EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: Cicada on December 05, 2018, 08:43:14 am
-
Hello
The Tektronix MSO 2024 have a sampling rate of 1GS/s and a bandwidth of of 200MHz. So that gives you 5 samples on a 200MHz sinusoidal wave.
Now
The Tektronix TBS1052B-EDU have a sampling rate of 1GS/s and a bandwidth of of 50MHz. So that gives you 20 samples on a 50MHz sinusoidal wave.
See page 1 in the manual. https://docs-emea.rs-online.com/webdocs/12bb/0900766b812bbde9.pdf
The question is why the reduction in bandwidth in the TBS1052B model if they both sample at the same rate.
Is it just product diversification to get different price points with the same hardware? i.e. analog front end bandwidth is deliberately limited
Or is there something else going on here.
If it is just an analog front end bandwidth limit it begs the question. How difficult is it to mod the scope to increase the bandwidth of the front end.
Thank you
-
If it's like the TDS it's probably software controlled. You may be able to upgrade it by patching firmware. But would require lots of reverse engineering.
See TDS thread here for the steps people went throught to get more bandwidth out: https://www.eevblog.com/forum/testgear/tektronix-tds1000b-series-hacks/ (https://www.eevblog.com/forum/testgear/tektronix-tds1000b-series-hacks/)
-
As the bandwidth is, ultimately, determined by the quality of the oscilloscope's vertical input/amplifier circuits (for repetitive waveforms), the TBS1052B-EDU, unless software limited, may simply have a cheaper 'Y' input stage.
-
You can always filter bandwidth away -- at any step in the chain, analog or digital -- but you can't put it back.
AFEs aren't hard to build since there are commodity chips for that (out to about 500MHz at least). Qualifying them can be tricky, since you need to check lots of conditions to make sure it's behaving, with consequences including tedious and time-intensive layout tweaks (or expensive PCB-level simulations). Or you can correct it digitally (instead of actually cleaning up the AFE, do it cheap and dirty and apply an inverse filter in the digital signal chain -- a calibration which can be even done per PCB, at test, just to make things that much scarier to the quality folks! ;) ).
It's always bugged me that modern budget scopes are almost exclusively ~1GSps and offer a tiny fraction of that in analog bandwidth, without any particular reason for it. Equivalent time sampling gets you the same bandwidth with a fraction of the ADC cost (unless it happens that 1G is a sweet spot; offhand, I see 8b 100MS for $10 on DK, then 8b/200M for $18, 12b/250M for $42, 8b/500M for $50, 8-16b/1G-500M for $111, and 8b/3G for $265; not that single-qty DK pricing is all that relevant in production, anyway). ET is perfectly fine for a great many things; it is rare that you need single shot acquisition at very high rates.
So, I suspect it's more about marketing. The oscilloscope version of Intel's late-00's push to ever-more-ridiculous clock frequencies, damn the power consumption and actual capability.
That said, I wonder if anyone has interest in cranking up the sample rate even more; there may simply not be <8b ADCs out there (flash type or otherwise) that are enough times faster to compensate for the ENOB, while being competitive with the current favorites. It would be perfectly possible to run at 3 or even 6GSps, with maybe 4 or 5 bits of sampling, and filter that down to a few 100MHz of analog bandwidth with an ENOB comfortably >= 8 bits.
I suppose the one thing with scopes is, you can actually see what it's capable of, so it's much harder to cheat the end user that way. Likewise, ET probably feels like a cheat, even though it's a perfectly valid and useful technique.
Go figure.
Tim
-
Equivalent time sampling gets you the same bandwidth with a fraction of the ADC cost (unless it happens that 1G is a sweet spot; offhand, I see 8b 100MS for $10 on DK, then 8b/200M for $18, 12b/250M for $42, 8b/500M for $50, 8-16b/1G-500M for $111, and 8b/3G for $265; not that single-qty DK pricing is all that relevant in production, anyway).
If ETS was cheaper, they would be using it in low end oscilloscopes. Implementing ETS requires a parallel analog signal conditioning path including analog pick-off and channel switching and a time-to-digital converter which is more expensive than the higher sample rate digitizer required for digital triggering once programmable logic is being used to process the output of the digitizer anyway.
Parts like the HMCAD1511 used in the Rigol DS1000Z series yield 1GS/s on one channel or 250MS/s on four channels for less than $50.