Rigol DS1054Z FFT

[engrguy42]

Something in the back of my mind about some folks complaining about the FFT on the DS1054Z?

Am I mis-remembering, or is it a real concern, or was it just some spec monkey, femto-amp accuracy thing that's mostly irrelevant? 

[Adrian_Arg.]

the ds1000z series FFT, serves to get you out of trouble, like most cheap oscilloscopes, if you need a good performance, buy something professional, for less than $ s1000. you can't expect much. It worked for me, like the i2c, isp, serial decoders, but now I see thaht I need a dedicated log analyzer with at least 16 channels. It all depends on what you need for your projects.

in spanish
el FFT de la serie ds1000z, sirve para sacarte de un apuro, como la mayoria de los osciloscopio baratos, si precisas una buena prestacion, compra algo profecional, por menos de u$s1000. no se puedes pretender gran cosa.  A mi me sirvio, al igual que los decodificadores i2c, isp, serie, pero ahora veo que preciso un anaizador logio dedicado de por lo menos16 canales. todo depende de lo que precises para tus proyectos.

[2N3055]

Something in the back of my mind about some folks complaining about the FFT on the DS1054Z?

Am I mis-remembering, or is it a real concern, or was it just some spec monkey, femto-amp accuracy thing that's mostly irrelevant? 

DS1054Z FFT is slow, max 16Kp (used to be worse ), but worst part about it is confused user interface where it is very hard and slow to setup FFT properly.
Could be used in a pinch, but a bit annoying.  If you want to use a FFT on a regular basis, some other scope would be better choice..

[engrguy42]

the ds1000z series FFT, serves to get you out of trouble, like most cheap oscilloscopes, if you need a good performance, buy something professional, for less than $ s1000. you can't expect much. It worked for me, like the i2c, isp, serial decoders, but now I see thaht I need a dedicated log analyzer with at least 16 channels. It all depends on what you need for your projects.

in spanish
el FFT de la serie ds1000z, sirve para sacarte de un apuro, como la mayoria de los osciloscopio baratos, si precisas una buena prestacion, compra algo profecional, por menos de u$s1000. no se puedes pretender gran cosa.  A mi me sirvio, al igual que los decodificadores i2c, isp, serie, pero ahora veo que preciso un anaizador logio dedicado de por lo menos16 canales. todo depende de lo que precises para tus proyectos.

Thanks, but what specifically do you mean by "good performance"? I'm curious about actual limitations, not generalizations. Does anyone have some actual data that shows where it misses the mark?

[engrguy42]

Something in the back of my mind about some folks complaining about the FFT on the DS1054Z?

Am I mis-remembering, or is it a real concern, or was it just some spec monkey, femto-amp accuracy thing that's mostly irrelevant? 

DS1054Z FFT is slow, max 16Kp (used to be worse ), but worst part about it is confused user interface where it is very hard and slow to setup FFT properly.
Could be used in a pinch, but a bit annoying.  If you want to use a FFT on a regular basis, some other scope would be better choice..

Thanks. By "confused and hard to set up" are you referring to the need to change from "trace" to "memory" to make sure it uses the max # of points? From what little I've used it so far that seems to be the only annoyance for me. That and trying to set up a frequency/division number that's a reasonable/intuitive number (rather than like 625kHz per division or something like that)

[2N3055]

That is not annoyance at all. But many other things are..

- You can't actually set number of bins. It won't show number of bins it magically chose.
- You set center freq and freq resolution by magic process of twiddling time base and several parameters..
 from  SCPI manual :
When the FFT mode is set to TRACe, the range of <cent> is from 0 to (0.4 x FFT
sample rate). Wherein, FFT sample rate equals to screen sample rate, that is,
100/horizontal timebase.
When the FFT mode is set to MEMory, the range of <cent> is from 0 to (0.5 x FFT
sample rate). Wherein, FFT sample rate equals to memory sample rate
(:ACQuire:SRATe?).
 Step = Horizontal Scale of the FFT operation result/50.

You figure that out. In the end you keep twiddling knobs until you get what you want.. Or not..

As I said, I managed to get usable FFT results few times but it wa painful.
Usable when used very rarely but not for everyday use.

[Fungus]

Something in the back of my mind about some folks complaining about the FFT on the DS1054Z?

It's definitely not great... but neither is anything else in the price range.

You can see where frequency peaks are, the worst part is the controls/adjustments.

