Rigol DHO804 Test and Compare Thread

[switchabl]

I'd be curious if all those (non-harmonic) spurs in the screenshot by mawyatt are actually still there if you download the data to a PC and do the FFT offline. I seem to recall that the FFT mode on the DHO4000 that Martin tested did some very weird resampling, potentially resulting in similar artefacts? I think it also showed the wrong units for the RBW. No units at all on the new one?

[2N3055]

I'd be curious if all those (non-harmonic) spurs in the screenshot by mawyatt are actually still there if you download the data to a PC and do the FFT offline. I seem to recall that the FFT mode on the DHO4000 that Martin tested did some very weird resampling, potentially resulting in similar artefacts? I think it also showed the wrong units for the RBW. No units at all on the new one?

That would be interesting to see. People reported that transfer to PC is much faster than on DS1000Z.
And I 'm afraid FFT offline on a PC is at the moment only way to actually take a look at how well ADC performs..

[mawyatt]

I'd be curious if all those (non-harmonic) spurs in the screenshot by mawyatt are actually still there if you download the data to a PC and do the FFT offline. I seem to recall that the FFT mode on the DHO4000 that Martin tested did some very weird resampling, potentially resulting in similar artefacts? I think it also showed the wrong units for the RBW. No units at all on the new one?

Those artifacts came directly from DHO814, as same source (AWG) was used for both scopes. I would expect they should show in data download as well, since they are displayed on screen, unless something really strange is going on with the FFT signal processing and/or display processing. Like 2N3055 mentioned we don't have a lot of direct control of the FFT engine it's very hard to compare and understand what's going on.

We posted both FFT results just to show the effects of spectrum averaging and amplitude and frequency table. The FFT averaging doesn't work properly on the DHO814 and messes up the signal amplitudes, whereas this works similar to doing spectral averaging on a SA with the Siglent SDS2000X+, giving the proper and executed results.

Here's the Siglent without spectral averaging to see if any artifacts show.

Best,

[switchabl]

Those artifacts came directly from DHO814, as same source (AWG) was used for both scopes. I would expect they should show in data download as well, since they are displayed on screen, unless something really strange is going on with the FFT signal processing and/or display processing. Like 2N3055 mentioned we don't have a lot of direct control of the FFT engine it's very hard to compare and understand what's going on.

I was asking specifically because something really strange seemed to be going on with the FFT on the DHO4000 on some settings. And I don't know if that was ever explained/fixed.

[2N3055]

Those artifacts came directly from DHO814, as same source (AWG) was used for both scopes. I would expect they should show in data download as well, since they are displayed on screen, unless something really strange is going on with the FFT signal processing and/or display processing. Like 2N3055 mentioned we don't have a lot of direct control of the FFT engine it's very hard to compare and understand what's going on.

I was asking specifically because something really strange seemed to be going on with the FFT on the DHO4000 on some settings. And I don't know if that was ever explained/fixed.

AFAIK it is still in the same condition.

[Martin72]

From memory, I can confirm that the FFT function of the 804 is no different from the 4204.
I just ran an FFT of a 10khz sine wave from the demo board, pictures to follow.
I can't find an Average function in the menu, so I guess it won't exist.
I just activated peaksearch, max 15 points, it doesn't matter where I put the threshold, all peaks "gather" around the 10khz fundamental.
It may be that the signal from the board is simply bad.

[mawyatt]

From memory, I can confirm that the FFT function of the 804 is no different from the 4204.
I just ran an FFT of a 10khz sine wave from the demo board, pictures to follow.
I can't find an Average function in the menu, so I guess it won't exist.
I just activated peaksearch, max 15 points, it doesn't matter where I put the threshold, all peaks "gather" around the 10khz fundamental.
It may be that the signal from the board is simply bad.

There's a average option under Menu button as Horizontal, Acquisition, Average. This seems to do an average with the waveform before FFT. We got this to work and this is the FFT result. No luck with Cursor Peaksearch in FFT tho. The Siglent does the FFT average after the FFT we believe, more like a SA.

