprings in 2016 with money it ultimately took), sheer clowniness of QMK's "NKRO" USB descriptors, necessity to reflash whole firmware to change one key in layout..
There are more - but what are yours?

PS: If you're within 500mi of Seattle or in Bay Area - I can lend you my 1054. I don't have much use for it currently.

DMA

11 Oct 2019, 09:06

OK. I spent ~12 hours today with a Tek MSO58. It is a WONDERFUL tool.

But turns out rigol 1054 is not that bad! The mayhem on the wire is indeed there, and rigol shows it pretty accurately (I'll post oscillograms in the MagValve docs section of the CS).

8 channels is an AWESOME thing, but far from a must. As is 15" display and RGB-highlighted vertical knobs.
6.25GS/s/channel is a good thing too, but again, not really a must. As is 1920x1080 screen resolution.

Although Rigol could've installed less cheap rotary encoders for the knobs. Tek's knobs are much more pleasant. But then again, $350 vs $35600..

User avatar
XMIT
[ XMIT ]

11 Oct 2019, 22:54

Subscribed! I'd love to dig into the electrical characteristics of Cortron magnetic valve switches.

DMA

12 Oct 2019, 02:58

I also, just because the device was there and there was a 12GHz diff probe, used a 25GHz Agilent for a bit, to record what happens on the switch.

I have pictures, will post after PTO. Picosecond resolution is nice but there's nothing special there, everything looks just as it looks in Tek.
But agilent SUCKS. The interface is openly hostile.

DMA

12 Oct 2019, 03:18

Tek (GPIO, pressed key, non-pressed key)
Tek001_008.png
Tek001_008.png (106.54 KiB) Viewed 78 times
Agilent (diff probe across key, end of the driving pulse)
Screen_2019-10-03000047.png
Screen_2019-10-03000047.png (90.08 KiB) Viewed 78 times
And this is a signal propagation delay - GPIO, PCB edge, mid-row and end-row. Yep, 8 nanoseconds delay. Somehow row tail responds before mid-row - but that's not an artifact, I've checked and rechecked connections and it indeed is.
Tek001_004.png
Tek001_004.png (106.03 KiB) Viewed 78 times
So, if anybody has an interesting non-conventional keyboard and wants to know what really happens within the matrix - ship it to me (or give it to me if you're in Bay Area or Seattle), I'm visiting the office which has this lab about about every 2 months, and that lab has some serious instruments.

PS: and yeah, 1054Z is perfectly fine even for this kind of job. Just don't get the two-channel version - remember, one channel is wasted on triggers anyway.

Post Reply

Return to “Workshop”