#96: Tutorial on Digital Oscilloscope sample rate, record length and data processing

This video gives a basic tutorial on how typical digital oscilloscopes (DSO) process data from the analog input to the display. Topics include Sample Rate, Sample processing / Acquisition Modes, Waveform Record Length, Display Processing and Measurements. Of course, this is a detailed topic, and every detail isn't covered in this short video, but it does give a general overview of the process. Every manufacturer has their own terminology and variations on how they specifically perform this processing. It is important for you to understand how your particular scope operates in order to get the most accurate results from your measurements. A copy of the notes in the video can be found here:
www.qsl.net/w/w...

[@koffibanan3099]

Fantastic explanation. After reading so many crappy posts on the internet, I now really understand. Thanks a lot!

[@Nabilphysics]

This men shows how you can make incredible engineering video without any editing.

[@anthonyheak3479]

Great explanation on dso parameters. Much appreciated it! You are a great teacher!

[@venkir1408]

This is the explanation I wanted and searching for many videos, fantastic. I'm a software guy designing hobby DSO. I got clear picture on many of my doubts here which other videos cannot answer. He also has many great videos collection. Thanks a lot for such a detailed videos. 👌👌👌👍👍👍

[@hubercats]

Another outstanding video. Thanks, Alan!

[@ernestb.2377]

You are doing a very good job educating like this. Thanks. 👌

[@x86cowboy]

Wonderful clarity. Thank you so much for your contribution. I've learned TONS from this channel. :)

[@PodeCoet]

You know it's going to be a good weekend when w2aew posts a tutorial video

[@Jnglfvr]

Most lucid explanation I have found on this topic.

[@petererasmus6460]

Thank You so much for this tutorial ,I am a bit late but only look found it today after many years as an Hobbyist I also now understand a bit more my HANTEK DSO ,clearly the problem with this entry level scopes is the Memory depth ,but it works fine for what I am doing .Thanks again for a brilliant video !!

[@AnujArya_Curious]

Precious video. Really appreciate the time and dedication for explaining each point.

[@jhillix8954]

I really enjoy these videos about oscilloscopes.

[@hassantj5775]

Simply excellent. A matchless instructor.

[@basketofkittens]

Another wonderful video with a clear and concise explanation! Thank you!

[@jrallen870]

Another great video.... wish you could post a video each day.... (I know impossible).. I"m always excited when I see you've posted another video as I know I'm in for some quality learning on my hobby subject.

[@vaggelismadias7481]

A very precise and thorough explanation! Thank you!

[@BrianHensleyRULES]

Great job Allen. Always love the engineering paper :)

[@timorfred1]

Thanks for such a clear and beautiful explanation, Double thumbs !

[@elliott1787]

Pretty much the same as mine, I'm jealous of your dual time base though. They are pretty decent scopes, you get a great 100Mhz analog and you get some basic digital features. I got mine for $40 off Craigslist, talked the guy down from $120 because the traces wouldn't show up in digital mode and it needed a couple of the shafts replaced for the "CAL" knobs. All it took to get the digital stuff to come to life was reseating the ribbon cables, which I was doing anyway to fix the knobs.

[@DAGODER]

Excellent video, thanks for the post!. I have downloaded the notes for reference.

[@sirousmohseni4]

Excellent video.

[@Taran72]

That was awesome! THANK YOU for the wonderful explanation.

[@SurfinScientist]

Thanks a lot! Wonderful tutorial!

[@RobB_VK6ES]

Again a clear and concise tutorial Alan and I still have 'scope envy with over that beast. I have a Chinese "copy" of the Tek with the a name. Whilst watching I had a notion. What if you could set the scope to record a sample only when the adjacent samples changed by some specified dV. In essence giving a variable bit rate, a bit like jpeg compression I'd imagine. I am sure this aproach has been considered by minds far greater than mine but is it do-able in real time / sane / cost effective.

[@cees1910]

Well great again, love all the videos your made Cees

[@censored9218]

Very informative!

[@ruhnet]

Excellent!

[@jeromequelin]

Thank you so much for this, it cleared things up tremendously!

[@J4e8a16n]

Thank you for this video.

[@chuck7592]

Excellent video! Thanks. I have a question as well. For a given sample duration, why would I NOT want to use the maximum sample rate? In your demonstration, your scope defaulted to 5MS/s. Why not just do maximum S/s for the given sample duration? Seems like sample memory is going unused.