Best,

[Martin72]

I had also discovered the averaging in the horizontal menu, but it didn't have a great effect now.
I must say, however, that I just went into the room and simply measured quick&dirty on it.
On the other hand, the scope also invites you to do this, because there is not much choice...
What I like again is the table display.
What I don't like is that "RBW" has no unit of measurement.
The signal is as mentioned 10khz, I had set to 100ms/div, the display is already quite slow, about like my HD in the same time base.
Attached pictures of the FFT menu.
On the one hand, I find this kind of presentation, everything at a glance, good.
On the other hand, you can no longer see too much of the signal "behind", or it irritates (yes, I know the transparency function).
Also not so great:
You can only change the vertical offset position of the FFT with a button, if the menu is active and you have selected the offset there.
Outside of the menu, you can neither move the FFT up or down with the vertical knob, nor are the 2 action buttons responsible for this, they change span and center.
Only with the fingertouch you can change the position.
The vertical actuator should be active, as with the Siglent.
My HD scope shows impressively how useful an average function is for the FFT, see the two pictures.
And before they say yes again, it is also much more expensive:
The DHO4204 also has no average mode for FFT.

[switchabl]

There's a average option under Menu button as Horizontal, Acquisition, Average. This seems to do an average with the waveform before FFT. We got this to work and this is the FFT result. No luck with Cursor Peaksearch in FFT tho. The Siglent does the FFT average after the FFT we believe, more like a SA.

Acqusition averaging can be a bit misleading because it doesn't just suppress noise but also tones that don't have a stable phase relationship with the trigger signal. In any case, the lack of trace averaging/"VBW" is somewhat annoying in practice.

For reference, the HDO4000 FFT bug: https://www.eevblog.com/forum/testgear/rigol-hdo1000-and-hdo4000-12bit-oscilloscopes-launched-in-china/800/
That seems to have been related more to windowing/leakage after all. I guess the spurs might just be sampling clock spurs (~200kHz from some switching converter maybe?). At <65dBc it's not really a concern but they would be expected to increase with frequency.

[Martin72]

The DHO4204 also has no average mode for FFT.

My memory ( when I had the scope in February) did not deceive me:
Functions and equipment of the FFT on the DHO4000 are absolutely identical to that of the DHO800 and now you can think for whom this is more embarrassing....
Not for the cheap DHO800, that's for sure.
Also the venerable MSO5000 has the same equipment.

[Performa01]

We do have the DHO814 and can say the FFT implementation is not quite there for our professional use, the SDS2000X+ implementation is useful and we have both DSOs with "hands on".

Thank you for demonstrating the difference between a measurement vs. a mess of noise and interference.

If we try to compare the two screenshots, we have all the information on the Siglent: FFT-sample rate, FFT-length and the Averaging mode. For convenience, we also get the frequency step - and in future versions of the FW we'll also get the RBW, which depends on the window function. About the only useful window for SA applications is Flattop, so we have to multiply the frequency step by ~3.8 and get a RBW of ~362 Hz in this case.

By contrast, the Rigol shows a much lower sample rate and claims a RBW of "20" - whatever that means. If we assume that the scope actually uses its maximum of 1 Mpts FFT, then the frequency step is about 30 Hz. How can we get a RBW of 20(Hz?) under these conditions? Only explanation could be the rather useless rectangle window, which can't be used for accurate measurements at all and doesn't have a constant RBW either - at the border between frequency bins it can be even much worse than a proper window function.

We can see the excessive noise and high internal spur levels. It does not look like it would measure the individual harmonics correctly. Considering the enormous difference in RBW (362/20), the dynamic appears to be inferior too. And don't forget, the Siglent is just an 8 bit DSO!

But then again, all this does not matter - it's incredibly fast after all!

[Antonio90]

Thank you for demonstrating the difference between a measurement vs. a mess of noise and interference.

If we try to compare the two screenshots, we have all the information on the Siglent: FFT-sample rate, FFT-length and the Averaging mode. For convenience, we also get the frequency step - and in future versions of the FW we'll also get the RBW, which depends on the window function. About the only useful window for SA applications is Flattop, so we have to multiply the frequency step by ~3.8 and get a RBW of ~362 Hz in this case.

By contrast, the Rigol shows a much lower sample rate and claims a RBW of "20" - whatever that means. If we assume that the scope actually uses its maximum of 1 Mpts FFT, then the frequency step is about 30 Hz. How can we get a RBW of 20(Hz?) under these conditions? Only explanation could be the rather useless rectangle window, which can't be used for accurate measurements at all and doesn't have a constant RBW either - at the border between frequency bins it can be even much worse than a proper window function.

