I know a guy who manually swapped uC from SIM card to satellite TV card (and it worked btw), now he's out of business, and for all his time he had Russian S1-73. Now I have it.
Small correction: It does not compute the general DFT algorithm, but the FFT is the algorithm that it uses. That is why the FFT-points are always in always a power of 2^n, due to the nature of the FFT algorithm. The general Discrete Fourier Transform does not have this limitation, but takes far more computing power.
Its intense, these days trying to find logic in the aspect of FTT and have it blasted on an OLED screen, and need to dial it down,, as the brightness and radiating heat is too much and likely not healthy for these organic subpixels. i.imgur.com/O6KLGXy.jpg (use a Miniware DSD212 as a crude signal gen)
As folks have mentioned, the DS1054Z is capable of a lot better - but it's a trick since you can't directly set the sample rate and adjusting the center frequency can take a lot of knob twisting. On Dave's signal you are shooting for a sample rate of 5Msps to see a symmetric plot or 2 if you want a better plot that just shows the left side (nothing past 1MhH). Use auto to get a nice stable waveform and fine adjust to fill the screen vertically. Under acquisition set the depth to the first value larger than 16k (depending on whether you have 12 or 24M of memory). The FFT memory mode supports 16K samples so there is no benefit to more depth. I can't see a difference between normal/ave/hires but peak doesn't seem to work well. Under the math function turn on FFT and select Hanning window and memory FFT (rather than screen FFT) on the second settings page. The trick at this point is to walk the sample rate down to 5Msps one step at a time by adjusting the horizontal timebase (CCW). Set FFT Hz/Div to its largest available value and the center frequency to 1Mhz. If you don't keep the Hz/Div at max or you skip steps it can take a million turns to move the center frequency. Always adjust the center even if it shows 1Mhz because it LIES and sometimes doesn't update when you change the timebase. Once you've gotten the sample rate down to 5Msps and the center at 1Mhz you can reduce your Hz/Div to 5k. If you get to a point where the center won't go to 1Mhz check that you are still in memory mode with Hanning window. It sometimes returns to defaults for no apparent reason. Check these again at the end just to be sure because the plot will be terrible if Hanning is off. Adjust the vertical size/position of the plot and you are done.
Great job as always. One need consider the cost of each DSO, parenthesis price is list price, otherwise discounted: Rohde & Schwarz HMO1202 Series;1232 $US2300 Tektronix MDO3000 as shown MDO3014 (US$4,110) Keysight 3000X Touch, MSO 3054T ($13,000) Lecroy WaveJet 354 Touch $US5000 GW Instek GDS-1104B $700 ($830) Rigol as 1054z ($US400) Rigol DS2000 as DS2202 ($US1790) You'd expect the more expensive DSOs to give you an FFT closer to a SA but it seems the bang for buck maybe the Instek as despite its poor appearing UI, its keystrokes are similar to the Tek FFT shown. The 2 CH or 50 MHz version street price for much under $400 and is identical except for firmware crippled bandwidth and reduced analog channels.
If a scope doesn't have enough computational power for a good DFT one thing you can do is transfer the waveform to a computer and do the DFT with LabVIEW, or some other software. It's been a long time since I've done this, but I think I remember it working really well.
Dave you need to go to page 2 on the FFT and set it to MEMORY Mode when using the FFT or your just looking at the TRACE data. The memory will have a tremendous amount more info to calculate the FFT off and will give a great result! Also set to 5ms on the horizontal!! This is for the Rigol DS1000z of course.
Nice video as always. One little technicality: at 4:35, you asked whether the FFT is done as a discrete Fourier transform (DFT). The answer should be yes no matter what, because FFT (Fast Fourier transform) is an algorithm that computes a DFT in a computationally efficient way. A DFT is by definition for discrete signals, i. e. lists of numbers. This is not 100% the same as a regular Fourier transform (defined on continuous signals), but of course mathematically these concepts are closely related.
Hi Dave, nice comparison. Just FYI, at 8:50 on HMO, you can actually toggle Horizontal control from FFT to time waveform by pressing on the Time/Div knob. Then you will notice, HMO can also do gated FFT. Btw, apart from the ease of use across different entry level scope, it would be nice to compare on FFT performance on accuracy, noise floor and bandwidth flatness.
Reuters latest. 21.07 GMT. Confirmed reports arrived this evening showing Dave Jones actually ''turning it on'' before taking it apart. A group of investigators from Interpol arrived in Sydney this morning and went directly to the offices of Mr Jones. Any repercussions will be reported here and it is expected he (Mr Jones) will defend his position with a stuffed wombat, as usual.
I found a way to make the rigols output much more usable: In the second menu panel of math go to "mode" and set it to "memory" instead of "trace". Then in the "acquire" settings menu, set the memory depth manually to a higher value. The result isnt top quality, but at least better than with the default settings
Dennis: Can you expand on this on FFT performance of the Rigol with any possible photos or other? I would like to see the Rigol (preferably the DS2000) compare more favorably to the other Scopes in this group. I also think Dave was a bit subjective in this review. Isn't Rigol giving him enough free stuff or what?
