New Tektronix TBS2000 oscilloscopes

[Wuerstchenhund]

Other interesting features (promised for the future) are the 16 bit high-res mode when most competitors only go to 12 bit (or less).

R&S offers 16bit (software) HiRes modes for RTE/RTO as option, but even on these scopes with fast ADCs the benefit over 11bit or 12bit modes is miniscule.

[linearphase]

Other interesting features (promised for the future) are the 16 bit high-res mode when most competitors only go to 12 bit (or less).

R&S offers 16bit (software) HiRes modes for RTE/RTO as option, but even on these scopes with fast ADCs the benefit over 11bit or 12bit modes is miniscule.

I would be surprised if the s/n ratio ratio ever would approach 16 bits also. High res  as implement in TEK scopes is really a type of boxcar integrator. It uses the extra samples that the A to D can produce beyond what is necessary for a given sweep speed and record length to compute a sort of a horizontal or spacial average. Thus the degree of improvement over non high res is most at lower (requested) sample rates and decreases to nothing when the sample rates approaches the max the scope can do.
For example if the sweep speed and record length you are using calls for 100MS/s and the scope can do 1 GS/s then you have ten times more samples than you need. What do you do with those samples?

In sample mode-nothing you simply display the 1st point
In high res mode-you add the the samples together calculate an average and put the average into the bin for display
In peak detect you keep the highest and lowest values and throw away the rest.

Thus at slow sweep speeds you have lots of extra samples and this increases both resolution and s/n a large amount. At fast sweep speeds where 1 GS/s is called for their is no improvement when using hi res.

This is greatly oversimplified but should illustrate the point.

[linearphase]

Judging by your comments you appear to be something of an industry expert.

I didn't represent that TEK currently has 50% DSO market share. I could not possibly know that. What I said was the last time I saw a market share report, and I only saw 1 independent report, that was the number I saw.

I do know there was essentially NO DSOs made at that time except those made by HP, TEK, a few Lecroy some by Gould and some by Nicolet. That tells me this was before Lecroy seriously entered the scope market.

Which LeCroy did in 1981 and back then there wasn't really a large market for DSOs anyways. My data doesn't go so far back but irrespective of Teks marketshare being 50% or not it's a lot easier to become dominant in an emerging market than in an established one, which is what the DSO market is today.

Now the market is loaded with cheap Chinese imports and I would expect that they would dominate the unit volume but probably not the dollar volume.

I doubt that they dominate the unit volume, at least not for units sold under their own brands. The hobbyist market seems to love them but in the commercial field they are still pretty much non-existent.

Since TEK became a part of Danaher in 2007 sales are no longer reported externally so no one outside of TEK/ Danaher really knows what their sales are. Indeed even employees below top management were not provided with sales information except in a general sense. I did hear some numbers from time to time that were " leaked". They seemed lower that what I though they should be. I can tell you that in my territory though I did annual market share analysis. I do not claim absolute precision but I had been in the territory for 31 years so I pretty much knew what everyone had. In my last year (2011) with TEK I estimated scope market share at around 40%. At that time I do not have responsibility for low end scopes- those below about $15K and so my numbers did not include them. My Agilent counterpart agreed with this analysis. He admitted that we were kicking his ass. I did not even know who the Lecroy guy was... Your stated market share in analog agrees with numbers I have heard, but did not see. I do not know how other territories or other parts of the world fared.

That's pretty much the opposite from what I have seen and also what our data shows. Tek sales have been dwindling since at least 2004, and in 2011 I very much doubt their market share was even close to 40%, much  less so for the mid-range and high-end segments, both areas where Tek hasn't really been competitive for ages. Things were a bit better in the low-end, predominantly thanks to the edu segment, which Tek courts in the hope that it hooks more people to their brand.

And frankly, that's not surprising. Unlike their analog scopes, Tek DSOs were rarely anything to write home about. Entry-level scopes with ridiculously low memory, mid-range and high-end scopes that were limited in performance and capabilities compared to its competitors while in general being painfully slow ("like wading through molasses" is a term often associated with TDS5000/6000/7000/70000 scopes, and rightfully so). Then there are some annoying limitations (like the DPO "high-speed" mode where measurements are disabled, or the mentioned memory sizes), and some really daft ideas (like the LCD shutter on early TDS scopes). To make matters worse, instead of coming up with some innovative new products Tek pretty much continued to push their stale products in a warmed-up form.

Tek may have had some interesting scopes like the 11000 Series back in the mid-'80s, and later introduced the lunchbox format and came up with intensity grading (on which HP and LeCroy worked as well), but that was back in the '90s. Technical advances of DSO technologies since then happened elsewhere (which isn't surprising because Tek has long lost most of its talent, not just thanks to the mind-numbing DBS).