It might help if we knew what you wanted to FFT.

[Fungus]

Thanks. By "confused and hard to set up" are you referring to the need to change from "trace" to "memory" to make sure it uses the max # of points? From what little I've used it so far that seems to be the only annoyance for me.

The major annoyance for me was the way it changes all your settings if you dare to touch the horizontal timebase knob.

[engrguy42]

2N3055, Not sure exactly what your issue is, but here's an example where I fed a 100kHz square wave into the DS1054Z and did an FFT. That's some yummy FFT goodness...

You can see it clearly shows the odd harmonics. And not sure if you're doing this but if you press down on the Intensity knob when you're in the Center setting it will automagically put the main/fundamental frequency in the center.

EDIT: Actually what it does is set the Center frequency at the same setting as the Hz/Div.

This was in Trace mode, BTW.

I'm guessing your issues are for some more complex waveforms or something?

 

[nctnico]

Something in the back of my mind about some folks complaining about the FFT on the DS1054Z?

It's definitely not great... but neither is anything else in the price range.

Actually there is. Maybe even with a spectrum analyser-ish interface. But even without you can get 1Mpts FFT for the same price.

In the end it depends on what you need but the more FFT points the more detail you can get. For example aliasing products which are very close together.

[2N3055]

2N3055, Not sure exactly what your issue is, but here's an example where I fed a 100kHz square wave into the DS1054Z and did an FFT. That's some yummy FFT goodness...

You can see it clearly shows the odd harmonics. And not sure if you're doing this but if you press down on the Intensity knob when you're in the Center setting it will automagically put the main/fundamental frequency in the center.

EDIT: Actually what it does is set the Center frequency at the same setting as the Hz/Div.

This was in Trace mode, BTW.

I'm guessing your issues are for some more complex waveforms or something?

Why are you like that? You asked people question. You got answers. And now you are trying to prove we are stupid because you think FFT  on DS1054Z is just fine in your opinion and everybody that doesn't think so are either not very smart (couldn't figure out how to use it) or are just "equipment snobs" that unnecessarily want something that is , in your mind, equivalent of audio-foolery.

Read my answer again. I said, twice, that FFT on DS1054z is perfectly usable, but cumbersome and slow to setup and has very limited frequency resolution (which does or doesn't have to be a problem).

And here is how it should look, and that took me whole 10 sec to setup:

So yes, if you need good FFT DS1054Z is not very good choice. If you need  FFT occasionally it will probably do just fine..
Regards,
Sinisa

[engrguy42]

2N3055, Not sure exactly what your issue is, but here's an example where I fed a 100kHz square wave into the DS1054Z and did an FFT. That's some yummy FFT goodness...

You can see it clearly shows the odd harmonics. And not sure if you're doing this but if you press down on the Intensity knob when you're in the Center setting it will automagically put the main/fundamental frequency in the center.

EDIT: Actually what it does is set the Center frequency at the same setting as the Hz/Div.

This was in Trace mode, BTW.

I'm guessing your issues are for some more complex waveforms or something?

Why are you like that? You asked people question. You got answers. And now you are trying to prove we are stupid because you think FFT  on DS1054Z is just fine in your opinion and everybody that doesn't think so are either not very smart (couldn't figure out how to use it) or are just "equipment snobs" that unnecessarily want something that is , in your mind, equivalent of audio-foolery.

Huh? Again, I'm trying to find out specifically what problems people are having. Specifically. You say "many other things" are an annoyance. And then you say you set something up in 10 seconds. And don't provide specifics about "many other things".

Are there actually technical issues or bugs? Is it just counter-intuitive with certain procedures? I'm trying to get past the meaningless generalizations like "good".

I'm not trying to prove anything, just find out what I'm missing. I never said FFT on DS1054Z is fine. As I said, I haven't used it much, and when I have it seems to work okay. Personally I don't really give a damn about the FFT since I rarely use it.

Geez people lighten up. 

[2N3055]

Read my answer and look at the image.
My Picoscope does FFT on 1M bins (2 Mpoints) with resolution bandwidth of 29,8 HZ on bandwidth of 31 MHz (62MS/s sampling rate), and does single FFT in 33,55 ms on that dataset.
Look at the dynamic range.. Also look how both axis are actually notated in meaningful numbers. Also I made a zoom in from 31 MHz instantaneous bandwidth to 1 MHz to better see area in question..

