New Tektronix scope - MSO 6 series

[fonograph]

Starts at 24.000$ for 1GHz 4channel.What I hate about this scope is that it uses this stupid FlexProbe rubbish.What is funny is that the last scope they released,the MSO 5 series had 8 channel option,this one is only 4 max.What this means is that you can have at maximum a 2 oscilloscope channels and 16 digital channels at once becose each of the retarded FlexProbe digital probes have only 8 digital channels while taking away the normal oscilloscope channel.Who ever came up with this FlexProbe crap deserves to be run over by ISIS tank,slowly,while being forced to listen Justin Beiber at 140db SPL.

While they describe it as 12bit 25 Gs/s scope,at 25 Gs/s its only 8bit.Strangely,at 12.5 Gs/s it becomes 12bit.How is that even possible? I have no idea what is happening there,correct me if I wrong but you need 4X oversampling to gain 6db (1bit) extra dynamic range,to increase it by 4 bits or 24db with only 2X oversampling thats weird,I wish it was 12bit all the way to 8GHz like the new Lecroy wavepro HD.I wonder,is this some kind of sigma delta ADC with high order noise shaping?That would explain the sudden dynamic range change due to steep noise increase in high frequencies.

It have one feature I like very much,the anti aliasing lowpass filter can be switched between superior bandwidth and superior impulse reponse.Sadly,they dont show what type of filters are these,what order,phase response,what attenuation at nyquist,how the impulse response looks like,that will be nice if they added this information to datasheet.

The DC Gain accuracy spec is confusing as well,it goes from 0.5% at 1M ohm full scale,to 4% at 50 ohm 1mv/Div.I cant tell if that is good or bad,other manufacturers give just one number and dont show at what settings so I cant compare it.My guess is they just slap best number there and call it a day while Tektronix went the extra mile and gave us the more complex and truthfull data,I like that.I think there should be some standardized way to rate DC Gain accuracy so its easier to compare.

They claim it have some new uber ASIC with half milion update rate per second,I wonder how is the update rate at full memory lenght of 250 mega points and with things like zone triggering,decoding and FFT on.I dont care if it have 29374667482994764^16384 googolplex updates per second with everything turned off but one channel with couple kb memory lenght,I wanna know how fast it updates when features are turned on and capture memory is increased.

[Eric_S]

When I look at this, I get the feeling that it was rushed out the door to compete with the new LeCroy. Take a 8 channel 5 series, interlace the channels down to 4 and OC the ASIC a bit, and hey. (I haven't looked at the specs on detail, so I may be unfair here)

Anyway, cool scope.

[tmbinc]

I understand your frustration about FlexProbe, but it's just offloading architectural compromises to the user. The demuxers can handle 4x25GSamples/s, and you can either "invest" them into analog or digital channels. If they allowed 4 analog + 32 digital, they would need twice the amount of memory bandwidth, and clearly they didn't have budget for that. I understand that you think 4 channels are just not sufficient, but as everything it's a cost/feature trade-off.

(And to be honest, the MSO5xxx MSO feature was more than crappy as well as it didn't match the analog channels in record length and sample rate)

[rx8pilot]

Wonder if they will have an adapter to breakout the MSO and still have access to all 4 analog channels.

[fonograph]

I understand your frustration about FlexProbe, but it's just offloading architectural compromises to the user. The demuxers can handle 4x25GSamples/s, and you can either "invest" them into analog or digital channels. If they allowed 4 analog + 32 digital, they would need twice the amount of memory bandwidth, and clearly they didn't have budget for that. I understand that you think 4 channels are just not sufficient, but as everything it's a cost/feature trade-off.

(And to be honest, the MSO5xxx MSO feature was more than crappy as well as it didn't match the analog channels in record length and sample rate)

I dont understand you becose I am noob.What are these demuxers? I know what mux is,but I dont know what demuxer is.Does it work in way that each channel have 8bit ADC made up of 8 parallel 1bit ADCs,and when you hook up 8 channel digital probe it becomes 8 1bit 25 Gs/s ADCs?

What do you mean MSO 5 didnt match analog channel sample rate? Explain it in more detail.