Today Tek is widely considered the bottom-of-the-barrel of the big brand scope manufacturers. The only ones buying Tek (aside from the edu market thanks to strong incentives) are people that don't know what else is out there, often with fond memories from back in the analog scope days (or they just spend someoneone else's money and thus don't care). And even then the preferrence only lasts until they have tried a modern Tek scope and scopes from other manufacturers.

Of all the labs we support, only one still buys Tek scopes, and this only for contractual reasons (limited to specific pieces of kit). We occasionally invite Tek for evaluations we do before making a procurement decision, but they regularly don't get the sale simply because their products weren't good enough and the pricing too unflexible.

Mind you that this is limited to scopes, not other categories of test instruments. And not all current Tek products are as poor as their scopes. And in regards to scopes, at least that new TBS2000 looks like a decent and well thought out unit, at least based on the data that is available so far. I'm sure the edu market will love it.

If you have current market share estimates those would be interesting.

Sure it would be interesting, we're often asked by manufacturers to provide access to our data. But that's not gonna happen.

I agree with some of your points and disagree with others. This is interesting to me. If you you would like to continue the discussion I am up for it but we should probably take it off line as we are getting off topic.

[nctnico]

Other interesting features (promised for the future) are the 16 bit high-res mode when most competitors only go to 12 bit (or less).

R&S offers 16bit (software) HiRes modes for RTE/RTO as option, but even on these scopes with fast ADCs the benefit over 11bit or 12bit modes is miniscule.

I would be surprised if the s/n ratio ratio ever would approach 16 bits also. High res  as implement in TEK scopes is really a type of boxcar integrator. It uses the extra samples that the A to D can produce beyond what is necessary for a given sweep speed and record length to compute a sort of a horizontal or spacial average.

Thus at slow sweep speeds you have lots of extra samples and this increases both resolution and s/n a large amount. At fast sweep speeds where 1 GS/s is called for their is no improvement when using hi res.

I'd be wary of putting too must trust into high-res modes other than doing unspecified filtering on the signal. For getting more bits from an A/D converter a lot depends on the linearity of them (and they are probably barely linear enough to support 8 bit resolution) and the amount of noise available to do oversampling. Think about feeding the scope with a very accurate and low noise DC signal. I doubt it can resolve beyond 8 bit. On the Tektronix TDS744A I used to own I could get the weirdest interpolation results in high-res mode with the bandwidth limiting on.

[linearphase]

I would be surprised if the s/n ratio ratio ever would approach 16 bits also. High res  as implement in TEK scopes is really a type of boxcar integrator. It uses the extra samples that the A to D can produce beyond what is necessary for a given sweep speed and record length to compute a sort of a horizontal or spacial average.

Thus at slow sweep speeds you have lots of extra samples and this increases both resolution and s/n a large amount. At fast sweep speeds where 1 GS/s is called for their is no improvement when using hi res.
[/quote]
I'd be wary of putting too must trust into high-res modes other than doing unspecified filtering on the signal. For getting more bits from an A/D converter a lot depends on the linearity of them (and they are probably barely linear enough to support 8 bit resolution) and the amount of noise available to do oversampling. Think about feeding the scope with a very accurate and low noise DC signal. I doubt it can resolve beyond 8 bit. On the Tektronix TDS744A I used to own I could get the weirdest interpolation results in high-res mode with the bandwidth limiting on.
[/quote]
The user is always in charge off his results. Hires in my experience works great. You can think of it as low pass filter. The approximate cut off frequency is sample rate divided by 2. But the response has a sinx/x characteristic. So if you use a 20MHz BW limit the sample rate would have to be around 50Ms/s to produces the best signal to noise ratio. If you signal had frequency components above 20MHz that were really strong this could cause the scope to alias at this low of a sample rate and that could cause the problems you describe. This is because BW limit filters on scopes are NOT brick wall. Once again, most scopes have a manual transmission. You have to shift the gears at the right time or bad things can happen. Judging by your tag line you, like many of us, have had that problem.
I am more than happy to explain ways to avoid this that might help you. With experience DSO's can be largely prevented from deceiving you!

One last point, there is always enough noise to dither the signal and make the Hires mode work. Tek does do dynamic characterization of the scope as a system in all modes, including hi res.
The 12 bit spec that you see so often is the result of ENOB testing that calculates the effective number of bits using well defined waveforms. You are right though about the " unspecified filtering" . This could be calculated by the scope but I know of no manufacturer that actually does this. So you are left with the rule of thumb mentioned above. This rule of thumb may not apply to other manufacturers as I don't know how they implement hires. In that case it is better to ask them what rules to use.

