Rigol HDO1000 and HDO4000 12bit oscilloscopes launched in China

[Martin72]

First test weekend is over..
Summary so far:

Pros:

-Good building quality
-The display is bright and clear and has the same high resolution as the screens of the R&S RTB2000 series.
-New UI, mostly intuitive to use
-Good response of the touchscreen
-Mostly fast response/performance in general

Cons:

-Brutal loud fans
-No bode plot
-Meager FFT features
-FFT is buggy
-No dot mode
-No deactivating sin(x)/x possible
-No complete screenshot possible (except via webserver)
-Auto memory mode not on all acquisiton modes avaible
-In average mode only max. 10M memory

Next things I´ll test are the decoder functions and I want to test the ultra acquire mode somehow to get an idea for what´s in real good for(there are a lot of limitations(refer to the manual) when this mode is active)
Further waveform update rate, just for curiosity.

[mawyatt]

Martin72,

Thanks for the testing. Any chance you could run a two tone IMD test on the Rigol and Siglent HD DSO for comparisons? Think you have a SDG2042X, so you could use the Waveform Combine feature. Maybe just 10KHz and 11KHz at 1VPP if you have the time.

Best,

[Martin72]

Hi,

Yes will do, maybe tomorrow, no problem.
The rigol can remain here for another 10 days before I have to send it away again.
We can play around with it until then, but I will have the most time for it again next weekend.
I will do the two-tone test earlier.

[TomKatt]

I'm not nearly smart enough to fully understand all this, but I always thought sin(x)/x was some kind of 'approximation' that restored curvature to the sample data.

According to this Agilent brief, sinx(x)/x can create an identical waveform as seen by the input...

https://siglent.fi/data/technical-common/Sin%28x%29x_Agilent.pdf

Digitizing real-time oscilloscopes provide the backbone of high speed time-domain measurements made in the
industry today. Modern oscilloscopes use high-speed digitizers to capture the input signal. An oscilloscope’s
sample rate is often touted as a banner specification for the instrument; higher being better. In reality, as long
as the rules of Nyquist are not violated, an oscilloscope can reconstruct a user’s signal identically. This
reconstruction process is often referred to as sin(x)/x interpolation. Whether the sample rate is 25x the Nyquist
frequency, or 2.5x the Nyquist frequency, interpolation can be used to reproduce the waveform exactly as it
appeared at the oscilloscopes input connector, removing all doubt about a signal’s behaviour between samples.

[Kleinstein]

The sin(x)/x interpolation gives the interpolation that has no signal in the higher Nyquist regions. So it would be the true signal after a brickwall filter at fs/2.
There is just no information what high frequencies could be there - so assuming no high frequencies is OK.

It can still help to just see the dots and no interpolation at all.

The simple linear interpolation is more a thing for faster drawing and no other real advanatge. So one can well live without that option if the scope is fast enough with sin(x)/x.

[Martin72]

I will do the two-tone test earlier.

Couldn´t wait...
Two-tone 10 and 11khz, 1vpp.
Rigol settings as it could get, siglent settings "decreasing" to the rigol settings for better comparison.
Be aware that the pics will be shuffled.EDIT : I hate this!!!
Next post will be siglent only with "all in", means FFT average mode and 2Mpoints.

[tautech]

Be aware that the pics will be shuffled.

BTW, meant to tell you to skip the first attachment and never use it then attachments will always retain their uploaded order.

[Martin72]

Siglent with FFt avg 8 and 2Mpoints, Memory Auto..

@Rob: Ah, thanks, it works!

[mawyatt]

Martin72,

Thanks 

The Rigol does look good, and appears to have better linearity under these conditions!!

So one can infer with this one test case that maybe they've done a good job with the input, and ADC designs, so this is encouraging!

Now if they can get their act together and get the firmware up to snuff, AND if the lower cost HD performs as well which remains to be seen, then this new HD deserves some serious consideration, especially at the intro price point!!

Anyway, thanks for the testing!! BTW if you want to run some additional tests at say 100K & 1MHz or whatever frequencies you like we won't complain 

Best,

[gf]

According to this Agilent brief, sinx(x)/x can create an identical waveform as seen by the input...

Only if the input signal sampled by the ADC was a priori band-limited.

Furthermore, perfect reconstruction assumes an ideal sinx(x)/x interpolation kernel, which has an infinite extent (in practice it needs to be truncated to a finite length, though).

[Fungus]

I'm not nearly smart enough to fully understand all this, but I always thought sin(x)/x was some kind of 'approximation' that restored curvature to the sample data.

Yes, sin(x)/x reconstructs the original signal from the sample points.

sin(x)/x isn't an approximation, per. se., but for perfect results you need an infinitely wide filter and nobody can do that in practice.

It also only works if the original signal doesn't have any frequencies above the Nyquist rate, which is rarely the case in real life. Almost everything has high frequency harmonics.

TLDR; sin(x)/x wants to be perfect but real life won't let it.

[richmit]

Some X-Y mode questions:

• I see in the manual you can pick a source for X & Y.  Can the source be a math result?
• Do cursors work in X-Y?
• Can you turn on color grade in X-Y?
• I see the "add a window" for X-Y.  So I can have a Y-T plot and a X-Y plot on the screen at the same time?
• Is the X-Y display impacted by the amount of memory used?  i.e. if you use more memory, do you get more points on the screen with a slower display update like the SDS2000X+?

