(Video) Rigol DSA1030A-TG3 Spectrum Analyzer Review and Experiments

[Hugoneus]

@ hugoneus
Compliments, this is how like a review to be. Good reference sources, solid explanation of what you do and we see. It is not often I have not found one single minor point.

I like your lab, very impressive and a very clean, ergonomic set up. I even more like your gear.
Is this your hobby lab ? Are you a nerdwork analyser fan too  i'm jalous for your cal kit. I would like a professional one too but I have no acces to a calibrated anlyser so I do not dare to buy a used one because an older  prof kit without known parameters can still be not very usefull.

I assume you do RF enginering for a living ( and some network analysis)
Keep up this good work.

Sorry, I just saw these messages!

Yes, this is my personal lab, 15 years in the making.

As for my professional life, I am a technical manager at Bell Labs in NJ.

[videobruce]

Quick review.
I have had this model with the TG for a couple of weeks to evaluate it to see if I would keep it. Unfortunately I didn't, I wound up sending it back. Briefly;

Pros;
1. Price (what else can be said here),
2. Size & weight,
3. Great US support,
4. Cost for the TG is more than reasonable unlike most others where you get gouged,
5. Preamp included, not a add on,
6. Thirty day return satisfaction return policy. They even pay for shipping both ways. I haven't heard of that anywhere before.

Cons;
1. The screen update rate is only 5Hz! I never heard of anything that slow.
2. Input level overload is 10db less than most others causing a problem in strong signal areas as where I am in,
3. Only four markers, or four marker pairs. (Sorry, I'm spoiled by having nine),
4. Position of the handle needs to be moved forward to balance the unit better. As it is, it leans back at the bottom when you lift the scope up, collapsing the tilt legs which is annoying. (Yes, that may be nit picking, but still a issue if you have to move it often).

Regarding the #1 deal breaker; I can't imagine why this was a cost issue. As I stated, even the cheapest consumer LCD monitor can show a display rate of 60Hz. What this means is what my eye sees the equivalent of a 100ms sweep or in same cases 75ms depending on settings. Anything below that, I could not see any difference in the display of the trace even though the sweep rate goes down to 10ms. In other words, what good are those slower speeds if you can't see them??

As soon as I powered this up with no input attached, I saw this, but kinda dismissed it for the time until I dug into this further. When I asked about this and was told the rewrite rate was only 5Hz I was dumbfounded. When I inquired about this I was told even the 1030 is this way. I threw in the towel and sent it back. I was ready to step up to the 1030, but no way can that be justified in that price range. I can maybe understand on a $1500 SA, but on a $5k model??
This is really too bad as I enjoyed conversing with their technical department.

(Corrected & clarified some terms.)

[tinhead]

1. The screen display rate is only 5Hz! I never heard of such a thing. Even the cheapest LCD display has a 60Hz refresh rate.

Regarding the #1 deal breaker; I can't imagine why this was a cost issue. As I stated, even the cheapest consumer LCD monitor has a display rate of 60Hz. What this means is the visual sweep rate can only go down to 100Hz or in same cases 75Hz depending on settings. Anything below that, you will not see any difference in the display of the trace even though the sweep rate goes down to 10ms. In other words, what good are those slower speeds if you can't see them??

wtf? seriously, what are you talking about? LCD refresh rate 5Hz?

[videobruce]

The update rate of the display.  The scope can sweep faster than the screen can update.

(Corrected incorrect terms.)

[tinhead]

 of course i know what refresh rate means, i would not say that "you have no idea what you talking about" .. but you know what,  go figure !

[videobruce]

The circuity between the actual SA and screen seems to be where the problem is with these scopes.

(corrected comments)

[nctnico]

I did see your original post which really spells out your mentality.
If you knew what a refresh rate is why did you ask? The video chipset can control what is finally delivered to the screen which is where the problem is with these scopes.

If it has an LCD screen the refreshrate of the display is probably somewhere between 40 and 70Hz. But that is completely irrelevant. The screen update rate is probably 5Hz (200ms) which is more than enough because if something is displayed for only 200ms you'll miss it if you blink your eyes.

[tinhead]

The video chipset can control what is finally delivered to the screen which is where the problem is with these scopes.

I haven't asked what refresh rate is but wtf are you talking about.

Fyi, there is no video chipset but an FPGA, the display is directly driven, the refresh rate is definitely 30-70Hz simply because it have to be within the display specs. I bet it runs at 60Hz. And as nctnico said, refresh rate have nothing to do with what you observed on screen.

[G0HZU]

I think what VB means is that the 'rewrite' rate (of new screen data) is very slow.

i.e. how often it writes new trace data rather than the native refresh rate of the LCD display itself which could be 60Hz.

You can see in the video that the analyser is often very slow even with a 100ms sweep time. Presumably there is a processing overhead that adds a big chunk of time on top of the sweep time it displays. It seems worse on large spans and presumably this is to do with the way it pieces together the large span as individual chunks of captured spectrum. On narrower spans where the whole span fits inside the digital IF the display rewrite speed looks a fair bit faster even though it still says 100ms sweep time.

So to me the analyser often looks really slow and I'm comparing against my analysers that are are over 25 years old but can process a 1000 x 1000 point display very quickly using 1980s tech.