[tmbinc]

"Demux" (in Tektronix terminology, not sure if this is standard) is the part between the ADC and the memory. I think this is now integrated into a single ASIC (finally...), but conceptually, it divides the "huge" ADC data rate (100 GByte/s in this case) into many different channels, so each channel can be handled with off-the-shelf memory. (See https://debugmo.de/2013/03/whats-inside-tektronix-dpo5034/ for example - I apologize for _still_ not having fixed the image links, use https://goo.gl/photos/dGy6WveTGwSdKgC4A directly if you want)

In the MSO/DPO5xxx (and DPO4xxx, and MSO4xxx and MDO4xxx), for example there are 2 ADC chips, each handling 10GS/s (from 1-2 channels); each ADC is connected to two Demux, each Demux has 4 memory chips. This means the data rate (20GByte/s in this case) is distributed to 16 memory channels; at 1.25GByte/s, running them at 800 MT/s @ 16-bit (DDR800) is sufficient to store the full rate. (The Demux does more - it can apply FIR filters, read back data etc.)

For the MSO5 and MSO6, Tektronix revamped their architecture and now has an integrated ASIC with ADC , Demux and memory controllers. I don't have more details than that, but overall it's still the same concept, just more tightly integrated.

So the sample rate bottleneck is not just the ADC, but also the memory bandwidth. More memory bandwidth would require more memory channels, and they are costly (mostly in terms of board space and ASIC pins).

With the ability to "bypass" the ADC, and directly feed the Logic signals to the Demux (slightly simplified - of course there are level shifter/comperators etc. involved), you can use the available bandwidth for either analog channels or digital channels. Each analog channel will replace 8 digital channels (which is the ADC depth at full speed). In either combination, the Demux are running at full rate. It's not matter of "breaking out" more pins. It's a matter of running into the primary (or secondary, if you count the ADC as primary) physical limitation of the device.

(Now of course we could be talking about running 4 Analog channels at half speed together with all digital channels at half speed, but - thankfully? - Tektronix didn't chose to go down that route.)


The MSO5xxx (not MSO5) MSO limitations are because the MSO is handled not by the main acquisition path (Demux -> DDR memory), but by a seperate path (Trigger ASIC -> plain old SDRAM), which is limited at 500MS/s (and has some other arcane limits).

[EEVblog]

Wonder if they will have an adapter to breakout the MSO and still have access to all 4 analog channels.

As I understand it, that's not possible, the underlying architecture they have chosen means the user is stuck choosing between analog or digital.

[rx8pilot]

Wonder if they will have an adapter to breakout the MSO and still have access to all 4 analog channels.

As I understand it, that's not possible, the underlying architecture they have chosen means the user is stuck choosing between analog or digital.

It carries are rather fat price tag to have such a limitation. 

[fonograph]

tmbinc Thank you for your detailed explanation,it was excellently written 

Hey everybody,does anyone understand whats up with that ADC bit number? Why it goes from 25 Gs/s 8bit to 12.5 Gs/s 12bit,that doesnt seem like conventional oversampling or interleaving.The new MSO 5 have same thing,how does that even work? Whats going on there? Why? I have never seen anything like this before,I have no idea why its like that,if someone knows please explain it to me.

[TheUnnamedNewbie]

Hey everybody,does anyone understand whats up with that ADC bit number? Why it goes from 25 Gs/s 8bit to 12.5 Gs/s 12bit,that doesnt seem like conventional oversampling or interleaving.The new MSO 5 have same thing,how does that even work? Whats going on there? Why? I have never seen anything like this before,I have no idea why its like that,if someone knows please explain it to me.

Don't know the details, but it could just be that their ADC's S/H simply doesn't track quickly enough to give you 12 bit of resolution at 25 GS/s. In addition, the noise at 8 GHz bandwidth might make it pointless to try and get 12 bits. I don't recall where the line is, but at some point you are limited more by clock jitter than by front end capability.


Regarding the digital/analog channel limitations: I'm not the target market, so I could be wrong, but I suspect that they assume few people will need 4x 8 GHz analog bandwidth and a large number of digital channels. But again, I could be wrong. And I seem to recall some people liking the system with the MSO5 as it gives you more flexibility for the price.

[tautech]

tmbinc Thank you for your detailed explanation,it was excellently written 

Hey everybody,does anyone understand whats up with that ADC bit number? Why it goes from 25 Gs/s 8bit to 12.5 Gs/s 12bit,that doesnt seem like conventional oversampling or interleaving.The new MSO 5 have same thing,how does that even work? Whats going on there? Why? I have never seen anything like this before,I have no idea why its like that,if someone knows please explain it to me.

