Rigol DS1052E nasty surprise!

[Lawsen]

I have the same 1052E oscilloscope on my cart table top.  It shows noisy, when nothing connected to it.  It is not as clean as the 2200 series analog Tektronix oscilloscopes in my memories.  Agilent oscilloscopes the noise is fuzz line, but not a with spikes as shown here in your posted pictures.  The 1052E does show what I need to see.  The input channel filtering capacitors are designed to leave the signal as original as possible or the probe?  I have not tried other 1X or 10X probes or earlier 1052E models.  I will not be able to, because I do not need two 1052E.

[alm]

Measuring the open circuit noise measures inherent noise, shielding and the amount of EMI in your environment. Amplifier/digitizer noise is usually measured with shorted inputs.

[A Hellene]

I have the same 1052E oscilloscope on my cart table top.  It shows noisy, when nothing connected to it.

This is normal, since it displays the internal self-generated noise of the instrument plus the noise generated by the unshielded switching power supply it has, or anything else the probes are picking up from the environment when connected. The truth is that the RP2200 probes are quite noisy; using better probes you should have cleaner waveforms, as I did.

[In] Agilent oscilloscopes the noise is fuzz line, but not a with spikes as shown here in your posted pictures.

As I have written in another thread, my first DS1052E was working quite well, in contrary to the second one I received, which I am currently discussing about, that came defective straight from the factory:

I wrote: "If this specific device was the first one I laid my hands on I would think that this is the average quality of those instruments and --probably-- I would not even bother complaining about it. BUT this is my second DS1052E and the first one was perfect in comparison. If you recall my very first message, where I posted the 2.05 SP2 hack, I was happy with the instrument, even though I called it "a quite noisy oscilloscope." The truth is that I was able to read almost perfect waveforms of a few millivolts p-p amplitude on shunt resistors of a switching power supply I was working on; and I was pretty happy with it. Now, I only have to touch the ground clip to the probe tip to literally fill the display with a ~100MHz noise garbage that prevents me from reading even the 3.0V calibration output..."

You can try to do a test, yourself: Set a single channel to maximum vertical sensitivity (20mV/div), probe & probe settings x10, no filtering and timescale 5.0 ms/div; you will see a noise level floor of a small fraction of one division, which is normal for those settings. Now, short the probe tip to the probe ground lead alligator: The noise level should decrease, in contrary to my defective unit where it literally jumps to a five divisions noise mess, as depicted in my snapshot called Noise 3.


-George

[torch]

You can try to do a test, yourself: Set a single channel to maximum vertical sensitivity (20mV/div), probe & probe settings x10, no filtering and timescale 5.0 ms/div; you will see a noise level floor of a small fraction of one division, which is normal for those settings. Now, short the probe tip to the probe ground lead alligator: The noise level should decrease, in contrary to my defective unit where it literally jumps to a five divisions noise mess, as depicted in my snapshot called Noise 3.

Ok, I tried it. And the results were pretty much as described. Then I tried touching the probe tip to the chassis ground just below the calibration contact: My screen then looked very similar to your screenshot -- 6 or 7 divisions of noise. I repeated this experiment with various other probes (the Rigol 150MHz ones that came with the scope, a set of 250MHz Coline probes, even a 100MHz 100x probe -- all with similar results (although the Rigol probes possibly had the greatest amplitude). I tried it against all the exposed BNC connector ground points too, with the same results.

I don't know enough about the insides of these things, but could your problem have something to do with matching the probes? I get the cleanest line by setting the coupling type to GND. Obviously that must by-pass something in the input or I would see similar noise to touching the probe tip to chassis ground. What happens if you set yours to GND? Perhaps that experiment might help localize the source of the problem?

[A Hellene]

[...]
Ok, I tried it. And the results were pretty much as described. Then I tried touching the probe tip to the chassis ground just below the calibration contact: My screen then looked very similar to your screenshot -- 6 or 7 divisions of noise.
[...]

I am sorry to read that, dear torch. It seems that not only my device has this noise problem... My previous (brand-new also) DS1052E, the one I had used a few months ago to do the hack (that is not in my possession anymore), did not have this problem: Shorting the probe input decreased the noise further instead of creating this ~100MHz noise mess that covers half the screen.

If you have a 50 Ohm terminator, connect it to the BNC input and repeat the test. This way you eliminate any possible probe-induced noise and errors.

Unfortunately, I cannot power-up my device right now, since I have de-soldered the FPGA (the Cyclone III) in order to create the PCB schematics of the digital section also. But I remember that setting the input coupling to GND, the residual noise disappeared completely, which means that a lot of circuitry and code functions were bypassed, since it is unnatural not to see on the screen the ADC digitation error noise of +/-1 LSB (the sum of the quantization errors introduced during the ADC conversions).