[@w2aew]

You could a faster sample rate and thus more waveform memory. The trade off is processing speed (or lack thereof).

[@uiticus]

This is an incredible scope. Just wondering if any malfunction should occur with the scope, how will you troubleshoot it, since the circuitry must be very complex?

[@jadaboras7523]

Thanks for great explanations

[@flipc6081]

Been scrolling the web for days trying to find a reasonable decent scope to simply graph the voltage of a car battery overnight. I don't use scopes on a regular basis but wondering if you can suggest one for around $400 or under. Can't seem to find one to do this. I have a small dso but its tiny as hell and only has 8k of sample storage and no adjustment for more than 2 sec div.

[@dianec.2920]

Very impressive. Thank you!

[@isurunuwan2575]

Good teaching. Thanks

[@michaelcostello6991]

Great video. Any video on the memory display/trigger at the top of a RIGOL scope.

[@DzpcbElectronicWorld]

Thank you Sir for this video. One questions: What kind of camera or camcorder you are using.? It seems to be good enough on low light.

[@w2aew]

+DZPCB This one was shot with an old Canon 130is digital still camera (I believe).

[@gamingSlasher]

Good tutorial but I really missed info about how most digital oscilloscopes can store more data than is displayed on screen and you can then scroll through the data both before the trigger point and long after. That is important for single shot events and I usually mess up and get too little time or too few sample points.

[@bradcfi2]

I am a ham and would like to purchase a scope. What do you recommend for under $600? Same for a SA for under $1,400. Is there a MDO that is decent for under $2,000?

[@clems6989]

Is it fair to say: this is why if you have a cheaper digital scope, it may advertise 1Gsample/sec but its memory depth will limit what you can display at a givin sec/div ? Correct ? (Sounds like a good reason to keep an analog scope around too..)

[@w2aew]

Yes, that is fair to say. Using it as a station monitor for example, to watch your RF output - when running SSB phone, you probably will want to set the horizontal to something like 500uS/div. That gives you 5ms across the screen. If your scope is using 10,000 points for the waveform, that translates to an effective waveform sample rate of just 2MS/s - which will be undersampling your HF RF signal. Sample processing like "peak" can help, but it does illustrate why, in this case, an analog scope is superior in this instance.

[@cees1910]

Forgot to ask, is it possible to increase the samples record lenght by hook up the computer with let say labview, or other program thats runs on computer?

[@J4e8a16n]

Hi, I am in trigger mode. Here is my problem. As I turn the ch 1 button the frequency goes up. Il I turn it counterclockwise it goes down. Why... JP

[@4abrownafrica939]

Is there a formula I can use to determine a square wave rise time that I can use to check the accuracy of the scope.

[@w2aew]

I'm not sure what you're looking for. If you have a squarewave with a very fast edge, you can estimate the scope bandwidth. I'm not sure what measure of accuracy you're trying to relate to risetime.

[@SuperZiutek123]

Thank you very much

[@Myclone]

That's really a great video! Unlucky there is just one thumb to set up!

[@akhilthomas3020]

Great

[@tbsmith77]

Why do you use 10 divisions in your calculations? If my scope has 15 horizontal divisions on the scale, would I use 15 in the calculations?

[@w2aew]

+tbsmith77 The vast majority of oscilloscope displays have 10 major divisions horizontally - that's why I used 10. I've never seen a scope that has 15 (or some other number) or horizontal divisions, but if your scope has a different number of divisions, use that.

[@tbsmith77]

+w2aew Thanks! My scope is a Hantek DSO1202E (handheld) and has 7 1/2 divisions on each side of center.

[@captainkirk16384]

gaming, having co-authored an open source WFMReader that reads and displays the data in a graph on your PC, I can tell you that is is nothing more than a WINDOW. Imagine the entire graph on a large wall. But all you can do is look through a window. You see LESS than what is actually stored on the wall. You can move around, left to right and see different areas (scrolling). Anyways, the tough part is synching the sample with a trigger, which is probably the key thing you need. Ext Trigger?

[@nabilazisputri]

what a great video, it's very helpful! by the way I'm having problem to set sampling frequency in Tektronix oscilloscope TDS 2014b, could you help me or give some advices? I'm kinda frustated,,,,, Thank you so much!!!!