[switchabl]

Have you tried Hi-Res mode yet? In the new MXO4 video Dave shows that the Rigol has 14 bit @ 100 MHz and 16 bit @ 50 MHz settings. That latter seems way overoptimistic, 2 more bits correspond to roughly 12 dB improvement in SNR while 2x decimation should give only 3 dB.

Capturing maybe >10 Mpts of noise at the max. sampling rate with and without Hi-Res for offline spectrum analysis should tell most of the story. If you have a levelled 1-2 GHz signal source, checking the frequency response at different analog bandwidth and Hi-Res settings might also be interesting.

[Martin72]

Hi,

I do not have a Ghz source here.

In the acquiremenu when you choose "Hi-Res", the bandwith will be displayed for 14 bit and 16 bit.
One channel only and full memory, 14 bit 200Mhz, 16 bit 100Mhz.
Drops down to 50Mhz and 25Mhz when all channel are active, drops further when you decrease the memory.
Pics of today:
FFT without source, 1mV/div, 50 Ohm, full bandwith, full Samplerate:Normal acquisition, Hi-res 14bit, Hi-res 16bit.
Last pic a 120Mhz sinewave with 1Vpp amplitude, Hi-res 16 bit active, amplitude decreased to under 700mVpp.

[switchabl]

Interesting! So presumably the 200 MHz bandwidth limiting on the 4204 is done digitally using the same filter as the High-Res mode. That might also contribute to the differences in step response that you saw when you enabled 2/4 channels (as the filter will need to be different when the sample rate changes).

Also, it seems that "14 bit" mode is at least semi-plausible (in the sense that if you round up, there is a 2 bit-ish improvement over the nominal 12 bit without the filter) and is in line with other manufacturers' marketing claims. "16 bit" mode on the other hand is essentially "alternative facts" territory (14.5 bits if we are being generous).

[Martin72]

By the way, something that may be interesting:
Note the pic with the FFT 16bit - You get the same result in normal mode when decreasing vertical resolution under 500µV(200µV, 100µV) and bandwith will be automatically limited to 20Mhz then.

[mawyatt]

Just noted the frequency is indicated as ~119MHz, shouldn't that be 120MHz?

Best,

[egonotto]

Last pic a 120Mhz sinewave with 1Vpp amplitude, Hi-res 16 bit active, amplitude decreased to under 700mVpp.

Hallo,

interestingly, after HIRES, the sample rate is still 4 GSa/s. So no points are decimated.

Best regards
egonotto

[Martin72]

Just noted the frequency is indicated as ~119MHz, shouldn't that be 120MHz?

Best,

Hm, you're right..
Will check it after work.

[2N3055]

Last pic a 120Mhz sinewave with 1Vpp amplitude, Hi-res 16 bit active, amplitude decreased to under 700mVpp.

Hallo,

interestingly, after HIRES, the sample rate is still 4 GSa/s. So no points are decimated.

Best regards
egonotto

It will be 4GS/s.. ERES and HIRES are running average filters.. they don't change number of sample points...Just lowpass filter it..
Also with Averaging, no resample happens, just average point by point between different triggers.

[Martin72]

Hm, you're right..
Will check it after work.

It is 120Mhz, but will measured between 119..121Mhz.
BUT: Rigol have a hardware (I guess) counter...
Following pics with 120Mhz and 1Vrms, look at the Vrms measure.

[Martin72]

Finally the siglent HD with 120Mhz 1Vrms signal..
The counter shows 120Mhz, the measurement 119.7, note the difference in the amplitude between rigol and siglent.
828mVrms to 961mVrms.
Second pic shows the FFt of the signal, the third the same but with ERES 3.0.
But it´s only for showing, note the measured amplitude of only 40mVrms.
Bandwith drops down to appx 12Mhz when ERES with +3 Bits is active.
(Note: Measures are taken with auto memory, when switching to fixed memory, ERES will be "deactivated" (switched to normal mode) )

[Martin72]

The last days are dawning, where I can still test the Rigol, actually it is only next weekend and at the latest on the following Wednesday I want to send it back.
I will still perform my tests as already described, otherwise possibly still fulfill wishes.
Who has wishes, may communicate them.
In general, I am somewhat surprised that there is probably no greater interest in the HDO4000.
One must not forget, the HDO1000 is almost identical in construction, except for the missing second chip and less memory.
I'm just saying...

[Fungus]

One must not forget, the HDO1000 is almost identical in construction, except for the missing second chip and less memory.
I'm just saying...

Yep. The HDO1000 is the most interesting one.

It's all very well comparing the expensive model with a fancy Siglent but the fact is HDO1000 will be identical for most usage cases and it's a third of the price.

In general, I am somewhat surprised that there is probably no greater interest in the HDO4000.

The fact that not many people are posting requests doesn't mean there's no interest. You're the expert, and you're the one who's got the hardware.

I'm more interested in how it is in daily use than the extreme edge cases that most people seem so obsessed with.

What does it do best?
What does it do worst?
What stands out?
What are the quirks?

How many thumbs up does it get?

[tautech]

One must not forget, the HDO1000 is almost identical in construction, except for the missing second chip and less memory.
I'm just saying...

Yep. The HDO1000 is the most interesting one.

It's all very well comparing the expensive model with a fancy Siglent but the fact is HDO1000 will be identical for most usage cases and it's a third of the price.

Comparing 2 different classes of instrument here but later this year there will an opportunity to compare two 1000 series HD scopes.