I do not know yet what causes this malfunction, which has also been reported in April 2010.


-George

[torch]

I am sorry to read that, dear torch. It seems that not only my device has this noise problem... My previous (brand-new also) DS1052E, the one I had used a few months ago to do the hack (that is not in my possession anymore), did not have this problem: Shorting the probe input decreased the noise further instead of creating this ~100MHz noise mess that covers half the screen.

I may not have been perfectly clear: I have very little noise (< 1/4 division), with the probe open. A touch more if there is a retractable hook mounted to the probe, and a bit of 60Hz sine wave overlaid if I touch it to my finger (mains are 60Hz here). Shorting the probe to the ground clip on the test lead itself (as you describe) reduces the noise further. It is only when I touch the probe to the chassis that the excessive noise appears.

If you have a 50 Ohm terminator, connect it to the BNC input and repeat the test. This way you eliminate any possible probe-induced noise and errors.

Ok, tried that (using a BNC tee and 50 ohm terminator as I don't have an in-line one) Connecting the terminator to the open BNC connector reduces the noise slightly (similar to grounding the tip with the lead ground clip). Installing the probe makes no difference. Touching the ground clip to the probe tip reveals a 3/4 division 130kHz signal, as the ground clip lead is acting as an antenna and picking up some RF. Shorting the tip with tinfoil confirms this, as the noise quiets and the 130kHz signal is not present then. Touching the tip to chassis ground produces 3-4 divisions of noise -- not as bad as without the terminator, but certainly present! Touching the chassis ground with the tip shorted in tinfoil does not produce any additional noise.

The difference in noise levels makes me think of the problems associated with having an unmatched cable on a radio transmitter. Years ago, I used to play with CB radios (~27MHz) and we would tune the cable length by cutting off a bit of co-ax at a time to find the sweet spot where reflections in the cable/antenna combination canceled each other out. I wonder if this noise could be tuned out with a variable inductor?

But I remember that setting the input coupling to GND, the residual noise disappeared completely,

Same with mine.

[A Hellene]

I am sorry for misreading your message, torch. It seems that I need to be sleeping some more...

From the description of the noise you are reading, your device seems to be working fine. The 130KHz signal you read when probing the grounded parts of the device is the switching PSU generated noise and if you play with the triggering point & edge you will be able to see the switching pattern of the TOP24x based PSU with the trailing ringing after the power pulses.

The noise my device has is of a very high frequency (~95..100MHz), as it is revealed in the "Noise 8" and "10" figures, which makes me associate it with the 100.0MHz device internal main time-base.


-George

[A Hellene]

Hehe! It is perfect!

Thank you both, flolic and tinhead, for your suggestions to use proper solder balls to reball the Cyclone III FPGA instead of improvising with a soldering iron and solder wire.

The FPGA 256-Pin FineLine Ball-Grid Array (FBGA) package has pads of 0.50mm diameter and 1.00mm pitch. Reballing was done by hand (without a stencil), using regular 63/37 alloy solder balls. It took me almost half an hour to do it, by applying gel flux to the cleaned chip pads and pushing every individual solder ball in place before baking the chip using 220°C low velocity hot air to avoid blowing the solder balls away. Surface tension magic did the solder balls alignment.

Before resoldering the FPGA back to the mainboard, in order to power the device up, I will have to verify the connections diagram I have drawn.
__________


Anyway, this is a quick preview of the DS1000X design (that may be partially erroneous since I have not yet drawn the final schematics, to have the whole picture available):

What I find kind of strange in the design is the use of two global buses; a 16-bit wide data bus and a 22-bit wide address bus, interconnecting almost all the stages: The BlackFin DSP to the Spansion boot FLASH RAM & the Hynix system SDRAM chips as well as the Lattice LUT that is a peripherals and memory manager, the Altera FPGA that handles the ADCs and the data acquisition storage Issi SRAM, the Philips USB controller and the optional logic analyser Logic Head.

The FPGA and the Logic Head share of the global address bus is 8-bits wide only, feeding the BlackFin with one screen of data (256 points of two channels data) at a time, on demand.

The DSP feeds the LCD controller with the display data using a local 8-bit wide data bus and talks serially to the Keypad PCB and to a 4Kbit FRAM system preferences memory.

The LUT is in charge of addressing the upper three MSB of the Spansion FLASH RAM and the 18-bit wide Issi SRAM local address bus. It also controls the analog front-end and the analog & triggering section parameters.

The Issi SRAM holds the ADC aquisition data read from & fed to by the FPGA that uses a 4x8-bit wide local data bus to talk to the memory.