On the X model Siglents that offer Eres 3 bit enhancement there is a sampling rate hit too and probably on scopes with Hires too. Dunno just how it works without study but it cleans up displayed noise to levels similar to averaging.

[Hydron]

Ugh, when they do the industrial design for a certain number of channels then release a model with fewer it almost always looks awful. This is not a dig at Tek specifically (see: e.g. 2-channel 3000 series Keysight models) and it's understandable wanting to save the cost of a redesign by sharing parts, but man, the empty space just looks completely terrible.
/rant

[fonograph]

Ugh, when they do the industrial design for a certain number of channels then release a model with fewer it almost always looks awful. This is not a dig at Tek specifically (see: e.g. 2-channel 3000 series Keysight models) and it's understandable wanting to save the cost of a redesign by sharing parts, but man, the empty space just looks completely terrible.
/rant

I agree but I think it doesnt matter much,its not made for test & measure device pageant competition,as long as it works I dont care,I would rather take scope looking like Rigol over fancy looking scope with FlexProbe cancer.

[nctnico]

Interesting move of Tektronix. I'd expected they introduced an MSO3 or MSO4 series but it seems their lower end portfolio still sells without needing an update.

[JPortici]

yeeaah a couple of months ago they told me to wait for 2019 *wink wink* but they forgot to mention this "MSO6" series.
Honestly, i saw it first in an AD from their facebook page. didn't receive any marketing email..

They also didn't mention this.
https://www.tek.com/oscilloscope/5-series-mso-low-profile

[Hydrawerk]

This DPO2000B is about 10 years old now. https://www.tek.com/oscilloscope/mso2000-dpo2000
Will they come up with a new scope of this price cathegory?
Tek still has no competitor to Keysight DSOX2000 series introduced in 2011.

[David Hess]

Wonder if they will have an adapter to breakout the MSO and still have access to all 4 analog channels.

As I understand it, that's not possible, the underlying architecture they have chosen means the user is stuck choosing between analog or digital.

It carries are rather fat price tag to have such a limitation. 

It would be even fatter if they included excess memory bandwidth to support it that most users did not require.  The alternative would be lowering the sample rate of the analog and digital channels.

Another way to look at it is that they are making the most of the limited memory bandwidth available for that price.

Don't know the details, but it could just be that their ADC's S/H simply doesn't track quickly enough to give you 12 bit of resolution at 25 GS/s. In addition, the noise at 8 GHz bandwidth might make it pointless to try and get 12 bits. I don't recall where the line is, but at some point you are limited more by clock jitter than by front end capability.

The tradeoff could also be in the ADC itself.  Some successive approximation converters had a "short cycle" mode to support a higher sample rate at a lower resolution with a similar tradeoff because smaller comparisons take more time than larger ones.  In the case of a flash pipelined converter, each stage has multiple aspects which limit accuracy at higher speeds; maybe they just drop the later stages which would be digitizing only noise anyway at the highest sample rate.

[snoopy]

It's a much higher sample rate than the MSO5 so I guess 4 channel is standard on the high end scopes.

Also these days does anyone need more than 8 channels of digital just to look at serial streams ?

[fonograph]

w2aew pls help! I need you bro      I have trouble understanding the MSO 6 datasheet,specifically the RMS noise in 50 ohm input at various bandwidths.The scope have one data for Hi Res mode and other for sampled mode.I know that Hi Res mode only works if you have bandwidth set to 4 GHz or lower,but you can turn off the Hi Res mode and still have lowered bandwidth.

That doesnt make no sense,why would anyone use this sampled mode if they can have hi res mode without any loss of frequency range? Also why does the ADC jump from 8 bits at 8 GHz to 12 bit at 4 GHz? Is it real 12bit ADC or 8 bit + oversampling to get 12 bits?

[rosin232]

I recently purchased one of these units MSO-64 1GHz with one digital probe it is pretty long lead time right now 6 weeks and so I'm still waiting.

Prior to my order I had a demo session with their sales and field engineers on campus to show me how it performs.

Pros and cons is IMO only and for type of work I'm doing also it was just an hour that I was testing it of non-production unit.