[David Hess]

The back shot shows it has ethernet socket. I think the wifi dongle is liable to go missing in a classroom. Hopefully generic chinese ones can replace it.

At this price you'd think WiFi would be built in - they're really pushing it as a selling point and it's only $1 in extra hardware. Adding the extra USB port at the back probably costs them nearly that much.

Or they wanted the capability to unambiguously disable wireless access for security reasons or wanted easy upgradability since WiFi standards progress so quickly.  I would worry about someone walking off with the WiFi dongle though.

In a classroom environment, I would hope that the wired ethernet is used instead.

Other interesting features (promised for the future) are the 16 bit high-res mode when most competitors only go to 12 bit (or less).

R&S offers 16bit (software) HiRes modes for RTE/RTO as option, but even on these scopes with fast ADCs the benefit over 11bit or 12bit modes is minuscule.

How do they handle the INL of the 8 bit digitizer though?  I would expect the measurement accuracy to be at best 8 bits and significantly less at higher frequencies where the ENOB drops no matter how the samples are processed into each bin without heroic self calibration.  Delta-Sigma ADCs avoid multibit quantization for a good reason if INL is important.

I see that nctnico is asking the same question.

I would be surprised if the s/n ratio ratio ever would approach 16 bits also. High res  as implement in TEK scopes is really a type of boxcar integrator. It uses the extra samples that the A to D can produce beyond what is necessary for a given sweep speed and record length to compute a sort of a horizontal or spacial average.

The signal to noise ratio is improved tremendously but the nonlinearity of the 8-bit digitizer trashes the SFDR and high resolution measurement accuracy which nctnico observed with his TDS744A.

Boxcar averaging is the worst common FIR filter but it does have the virtue of only using 0 and 1 coefficients making it easy to implement in hardware for maximum performance.

High resolution mode is a great simple feature for lowering noise to produce a clearer display at the cost of bandwidth though.  Just do not be fooled into thinking that DC measurements are improved as well.

[nctnico]

The user is always in charge off his results. Hires in my experience works great. You can think of it as low pass filter. The approximate cut off frequency is sample rate divided by 2. But the response has a sinx/x characteristic. So if you use a 20MHz BW limit the sample rate would have to be around 50Ms/s to produces the best signal to noise ratio. If you signal had frequency components above 20MHz that were really strong this could cause the scope to alias at this low of a sample rate and that could cause the problems you describe. This is because BW limit filters on scopes are NOT brick wall. Once again, most scopes have a manual transmission. You have to shift the gears at the right time or bad things can happen. Judging by your tag line you, like many of us, have had that problem.

I'm very well aware of aliasing problems! It is just that you can't create extra bits from an AD converter in a meaningfull way in an uncontrolled environment like an oscilloscope.
BTW my tag line is more about the effects of limited bandwidth + probes which by definition make what you see on your screen an interpretation of the reality within the limits of your test setup.

[linearphase]

The back shot shows it has ethernet socket. I think the wifi dongle is liable to go missing in a classroom. Hopefully generic chinese ones can replace it.

At this price you'd think WiFi would be built in - they're really pushing it as a selling point and it's only $1 in extra hardware. Adding the extra USB port at the back probably costs them nearly that much.

Or they wanted the capability to unambiguously disable wireless access for security reasons or wanted easy upgradability since WiFi standards progress so quickly.  I would worry about someone walking off with the WiFi dongle though.

In a classroom environment, I would hope that the wired ethernet is used instead.

Other interesting features (promised for the future) are the 16 bit high-res mode when most competitors only go to 12 bit (or less).

R&S offers 16bit (software) HiRes modes for RTE/RTO as option, but even on these scopes with fast ADCs the benefit over 11bit or 12bit modes is minuscule.

How do they handle the INL of the 8 bit digitizer though?  I would expect the measurement accuracy to be at best 8 bits and significantly less at higher frequencies where the ENOB drops no matter how the samples are processed into each bin without heroic self calibration.  Delta-Sigma ADCs avoid multibit quantization for a good reason if INL is important.

I see that nctnico is asking the same question.

I would be surprised if the s/n ratio ratio ever would approach 16 bits also. High res  as implement in TEK scopes is really a type of boxcar integrator. It uses the extra samples that the A to D can produce beyond what is necessary for a given sweep speed and record length to compute a sort of a horizontal or spacial average.