Set that up on DS1054Z and get back to me.

So, yes in comparison, FFT on DS1054Z is not very good. And again, i did use it few times, when I had it and it helped me find a problem. So, in a pinch it was useful and definitely good to have.

[engrguy42]

Read my answer and look at the image.
My Picoscope does FFT on 1M bins (2 Mpoints) with resolution bandwidth of 29,8 HZ on bandwidth of 31 MHz (62MS/s sampling rate), and does single FFT in 33,55 ms on that dataset.
Look at the dynamic range.. Also look how both axis are actually notated in meaningful numbers. Also I made a zoom in from 31 MHz instantaneous bandwidth to 1 MHz to better see area in question..

Set that up on DS1054Z and get back to me.

So, yes in comparison, FFT on DS1054Z is not very good. And again, i did use it few times, when I had it and it helped me find a problem. So, in a pinch it was useful and definitely good to have.

Got it. So from your perspective, the DS1054Z may not be a good choice for those who need greater/finer frequency resolution, and are willing to pay for it. But there are apparently no major/known bugs with the FFT, and for those who don't need that resolution (and/or aren't willing to pay for it) it might be fine. Albeit a bit cumbersome and/or annoying.

Thanks. Anyone else?

[engrguy42]

BTW, has anyone tried taking the CSV output of the DS1054Z and used better FFT software on a PC? I wrote an FFT software in C# not long ago, but with minimal features.

Of course that won't change the resolution limitations, but I'm wondering if it might provide a way to overcome some cumbersome aspects of the 1054Z.

Just curious...maybe it's just not worth the trouble....

[2N3055]

BTW, has anyone tried taking the CSV output of the DS1054Z and used better FFT software on a PC? I wrote an FFT software in C# not long ago, but with minimal features.

Of course that won't change the resolution limitations, but I'm wondering if it might provide a way to overcome some cumbersome aspects of the 1054Z.

Just curious...maybe it's just not worth the trouble....

That is good question.
There was an effort few years ago, that completely sidestepped built in FFT, but was simply taking long captures to PC and did all the processing on PC.
take a look at:
https://hackaday.com/2015/09/22/a-better-spectrum-analyzer-for-your-rigol-scope/
People had some decent results with it. And yes, in that config, you get everything better, including RBW, because you can grab loong capture and do FFT on it.....

[engrguy42]

Here's a shot of the FFT app I wrote in C#. Input signal on the top and FFT on the bottom. The intent was solely to measure harmonics of 60Hz from a data acquisition unit. It just grabbed a signal via USB from the data acquisition unit then did the FFT.

Hmm....I wonder if I could tweak that for some real business... 

But then I'd have to figure how to grab the data via LAN from the DS1054Z. Though I've got that crappy UltraSigma/UltraScope stuff working...I wonder how easy it would be to catch the stream from the DS1054Z...

Hmm...   

[Mechatrommer]

I'm not trying to prove anything, just find out what I'm missing. I never said FFT on DS1054Z is fine. As I said, I haven't used it much, and when I have it seems to work okay. Personally I don't really give a damn about the FFT since I rarely use it.

if you ask me, FFT in DS1054Z is nothing more than a toy that hogs DSO resources... a DS1054Z owner here. thats why i built my VisaDSO, with some basic measurements capability, motion signal recording, view long memory signal, simple project reporting, text annotation and images, 8 and 16 bytes CSV import/export etc. above all, with capability of up to 200Mpts FFT made virtually any 1Mpts FFT scopes out there moot. shown is 20Mpts FFT DC - 500MHz span, 50Hz resolution. the 24Mpts data was captured from DS1054Z. ps: am currently doing a specialized utility SW to process signal, fft, post processs, inverse fft and export the data out in CSV format compatible for VisaDSO import/viewing.

[Fungus]

BTW, has anyone tried taking the CSV output of the DS1054Z and used better FFT software on a PC? I wrote an FFT software in C# not long ago, but with minimal features.

Of course that won't change the resolution limitations, but I'm wondering if it might provide a way to overcome some cumbersome aspects of the 1054Z.

Just curious...maybe it's just not worth the trouble....

The frame rate will be horrible. 

[Someone]

BTW, has anyone tried taking the CSV output of the DS1054Z and used better FFT software on a PC? I wrote an FFT software in C# not long ago, but with minimal features.

