Author Topic: DSO Record Length  (Read 2292 times)

0 Members and 1 Guest are viewing this topic.

Offline ApochromTopic starter

  • Contributor
  • Posts: 28
  • Country: de
DSO Record Length
« on: April 13, 2019, 11:32:03 am »
Hello,

I am a beginner in electronics, but I have a background in physics and basic knowledge in electricity. For many years I have had some electrical measuring instruments, e.g. multimeters and also a 40 year old Oszilloscope. (to be honest I have a lot of multimeters, which Dave Jones is not completely innocent of... ;) ) But now I have decided to do more in the field of electronics and since half a year I'm doing more information. The first Arduino projects I have already worked on.  Here in the eevblog I can read so much that I am busy for years...
Now I bought a DSO (Siglent sds1202x-e) three weeks ago. It's a lot of fun to try everything I can with it and I constantly get new ideas here in the forum and on the linked Youtube tutorials.
Now I was visiting a friend of mine who is a professional electronics engineer. At work he probably works exclusively with Tektronix Oscilloscopes. For his small private laboratory he also bought a Tektronix, namely a Tektronix TBS1154.
Now finally to the actual question.
I could have a look at this device and my friend showed me some details of it. As a beginner I wasn't able to recognize the differences between Tektronix and my beginner device so fast, but of course it's clear to me that Tektronix belongs to the professional league and is therefore much more expensive than my Chinese beginner device. He paid 1800€ for it. I do not want to compare the two DSO's at all and price comparisons always limp. But what I didn't understand at all is that the device has a record length of 2,5kpt.  When I asked him, that this seems to me in the comparison with other current DSO's extremely little, he said that one cannot compare this. Tektronix have a completely different technical approach and therefore the 2,5kpt are more than sufficient. But he didn't say more about that.
But googling hasn't helped me figure out what this other approach at Tektronix is in terms of record lenght. Therefore my question here, what do the 2.5kpt of the Tektronix TBS11154 mean and what is the difference to the 14Mpt of my DSO?

Thanks, Jürgen
 

Offline DDunfield

  • Regular Contributor
  • *
  • Posts: 173
  • Country: ca
Re: DSO Record Length
« Reply #1 on: April 13, 2019, 02:44:59 pm »
I don't think Tek is doing anything special.

DSO's from not all that long ago has much lower buffer depth.

My Tek2032 has 4k, TDS210 has 2.5k , TDS380 has 1K.

Traditional storage capacity was limited by the memory having to reside in the acquisition module and it was expensive. Things have improved and modern DSO's often have very deep memory. My DS1054Z has 24M (equivalent to 12M compared to above as they are two separate channels).

Dave
 
The following users thanked this post: Apochrom

Offline Doctorandus_P

  • Super Contributor
  • ***
  • Posts: 4004
  • Country: nl
Re: DSO Record Length
« Reply #2 on: April 13, 2019, 04:25:22 pm »
Siglent makes pretty good scopes nowadays.
I had a short look at the TBS1154, and it does not only have a very short buffer depth, but also a pretty small screen. These are both indications of an old desgign, while the Siglents are based around newly developed hardware.
The capabilities of scopes has made tremendous leaps tni the last 10 to 15 years.

I did not look very deep into it but If I were to choose I would almost certainly choose the Siglent over the Tek, even if the price difference was no issure.
The new Siglent scopes are not "Simple Cheap Chinese Scopes". They are excellent pieces of modern equipment and your scope wil probably give you many years of excellent performance.

I can understand your friend wanting to have a Tek for at home if he also works with them. Switching between user interfaces is a bit annoying for some people, especially for the more osbsure capabilities of the scope.

It may be fun if you invite your friend for some measurements brainstorming and compare the scopes. This is probably also educational for you to see some different measurements and get a better idea of how to use an scope and how to use it.

I have a "lowly" Rigol DS1052-E and I was thingking about upgrading to a 4-channel Siglent, but for most of my measurements the old Rigol is already more Oscilloscope than I need.
The built in logic analyser of the Siglent is a nice bonus, but for a logic analyser I still prefer an USD10 FX2 device combined with Sigrok & Pulseview. A big PC monitor with mouse is an excellent interface for a Logic Analyser (Not so for a scope, for a scope I prefer all the buttons on the front panel).

 
The following users thanked this post: Apochrom

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 4135
  • Country: cn
  • Born in Finland with DLL21 in hand
Re: DSO Record Length
« Reply #3 on: April 13, 2019, 07:21:50 pm »
If look only this question about one single acquisition capture length.

2.5k sample and 14M sample difference is huge.