The signal to noise ratio is improved tremendously but the nonlinearity of the 8-bit digitizer trashes the SFDR and high resolution measurement accuracy which nctnico observed with his TDS744A.

Boxcar averaging is the worst common FIR filter but it does have the virtue of only using 0 and 1 coefficients making it easy to implement in hardware for maximum performance.

High resolution mode is a great simple feature for lowering noise to produce a clearer display at the cost of bandwidth though.  Just do not be fooled into thinking that DC measurements are improved as well.

Agreed. Isn't the high res filter implemented in hardware so it is essentially processing samples as fast as they come in?

[Wuerstchenhund]

I would be surprised if the s/n ratio ratio ever would approach 16 bits also. High res  as implement in TEK scopes is really a type of boxcar integrator.

Yes, as on pretty much every other scope that offers such mode, aside from LeCroy (it's ERES function is actually a bellcurve FIR filter, not just a simple boxcar filter).

I'd be wary of putting too must trust into high-res modes other than doing unspecified filtering on the signal. For getting more bits from an A/D converter a lot depends on the linearity of them (and they are probably barely linear enough to support 8 bit resolution) and the amount of noise available to do oversampling. Think about feeding the scope with a very accurate and low noise DC signal. I doubt it can resolve beyond 8 bit. On the Tektronix TDS744A I used to own I could get the weirdest interpolation results in high-res mode with the bandwidth limiting on.

Indeed, plus it's prone to ringing. It's also an destructive mode, i.e. you lose the original sample data. At least with ERES I can do other stuff with the original waveform alongside the ERES trace.

But the processing requirements for ERES are way to high for an entry-level scope.

[Wuerstchenhund]

I agree with some of your points and disagree with others. This is interesting to me. If you you would like to continue the discussion I am up for it

Any time but then probably in a separate thread.

but we should probably take it off line as we are getting off topic.

Agree.

[Hydrawerk]

Tektronix should do an upgrade of DPO2000 / MSO2000 with that 480×240 WQVGA LCD.
http://www.tektronixoverachiever.com/

[timb]

Tektronix should do an upgrade of DPO2000 / MSO2000 with that 480×240 WQVGA LCD.
http://www.tektronixoverachiever.com/

Yes, despite their age they are nice little basic scopes. I have an MSO2024B as my daily driver. A bigger screen *would* be nice!

[David Hess]

The signal to noise ratio is improved tremendously but the nonlinearity of the 8-bit digitizer trashes the SFDR and high resolution measurement accuracy which nctnico observed with his TDS744A.

Boxcar averaging is the worst common FIR filter but it does have the virtue of only using 0 and 1 coefficients making it easy to implement in hardware for maximum performance.

High resolution mode is a great simple feature for lowering noise to produce a clearer display at the cost of bandwidth though.  Just do not be fooled into thinking that DC measurements are improved as well.

Agreed. Isn't the high res filter implemented in hardware so it is essentially processing samples as fast as they come in?

That is exactly how this type of high resolution mode is done although it sounds like some R&S DSOs are more sophisticed.  Peak detection and DPO histograms operate in the same way; samples are processed in real time before being stored in the acquisition record or acquisition histogram in the case of DPO operation.

This requires custom logic to get enough performance but some early DSOs with low sample rates used discrete SSI logic to accomplish it so if you want an example with full documentation of peak detection done this way, check out the Tektronix 2230 service manual.  The faster 2232 with 5 times the sample rate moved this function to a custom ASIC which also operates as multiplexer between the digitizers and memory.

It has been a long time but I do not remember a set of FIR coefficients which would produce acceptable results compared to boxcar averaging using only shifts.  Fixed point processing performance is probably good enough to make a real FIR filter but I doubt the improvement would be worthwhile considering the extra complexity and power requirements.

[Hydrawerk]

I like that big Multipurpose knob of TBS2000. It is much better than Keysight, Rigol or Siglent.

[Muxr]

Looks nice. I always liked that Tek color. Would love to see Dave take it apart in a video, the price doesn't seem bad either.

[Hydrawerk]

For schools I prefer DSOX2000 / DSOX3000 series. These scopes are used in industry, too.

[Wuerstchenhund]

For schools I prefer DSOX2000 / DSOX3000 series.

Why? Both are very expensive (especially the DSOX3k Series), the DSOX2k comes with a silly 1Mpts memory, and both don't even come with something so basic as a LAN port (and forget about WiFi).

These scopes are used in industry, too.

*Any* big brand scope is used in industry. Simply because that is where they customers are.

[pascal_sweden]

I like that big Multipurpose knob of TBS2000. It is much better than Keysight, Rigol or Siglent.