Of course that won't change the resolution limitations, but I'm wondering if it might provide a way to overcome some cumbersome aspects of the 1054Z.

Just curious...maybe it's just not worth the trouble....

The frame rate will be horrible. 

The example linked above (https://hackaday.com/2015/09/22/a-better-spectrum-analyzer-for-your-rigol-scope/) had it running at 2 updates per second for 16k data length. Compared to the internal speed its similar:

DPO4000 1M point FFT, 0.08 update/second (greater than 13 seconds between FFT completions!)

DPO4000 100k point FFT, 1 update/second

DSOX1000 64k point FFT, 3 updates/second
DS1054Z 64k point FFT, 1 update/second

DPO4000 10k point FFT, 10 updates/second

DSOX1000 1k point FFT, 60 updates/second
DS1054Z 1k point FFT, 3 updates/second
DS4000 1k point FFT, 8 updates/second
DPO4000 1k point FFT, 24 updates/second

So the benefit there is a much richer interface and possibility to customise/enhance the capabilities.

[technogeeky]

The 1054z FFT is serviceable. It can be difficult to get the settings right; you do not have any control of some things you'd normally want to control (like RBW, or the number of bins).**  When you change the horizontal timebase, you may as well just reset the configuration and start over. You will sometimes need to intentionally turn up the gain and lower the noise floor so low that it's off-screen so you can see enough detail in peaks.

The Trace mode is OK, and the Memory mode is much nicer but is not available in all timebases. I often turn the Memory mode on, change the horizontal timebase until it is turned back off, and then go back and operate in Memory mode.

If you are capturing the screen to a computer, the FFT won't actually be recalculated if you capture frames as fast as possible. If you set the rate a little slower, you can give the oscilloscope time to update the FFT and capture at the right speed (capture taken every 600ms as below).



I have been happy with it. If you need to catch faster transient signals, it's the wrong instrument. If you need to have good resolution at low frequencies -- you probably have the wrong instrument.


**  I'm surprised none of the 1054z firmware hacks have addressed this yet.

[gf]

The frame rate will be horrible. 

Why do you care too much about frame rate in this context?

If you calculate FFT you imply a periodic signal. Both, the time domain display and the Fourier spectrum are not supposed to change from frame to frame when the signal is periodic (the only expected difference is the random noise component which is not periodic). In this case, even a single-shot capture is fine - just capture as many samples as feasible (or required for the particular use case) and calculate the FFT once.

[ If the FFT plots in subsequent frames happen to differ significantly, then I'd rather say that the prerequisite requirements for a valid FFT are not fulfilled - either the signal is not periodic, or the FFT was calculated from a too small number of points. IMO it is always better calculate the FFT once from a large number of samples, than calculating it again and again from a too small number of samples. ]

Short-time FFT is a different issue, of course, deliberately aiming at a spectrogram where the spectrum is a function of time, by explicitly using shorter buffers, thus explicitly renouncing frequency resolution.

[Mechatrommer]

The frame rate will be horrible. 

Why do you care too much about frame rate in this context?

maybe he was refering to "CSV download" time. trying to export full 24MPts CSV format into USB drive from DS1054Z is not that horrible, its just that abysmally take ages to complete. downloading raw byte data to/from PC SW will take about 30-60 seconds, so its 1-2 fpm (frames per minute) is not so bad if smaller FFT is adequate such as maybe 10K to 1Mpts, then fpm/fps can greatly being improved using the PC SW

[gf]

maybe he was refering to "CSV download" time. trying to export full 24MPts CSV format into USB drive from DS1054Z is not that horrible, its just that abysmally take ages to complete. downloading raw byte data to/from PC SW will take about 30-60 seconds, so its 1-2 fpm (frames per minute) is not so bad

OK, I understand...

[Mechatrommer]

but you have a valid point on a device should output a stable spectrum with very little noise variation, so hobby FFT work on DS1054Z should be feasible even at very slow rate. but try to download large amount of CSV data to USB drive, you'll know its simply unacceptable. but there are time when fast update rate is required, when development time is critical usually in pro/paid work, when you change input signal, you need to see output spectrum instantly, or things like intermittent signal or those WiFI/radio digital communication signal (thesignalpath covered this in his video), but for this, DS1054Z is completely out of the league. we'll need a proper SA device, even sometime a normal cheap/used SA is incapable, large RBW SA is required for modern WIFI/broadband signal analysis (OT).