(of course there is also other differences - lot of)

One advantage is in this TBS1000 series. It have this maximum 1GSa/s for all channels simultaneously.
SDS1004X-E can 1GS/s only for two channels simultaneously and 500MS/s for all four channels. And 2 channel models only for one channel and half speed for two channel.

One other basic things big difference is that this Tek have analog trigger circuit when Siglent use full digital side trigger system what offer lot of more advanced triggering with also better trigger accuracy.  But it need remember TBS is quite old design. So it can bit more understand. Today it is really "outdated" and in most of things example this tiny cheap Siglent  outperform it hands down nearly in everything. But not start this debate at all.

Here image what may tell part of story about short vs bit longer memory.

As can see with long memory we can use lot of higher samplerate when go to slower timebases.
Example with 1ms/div still full samplerate 1GS/s when  TBS1000 have only 250kS/s   4000 times less.

Quote
When I asked him, that this seems to me in the comparison with other current DSO's extremely little, he said that one cannot compare this. Tektronix have a completely different technical approach and therefore the 2,5kpt are more than sufficient.
But he didn't say more about that
.
Quite easy to understand he didn'd say more about it...  :-DD
« Last Edit: April 13, 2019, 07:32:14 pm by rf-loop »
EV of course. Cars with smoke exhaust pipes - go to museum.
Wises must compel the mad barbarians to stop their crimes against humanity. Where have the (strong)wises gone?
 
The following users thanked this post: Apochrom

Offline ApochromTopic starter

  • Contributor
  • Posts: 28
  • Country: de
Re: DSO Record Length
« Reply #4 on: April 13, 2019, 08:30:45 pm »
Many thanks for the answers!

The list is great! It was only now that I really understood the effect of the memory size.
Unfortunately my friend doesn't live nearby and we see each other only once or twice a year. I don't want to embarrass him with this DSO comparison either. He is a big fan of Tektronix and has two old Tektronix analog oscilloscopes that really impress me. He also has a unbelievable amount of knowledge about electronics and I am a very small beginner. That's why I won't talk to him about the TBS1154 any further.

Jürgen
 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: DSO Record Length
« Reply #5 on: April 13, 2019, 09:27:05 pm »
Tektronix has long been the gold standard to which all other oscilloscopes are compared, they practically invented the modern scope. Technology marches on and formerly exotic features become commonplace and the performance of modern budget instruments can exceed that of high end instruments from a few years ago. I'm still a fan of Tektronix scopes though, they're made in the USA not far from where I was born, they're well built and have a nice familiar user interface. They're not magically superior in every way to lower cost instruments though.
 
The following users thanked this post: Apochrom

Offline pwlps

  • Frequent Contributor
  • **
  • Posts: 372
  • Country: fr
Re: DSO Record Length
« Reply #6 on: April 14, 2019, 10:37:36 am »
Tektronix has long been the gold standard to which all other oscilloscopes are compared, they practically invented the modern scope. Technology marches on and formerly exotic features become commonplace and the performance of modern budget instruments can exceed that of high end instruments from a few years ago. I'm still a fan of Tektronix scopes though, they're made in the USA not far from where I was born, they're well built and have a nice familiar user interface. They're not magically superior in every way to lower cost instruments though.

I was also an unconditional Tektronix fan but about 10 years ago I tried a DL1640 from Yokogava and discovered several features that none of Tektronix had (at least in the same price range).  This scope has a 1M (extendable to 32M) memory depth and offers, via oversampling, very nice features like enhanced resolution (up to 14 bits, obtained by decimation) or a digital lowpass filter.  It has a dedicated DSP per channel so that the performance does not depend on how many channels are active.  If you are only looking at the captured signals on the scope screen then these features are not so important, but I also used it for acquisition, then the memory depth and the resolution become important.
One thing you can't find in the specs is the quality of the input amplifier. Sometimes I had to measure a tiny signal a few microsecond after a 1V pulse, then my old TDS had problems (below 50mV/div there was a long saturation tail) but the DL1640 could measure it properly down to 2mV/div (btw. Tektronics doesn't go below 5mV/div).   
Another thing you never find in the specs: the acquisition speed in the "average" mode, another parameter my measurements rely on.  This is not the same as the number of traces displayed per second (can be much lower).  I remember once a Lecroy representative  was presenting us a new model with very fast update rate, some thousands of traces per second. But when I set up a test in the average mode it could not average more than some hundred traces per second. Here the TDS speed was strongly dependent on the number of points (3k/second for 512 point trace, <1k/s for 10ksamples traces), and the DL1640 beat them all by a very large margin (certainly due to its 4-DSP configuration).
So in general the best choice depends on the application and you can only know if you borrow one and test. Also, when I'm searching for a specific feature I first dowload and read the manual.