Just one nice button does not justify that you have to break the bank to buy this scope.
Worst case, just add yourself a bigger button on your Rigol or Siglent oscilloscope.

Probably that big button is about the only good thing about that scope. The dual color handle looks ugly. The feet make it look like a horse that is about to leave the stable. The button placement is completely different on the 4 channel series than on the 2 channel series, so if they have both scopes in the university, students will have to learn two separate layouts, in case for some labs they have to use the 2 channel scope and for other labs they have to use the 4 channel scope. Very smart move Tektronix!

[Hydrawerk]

the DSOX2k comes with a silly 1Mpts memory, and both don't even come with something so basic as a LAN port (and forget about WiFi).

This might often be no problem. I did not see any scope with LAN being used at my university.

[Wuerstchenhund]

the DSOX2k comes with a silly 1Mpts memory, and both don't even come with something so basic as a LAN port (and forget about WiFi).

This might often be no problem.

True, but you're still paying a premium for a scope with miniscule memory.

I did not see any scope with LAN being used at my university.

Maybe a missed opportunity? But then it well be because Keysight wants extra $$$ for that interface.

[Hydrawerk]

Yes, if you want a low-cost middle range scope with a long memory (at least 5MPoints), you should stay away from Keysight.
Infinii Vision 6000X has 4Mpoints, but

Maximum analog channels record length
? 2 GSa/s 4 Mpts half channel interleaved, 2 Mpts all channel
> 2 GSa/s 1 Mpts half channel interleaved, 500 kpts all channel
Maximum digital channels sample rate 2 GSa/s half pods interleaved, 1 GSa/s all pods
Maximum digital channels record length 4 Mpts half pods interleaved, 2 Mpts all pods

http://literature.cdn.keysight.com/litweb/pdf/5991-4087EN.pdf?id=2456396
http://www.keysight.com/en/pcx-x205214/infiniivision-6000-x-series-oscilloscopes?nid=-55355.0&cc=CZ&lc=eng
This is a 15000USD scope. 

OK, for a long memory you have to buy this 21000USD beast. http://www.keysight.com/en/pcx-x205213/infiniium-s-series-oscilloscopes?nid=-55369.0&cc=CZ&lc=eng
http://www.testmart.com/estore/unit.cfm/DIGOSC/KEYSIG/DSOS104A/390339015/8.html

Keysight has no scope like Tek MDO3000. It is not expensive as fuck, but has 10Mpoints per channel.
http://www.tek.com/oscilloscope/mdo3000-mixed-domain-oscilloscope

[pascal_sweden]

It's only normal that an A-brand scope costs more than a Chinese scope.

But it would be more acceptable if the cost factor is 2x, and not 4x or 5x.

[Lukas]

It's teardown time! Since none of has the hardware at the moment, let's take a look at the firmware. As to be expected, this scope runs linux as well as Tek's other non-windows scopes. Looking at the firmware update, things are a bit different than they used to be. Instead of some ext2 image, you now get a "TBS2KB.TEK" file.
Binwalk tells us that there's a Cramfs at offset 340 in there, looks interesting. Unfortunately, mounting the cramfs image doesn't work as does using fsck.cramfs for unpacking. After staring at the blob and the cramfs documentation I figured out that tek for some reason decided to hack cramfs in interesting ways. Instead of the standard 12 byte inodes, this one's got 16 byte inodes, 4 bytes more! I couldn't patch fsck.cramfs to understand tek's crazy format, so I came up with this http://pastebin.com/BW7nYDWi minimum-brains python script that doesn't understand paths and simply extracts every file it finds in one directory.

Some notable things on the extracted files:
 - There are two monster executables, ULPP and tekapp, totalling over 40MB. Fore some reason, they left the map file for the ULPP binary in the image.
 - It really seems to be based on the AM33xx, there's a programm called gpmcdma
 - There are many libraries referring to other tek instruments,
 - There's a 16MB big Opentype font, seems a bit like an overkill

I don't have the time to take a closer look at the extracted files right now, maybe someone of you is interested...

There's no indication whatsoever on the tek website of them integrating GPL'ed software in there products I had hoped that tek would do better than some random chinese CCTV camera manufacturer...

[pascal_sweden]

How many points are used in the FFT? In the video from Tektronix it says 2000 points.
In that respect the FFT functionality isn't better than the one you can find in Siglent or Rigol.

[tautech]

How many points are used in the FFT? In the video from Tektronix it says 2000 points.
In that respect the FFT functionality isn't better than the one you can find in Siglent or Rigol.

Not for long.