[@w2aew]

+Nabila Ardhana The older scopes like these TDS2000 series have a fixed waveform record length - thus the waveform sample rate is simply a function of the horizontal scale setting.

[@nabilazisputri]

+w2aew so it means we cannot change the record length, can we? but can we set the horizontal time scale for the older series Tds 2000??

[@w2aew]

The Time/div control is the horizontal scale control.

[@nabilazisputri]

+w2aew so we cannot take 25000 record length within 5 kHz sample rate? and what is the meaning of its feature, 1GS/s? thank youuu

[@w2aew]

+Nabila Ardhana The TDS2014B has a fixed record length of 2.5k samples. You can't adjust or change this, it is always 2.5ksamples. The scope samples the input at 1GS/s (one billion samples per second). However, at a horizontal setting of

[@friedmule5403]

Thanks for making it so easy to understand! :-) So that's why an analog oscilloscope can be a lot better for wary fast changing signal?

[@w2aew]

...yes, especially at slow sweep speeds.

[@AlainHubert]

Thanks for this Alan. Very informative. I have an additional question: some scopes, like my Rigol DS1054Z, have a function called "anti aliasing" in the acquire menu. What exactly does it do ? When I press the help button on the scope it simply states: "Use this feature so that the oscilloscope can avoid the signal aliasing". I've used it but didn't see any change. I suppose it's when a very high frequency analog signal is input that could suffer from digital sampling errors due to the sampling frequency being too low for such a high frequency ?

[@w2aew]

Aliasing can occur when you use a slow horizontal scale when looking at a high frequency signal. I don't know what the RIgol anti-aliasing function does, but I suspect is makes several acquisitions and performs some type of signal processing to determine if the sample rate is too low for the given input signal. Unfortunately, the Rigol manual provides no detail on the process.

[@AlainHubert]

@@w2aew Thanks Alan. That makes sense. Yep, the manual only mentions a reduced sampling rate and a "dedicated display algorithm" to avoid aliasing. It's probably a proprietary software. But I've yet to see what it does, since the scope is a 100 MHz (fully unlocked from 50 MHz initially), but I don't have such high frequency signals to test it with.

[@allesklarklaus147]

AlainHubert If you just put in a signal with fast slim pulses (probably under 100ns rise/fall time) but at a say 1kHz repetition rate you see the problem. Maybe at a 1ms/div timebase setting the pulses will (partially) disappear and do some funny things. This is because your scope reduces sampling rate and the pulses are too fast. Most older digital scopes like my tek tds310 do a horrible job at this since the sample memory is so small. But many modern ones deal with it very nicely. Aliasing occurs when you feed an ADC with signals above half it's sampling frequency and is usually undesirable. (see nyquist theorem for example)

[@allesklarklaus147]

AlainHubert You could also just feed in a 10MHz sine wave and turn the horizontal to 10ms/div or slower and the waveform be going completely out of shape but that is not very close to showing you the real problem of this since nobody does that.

[@hexdragon_]

Just bought my first digital scope - Hantek DSO2D10 and it seems that it measures from the screen, but not from the sample buffer like the Tektronix does. When I measure rise- and fall- times of a square wave and change memory depth 4k-40k-400k-1m-..., the time values remain about the same, but changing horizontal screen scale (time/div) - measures become more accurate with faster time setting. This is not surprising considering the cost of 250$. But if it is so, what is the point of having different memory depths in such scopes as mine if it converts the sample buffer (even the smallest of 4k) to a constant screen resolution length (800 points) anyway and only then measures the values? Thank you!

[@w2aew]

This is precisely why Tek scopes always use the waveform points and not the display points for measurements. The reasons other manufacturers (including some BIG names) make the measurements on the display points are 1) it is much faster, giving the appearance of a much snappier response, and 2) they assume that detailed measurements would be made when the waveform is adjusted to visually see the detail. The deeper memory can be very useful to look at long term trends, capture long duration events, as well as being able to zoom in and make the more precise measurements.

[@hexdragon_]

@@w2aew Thanks for your answer. This really makes Tektronix to stand out and this is how measures should be acquired. Just fiddled with my Hantek and found out that frequency cannot be measured if there is no full period on screen, no maximum or minimum values or rms if the extremes are partly off screen, etc etc. Average acquiring mode is also going mad when the waveform is partly cut by the screen top or bottom, which I assume means that it also averages from the screen buffer(!). I sold my two old 10Mhz analog scopes to buy this Hantek to be able to capture the transients, bought it yesterday and I already want Tektronix👍

[@passanger3e]

If i have hantek dso5202p (which has 40kbts memory, 1gs/s) by setting 25us/div and appling 20k sine. signal ,would it be look like a sine wave?

[@w2aew]

Provided you have the trigger setup properly, then yes. You have 5 sine periods.

[@vishnudas9476]

But because the memory is limited to 40000bits he wont get the full benefit of 1Gs/second sampling. The scope will throw away the extra samples to fit in the 40Kb memory.There is a chance that some spike or glitch in the signal is lost due to decimation. To get full benefit the memory of tthe scope should be 1gs/s*250uS=250000 bits, ie 250Kb. Is my understanding of the concept right ?

[@msk19991]

Any info about the 2khz signal source circuit? Or is it a function gen? Thanks Allen

[@vishnudas9476]

Hi allen I have a doubt. How does TEK TDS 430A WITH 400Mhz analog bandwidth uses only 100Msa/s sample rate. How it prevents aliasing when measuring high frequencies ?

[@alanwolke6253]

Because it uses equivalent time sampling at the rated bandwidth. The waveform is built over several acquisitions. The single shot bandwidth (where you must meet nyquist criteria, is well under 100MHz.

[@vishnudas9476]

Thank you. 😁

[@nicolaslopeza2251]

Great video! But i have a question. Why would you set the memory to 10k when you have the chance to use 20M. It seems to me it's a waste of precision, but i guess there is a reason to let you tune that parammeter too.

[@w2aew]

It really comes down to responsiveness. Moving and processing 20Msamples of data (vs. something smaller like 10Ksamples) can take a fair amount of time. Many scope manufacturers will let you set the memory deep, but only process the points that are going to the display. The *big* downside to this is that measurements are made on this highly decimated set of points. The Tektronix scopes always make measurements, etc. on the acquired data points, not the displayed data points - so there is always processing happening on the complete acquired memory set. This is why Tek gives you the option to adjust the memory depth as needed (shorter when you don't need the sample resolution, deeper when you do).

[@nicolaslopeza2251]

@@w2aew Thanks for the fast reply. I'm an electrical engineering student from Argentina and your videos are really helping me out. I'm looking forward to buy an oscilloscope for some small DIY projects and this video was exactly what i needed. The back to basics videos are awesome too!

[@samuelolisa6269]

Nice video. However, I think the 1 million Samples should be 10,000 Samples. Could you cross-check the calculation you made?. Thanks.

[@w2aew]

5GS/s * 20uS/div * 10div = 1 million. 5e9 *20e-6 *10 = 1e6

[@jimsutherland501]

Thank you!

[@waynegram8907]

when measuring bits per second logic signals or baud rate signals which these signals are logic words of packages of data. If you Oscope can't trigger on these types of signals what can you do to get a stable display for bits per second and baud rate signals? The baud rate value/frequency is determining the rate of which the logic words of packages are being sent over the serial data transmission buss?

[@w2aew]

Check out my video on eye patterns - this should give you an idea of the baud rate.

[@waynegram8907]

@@w2aew yes that eye pattern video was confusing. I don't know why baud rate is sometimes called eye pattern

[@w2aew]

@@waynegram8907 The baud rate is NOT called an eye pattern. The baud rate can be CALCULATED from the eye pattern. Measure the time between the crossings (width of the eye) and invert it to arrive at the baud rate.

[@waynegram8907]

@@w2aew I will watch your eye pattern video again because i don't think you showed examples of measuring different baud rates of logic signals using the eye pattern. You will see the logic word packet of data but then you measure the Pulse width of the whole logic word packet which the logic package if a bunch of logic on/off signals at random pulse width intervals. Then you measure the time between "dead space" from logic word package to the next logic word package. The Dead Space is the Width of the eye?

[@w2aew]

@@waynegram8907 Yes, 1/dead_space equals the the baud rate. More accurately, it is 1 over the time between the crossings.

[@michaelhawthorne8696]

Hi Alan. Nice video. I have a THS720A that I use at work... Pretty old I know but is still in calibration.... I keep seeing a problem that I can't resolve so as a Tek Tech I thought I would pass this by you. When I look at the waveform on the scope screen, I notice the update rate of the waveform goes fast and then pauses for a fraction of a second. The sample rate displayed doesn't change but the update of the waveform on the screen does. To illustrate further, If I have the Horizontal timebase set to say 500ms/Div, the display would traverse the screen in short bursts almost leap frogging. By selecting auto ranging the problem disappears only to return when I make a manual setting myself which cancels auto ranging. I can't find any answer to this, I can still use the scope but this problem is rather off putting. Do you have any ideas?

[@w2aew]

+Michael Hawthorne Interesting... I"ve never laid my hands on a THS720A, so I can't comment from any first-hand experience. it is odd. If anything, I would've expected the inconsistent update to occur in Auto mode (due to to the scope taking time to figure out the right settings), but that is not the case. I'm sorry I can't offer any insight. I'll ask around with some of my colleagues...

[@michaelhawthorne8696]

That would be great and thanks for responding to an old post of yours.

[@elliott1787]

Better than me, Tektronix 2221A with a whopping 4K. I do use the digital feature occasionally, but most of the time it stays in analog mode.

[@cees1910]

Hi Are you capable to explain maxwell law? Cees I cannot thats why!

[@RobB_VK6ES]

That should have reas ... with a similar name. the joys of editing to fit the limited number of chars for a reply :)

[@kasra7907]

Thanks

[@schwinn434]

Why not just always sample at the max - e.g. 5 GS/s?

[@w2aew]

Because of memory limitations. Waveform memory is filled faster at the faster sample rate. So, when you need to acquired a long duration event, the sample rate has to be reduced in order to fit the waveform in memory.

[@gunpaljain8350]

What is memory depth of oscilloscope?

[@w2aew]

Digital oscilloscopes sample the incoming waveform and store those samples in memory. The term "memory depth" refers to how much memory is available to store waveforms. Early digital scopes had very small memory, like 1kbyte. Modern oscilloscopes have much more memory - 10's of MBytes or more, some with up to 1GBytes.

[@eazyrotterdam]

i dont understand what you guys do with these machines , can someone explain please ?

[@w2aew]

Well, there are well over 100 videos on this channel alone that show various ways that a scope can be used.

[@mnovo]

lmgtfy.com/?q=How+to+use+an+oscilloscope+video please click here

[@eazyrotterdam]

ok thanks ill check it out

[@jasoneyes01]

Hmm. Does an analog scope have infinite sampling rate?

[@w2aew]

Analog (CRT-based) scopes don't sample the signal at all...

[@Roy_Tellason]

A hell of a sales pitch for something that I'll never be able to afford. :-) I don't own a digital scope at this point in time, don't see a real need for one in my hobby pursuits for the near-term future, either. I have always liked tek stuff though. I look at those cheap asian "digital scopes" and laugh, and figure that cheap as they are, they're really not worth pissing that bit of money away on, because they'll tell you *nothing* about a lot of what you touched on here.

[@J4e8a16n]

turningCH1Knob is on its way.

[@sammin5764]

🎖🎖🎖🎖🎖👍

[@supertruckertom]

I scored a 2230 a few years ago for $79.

[@tunicana]

i have a cheap oscilloscope with only 40k of memory, but for me it's doing the job.lol

[@chrisengland5523]

Normally, I take everything that Alan says very seriously, but this time he made me laugh: "So that's just a brief overview of what goes on in a typical digital oscilloscope". Typical? Really? That scope, the Tektronix MDO4104-6, retails in the UK at about £12,500. The only thing typical about it is the cream colour of the case and the only thing it's not very good at is making coffee.

[@w2aew]

Well, most of what is discussed here applies equally well to lower-cost digital scopes as well.

[@chrisengland5523]

@@w2aew I have a 30 year old analog Iwatsu 5710 that seems to have been modelled on Tek scopes. It still works very well. And I recently bought a Siglent SDS 1202X-E. Your video has clarified the trade-offs involved in setting the sample rates etc. - Keep up the good work; it's much appreciated.

[@hulladek3]

Dear Santa I need a 29k$ MDO4104-6 for Christmas.

[@PeterWalkerHP16c]

I guess you want a lot of oscilloscope for $20,000 AUD!!

[@kasra7907]

Thanks