« Last Edit: April 14, 2019, 10:39:14 am by pwlps »
 
The following users thanked this post: Apochrom

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3516
  • Country: us
Re: DSO Record Length
« Reply #7 on: April 14, 2019, 12:51:31 pm »
As noted previously, older DSOs had the memory in the acquisition unit.  They used SRAM which is faster, but also much more expensive.  An  MCU then read data from the DAQ and wrote it to the display.  You can have large memory or you can have fast memory.  But you cannot have both.  The address line capacitance sets the achievable speed and size.

Modern DSOs are built around an FPGA and MCU.  The Siglent SDS1000X-E and Instek GDS2000E use the Xilinx Zynq chip.  Thus they are able to use cheap DDR DRAM.  Rather than an MCU reading SRAM to display data stored in the acquisition system, the ADC feeds into the FPGA which implements additional processing and writes the result to DRAM.  This is now the norm in DSOs and high end models from the top tier OEMs feature over 1 GB/channel.

Whether having deep memory is important or not depends upon what you want to do.  For example, if you want to do vector network analysis by time domain reflectrometry, deep memory is very important for increasing the dynamic range from 48 dB by summing properly aligned segments of the trace record.  It dramatically reduces the number of records that need to be transferred to the computer for analysis.

FWIW I currently have an Instek MSO2204EA, Owon XDS1202A (12 bit ADC), Rigol DS1102E.  I plan to buy a Siglent SDS1202X-E or SDS1104X-E and make detailed comparisons of results for the same experiments.   My hope is to embarrass vendors into fixing at least some of the bugs and improving the overall FW designs.

I'm a retired research level geophysicist (reflection seismology), so very heavy physics and DSP background.  I'm am so frustrated by the poor quality of DSO FW even in $20K instruments from Keysight and Rohde & Schwarz that I am engaged in a project to write a FOSS DSO FW stack for Zynq and Cyclone V based instruments.  I expect the project to take 2-3 years to complete.  I have the needed FPGA dev boards and bought my Tek 11801 so I can *measure* bit skew in DSP filter pipelines rather than simply rely on hoping that the vendor software calculates the timings correctly.

The ultimate goal of the FOSS DSO FW project is to  produce a portable replacement for the OEM DSO FW that sets a very high standard for DSO features and FW quality.

 
The following users thanked this post: james_s, Apochrom

Offline ApochromTopic starter

  • Contributor
  • Posts: 28
  • Country: de
Re: DSO Record Length
« Reply #8 on: April 14, 2019, 04:25:40 pm »


FWIW I currently have an Instek MSO2204EA, Owon XDS1202A (12 bit ADC), Rigol DS1102E.  I plan to buy a Siglent SDS1202X-E or SDS1104X-E and make detailed comparisons of results for the same experiments.   My hope is to embarrass vendors into fixing at least some of the bugs and improving the overall FW designs.

I think this will be a comparative test that will interest many hobbyists. I am already very curious!


  I'm am so frustrated by the poor quality of DSO FW even in $20K instruments from Keysight and Rohde & Schwarz that I am engaged in a project to write a FOSS DSO FW stack for Zynq and Cyclone V based instruments.  I expect the project to take 2-3 years to complete. 

So far I have only seen an expensive Rhode and Black in action, but could not operate it myself but only watch it.
I was very impressed by the huge touch-display and the possibilities, but the device costs about 5000€. I thought this device class of Rhode and Schwarz and Keysight are the top class of professionals. Can you give an example why you so frustrate with the low quality of these DSO's?

Jürgen
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3516
  • Country: us
Re: DSO Record Length
« Reply #9 on: April 14, 2019, 07:30:52 pm »
https://www.eevblog.com/forum/testgear/compensating-bandwidth-with-software/msg2342280/#msg2342280

I was absolutely stunned at how bad the FW was.  I was so affected by the Keysight for which I had paid $10K that I had to lie down for an hour or more.  And very relieved when I realized I could return it.

I used a vulgar Anglo-Saxon word when requesting a return.  I called it "a gold plated turd".  Gold plated excrement for non-native English speakers.

The RTM3104 is a $20K instrument with the options I wanted.  I am still appalled by the FW update fiasco.  R&S are selling their customers beta test machines.  I kept quiet about it for a long time, but as they have never provided an explanation and documentation of a fix, I have decided it's time to call a spade a spade.  The last time  Rich responded, I got an "I'm not ignoring you, I'm busy" email.  I've not gotten *any* replies since. Sadly I did not have the wit to save a screen shot of the <3% overshoot.  But I've got plenty of shots showing the 10% overshoot and a long list of other bugs, all of which were verified by R&S.