@@aklankrisz Rigol uses software -- of what's on the display. That said, the Rigol DS2000 certainly shows a better display than what was shown in this video. Dave is ignoring controls right in front of him. That's what happens when you are a 'reviewer' as opposed to a 'user' and/or 'owner'.
The somewhat higher end Rohde & Schwarz RTE and RTO scopes really blow all these out of the water with their FFT. It was one of the main selling points that made me settle on an RTE for my lab.
I believe the Textronics does the FFT on ALL of the sample points. If you reduce the number of sampling points, then the resolution of the FFT is decrease and the speed is increased. I believe 100K works rather well.
I'd be interested in seeing an updated version of this topic. I was in the market for my first scope recently and got a siglent sds812x hd which will be delivered day after tomorrow. It claims to have a 2Mpts fft which i can't wait to dive right into to test some of my radios.
We used to have some old school 400 MHz scopes when I was in the military. Loved those old CRT displays. I have worked on modern FFT digital scopes...but, not as much raw fun as the old ones. Cheers!
Johann Carl Friedrich Gauss - pure genius discovered FFT, but didn't have much application for it. FFT was rediscovered for cold war nuclear disarmament verification and useful for a million other things! Very in formative as usual, Dave - we all owe you big time - Merry Christmas!
How does FFT help the verification of cold war nuclear disarmamament? Do they check for radioactive stuff in the atmosphere and show it on a FFT graph to see if anyone tested their nukes? This seems quite interesting
At least the LeCroy got the name of the window right :) You know, the one named after Julius von Hann, and doesn't confuse it with the one named after Richard Hamming
Great comparison, even when you called my scope a toy. I wonder how fast it would get if use a computer loading data through USB and doing fft on the full sample window.
The Rhode & Schwarz claims to use a "Hanning" window. There is no "Hanning" window. There is a Hann window and a Hamming window, but no "Hanning" window.
+rfmerrill en.wikipedia.org/wiki/Window_function#Hann_.28Hanning.29_window. # You say "Hann" window, I say "Hanning" window... let's call the whole thing off!
Dave creams himself over the 128kpt R&S FFT and takes a huge dump on the GW Instek 1Mpt FFT. I'm guessing Dave is looking to score some sweet sweeet R&S free test gear for his lab.
Dave, you do not need to turn off FFT on HMO, you can push the Time/Div knob then you will get to the upper time domain window, push again you can set gated FFT window size and possition, push third time goes back to the bottom spectrum window.
I've got the 64Xi model which also calls the Hanning for VonHann. It must be some LeCroy jargon. You can get used to it once you know what it is, but yeah, it is weird for the first time.
I suspect one of the reasons the Rigols don't look very good is the poor phase noise on the sampling clocks (#683, #699). Would also explain the jumpiness on the 2202. Would be nice if Dave could do some tests to show how stable the sampling clocks are across different scopes. Even the DSA815's phase noise seems poor - note that the noise floor is only -55 dBc at low offsets, while the MDO3104 (in RF mode) had better than -80 dBc.
John Ridley Well, the problem doesn't seem to be the memory. It's not making use of the memory it has. Currently you can either use the displayed waveform or you can use all datapoints (of the waveform itself) in memory. But if you use memory, it won't let you zoom in anywhere near as far, so it's mostly useless. What you CAN'T do is use a big chunk of the waveform memory, and evaluate it at enough frequencies to get a detailed output from the FFT. That probably has to do with a lack of processing power, but even at 1 update per 10 seconds, it would still be more useful than what we have. I'm not THAT salty about it, but it's a feature I will likely want to use in the very near future.
You mentioned that a manufacturer would implement the DFT in different ways, but shouldn't the FFT of any method take the least processing power? I can only imagine it would be small variations on that particular theme.
RE: MDO3104 being slow, you might have a really wide frequency bandwidth for an FFT? I just sold my BladeRF to a friend (wasn't using it), it had very nice FFT output.
+Jesus von Nazaret it is some work around for ds1054z owners and works not only for audio. Save waveform to usb flash and use FFT Viewer programm to analyse signal
+Jesus von Nazaret The test signal in this video was a reasonably demanding "RF" case -- it wasn't baseband, it was up at 1 MHz, yet the FM frequency was up at 5kHz. So you need a long timebase to see the 5kHz behaviour, but you need a high sample rate to see the 1MHz signal. These two factors combined mean that a large number of FFT points are required in order to catch the 1MHz while seeing enough to catch the 5kHz. Audio, on the other hand, is a baseband signal. You're not interested in anything beyond 20kHz, so your 1024-point FFT can be spread out nice and wide, allowing you to catch 40Hz signals at the same time. That's basically the entire audio spectrum on a single screen. You can measure any frequency component with an accuracy of 0.1% (roughly speaking), provided you optimize your timebase. Incidentally, MP3 audio compression used 1024-point FFTs, so anything that survives highest-quality MP3 compression is, loosely speaking, fair game for the Rigol to analyze. What *specific* things do you want to do with "audio stuff"?
***** Interesting, I stand corrected. Didn't consider that a totally different circuit was being switched in, which obviously throws any trends out the window. How did you get 24 meg points of data out? Ethernet, USB-computer, or USB-drive? I find it very difficult and/or insanely slow no matter how I do it? I've noticed the same thing about Hi-Res mode, you only get 8 bits worth in stored data. I presume you get given the data fit for the 2^8 high screen, the extra bits are (correctly) considered worthless for on screen display and are thus discarded ASAP. Annoying! Your noise floor should be pretty sweet with 24M FFT, though?
Dave, did you check to see if every scope was running the latest firmware /os before the test?,,because there could be a lot of changes with that if the fft run in the cpu
hi, was your rigol original or "hacked", so with full bandwidth? or doesn't the number of points have anything to do with the limited functions of the "base" unit?
A 128k point FFT implemented in the usual fashion would require 17 passes over the data for 2.2 million operations. That said, you can parallelize it enormously. You could potentially run as many as 32k threads in parallel with no contention and little in the way of synchronization.
Hey Dave, are you running the latest firmware on that Tek? I remember there was a firmware update to fix RF self-test failure if there was a signal present on startup.
Great Video still but the Rigol was not treated fairly! You needed to turn on "Deep Memory FFT" on the DS1054Z and the FFT will come to Life. This feature is on the next page in FFT mode. You most likely had it set to TRACE and only showed the data points that were displayed in the trace window. This upgrade to the firmware was released 1 year before your video. I know It's hard to know all the in's and out's of all the equipment!
How useful would DS1054Z be in the 20-20khz range, I understand with 1mhz center point example..but would it show data points in the lower frequency range? Or would it again only show the carrier frequency and not much detail? Interested in using it just for the audio spectrum, so curious how it works between 20hz to 20khz ...
Dave on that Rigol 1000 zed did you have all the upgrades and the latest firmware? If not is it possible you could do the upgrades and firmware and show FFT again please.
Dave, visually the carrier frequency and sidebands seem fine, but how accurate are the amplitudes on each scope? Any different detector settings on those scopes? What are the noise floor levels on each scope? Kudos on the video :)
one of the performance comparison that I want to see is the maximum frequency that these oscilloscopes can do ... because some can go (as nyquist says) to 2.5Ghz... (I know whit atenuation) but I think that it is on propose limited ... ex:I used to have an agilent DSO-3202 that had best fft performance (in frequency range) than an DSO-1024a that I have now .. but unfortunately I didn't make any specific comparison (maby just impression ???) If you coud let us know the maximum allowed frequency they show) best regards from Brazil ...
Hi Dave, not sure if you have a Siglent scope. How does this one hold up WITH FFT. second question, I am a electronics hobbyist and software engineer by trade. What oscilloscope would you recommend for us hobbyists. Thanks keep up the good work. PS apologise if you have covered siglent scope already, if so point me in the right direction.
It would be nice to hear more on windowing in FFT and what's best for what i know i just use hanning for everything Also i must say i still hate the Z on the 1054Z ... it looks like someone took a marker to it
FFT stands for Fast Fourier Transform, its an algorithm for calculating the DFT (Discrete Fourier Transform) of a signal. The DFT is essentially the digital version of a Fourier transform, which if you apply it to a signal that is measured with respect to time gives you the frequency components of that signal (or more precisely what combination of sine waves at each frequency would add up to the original signal). The frequency domain representation of a signal is often referred to as its spectrum hence where spectrum analyzers got their name. In short an FFT is an algorithm for getting the spectrum of a digital signal.
Dave, great vid as always. I picked up a 1054Z last year but wasn't aware of its limited FFT function. With regards to audio, 20Hz-20kHz, is this scope still lacking severely or is it suitable for analyzing components within an audio circuit..ie, cables, etc? Pr should I look to a dedicated spectrum analyzer. Thanks
Thanks for disappointing me about my Rigol 1054Z's FFT capability, Uhhh, I think. :/ The cheap SDRs seem to produce a much better display resolution, so why these scopes are so wimpy is a big question.
Just a slight nitpick - WHITE balance has to do with COLOUR (hence the name, hint-hint); the term you're looking for is EXPOSURE (and/or APERTURE), i believe, which has to do with the amount of LIGHT :) en.wikipedia.org/wiki/Color_balance en.wikipedia.org/wiki/Exposure_value en.wikipedia.org/wiki/Aperture
I know this comparison is ancient but No waterfalls? But even in 2024 I could find no comparisons or reviews of new models' waterfall performance. And yet, the new 2024 Uni-T UPO1204 has waterfall however I could only find one review of that feature and in that review, they had bandwidth maxed out so the waterfall line rate was many seconds which would be useless for measuring the dynamic response of structural modes (from strain gauges) to a transient input. A simple bench test would be a microphone input and then striking a tuning fork or anything that "rings." If that UniT has a functional waterfall feature, it's game over, we have a winner.
Smells like an Update for the Rigols. But I think you are a bit unfair to the rigols, especially the 1054Z... The cost should always be something you have to bear in mind... But what I really don't understand is why those scopes have so little memory and so little processing power - both should be rather cheap these days... I really don't understand why they don't use some multicore 1GHz ARM processor with at least 2GiB of Memory...
+Stefan Payne They probably do not do the FFT in the ARM processor but in an FPGA. But even if they did it on the ARM core they would probably not just go for the next 5$ chip they can find on the local Shenzhen market because they have to run their software for years and don't always want to reinvent the wheel (drivers etc). If you look at ARM chips that are supposed to have a long term availability there is not that much left and what is left is expensive, in particular if you want the quad core version, e.g. the iMX6 from Freescale.
+Alexander Gosselin So what do you actually DO with it? Dave does a lot of scope reviews, but I never really grasp what you would need one for. It can see voltage over time, what can that tell me? Pulses of voltage? What is an average day to day use for a scope?
+Oddzball3 An average day-to-day use is doing electronics. Electronics is the art/science of making signals on wires behave the way you want. Without an oscilloscope, you can't see what those signals look like -- without an oscilloscope, you're quite literally working blind.
+Alexander Gosselin Definitely explains it a bit, thanks for the details. I guess my experience with electronics ends at basically my little radio shack electronics learning lab, and a few simple circuits I have built over the years, so I just dont have the experience with scopes to understand because Ive never worked with complicated electronics.
I can't understand the word before "digital FFT rubbish", but DSA815 has "all digital IF technology"...to make it short, I think it's doing the same thing, just per partes for the whole RF spectrum. Btw you could also link en.wikipedia.org/wiki/Fast_Fourier_transform
Gee look there a shiny new SDG 2122X just begging for a tear down video :) It the Rigol you are using one with or without the firmware update to "help" the noise problem on the 1 GHz clock (5, 10, 20 ms jitter issue) ? The same spurs that nuke the jitter will be an issue doing an FFT ...
![Lp. Сердце Вселенной #60 РОЖДЕНИЕ ЛОЛОЛОШКИ [Финал] • Майнкрафт](http://i.ytimg.com/vi/YoR0pAV9FVQ/mqdefault.jpg)
5 scopes, spectrum analyzer, signal generator. And i'm sitting here with my tiny soviet analog scope.
I know a guy who manually swapped uC from SIM card to satellite TV card (and it worked btw), now he's out of business, and for all his time he had Russian S1-73. Now I have it.
Not to mention how much some of his scopes cost.
14:41 Isn't that what dB stands for ???
LMFAO
exactly and that's where terms like
3dB UP or down come from
hehe
14:40 - "It looks like dick and balls". You don't make me laugh very often but when you do, I laugh hard.
+ElGatoLoco698 Reason enough to click that subscribe button for me!
+ElGatoLoco698 If Dave ever does a new trailer for his channel, this needs to be in it.
+ElGatoLoco698 Yes, I'm glad I finished swallowing my diet cola, or I'd have a screen to clean.
+ElGatoLoco698 To me that waveform appeared like a human hand flipping the bird.
+ElGatoLoco698 Good thing i wasn't drinking any liquids, otherwise i would have ruined my keyboard =)
Small correction: It does not compute the general DFT algorithm, but the FFT is the algorithm that it uses. That is why the FFT-points are always in always a power of 2^n, due to the nature of the FFT algorithm. The general Discrete Fourier Transform does not have this limitation, but takes far more computing power.
We have 5 scopes, a spectrum analyzer and a function generator, its dark and we are wearing sunglasses... lets go
+Heads Tails Why didn't I think of that intro?!
+EEVblog Fix the cigarette lighter
+Heads Tails Hit it!
A microphone, yeah i can see that
Its intense, these days trying to find logic in the aspect of FTT and have it blasted on an OLED screen, and need to dial it down,, as the brightness and radiating heat is too much and likely not healthy for these organic subpixels.
i.imgur.com/O6KLGXy.jpg (use a Miniware DSD212 as a crude signal gen)
As folks have mentioned, the DS1054Z is capable of a lot better - but it's a trick since you can't directly set the sample rate and adjusting the center frequency can take a lot of knob twisting. On Dave's signal you are shooting for a sample rate of 5Msps to see a symmetric plot or 2 if you want a better plot that just shows the left side (nothing past 1MhH). Use auto to get a nice stable waveform and fine adjust to fill the screen vertically. Under acquisition set the depth to the first value larger than 16k (depending on whether you have 12 or 24M of memory). The FFT memory mode supports 16K samples so there is no benefit to more depth. I can't see a difference between normal/ave/hires but peak doesn't seem to work well. Under the math function turn on FFT and select Hanning window and memory FFT (rather than screen FFT) on the second settings page. The trick at this point is to walk the sample rate down to 5Msps one step at a time by adjusting the horizontal timebase (CCW). Set FFT Hz/Div to its largest available value and the center frequency to 1Mhz. If you don't keep the Hz/Div at max or you skip steps it can take a million turns to move the center frequency. Always adjust the center even if it shows 1Mhz because it LIES and sometimes doesn't update when you change the timebase. Once you've gotten the sample rate down to 5Msps and the center at 1Mhz you can reduce your Hz/Div to 5k. If you get to a point where the center won't go to 1Mhz check that you are still in memory mode with Hanning window. It sometimes returns to defaults for no apparent reason. Check these again at the end just to be sure because the plot will be terrible if Hanning is off. Adjust the vertical size/position of the plot and you are done.
Great job as always. One need consider the cost of each DSO, parenthesis price is list price, otherwise discounted:
Rohde & Schwarz HMO1202 Series;1232 $US2300
Tektronix MDO3000 as shown MDO3014 (US$4,110)
Keysight 3000X Touch, MSO 3054T ($13,000)
Lecroy WaveJet 354 Touch $US5000
GW Instek GDS-1104B $700 ($830)
Rigol as 1054z ($US400)
Rigol DS2000 as DS2202 ($US1790)
You'd expect the more expensive DSOs to give you an FFT closer to a SA but it seems the bang for buck maybe the Instek as despite its poor appearing UI, its keystrokes are similar to the Tek FFT shown. The 2 CH or 50 MHz version street price for much under $400 and is identical except for firmware crippled bandwidth and reduced analog channels.
Rigol DSA815 with a Rigol DS1054Z can now be had for £1,300 - that's for both.
If a scope doesn't have enough computational power for a good DFT one thing you can do is transfer the waveform to a computer and do the DFT with LabVIEW, or some other software. It's been a long time since I've done this, but I think I remember it working really well.
Dave you need to go to page 2 on the FFT and set it to MEMORY Mode when using the FFT or your just looking at the TRACE data. The memory will have a tremendous amount more info to calculate the FFT off and will give a great result! Also set to 5ms on the horizontal!! This is for the Rigol DS1000z of course.
Nice video as always. One little technicality: at 4:35, you asked whether the FFT is done as a discrete Fourier transform (DFT). The answer should be yes no matter what, because FFT (Fast Fourier transform) is an algorithm that computes a DFT in a computationally efficient way. A DFT is by definition for discrete signals, i. e. lists of numbers. This is not 100% the same as a regular Fourier transform (defined on continuous signals), but of course mathematically these concepts are closely related.
The Department of Redundancy Department would like to speak with you regarding your FM modulation.
Hi Dave, nice comparison. Just FYI, at 8:50 on HMO, you can actually toggle Horizontal control from FFT to time waveform by pressing on the Time/Div knob. Then you will notice, HMO can also do gated FFT.
Btw, apart from the ease of use across different entry level scope, it would be nice to compare on FFT performance on accuracy, noise floor and bandwidth flatness.
14:40 "It looks like dick and balls!" I laughed way too hard at that :D
Reuters latest. 21.07 GMT. Confirmed reports arrived this evening showing Dave Jones actually ''turning it on'' before taking it apart. A group of investigators from Interpol arrived in Sydney this morning and went directly to the offices of Mr Jones. Any repercussions will be reported here and it is expected he (Mr Jones) will defend his position with a stuffed wombat, as usual.
the lecroy one seems to show a middle finger instead of the real deal
I found a way to make the rigols output much more usable: In the second menu panel of math go to "mode" and set it to "memory" instead of "trace". Then in the "acquire" settings menu, set the memory depth manually to a higher value. The result isnt top quality, but at least better than with the default settings
Dennis: Can you expand on this on FFT performance of the Rigol with any possible photos or other? I would like to see the Rigol (preferably the DS2000) compare more favorably to the other Scopes in this group. I also think Dave was a bit subjective in this review. Isn't Rigol giving him enough free stuff or what?
@@tkarlmann Rigol's FFT is simply not good.
@@aklankrisz Rigol uses software -- of what's on the display. That said, the Rigol DS2000 certainly shows a better display than what was shown in this video. Dave is ignoring controls right in front of him. That's what happens when you are a 'reviewer' as opposed to a 'user' and/or 'owner'.
Is there a chance to see on the picture what you are saying?
@@aklankrisz That's enlightening!
The somewhat higher end Rohde & Schwarz RTE and RTO scopes really blow all these out of the water with their FFT. It was one of the main selling points that made me settle on an RTE for my lab.
I believe the Textronics does the FFT on ALL of the sample points. If you reduce the number of sampling points, then the resolution of the FFT is decrease and the speed is increased. I believe 100K works rather well.
I'd be interested in seeing an updated version of this topic. I was in the market for my first scope recently and got a siglent sds812x hd which will be delivered day after tomorrow. It claims to have a 2Mpts fft which i can't wait to dive right into to test some of my radios.
We used to have some old school 400 MHz scopes when I was in the military. Loved those old CRT displays. I have worked on modern FFT digital scopes...but, not as much raw fun as the old ones. Cheers!
Yep, carrier nulls at mod index values of 2.405 and others - great way to set transmitter deviation and to calibrate deviation meters. Good demo.
I have been going back and watching some of the older EEVblog educational video's as they are great.
I know these EEVblog videos
Johann Carl Friedrich Gauss - pure genius discovered FFT, but didn't have much application for it.
FFT was rediscovered for cold war nuclear disarmament verification and useful for a million other things!
Very in formative as usual, Dave - we all owe you big time - Merry Christmas!
How does FFT help the verification of cold war nuclear disarmamament? Do they check for radioactive stuff in the atmosphere and show it on a FFT graph to see if anyone tested their nukes? This seems quite interesting
At least the LeCroy got the name of the window right :) You know, the one named after Julius von Hann, and doesn't confuse it with the one named after Richard Hamming
+Rob Bowman I was waiting for the VonHann fanboys...
fanboy lol -it's his name dammit! Nice video though, Doug
+Rob Bowman Yes, a shame even R&S got that wrong. "von Hann" makes it both more correct and less prone to confusion with Hamming.
Great comparison, even when you called my scope a toy. I wonder how fast it would get if use a computer loading data through USB and doing fft on the full sample window.
On the Rohde & Schwarz, in FFT mode, just push the timebase knob to switch between frequency view, signal view and window adjust.
I do believe the FFT on the Rigol DS1054Z has had a bit of a firmware upgrade since this video was made or am I mistaken?
Any details on what changed?
@Marcin Berman indeed yes
Dave, how about a video for probing techniques video?
Most guys like me don't know which probing method to use in which situation.
The Rigol is displaying at 25kHz /Div (300kHz/Span) as opposed to 5kHz (50kHz) with the Rode und GW.
Why?
Number of operations in an FFT is not linearly dependent on sample number. So 2 MegaSamples should take more than twice the time of a 1 MegaSamples...
The Rhode & Schwarz claims to use a "Hanning" window. There is no "Hanning" window. There is a Hann window and a Hamming window, but no "Hanning" window.
+rfmerrill en.wikipedia.org/wiki/Window_function#Hann_.28Hanning.29_window. # You say "Hann" window, I say "Hanning" window... let's call the whole thing off!
Dave creams himself over the 128kpt R&S FFT and takes a huge dump on the GW Instek 1Mpt FFT.
I'm guessing Dave is looking to score some sweet sweeet R&S free test gear for his lab.
Dave, you do not need to turn off FFT on HMO, you can push the Time/Div knob then you will get to the upper time domain window, push again you can set gated FFT window size and possition, push third time goes back to the bottom spectrum window.
18:55 - Rigol DS1004Z with Hz/Div of 250kHz is good, why change it to 25kHz and messy the graph?
wow thank you dave you have cured me of my lust for a tek 3000 mdo scope one less thing 2 want
Those Hameg 'scopes are really good, the FFT arent that bad either, at least not for low frequency use :)
VonHann was fun, what were they thinking?
Have you ever looked at Owon 'scopes ?
Dont waist you time
I've got the 64Xi model which also calls the Hanning for VonHann. It must be some LeCroy jargon. You can get used to it once you know what it is, but yeah, it is weird for the first time.
Dave, we're looking forward to a hover board tear down. They're around $200 USD on eBay nowadays. Thanks!
+Distant vErvE Illegal to ship explosives to Oz.
I suspect one of the reasons the Rigols don't look very good is the poor phase noise on the sampling clocks (#683, #699). Would also explain the jumpiness on the 2202. Would be nice if Dave could do some tests to show how stable the sampling clocks are across different scopes. Even the DSA815's phase noise seems poor - note that the noise floor is only -55 dBc at low offsets, while the MDO3104 (in RF mode) had better than -80 dBc.
I don't know what any of the words you say mean but I can't stop watching these videos.
+Noah LeFoot You will turn to the dark side soon.
+EEVblog And we have cookies.
Aww! My poor little 1054Z!
But yea, after I unlocked it, I started playing with the FFT. It's not particularly useful as is.
+Falcrist Maybe Rigol will watch this, and a future firmware flash is all it needs...
John Ridley Well, the problem doesn't seem to be the memory. It's not making use of the memory it has.
Currently you can either use the displayed waveform or you can use all datapoints (of the waveform itself) in memory. But if you use memory, it won't let you zoom in anywhere near as far, so it's mostly useless. What you CAN'T do is use a big chunk of the waveform memory, and evaluate it at enough frequencies to get a detailed output from the FFT.
That probably has to do with a lack of processing power, but even at 1 update per 10 seconds, it would still be more useful than what we have.
I'm not THAT salty about it, but it's a feature I will likely want to use in the very near future.
+Falcrist Yeah, sadly my DS4000 series is crap as well :(Though it seems to be ok being used for audio filters etc.
I wish I had just one of those flat screen scopes. I have a Tektronix 465 that doesn’t work and a 475 that needs alignment.
You mentioned that a manufacturer would implement the DFT in different ways, but shouldn't the FFT of any method take the least processing power? I can only imagine it would be small variations on that particular theme.
the scope lags when u try to scoll it, and we understand, but whay if it just simply takes a snap shot of the current out put when u are scrolling?
in rigol you can manualy set number of points so 17:30 i do not agree
RE: MDO3104 being slow, you might have a really wide frequency bandwidth for an FFT? I just sold my BladeRF to a friend (wasn't using it), it had very nice FFT output.
I guess the Rigol 1054Z's FFT would not be good enough for audio stuff?
+Jesus von Nazaret it is some work around for ds1054z owners and works not only for audio. Save waveform to usb flash and use FFT Viewer programm to analyse signal
+Jesus von Nazaret The test signal in this video was a reasonably demanding "RF" case -- it wasn't baseband, it was up at 1 MHz, yet the FM frequency was up at 5kHz. So you need a long timebase to see the 5kHz behaviour, but you need a high sample rate to see the 1MHz signal. These two factors combined mean that a large number of FFT points are required in order to catch the 1MHz while seeing enough to catch the 5kHz.
Audio, on the other hand, is a baseband signal. You're not interested in anything beyond 20kHz, so your 1024-point FFT can be spread out nice and wide, allowing you to catch 40Hz signals at the same time. That's basically the entire audio spectrum on a single screen. You can measure any frequency component with an accuracy of 0.1% (roughly speaking), provided you optimize your timebase. Incidentally, MP3 audio compression used 1024-point FFTs, so anything that survives highest-quality MP3 compression is, loosely speaking, fair game for the Rigol to analyze.
What *specific* things do you want to do with "audio stuff"?
+korram Do you mean distortion or noise? Reducing signal size is the classic way to *reduce* distortion...
***** Interesting, I stand corrected. Didn't consider that a totally different circuit was being switched in, which obviously throws any trends out the window.
How did you get 24 meg points of data out? Ethernet, USB-computer, or USB-drive? I find it very difficult and/or insanely slow no matter how I do it?
I've noticed the same thing about Hi-Res mode, you only get 8 bits worth in stored data. I presume you get given the data fit for the 2^8 high screen, the extra bits are (correctly) considered worthless for on screen display and are thus discarded ASAP. Annoying! Your noise floor should be pretty sweet with 24M FFT, though?
Dave, did you check to see if every scope was running the latest firmware /os before the test?,,because there could be a lot of changes with that if the fft run in the cpu
Indeed the Tek MSO4032 is quite slow too. I don't remember it being THAT slow, but I haven't used it in a while....
The Keysight version apparently can display Min Hold, Max Hold and Average from your screenshot. Was hoping that you can demonstrate that though.
Dave, you've gotta lock the focus when filming screens! Just a tip :)
hi, was your rigol original or "hacked", so with full bandwidth? or doesn't the number of points have anything to do with the limited functions of the "base" unit?
the exposure thingy setting actually looks really good to me! isn't that's how we see things?
A 128k point FFT implemented in the usual fashion would require 17 passes over the data for 2.2 million operations. That said, you can parallelize it enormously. You could potentially run as many as 32k threads in parallel with no contention and little in the way of synchronization.
Hey Dave, are you running the latest firmware on that Tek? I remember there was a firmware update to fix RF self-test failure if there was a signal present on startup.
+SCR2IGBT Oh, ok, thanks. Have not updated the firmware.
crikey!
Great Video still but the Rigol was not treated fairly!
You needed to turn on "Deep Memory FFT" on the DS1054Z and the FFT will come to Life. This feature is on the next page in FFT mode.
You most likely had it set to TRACE and only showed the data points that were displayed in the trace window.
This upgrade to the firmware was released 1 year before your video. I know It's hard to know all the in's and out's of all the equipment!
How useful would DS1054Z be in the 20-20khz range, I understand with 1mhz center point example..but would it show data points in the lower frequency range? Or would it again only show the carrier frequency and not much detail? Interested in using it just for the audio spectrum, so curious how it works between 20hz to 20khz ...
Dave on that Rigol 1000 zed did you have all the upgrades and the latest firmware? If not is it possible you could do the upgrades and firmware and show FFT again please.
We like it when you waffle on Dave it shows real passion for what you know and love
Dave, visually the carrier frequency and sidebands seem fine, but how accurate are the amplitudes on each scope? Any different detector settings on those scopes? What are the noise floor levels on each scope?
Kudos on the video :)
one of the performance comparison that I want to see is the maximum frequency that these oscilloscopes can do ... because some can go (as nyquist says) to 2.5Ghz... (I know whit atenuation) but I think that it is on propose limited ...
ex:I used to have an agilent DSO-3202 that had best fft performance (in frequency range) than an DSO-1024a that I have now .. but unfortunately I didn't make any specific comparison (maby just impression ???)
If you coud let us know the maximum allowed frequency they show)
best regards from Brazil ...
Do any of them display the phase information of the DFT? (as opposed to just the magnitude of the complex transform?)
Hi Dave, not sure if you have a Siglent scope. How does this one hold up WITH FFT. second question, I am a electronics hobbyist and software engineer by trade. What oscilloscope would you recommend for us hobbyists. Thanks keep up the good work. PS apologise if you have covered siglent scope already, if so point me in the right direction.
It would be nice to hear more on windowing in FFT and what's best for what
i know i just use hanning for everything
Also i must say i still hate the Z on the 1054Z ... it looks like someone took a marker to it
I wonder if anyone really uses the scope FFT for something other than a quick test of the feature?
+VoltLog Yes. I successfully tested, that you can catch FM radio signal on a piece of wire and see it on the FFT plot (with averaging).
Has anyone ever tested the Hantek DSO5202P FFT for comparison? Thanks.
DAVE , what about a review of RIGOL 4000 series? any plans to do one?
Hi Dave, what is that piece of rack mount equipment on your shelf?
Fundamentals Friday: how works a spectrum analyser and digital FFT. I love your Videos .
Hi Dave, if you had to build the ultimate lab, what stuff would you buy, (stuff that worth the money), what your lab dream would contain
Does anyone know how well a siglent SDS1202X-E fares in this regard?
For us semi newbs it would be really helpful if you started a video likes this with what an FFT is... Im still not sure after watching the video.
FFT stands for Fast Fourier Transform, its an algorithm for calculating the DFT (Discrete Fourier Transform) of a signal. The DFT is essentially the digital version of a Fourier transform, which if you apply it to a signal that is measured with respect to time gives you the frequency components of that signal (or more precisely what combination of sine waves at each frequency would add up to the original signal). The frequency domain representation of a signal is often referred to as its spectrum hence where spectrum analyzers got their name. In short an FFT is an algorithm for getting the spectrum of a digital signal.