The phase noise is quite poor on that analyser too. At 21:12 in the video the Miteq phase noise is 'measured' as about -107dBc/Hz at 100kHz offset. I don't know what datasheet is being referred to in the video but I would expect the Miteq test oscillator to achieve -130dBc/Hz at 100kHz offset and NOT -107dBc/Hz.
If we assume the Rigol spec is typically -88dBc/Hz at 10kHz offset then it will be roughly 20dB better at 100kHz and this will be -108dBc/Hz which is close to what is seen in the video.

I think the 600MHz Miteq oscillator will easily meet -130dBc/Hz at 100kHz offset because it is a fixed frequency reference. My 1980s Advantest analyser has a typical carrier noise of -128dBc/Hz at a 100kHz offset so would get closer to measuring the true noise of the Miteq. But realistically you would need a signal source analyser to measure the Miteq properly. eg an Agilent E5052A.

[tlu]

I think what VB means is that the 'rewrite' rate (of new screen data) is very slow.

i.e. how often it writes new trace data rather than the native refresh rate of the LCD display itself which could be 60Hz.

You can see in the video that the analyser is often very slow even with a 100ms sweep time. Presumably there is a processing overhead that adds a big chunk of time on top of the sweep time it displays. It seems worse on large spans and presumably this is to do with the way it pieces together the large span as individual chunks of captured spectrum. On narrower spans where the whole span fits inside the digital IF the display rewrite speed looks a fair bit faster even though it still says 100ms sweep time.

So to me the analyser often looks really slow and I'm comparing against my analysers that are are over 25 years old but can process a 1000 x 1000 point display very quickly using 1980s tech.

The phase noise is quite poor on that analyser too. At 21:12 in the video the Miteq phase noise is 'measured' as about -107dBc/Hz at 100kHz offset. I don't know what datasheet is being referred to in the video but I would expect the Miteq test oscillator to achieve -130dBc/Hz at 100kHz offset and NOT -107dBc/Hz.
If we assume the Rigol spec is typically -88dBc/Hz at 10kHz offset then it will be roughly 20dB better at 100kHz and this will be -108dBc/Hz which is close to what is seen in the video.

I think the 600MHz Miteq oscillator will easily meet -130dBc/Hz at 100kHz offset because it is a fixed frequency reference. My 1980s Advantest analyser has a typical carrier noise of -128dBc/Hz at a 100kHz offset so would get closer to measuring the true noise of the Miteq. But realistically you would need a signal source analyser to measure the Miteq properly. eg an Agilent E5052A.

GOHZU, can you explained a bit more on the phase noise and how it affects a measurements? I'm planning on getting an dsa815-tg and want to understand a bit more before making a purchase.

[videobruce]

The screen update rate is probably 5Hz (200ms) which is more than enough because if something is displayed for only 200ms you'll miss it if you blink your eyes.

I didn't remember the exact term Rigol used, so I used the term "screen display rate" instead of the probably correct "screen update rate". I also incorrectly used the term "Hz" twice instead of "ms" (which was corrected).
On my IFR2399A, when I drop the sweep rate below 100ms to 20ms, I clearly see the difference on how the trace reacts. I do the same test here and there is no change below 100ms which means to me the displayed sweep capability only goes down to 100ms, not 10ms.

.

I think what VB means is that the 'rewrite' rate (of new screen data) is very slow.
i.e. how often it writes new trace data rather than the native refresh rate of the LCD display itself which could be 60Hz.

Exactly. 

Bottom line, that was unacceptable to me even at $1500 and surely not acceptable at $5k for the 1030 which I would of stepped up to. 

[G0HZU]

GOHZU, can you explained a bit more on the phase noise and how it affects a measurements? I'm planning on getting an dsa815-tg and want to understand a bit more before making a purchase.

Hi tlu
The significance of the high phase noise on the analyser will only affect users who want to make high dynamic range measurements on fairly narrow spans or people who want to measure carrier noise performance of local oscillators etc.
.
Usually this means professional users who want to do design or strict verification work. eg they may want to look for very small signals that are very close to very large signals and if the phase noise is high it can hide the small signal in the (Rigol increased) noise sidebands of the large signal. But relatively few people need to be concerned about this.

So I think Rigol have targeted this analyser at people who will never want to do these types of measurement and so the cost vs phase noise issue isn't a problem.

For my (design) needs the phase noise spoils the analyser performance but I accept why the noise is high because the price of the instrument is so remarkably low.

I imagine that many users of this analyser will simply accept what noise they see on the display and will assume it comes from the device they are testing. In many cases it doesn't matter either way because many users will probably use the analyser for basic spurious testing on a transmitter or for looking at the RF spectrum via an aerial or for basic filter response tests using the tracking generator.

[1design]

While playing around with the DSA1030-TG I have on test I noticed that there seems to be a internal leek of the 10MHz reference signal into the RF chain. The issue was also forwarded to a Rigol representative, now lets see what do they have to say about that. It would not have bothered me that much if it wasn't such a specific and quite often used frequency.

Regards.

[1design]

Could other owners of this instrument check if they notice this signal on their devices so we can determine if this is an isolated case or not? Thank you!

EDIT: Rigol support was very quick and today I already received their answer, as it seems the unit is specified to have a DANL of only -85dBm @10MHz, which corresponds to the position of the spur with the attenuator at 0dB and no preamp. So even though the actual noise floor is around -117dBm they prey prefer to state -85dBm instead of fixing the leakage issue