I *had* thought if I spent the money I could buy my way out of buggy DSO FW. I now know that is not possible.  As I am old and unemployable, I decided to do something about it.  Having worked with over a half dozen DSP systems over a period of 37 years, both as a user and primarily a programmer, I know how to do it.  I've fixed other people's mistakes in over 2 million lines of code.

I soured on Siglent after a run in with an SSA-3021X.  I now have a much better grasp of how bad the situation is. They now look as if they might be the best.   I'm *extremely* curious what the comparison will show.  Enough so to buy a $400-500 DSO I don't need.  I also intend to document what the correct display should be and how to produce it.

What I intend to write will be an order of magnitude better than anything I've touched. DSP was done in reflection seismology  for roughly 20 years before anyone else did it.  We only needed DC to 125 Hz and there were strong financial incentives to pay for the R&D work needed.
 
The following users thanked this post: Apochrom

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: DSO Record Length
« Reply #10 on: April 14, 2019, 08:12:56 pm »
The RTM3104 is a $20K instrument with the options I wanted.  I am still appalled by the FW update fiasco.  R&S are selling their customers beta test machines.  I kept quiet about it for a long time, but as they have never provided an explanation and documentation of a fix, I have decided it's time to call a spade a spade.  The last time  Rich responded, I got an "I'm not ignoring you, I'm busy" email.  I've not gotten *any* replies since. Sadly I did not have the wit to save a screen shot of the <3% overshoot.  But I've got plenty of shots showing the 10% overshoot and a long list of other bugs, all of which were verified by R&S.


It's the nature of much of the software industry now, enabled by the ease of providing software updates. Ship the minimum viable product and update it later, you might have one QA guy on a team of developers and rely mostly on automation and customer feedback for testing. The result is predictable, crap quality with new bugs appearing as often as fixes.
 

Offline ApochromTopic starter

  • Contributor
  • Posts: 28
  • Country: de
Re: DSO Record Length
« Reply #11 on: April 14, 2019, 08:34:03 pm »
Well, I'm very surprised. For a beginner and hobbyist with a small budget these DSO's from Rohde and Schwarz or from Keysight or lecroy are unattainable.
Before I bought the Siglent I studied many reviews and videos about DSO's and followed many discussions in the forum before I decided for the Siglent 1202x-e. So I was astonished with how many firmware problems (and Siglent had a hardware problem with its predecessor) some DSO's from Siglent/Rigol and Owon went on sale.  Through my research I knew, that especially the Siglent 1202x-e and the Rigol 1054z got some firmware updates in the meantime and most of the problems are probably fixed. Since I don't need four channels and I think it's good that you don't have to hack the Siglent to get all options and the full 200Mhz BW I finally decided for the Siglent.  I thought, okay if you have to work professionally with the DSO's or have the necessary money, you better get a device of the renowned brands Keysight, LeCroy or Rohde and Schwarz.
I wouldn't have thought that these brands also have such big FW problems and that these devices are brought to the market too early and not fully developed. Madness - somehow it destroys my dream image of these brands...

regards, Jürgen
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: DSO Record Length
« Reply #12 on: April 14, 2019, 10:53:43 pm »
There are a couple of things which help to make up for such short record lengths.  Peak detection, and then later DPO (digital phosphor oscilloscope) operation which produces a complete histogram, capture the input at the maximum sample rate at any sweep speed.  Delayed sweep and delayed triggering allow moving the acquisition window to a later point (1) without compromising sample rate.  Delta delayed sweep allows two separate delayed acquisition windows.

Older SRAM based Tektronix DSOs actually had larger record lengths than 2.5 kSamples but their much faster CCD based instruments were always very limited in comparison and I think this carried over into their later designs because they knew how to make a usable DSO with a short record length.  Their early RAM based DSOs like the 2230/2232 had 4 kSamples while their much faster CCD based 2440 series (2) had only 1kSamples.  Later RAM based TDS DSOs might have 8 MSamples through an option which is a respectful even for now and CCD based TDS DSOs had only 15 kSamples in 1998.

(1) Technically it is possible to move to an earlier point as well by phase locking to the trigger but the only oscilloscopes I know of which did this were some early sampling oscilloscopes where it is a big advantage for preserving their very high bandwidth.

(2) The 2440 series is the only example I know of where a DSO supports delta delayed sweep and in a weird way, that feature sort of doubles its record length although that was not the intention.  I suspect they included it exactly to make up for its short record length even for that time and it does do that in most cases.
 
The following users thanked this post: Apochrom


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf