Analog vs digital X-Y mode

[Miti]

Just for fun, I guess no digital can beat an analog scope at X-Y.

[tautech]

Link to file please so we can all play. 

[james_s]

File? Isn't it a piece of hardware? I have a scope clock that I built from a kit years ago, the particular one I have is a standalone device but I've seen boards meant to connect to an existing scope.

As much as I love my Tek DSO, the XY mode is pretty hopeless, that is one of the few things where analog wins hands down.

[Miti]

It's a kit.

http://www.dutchtronix.com/ScopeClockH3-1-Enhanced.htm

[Tom45]

So why do digital scopes do so poorly in X-Y mode?

[ataradov]

So why do digital scopes do so poorly in X-Y mode?

Pixels are much more visible. I'm sure some of those $5000 R&S scopes will look much nicer.

And I'm pretty sure scope manufacturers simply don't care, just like with FFT function. I personally never had to use XY mode for anything in my entire life.

[egonotto]

Hello,

I think most DSO are good in streaming. They capture many samples and after they postprocess the samples. In x-t-mode they "know" the best samplerate.
But in x-y-mode they dont know the best  sample rate. If you have 10 Hz than a sample rate of 20kHz would be good, but if you have  10Mhz than you need a much higher samplerate.
Perhaps they take 1 sample and process this sample and so on, but this slows the whole.

Best regards
egonotto

[Performa01]

Setting up X/Y mode is different for analog vs. digital.

For analog, we only need to take care of the channel gain.

For digital, we also need to set the timebase to make sure at least one full signal period is captured (but also not too many more). This also determines the sample rate which is usually plenty high enough, hence not a problem.

What is a problem though is the low amplitude resolution of an 8-bit digital scope.

8 bits means 256 steps, but most scopes allow some overscan and e.g. only 200 steps are visible on the screen. On a 400 pixel high display, this means two pixel per amplitude step. In X/Y mode, the same is true for the x-direction. A picture composed of 2x2 pixel clusters on a low resolution 400x400 screen area just cannot look pretty, full stop.

A DSO with at least 10 bits and high-res display should be able to produce reasonably nice X/Y figures.

[tautech]

We'll all know when DSO's have really come of age when they can handle Jerobeam Fenderson's oscilloscope music.
https://www.youtube.com/user/jerobeamfenderson1

Yummy stuff.   

[rf-loop]

Also with digital scopes there is one thing what also may affect result. There is some different methods to do X-Y
Some DSO's may use direcft XY plot method (read X and Y ADC and plot point and repeat)
Some DSO may use normal acquisition to sampling buffer (just as in YT mode) and  then process it for X-Y "image")
This last method make possible to do many automatic measurements also in XY mode. Rise times, pos and negative width, periods - whole buch of measurements including full sample resolution rise and fall times, pulse widths etc etc..


I do not know any reason to plot analog clock to oscilloscope display - exept fun. Oscilloscope is tool mainly for inspect more or less unknown signal. But of course if want entertainment box there is lot of many kind goods for this purpose. Much better than oscilloscope. I do not even know where XY is useful today. For what?

Yes I have played with clock and other images with  my previously owned couple of Tek7000 series and some Tek portable tool as 2465 etc.

[Fungus]

We'll all know when DSO's have really come of age when they can handle Jerobeam Fenderson's oscilloscope music.
https://www.youtube.com/user/jerobeamfenderson1

Yummy stuff.   

My cat loved it. 

[Tom45]

Also with digital scopes there is one thing what also may affect result. There is some different methods to do X-Y
Some DSO's may use direcft XY plot method (read X and Y ADC and plot point and repeat)

Or instead of plotting points,  draw a vector from X_n-1, Y_n-1  to X_n, Y_n

There is still the question of what sample rate to use. But I don't see any downside to oversampling. So why not sample as fast as possible.

But after thinking about it, perhaps the tricky part is how to emulate phosphor persistence.

[David Hess]

So why do digital scopes do so poorly in X-Y mode?

DSOs have six problems with X-Y mode:

1. Non-index graded displays cannot be anti-aliased and most DSOs with index graded displays do not anti-alias anyway or anti-alias poorly; this becomes an additional source of quantization noise.  Note that the analog examples above show a raster displayed on a vector display device producing further aliasing on the DSOs.  A true vector implementation would look better on both types of oscilloscopes.
2. DSOs suffer from quantization noise from the digitizer which is usually handled poorly in X-Y mode.  High resolution mode should solve this.  Average acquisition mode should also solve this unless it lowers the update rate.
3. Most DSOs produce display updates by batching record length groups of samples.  If the vector generation time does not fill the record length evenly, then the display updates either miss or overlap sections.  "Phosphor" type DSOs still update the display in batches determined by the number of samples between display updates.
4. DSO display refresh and update rate is orders of magnitude slower than the continuous update rate of an analog oscilloscope.  This should not matter above about 60 Hz but combined with 3 above it becomes a problem.
5. Almost all DSO displays have a much lower resolution per area than an analog CRT.  Many old CRT DSOs operated at 254dpi (100dpc) and could produce a pretty good vector display in X-Y mode except for the record length and update rate issues described above despite lacking anti-aliasing.
6. Most DSOs do not support a Z-axis (intensity) input so all of the vectors run together.  The Rigol example above shows this.

These issues with DSO XY mode could be fixed to produce a display limited only by the display device itself but it is hardly a high priority.

[T3sl4co1l]

Hey, I'm impressed those did as well as they did.

X-Y is a notoriously hard problem, and low on the list of priorities to address.

Tim

[james_s]

I use XY when working on vector arcade boards, a scope is more convenient than having one of the (hard to find) 19" vector monitors and a power supply producing the multiple oddball voltages they require on the bench. At least I did before I built a 5" vector monitor for one of my other projects.

[JohnnyMalaria]

I have to monitor an I-Q pair of a demodulated signal while the signals are also being captured. My Rigol DSO is great for Y-T plots and I love the compact size and weight. But the X-Y function absolutely sucks (the FFT isn't all that great, either.) I use my 1974 vintage Tektronix dual 100MHz just for the X-Y plots and they are gorgeous. I can tell what's going right/wrong with my instrument. With both 'scopes, I get to look at 6 signals which happens to be how many I have

[Scratch.HTF]

Despite the retrace lines on the digital scope, it still looks impressive.

[vk6zgo]

Just for fun, I guess no digital can beat an analog scope at X-Y.

Interesting that the old "tedious"210 does better than the much more capable DS1054Z.

[ataradov]

Interesting that the old "tedious"210 does better than the much more capable DS1054Z.

210 has a horribly slow display, so it probably contributes quite a bit to the "analog" feel.

But it seem like DS1054Z does not even attempt to blank the transitions, so may be some configuration of intensity will improve things.

[JohnnyMalaria]

Hopefully the linked videos will convince the nay-sayers that analog X-Y rocks  (They are too big to attach. The links are to OneDrive).

https://1drv.ms/v/s!AhmR5if7W0HCjcJG8SPwrOfTWF-grg

https://1drv.ms/f/s!AhmR5if7W0HCjcJFalJzpym6hx4C_A

A bit of description is probably necessary. There are two videos of a real/imaginary pair of signals from a photodetector (after demodulation) for light scattered by electrically charged nanoparticles moving in a liquid. A laser illuminates the sample. The first video shows just random Brownian motion from the particles. In the second video, an electrical field is turned on and off every five seconds or so. When it is on, the nanoparticles move toward the electrode of opposite charge. Movement of the particles causes a Doppler shift of the laser light. The X-Y trace indicates the amplitude and the optical phase difference of the light. The phase is proportional to the position of the particles. When the field is on, you can see "circles". One complete circle is one hertz. In terms of position, it is approximately 1 micron. The frequency of the circles is a perhaps 5 to 10Hz. That's a change of 5 to 10 parts in 100 trillion.

Rock on, 48-year old 475 scope!!!

(Sorry the videos are upside and grainy).

[nctnico]

Are there any loss-less wav files available for the clock so people can try without needing the actual hardware?

[CopperCone]

for constallation diagrams

[rstofer]

Over in a distant corner of the electronics sandbox, a few of us play with analog computing.  Not many people over here...

X-Y mode is critical.  We usually use one integrator to provide the X axis and the output drives the Y axis.  There are cross-coupled differential equations that draw a kind of egg-shaped pattern.  Not a circle, more like one that has been stomped on.  It's interesting to print two channels of Y(t) but more interesting to plot the simultaneous values of X and Y.

FWIW, the classic example is the Predator Prey problem.  As the rabbit population increases, eventually there will be more foxes.  More foxes will drive down the rabbit population which will reduce the fox population allowing the rabbit population to increase, and so on.

When I want X-Y done well, I use my Analog Discovery 2.  Among other things, I get a better screen shot.

[mtdoc]

FWIW, the classic example is the Predator Prey problem.  As the rabbit population increases, eventually there will be more foxes.  More foxes will drive down the rabbit population which will reduce the fox population allowing the rabbit population to increase, and so on.

Lotka–Volterra equations

[rstofer]

FWIW, the classic example is the Predator Prey problem.  As the rabbit population increases, eventually there will be more foxes.  More foxes will drive down the rabbit population which will reduce the fox population allowing the rabbit population to increase, and so on.

Lotka–Volterra equations

Exactly!  I need to read through that document.  About half way down they give the X-Y plot that I was discussing above.  Both populations are displayed on a single graph.

These equations don't have to be put into an actual analog computer.  MATLAB does a fantastic job of modeling the required integrators (2) and gain stages.  Simulink  gives the same answers without all the op amps.  But I like the old-school approach.  There's just a warm fuzzy feeling as the integrators integrate, the multiplier multiplies and the gain stages do their thing.

Attached is the Simulink model

Note that the model is pretty superficial.  As the article points out, the assumptions are probably not realizable.  Notably that the Prey can grow without bound because there is always adequate food.  The model could be improved to account for an exponential growth asymptotic as some logistic limit.

In my view, the X-Y mode more appropriately displays the interaction between the populations.  The Y(t) mode is useful but the X-Y mode drives the point home.

[chirality]

I'm curious, of all the <$2000 DSOs on the market as of 2018, which make/models do XY-mode the best (by best, I mean most analog-like)? Keysight? Tek? Rhode & Schwarz? Siglent?

[tautech]

I'm curious, of all the <$2000 DSOs on the market as of 2018, which make/models do XY-mode the best (by best, I mean most analog-like)? Keysight? Tek? Rhode & Schwarz? Siglent?

In a few days we'll know how well the SDS1*04X-E XY mode has been improved as it's just been enhanced in the latest firmware.

[chirality]

In a few days we'll know how well the SDS1*04X-E XY mode has been improved as it's just been enhanced in the latest firmware.

That's great to hear. Please keep me updated! I'm in the market for a DSO and the Siglent is attractive. I just definitely don't want to see an XY mode like this:

[tautech]

In a few days we'll know how well the SDS1*04X-E XY mode has been improved as it's just been enhanced in the latest firmware.

That's great to hear. Please keep me updated! I'm in the market for a DSO and the Siglent is attractive. I just definitely don't want to see an XY mode like this:

Yes, to properly evaluate it it needs be videoed.

The new FW I've mentioned is discussed here:
https://www.eevblog.com/forum/testgear/siglent-sds1104x-e-in-depth-review/msg1555936/#msg1555936

It's not up on the Siglent websites yet but should be in the next few days.

[CopperCone]

In a few days we'll know how well the SDS1*04X-E XY mode has been improved as it's just been enhanced in the latest firmware.

That's great to hear. Please keep me updated! I'm in the market for a DSO and the Siglent is attractive. I just definitely don't want to see an XY mode like this:

That looks like some kind of 1970's casino sign from Las Vegas. 

[Rich@RohdeScopesUSA]

I'm curious, of all the <$2000 DSOs on the market as of 2018, which make/models do XY-mode the best (by best, I mean most analog-like)? Keysight? Tek? Rhode & Schwarz? Siglent?

There are several videos of the RTB2000’s XY mode available from 3rd party reviewers, including one pretty thorough one here:



Mikes Electric Stuff has also played around with it a bit with some unique combinations:

https://mobile.twitter.com/mikelectricstuf/status/959025632830197761

-Rich

[nfmax]

Last year I had occasion to use XY mode on my Keysight MSO-X 2024A, to characterise the transfer function of an opto-isolator circuit (tweaking for best noise margins with lowest power dissipation). It worked beautifully, including a simulated 'spot' following the signal - which was a triangle output from the built-in generator. First time I'd ever tried XY mode, though!

[tautech]

In a few days we'll know how well the SDS1*04X-E XY mode has been improved as it's just been enhanced in the latest firmware.

That's great to hear. Please keep me updated! I'm in the market for a DSO and the Siglent is attractive. I just definitely don't want to see an XY mode like this:

Yes, to properly evaluate it it needs be videoed.

The new FW I've mentioned is discussed here:
https://www.eevblog.com/forum/testgear/siglent-sds1104x-e-in-depth-review/msg1555936/#msg1555936

It's not up on the Siglent websites yet but should be in the next few days.

Well I've managed to find a copy of the new SDS1004X-E series FW and the XY mode is dramatically improved.
With lowest memory depth set it really flies for a cheap DSO. 
But unfortunately even just a 10s video is 30+MB on my phone so it's too big to post here. 
Tonight might be YT A/c setup night.  Done.

https://youtu.be/wUCdaVq3kL0

[borjam]

I'm curious, of all the <$2000 DSOs on the market as of 2018, which make/models do XY-mode the best (by best, I mean most analog-like)? Keysight? Tek? Rhode & Schwarz? Siglent?

In a few days we'll know how well the SDS1*04X-E XY mode has been improved as it's just been enhanced in the latest firmware.

Any plans for the two channel SDS1202X-E? Seems that despite belonging to the same "family" firmware development is diverging a lot.

[tautech]

In a few days we'll know how well the SDS1*04X-E XY mode has been improved as it's just been enhanced in the latest firmware.

Any plans for the two channel SDS1202X-E? Seems that despite belonging to the same "family" firmware development is diverging a lot.

The last 1202X-E FW was back in Jan so any improvements it supports that's been added to the 4ch X-E should follow in time. The divergence is due to the substantial additional functionality the 4 ch models provide and the need to get all that stuff right.

[bd139]

Someone should try this on an HP 546xx. They have a superior X-Y mode. Not bad for 1994...

https://youtu.be/D2uEvRPd4w8

[chris_leyson]

Nice tutorials from Jerobeam Fenderson

[nctnico]

At least some DSOs can do color grading in XY mode:

[JohnnyMalaria]

I've put together a short sequence of videos to show why an ancient analog scope beats the crap out of a budget DSO in XY mode. There's no audio but there's a description which I've included here, too.

Budget digital storage oscilloscopes offer many advantages over analog CRT oscilloscopes but one area they suck at is XY mode.

This mode is important for looking at, for example, rotation of quadrature signals in the complex plane. This short video has back-to-back clips of an XY signal rotating at different angular rates. The analog oscilloscope is a Tektronix 475 manufactured in 1974. The DSO is a Rigol DS1104Z manufactured in 2017.

The first two clips are slowly rotating and the two oscilloscopes compare very well.

When the rotation speed of signal in the second example is increased, the DSO simply gets in wrong. The direction of the rotation is incorrect as is the speed. This is because the DSO captures a finite amount of the signal and displays it once complete.

In contrast, the analog oscilloscope displays the signal in real-time without interruption. A good analogy is the effect of video or film frame rate on the appearance of rotating aircraft engine propellers or rotating spoked car wheels. There's no way to tell if the observed rotation speed and/or direction is correct.

At much faster rotation speeds, the DSO yet again presents misleading information (clip 4).

Clip 5 shows how the DSO XY mode doesn't display significant amounts of the signal and the signal it does display lacks the extraordinary detail shown on the analog oscilloscope.

Finally, clip 6 shows that the DSO displays all of the signal but at a much lower resolution than the analog oscilloscope and the rotation is in the wrong direction.

It is common to compare these types of oscilloscopes using simple Lissajous figures but, frankly, the only real application of this is in 1950's sci-fi movies. Looking at changes of XY signals in the complex plane involves much more complex signals with intricate detail.

The analog oscilloscope may be 43 years older than the DSO but for budget DSOs, the analog one beats them by miles.

Professional grade DSOs generally offer far superior XY rendering than the budget ones but you can't beat the real-time, continuous and varying intensity of the analog equipment.

[alsetalokin4017]

Well, to be fair, it is possible to get a _slightly_ better display of the ScopeClock on the Rigol DS1054Z than what was shown in the OP.

But it definitely looks better on the analog scopes.

[joeqsmith]

I had someone ask me about the X-Y mode on one of my old scopes.  I was not sure what they were asking so I ran is various settings. One comment I had was about the sparse amount of data.  I suspect they were thinking audio and I was looking at much higher frequencies.   

Is the goal here just use use a scope for art?   I watched some of these fancy YT videos like the one linked above.   I then tried using Labview to put together a simple program to use the sound card but it was far too slow.   I ripped some of the audio tracks from YT and then tried it with some of the dedicated PC sound card scopes.  Looks like they may do a decent job.   YT compresses the audio so the graphics don't look very impressive but the dedicated programs far exceed what I could do with Labview.   I tried playing these files into one of my scopes and could post a clip but it's not very impressive with the YT compression.   At best you would have some idea on the update rate of my old scopes. 

As mentioned, I think if the goal is to benchmark the visual effects, a standard audio clip may be a good place to start.       

The letter 'D'

[JohnnyMalaria]

Is the goal here just use use a scope for art?   I watched some of these fancy YT videos like the one linked above.

Sometimes science is art

I created the clips with preset waveforms from my function generator (such as chirp, FM modulation etc). The signals were demodulated about a fixed frequency to generate a quadrature pair which are the XY signals (like a lock-in amplifier).

My real-world application is demodulation of a modulated photodetector signal from nanoparticles moving in a liquid. The phase information in the signal relates to the position of the particles. If the particles are just moving randomly, the XY trace will show a random walk. If there is linear motion such as due to movement in an electric field, the phase will change linearly and the XY trace will rotate. Observing this directly and live on a CRT scope in XY mode is tremendously valuable. It tells me if there are issues with sample or hardware and I can "see" when the particles are moving due to the electric field (which is the primary purpose of the experiment). Without this, I'd be working "blind". It is analogous to the use of vectorscopes for video. For my particular application, the frequencies are in the audio band.

[joeqsmith]

Someone should try this on an HP 546xx. They have a superior X-Y mode. Not bad for 1994...

https://youtu.be/D2uEvRPd4w8

Hunting around this site, I found were this same post was made in 2012 or so.  Someone had the raw file but the links were dead.  Doing a quick Google search, the are still available if we wanted to use this as some sort of benchmark.    Just search for oscillofun.flac. 

Is the goal here just use use a scope for art?   I watched some of these fancy YT videos like the one linked above.

Sometimes science is art

I created the clips with preset waveforms from my function generator (such as chirp, FM modulation etc). The signals were demodulated about a fixed frequency to generate a quadrature pair which are the XY signals (like a lock-in amplifier).

My real-world application is demodulation of a modulated photodetector signal from nanoparticles moving in a liquid. The phase information in the signal relates to the position of the particles. If the particles are just moving randomly, the XY trace will show a random walk. If there is linear motion such as due to movement in an electric field, the phase will change linearly and the XY trace will rotate. Observing this directly and live on a CRT scope in XY mode is tremendously valuable. It tells me if there are issues with sample or hardware and I can "see" when the particles are moving due to the electric field (which is the primary purpose of the experiment). Without this, I'd be working "blind". It is analogous to the use of vectorscopes for video. For my particular application, the frequencies are in the audio band.

If you wanted to detail a setup, we could use this as a test case as well.   Or if you wanted to model it in software and create a raw WAV file, that may make things even easier and possible allow more people to run it for you.   

[joeqsmith]

Link to the oldest thread I could find showing oscillofun.

https://www.eevblog.com/forum/suggestions/audio-signals-quality-in-x-y-mode-what-is-important-what-is-not/msg164751/#msg164751

[james_s]

I use the XY mode on my 465B scope while working on vector arcade game boards, it's a lot more convenient than trying to rig up a fullsized vector monitor which are rare and valuable in addition to being bulky, fragile and requiring some odd voltages.

[joeqsmith]

Oscillofun does appear to a much cleaner source.   I took a few screen shots while it was playing.   

[JohnnyMalaria]

Audio is too restrictive because of the limited bandwidth. It precludes the ability to compare technologies in the 100kHz+ range and sub ~20Hz. It also introduces variation due to different soundcard characteristics such as THD, equalization, DAC jitter etc.

A 2-channel DDS arbitrary waveform generator allows for better control, consistency and wider bandwidth (DC to many MHz). In my case, I have a FeelTech FY2300 that goes to 20MHz. It's basically the same as the 6600 but with a lower max. frequency and a different physical form.

The simplest test is to set up a sine wave on both channels at the same frequency somewhere in the 10s of Hz region with a 90deg phase difference between them. Then introduce a small frequency difference (say 1Hz). In XY mode, an ellipse should appear to "rotate" at 1Hz. Increase the frequency difference and compare the scopes. At some point, the DSO will show its weakness due to the buffered acquisition. Increase the frequency of both channels to the next decade and repeat the frequency difference observations. Repeat this for each decade. You could also have one channel at 3 times the frequency of the other to generate a Lissajous figure and then add small frequency differences to cause rotation.

Now choose an arbitrary waveform for one channel and keep the other as a sinusoid. Do a similar investigation of decade and frequency difference. At lower frequencies, the CRO will prevail but at higher decades the DSO will start to show detail that is too fast for the human eye to pick out on the CRO. This is because the weakness of the DSO at lower frequencies becomes a benefit at higher ones.

In my case, I have a '74 Tektronix CRO and a 2017 Rigol DS1104Z.

In essence, you need to compare both technologies at the frequency range that is important to you. For me, it's a no brainer to use both and pick the most suitable one for your application.

Playing around with different combinations of arbitrary waveforms on each channel as a function of decade and frequency difference is a revealing exercise.

I am curious to learn how the more expensive DSOs fair across the frequency range.

[alsetalokin4017]