We can see the excessive noise and high internal spur levels. It does not look like it would measure the individual harmonics correctly. Considering the enormous difference in RBW (362/20), the dynamic appears to be inferior too. And don't forget, the Siglent is just an 8 bit DSO!

But then again, all this does not matter - it's incredibly fast after all!

FFT has always been an Achilles heel of cheap Rigol scopes if I'm not mistaken. Also, the screenshots are comparing a 500€ scope with a 1500€ one (assuming 4 channels in both cases). For many, that's peanuts. I don't work in an engineering environment, but, for my job, 1000€ is peanuts if it gets more work done, faster and more reliably.

For a hobbyist that can be the difference between 1 year of savings and 3. That's 3 years without a scope. If you extend that to the multimeters, power supply, AWG, logic analyzer, soldering iron, lab computer etc. it gets out of hand pretty fast. Ask me why I know.

There are a lot of problems, IMHO, with the DHO900, which roughly amounts to paying more for the same, as neither the bode plot nor LA work properly.

But the 800 is really good. It's cheaper than all the alternatives, Siglent, GWInstek or Micsig, its basic functionality is good, high resolution, reasonably low noise, fast interface, touchscreen, mouse support, HDMI out, fast data out to PC, good web interface, and a lot of other stuff.

I really don't get the hate. I mean, user 2N3055's criticism to the DHO900 is well deserved, I think. I would be really pissed with the FRA myself.
But why bash a bottom-of-the-barrel scope (in terms of it's price) comparing it with scopes 3 times (or even 7 with the 2000HD) the price?

The software is not finished, indeed, but everyone expected that, because that's how Rigol operates, at least with their entry-level scopes.

So, yeah, it's the best standalone scope you can buy for <500€. And yes, I don't think anybody would chose this over an SDS2000X+ or HD.

[2N3055]

Thank you for demonstrating the difference between a measurement vs. a mess of noise and interference.

If we try to compare the two screenshots, we have all the information on the Siglent: FFT-sample rate, FFT-length and the Averaging mode. For convenience, we also get the frequency step - and in future versions of the FW we'll also get the RBW, which depends on the window function. About the only useful window for SA applications is Flattop, so we have to multiply the frequency step by ~3.8 and get a RBW of ~362 Hz in this case.

By contrast, the Rigol shows a much lower sample rate and claims a RBW of "20" - whatever that means. If we assume that the scope actually uses its maximum of 1 Mpts FFT, then the frequency step is about 30 Hz. How can we get a RBW of 20(Hz?) under these conditions? Only explanation could be the rather useless rectangle window, which can't be used for accurate measurements at all and doesn't have a constant RBW either - at the border between frequency bins it can be even much worse than a proper window function.

We can see the excessive noise and high internal spur levels. It does not look like it would measure the individual harmonics correctly. Considering the enormous difference in RBW (362/20), the dynamic appears to be inferior too. And don't forget, the Siglent is just an 8 bit DSO!

But then again, all this does not matter - it's incredibly fast after all!

FFT has always been an Achilles heel of cheap Rigol scopes if I'm not mistaken. Also, the screenshots are comparing a 500€ scope with a 1500€ one (assuming 4 channels in both cases). For many, that's peanuts. I don't work in an engineering environment, but, for my job, 1000€ is peanuts if it gets more work done, faster and more reliably.

For a hobbyist that can be the difference between 1 year of savings and 3. That's 3 years without a scope. If you extend that to the multimeters, power supply, AWG, logic analyzer, soldering iron, lab computer etc. it gets out of hand pretty fast. Ask me why I know.

There are a lot of problems, IMHO, with the DHO900, which roughly amounts to paying more for the same, as neither the bode plot nor LA work properly.

But the 800 is really good. It's cheaper than all the alternatives, Siglent, GWInstek or Micsig, its basic functionality is good, high resolution, reasonably low noise, fast interface, touchscreen, mouse support, HDMI out, fast data out to PC, good web interface, and a lot of other stuff.

I really don't get the hate. I mean, user 2N3055's criticism to the DHO900 is well deserved, I think. I would be really pissed with the FRA myself.
But why bash a bottom-of-the-barrel scope (in terms of it's price) comparing it with scopes 3 times (or even 7 with the 2000HD) the price?

The software is not finished, indeed, but everyone expected that, because that's how Rigol operates, at least with their entry-level scopes.

So, yeah, it's the best standalone scope you can buy for <500€. And yes, I don't think anybody would chose this over an SDS2000X+ or HD.

I agree with you that cheap scopes don't have unlimited development budget..
And I can see that it can seen as bashing when someone has large list of problems to complaint about.
In addition to that, I'm old enough that I remember well how it is if you simply cannot afford 50€ more..

Problem is:
DHO4000 that is NOT cheap scope has equally bad BODE plot and FFT implementation as cheap DHO900/800.
Actually Rigol here developed a common platform. If they make excellent FFT for DHO4000, users of DHO800 could get same excellent FFT.
This is happening with Siglent: many features from Siglent's 10000 € scopes are available on much cheaper touchscopes...

Also cheap GW-Instek has very good FFT implementation, and slightly more expensive model has full realtime spectrum mode, well implemented.
Trust me, low price is not reason for shoddy FFT implementation.

So if all the problems, and unknowns are put together, DHO800 is not best scope you can buy today for 500ish €.
12 bit makes not such a big advantage to trade off that for many other (maybe necessary stuff, people need to decide for themselves what they need) that is missing or does not work well yet.

It does have capability to become good buy, when Rigol actually makes it work properly.
I can see future in which that statement might be true. But right now, today, no.

Of course, people, fully aware of current unfinished status, might chose to consciously buy it in this state of development and simply wait for as long as needed for Rigol to debug and finish it. We all have our own will, priorities, etc. But make sure you understand you are buying unfinished product and sort of promise that Rigol will eventually finish it in due time. If that is OK with you, it's perfect.

I must admit I tend to project a different perspective: I cannot afford to have a measurement instrument that I cannot trust. Even with very expensive instruments I sometimes measure with two different ones to verify.. Keysight has bugs too... R&S and Tek too...
My experience have proven that it is best to avoid ANY product for at least 1 year after initial release, if you actually need to depend on it...
With that in mind, my opinions are obviously influenced by that.

Best,
Siniša

[Fungus]

FFT has always been an Achilles heel of cheap Rigol scopes if I'm not mistaken.

The DS1054Z initially did the FFT with just the data visible on screen, 1200 points. It wasn't great...

Later on they fiddled the firmware to use 64k points of data. It was a bit better but much slower...

These are 12-bit and up to 1Mpts. The "12 bits" seems to make about 10dB difference.

[mawyatt]

FFT has always been an Achilles heel of cheap Rigol scopes if I'm not mistaken. Also, the screenshots are comparing a 500€ scope with a 1500€ one (assuming 4 channels in both cases). For many, that's peanuts. I don't work in an engineering environment, but, for my job, 1000€ is peanuts if it gets more work done, faster and more reliably.

For a hobbyist that can be the difference between 1 year of savings and 3. That's 3 years without a scope. If you extend that to the multimeters, power supply, AWG, logic analyzer, soldering iron, lab computer etc. it gets out of hand pretty fast. Ask me why I know.

There are a lot of problems, IMHO, with the DHO900, which roughly amounts to paying more for the same, as neither the bode plot nor LA work properly.

But the 800 is really good. It's cheaper than all the alternatives, Siglent, GWInstek or Micsig, its basic functionality is good, high resolution, reasonably low noise, fast interface, touchscreen, mouse support, HDMI out, fast data out to PC, good web interface, and a lot of other stuff.

I really don't get the hate. I mean, user 2N3055's criticism to the DHO900 is well deserved, I think. I would be really pissed with the FRA myself.
But why bash a bottom-of-the-barrel scope (in terms of it's price) comparing it with scopes 3 times (or even 7 with the 2000HD) the price?

The software is not finished, indeed, but everyone expected that, because that's how Rigol operates, at least with their entry-level scopes.

So, yeah, it's the best standalone scope you can buy for <500€. And yes, I don't think anybody would chose this over an SDS2000X+ or HD.