Dave, great vid as always. I picked up a 1054Z last year but wasn't aware of its limited FFT function. With regards to audio, 20Hz-20kHz, is this scope still lacking severely or is it suitable for analyzing components within an audio circuit..ie, cables, etc? Pr should I look to a dedicated spectrum analyzer.
Thanks
Do you have more knowledge since you have wrote this comment?
Hey Dave, ever got your hands on a Yokogawa DLM2000 type scope? What's your opinion about that?
+dakhaas300 Only used an old model a few times at a test facility. Bit clunky, but had decent performance IIRC.
Thanks for disappointing me about my Rigol 1054Z's FFT capability, Uhhh, I think. :/ The cheap SDRs seem to produce a much better display resolution, so why these scopes are so wimpy is a big question.
Just a slight nitpick - WHITE balance has to do with COLOUR (hence the name, hint-hint); the term you're looking for is EXPOSURE (and/or APERTURE), i believe, which has to do with the amount of LIGHT :)
en.wikipedia.org/wiki/Color_balance
en.wikipedia.org/wiki/Exposure_value
en.wikipedia.org/wiki/Aperture
Is the RS Pro 1100 series the same as the Gw Instek here?
more scopes?
I know this comparison is ancient but No waterfalls? But even in 2024 I could find no comparisons or reviews of new models' waterfall performance.
And yet, the new 2024 Uni-T UPO1204 has waterfall however I could only find one review of that feature and in that review, they had bandwidth maxed out so the waterfall line rate was many seconds which would be useless for measuring the dynamic response of structural modes (from strain gauges) to a transient input. A simple bench test would be a microphone input and then striking a tuning fork or anything that "rings." If that UniT has a functional waterfall feature, it's game over, we have a winner.
Hey this is not related to this topic, but what did you do with ya old pc ??
Smells like an Update for the Rigols.
But I think you are a bit unfair to the rigols, especially the 1054Z...
The cost should always be something you have to bear in mind...
But what I really don't understand is why those scopes have so little memory and so little processing power - both should be rather cheap these days...
I really don't understand why they don't use some multicore 1GHz ARM processor with at least 2GiB of Memory...
+Stefan Payne They probably do not do the FFT in the ARM processor but in an FPGA. But even if they did it on the ARM core they would probably not just go for the next 5$ chip they can find on the local Shenzhen market because they have to run their software for years and don't always want to reinvent the wheel (drivers etc). If you look at ARM chips that are supposed to have a long term availability there is not that much left and what is left is expensive, in particular if you want the quad core version, e.g. the iMX6 from Freescale.
WOW YOU KNOW OscilloscopeS.. I'D LIKE TO SEE SOME OPERATIONS NOW. Hold on ok, I'll be right back?
yes Felix I agree but no body working in a repair workshop would ever use these functions
hey, Dave this is completely off topic but I watched your favorite movie last night, hmm wonder what it could be?
I personally like the rohde&schwarz :-D
me too
Where is siglent sds1202x 200MHZ osliscope?
I have question:
What happen if I measure 100MHZ
With 50MHZ osliscop?
This might sound stupid but what is a oscilloscope I should probably Google it
+super awsome minecrafter I'm with you. If I had to guess it's ghost hunting equipment
+Alexander Gosselin So what do you actually DO with it? Dave does a lot of scope reviews, but I never really grasp what you would need one for. It can see voltage over time, what can that tell me? Pulses of voltage? What is an average day to day use for a scope?
+Oddzball3 An average day-to-day use is doing electronics. Electronics is the art/science of making signals on wires behave the way you want. Without an oscilloscope, you can't see what those signals look like -- without an oscilloscope, you're quite literally working blind.
+Alexander Gosselin Definitely explains it a bit, thanks for the details. I guess my experience with electronics ends at basically my little radio shack electronics learning lab, and a few simple circuits I have built over the years, so I just dont have the experience with scopes to understand because Ive never worked with complicated electronics.
POST failure might be a bad CMOS battery failure?
Im after a unit that will allow me to test the SWR of antennas up to 6Ghz. Can anyone recommend a unit that wont cost $10k?
22:15 Err... No thanks Dave!
I can't understand the word before "digital FFT rubbish", but DSA815 has "all digital IF technology"...to make it short, I think it's doing the same thing, just per partes for the whole RF spectrum.
Btw you could also link en.wikipedia.org/wiki/Fast_Fourier_transform
"It should be indicative" - Dave Jones 2017
thanks for great info as usual sir....
What is the cost of those? New and used?
لماذا يستخدم هذا الجهاز
لدي مثله ولم اعرف لما هو؟؟
اجابة لو سمحتو
voce nao pode doar um osciloscopio aqui pro Brasil ???
My poor Rigol. :)
Gee look there a shiny new SDG 2122X just begging for a tear down video :)
It the Rigol you are using one with or without the firmware update to "help" the noise problem on the 1 GHz clock (5, 10, 20 ms jitter issue) ? The same spurs that nuke the jitter will be an issue doing an FFT ...