Pros:
* Individual 12bit ADC per channel
* Fast acquisition 25 GSPS per channel
* Low AC RMS noise
* All hardware is already built-in everything is upgradable with software key which is huge for me since I don't like people to poke at my scope if I need an upgrade.
    * AFG
    * Memory - 250M max per channel
    * BW upgrades are done via software code and they send you new sticker and you have to run SPC after that.
* Linux UI is very fast and smooth
* Quite and cool - surprisingly it was very quite and very cool high end scopes that I saw from other vendors were running pretty hot and loud.
* Huge touch screen display and scope is pretty short so can be used as a computer and won't take much space on your bench.

Cons:
* Sacrifice analog channel for digital
* Linux locked and not accessible
* Win 10 is an option
* UI - felt a bit raw/naked and uncooked
* 250MS/channel memory cannot be combined and a bit too little for 25GSPS scope especially when competitors offer 2GS/channel or 500MS/channel
* FFT was ridiculously basic very few configuration options(e.g. no way to select window size) and very basic visualization IMO it should be much better in such high end oscilloscopes especially if we are talking about HF domains > 1GHz good FFT is a must. I had a chance to compare it against R&S RTO series and their FFT is fantastic not without its own issues but visualization is very nice. Am I asking too much of RF features?
* Basic bus  (UART, SPI/I2C) protocol triggering and analysis upgrades are ridiculously expensive 1850$ as expensive as actual TLP058 digital probe
* Waveform refresh is slow when you set acquisition to 100M but want to use zoom and display only a chunk of it it will be still as slow as if they display 100M samples which shouldn't be the case. Their field engineer was trying to tell me that samples still have to go into memory and will take time ... I was like ok lol ...

In attachment baudline spectrogram and FFT example of 50ohm termination 1M sample with nothing attached. You can see nasty harmonics from ADC interleaving I think they have 4 x 6.25GSPS ADCs inside each of them but it was still pretty low noise and also QUAL and not production unit.

In the end I still went with it because of fantastic HW that they put in there and it will last me for long.

Disclaimer: These are my opinions only and they may be inaccurate also I'm not associated with Tek and wasn't sponsored by any means to write this so people shouldn't get any assumptions I'm advertising here.

PS: Hopefully I can get it early October and will update this thread with pictures and more test results.

[rosin232]

but man, the empty space just looks completely terrible.

100% agree I think they will come up 6-channels model and working on it trying to figure out how to fit it in there. You can see here https://youtu.be/sA_ijGl2zzo?t=1m8s there are two channels missing in the center.

[fonograph]

Weak FFT is deal breaker for me.I like the Keysight S style where it works like pseudo spectrum analyser.

[mikeselectricstuff]

I recently purchased one of these units MSO-64 1GHz with one digital probe it is pretty long lead time right now 6 weeks and so I'm still waiting.

..and they didn't leave you a loaner til it arrives.. that's a bit poor. 

[rx8pilot]

Weak FFT is deal breaker for me.I like the Keysight S style where it works like pseudo spectrum analyser.

The S-series is really nice in that area. The FFT is so fast.

[Proto]

I talked about what I thought Tek could do with the 6-Series design in an extensive, separate discussion thread.  My data source was the video and comments from Tek engineers on their expert blog.  No, No, No -  Tek DOES NOT talk about any of the things I discuss like 6 channels or logic-only channels, OR 6-channel ready design, etc.  But, I pretty much said the 6-channel design is done.  The current main board appears to have a stuffing option for 2 additional Tek061 preamps and they seem to be pre-routed to the 100GS/s ASICs.  There are 2 ASICs, each of which can handle 4 25GS/s FlexChannels so there is spare capacity in the 4 channel designs already.  In fact, they could have used just 1 ASIC but that would have meant a new mainboard for a 6-channel version.  I don't believe that 6 channels would be a retrofit for existing customers.  But, add 2 preamps and alternative faceplace and subpanel components and you're done.  It's a market timing decision.

Tek could also add 2x 8-bit logic channels to a 4 or 6 channel 6-Series and de-feature analog functionality for these additional channels.  If I were them, I would do this to reduce the incremental cost they would otherwise charge for full FlexChannels.  Vendors naturally charge a higher per-channel cost with increasing frequency and this scales with analog channel count making pricing for extreme combinations of BW and channel count excessive.  So I would not expect a 6-Series with 8 FlexChannels but one with 6 FlexChannels and 2 Logic only 8-bit channels might be reasonable.  In this hypothetical scenario, you would buy the new logic probe and use it in  2 channel positions without analog input capability.

Take a look at my 6-Series discussion topic for a bunch of observations and assertions gleaned from the current 6-Series design.