BTW the SDS2000X+ was available on sale for ~$1000 awhile back, good deal if acquired. Maybe this sale will reappear!!

Since "Best" and "Worst" are highly subjective, we tend to refrain from such, and our "Limited" opinion, limited in the sense we've only had a week to play around with such, the DHO814 has shown to be a good performer for General Purpose tasks, and has some unique/interesting properties that one might find useful.

Best,

[ebastler]

BTW the SDS2000X+ was available on sale for ~$1000 awhile back, good deal if acquired. Maybe this sale will reappear!!

The SDS2104X+ is 1200€ before tax here at the moment. But we get a special deal on an "option bundle" -- 235€ for the LA and AWG license, and the logic probe kit which normally sells for more than that alone. That's 1435€ total, compared to 1000€ for the DHO914S with logic probes.

Proper 8-bit scope with large screen vs. corner-cutting 12 bit scope with better portability -- ah, decisions, decisions... I'll still wait to see the price for the SDS1000X HD (but am prepared to be put off by it). 

[Antonio90]

I agree with you that cheap scopes don't have unlimited development budget..
And I can see that it can see as bashing when someone has large list of problems to complaint about.
In addition to that, I'm old enough that I remember well how it is if you simply cannot afford 50€ more..

Problem is:
DHO4000 that is NOT cheap scope has equally bad BODE plot and FFT implementation as cheap DHO900/800.
Actually Rigol here developed a common platform. If they make excellent FFT for DHO4000, users of DHO800 could get same excellent FFT.
This is happening with Siglent: many features from Siglent's 10000 € scopes are available on much cheaper touchscopes...

Also cheap GW-Instek has very good FFT implementation, and slightly more expensive model has full realtime spectrum mode, well implemented.
Trust me, low price is not reason for shoddy FFT implementation.

So if all the problems, and unknowns are put together, DHO800 is not best scope you can buy today for 500ish €.
12 bit makes not such a big advantage to trade off that for many other (maybe necessary stuff, people need to decide for themselves what they need) that is missing or does not work well yet.

It does have capability to become good buy, when Rigol actually makes it work properly.
I can see future in which that statement might be true. But right now, today, no.

Of course, people, fully aware of current unfinished status, might chose to consciously buy it in this state of development and simply wait for as long as needed for Rigol to debug and finish it. We all have our own will, priorities, etc. But make sure you understand you are buying unfinished product and sort of promise that Rigol will eventually finish it in due time. If that is OK with you, it's perfect.

I must admit I tend to project a different perspective: I cannot afford to have a measurement instrument that I cannot trust. Even with very expensive instruments I sometimes measure with two different ones to verify.. Keysight has bugs too... R&S and Tek too...
My experience have proven that it is best to avoid ANY product for at least 1 year after initial release, if you actually need to depend on it...
With that in mind, my opinions are obviously influenced by that.

Best,
Siniša

BTW the SDS2000X+ was available on sale for ~$1000 awhile back, good deal if acquired. Maybe this sale will reappear!!

Since "Best" and "Worst" are highly subjective, we tend to refrain from such, and our "Limited" opinion, limited in the sense we've only had a week to play around with such, the DHO814 has shown to be a good performer for General Purpose tasks, and has some unique/interesting properties that one might find useful.

Best,

You are right. "The best scope" does not exist, and my assertion was baseless without taking individual needs into account.

However, in this thread and the other long thread about this same series, there are quite a few examples about how things should be done, and almost all of them, if not all, come from instruments of a different class, in capabilities and price, and with older and more mature firmware. And I'm not quite sure what should someone with 500 bucks to spend learn from it.

I do still think that it is a really good scope for the price, and will probably set a new standard for budget scopes. Just think about having a 12bit scope for less than 500 bucks five years ago. But yes, it is also buying into an implicit promise of future improvement and development.

$1000 for the SDS2000X-Plus is a really good price, and I might think about selling my scope if that offer was available in EU.

[Martin72]

And I'm not quite sure what should someone with 500 bucks to spend learn from it.

At least my examples with my SDS2504X HD are not intended for direct comparison in the sense of "competition"
Because that really doesn't make sense, given the price difference.
But that should be clear to everyone...
No, at least my intention is to show how it could work.
The actual comparison is still to come....

[ebastler]

The actual comparison is still to come....

... provided that Siglent wants to play this game at all. 

[rpro]

What I don't like is that "RBW" has no unit of measurement.

Comparing spectra with my MSO5000, the RBW seems to be in Hz. 

You can only change the vertical offset position of the FFT with a button, if the menu is active and you have selected the offset there.
Outside of the menu, you can neither move the FFT up or down with the vertical knob, nor are the 2 action buttons responsible for this, they change span and center.
Only with the fingertouch you can change the position.

Maybe this is what you meant, but upon clicking on the Offset or Scale input fields in the FFT menu window, the two small yellow 1 & 2 hexagons appear at their corners, indicating you can change their values with the respective two "action encoders".

[rpro]

No luck with Cursor Peaksearch in FFT tho.

By clicking on the cursor panel and going into "Setting" you can change the source for the cursors to the FFT math channel, and then manually move the cursors with the respective 2 "action encoders", e.g. to match displayed frequencies on the peaksearch table. Far from ideal, but usable...

[KedasProbe]

I received my DHO914S this week, will do some tests this weekend.
I see there is an bandwidth option that is now showing "Limit" all other options show "Forever"

Does someone know how long the limited NW upgrade option will stay present?

I assume all current tests are done with DHO800 at 100MHz and DHO900 at 250MHz.

edit: see attachment

[Fungus]

I wondered about that. Mine has a storage depth option with "Limit" but storage isn't an option on a DHO804.

[Mechatrommer]

how do you know its boxcar averaging?

This is quite easy: When doing a single shot trace of preset sample count and analyze the individual samples, you can easily see the quntization of the values (like the "bins" in a histogram). At 1.25MSa/s, 1V/div you will find a quantization of approx. 2.31mv, which, assuming a quantization of 12bits (as per Rigol's specs), results in a total peak-to-peak range of approx. 9.5V which easily matches the 8 vertical divisions (8V) of visible vertical range.

The same test with a sampling rate of 100kSa/s has a quantization of approx. 0.2875mV which leads to a resolution in the ballpark of 15 bits. Since in a single shot, the calculation of a classic "isotemporal" average isn't possible, and the sampling engine itself can still be kept running at its "native" speed, it's a reasonable approach to average the "raw" samples that fall between two "shadow memory" (software) samples, into each adjacent s/w sample, resulting in the observed increase in resolution. And that's just boxcar averaging.

So to cut a long story short, just the fact to find a higher than hardware resolution in down-sampled single shot traces indicates that some kind of boxcar averaging has been applied to decimate the ADC raw data.

this post is continuation from...

...
I think people are confusing the several "sample rates" within a scope. The ADC sampling rate and acquisition sampling rate (which is the data stored to memory) are not always equal. And what happens between those steps is not consistent between difference scopes/brands.

The megazoom "issue" that you've pointed to (out of context) is those scopes reducing the acquisition sample rate when in non-8bit modes, (peak detect, averaging, high resolution) as the acquisition rate changes to fill the available aquitision memory (keeping the horizontal timebase the same). The ADCs keep running at their full rate for all acquisition modes, just as the Rigol do, tuning on more channels increases the multiplex to the ADCs and drops the per channel ADC sample rate (XXGS/s to 0.5*XXGS/s).

switchabl is correct that the ADC sample rate is determining the peak detect capture window (unless some scopes have analog domain peak detect?).

With "normal" sampling configuration, the DHO900 is actually boxcar averaging when the "Shadow Memory Sampling Rate" is lower than the ADC rate which apparently stays the same all the time (only activating additional channels multiplexes this sample rate over the enabled channels). See here: https://www.eevblog.com/forum/testgear/rigol-dho804-test-and-compare-thread/msg5109618/#msg5109618

how do you know its boxcar averaging?

i tried the test you described. i got same bin size of around 2.2-2.4mV (very close to 12bits using your metric of 9.5Vpp) for both sample rate, so my setup is nowhere near your 20500 bins. maybe you've set other parameters/setting in your scope? Average acquisition maybe? or different FW? attached are my data and running FW version...

[Martin72]

... provided that Siglent wants to play this game at all. 

No I meant the scopes I have here or can loan "immediately" (SDS1104X-E, DS1054Z).