The FPGA configuration scheme is set to Active Serial Interface mode (MSEL2:0 = 0b100) and the the BlackFin acts as a passive serial configuration device during FPGA configuration.

Finally, the JTAG ports run on 3.3V
More to come, soon.


-George



EDIT: After a closer examination of the picture, it seems that the C16 solder ball (the third one up, counting from bottom right) needs to be replaced...

EDIT 2: A few corrections and additions.

[saturation]

Wow, that's great hand work George, it takes balls to fix those balls 

I'd be afraid to touch chips of that type, I fear damaging something permanently that I just paid $400 for, but your work looks fantastic.

This is a great unfolding story, thanks for posting.  Its also fun to read the enthusiasm you project in your writing as you go through this scope dissection.

[A Hellene]

Thank you, dear Saturation!

Your first line gave me a really big smile!
Thank you, also, for your nice words. You are very kind.

You are right; if this was a commercial job I would think twice before touching such components. On the other hand, this was a calculated risk I took: This $15 worth FPGA chip keeps no secrets inside of it; when powered off it becomes 100% identical to a brand new one. So, the fear of destroying the chip is actually reduced to a mere concern of having to replace it in the worst case.

Dissecting the DS1052 without having the complete service literature for it is, at least, a sacrilege. Of course, throwing $400 away is not a constructive nor an enjoyable thing to do. But, since I purchased a malfunctioning device without having the safety net of the official after sales support, viewing the same situation from another angle it certainly becomes an exciting challenge and one hell of a learning experience, as well as a relentless spare time killer! The bottom line is that even if I fail to restore the DSO to an acceptable working condition, I will have only thrown €240 out of the window; the learning experience (not to mention the game itself --you know, boys with toys!) is always a bonus!

You see, according to my personal views, everything we do, or we not do, has its risks, its cost and its returns...


-George

[rigol_owner]

Hi George,

I have found this thread by googling. I have recently bought a Rigol DS1052E.
I am also experiencing strange, quite strong 100MHz-or-so-ish noise when shorting the probe and touching ground somewhere. My scope is very quiet with an open (nothing connected) BNC input. Noise is then about 200uV RMS.

I have used the FFT function to view the spectrum of the strange noise signal.
Surprisingly, the spectrum shows that most of the energy is in the 88-108MHz band. I my country (Netherlands); that is the FM broadcast band! I also see a lot of energy in the 6-7MHz region. Don't know what that is, maybe SW radio?

Anyhow, my guess is that the noise is caused by external RF pickup. Maybe there is something wrong with the grounding of the BNC plug inside the scope. I haven't opened my scope yet as I've bought mine through the official distributor but maybe this is something to check on your scope?

Regards,

Rigol_owner

[Lawsen]

The Rigol 1052e is the standard oscilloscope for many.  We should take survey of which oscilloscope is our favorite out of the many.  It is like the sciences, which microscope is your favorite?  If one of you is an amateur astronomy, then the conversation is which telescope is the favorite?  From my experience, all digital oscilloscope have some noise.  The Atten 1102CAL has noise, but if your signal is larger than the noise, then it is no problem.  For my application for checking signals inside a giant plotter, the signal is larger than the noise, then no problem.  Those working in product development like low voltage signals, then that might be a problem.  What matters here is signal to noise ratio.  As long as the signal is larger than the noise, that is not a problem with my interests and needs. 

Oscilloscopes that I have experienced with more noises are :
Atten 1102CAL
Rigol 1052e
Tektronix 2002c

The oscilloscope with the least noise is the Agilent 2002a at 70 MHz and two channels.

The most used oscilloscope in my place and for the on call jobs with giant plotters are the Rigol 1052e, smaller and compact and inexpensive to use.  I do not use the Agilent 2002a much for fear of dropping it or breaking it.  I cannot afford to have it repaired, if broken.  I do not have a Tektronix 2002c, but dream of trying it.  The Tektronix 2002c feels like an Atten 1102CAL.  I have tried the Tektronix 2002c like, on display in the Fry's Electronic store in San Jose, California.  It has that noise.  The smoothest oscilloscope is the analog oscilloscope like older Tektronix, Hameg, and Hitachi oscilloscopes.  I do not have any of these.  Analog oscilloscopes simply feel different and smooth.  It cannot record waveforms, that will need a camera film or digital to take a picture of the screen with a camera macro lens. 

[alm]

Try to use a proper RF short (eg. BNC shorting cap), not an antenna if you want to check scope noise performance. Clipping the ground lead to the tip is not a short at 100MHz. A BNC 50ohm terminator works fairly well in my experience, since it has a very small loop area and typically low inductance.

One reason why analog scopes may appear less noisy is that a cheap DSO without any DPO-like features only has a 1-bit intensity resolution. That means that a very rare noise peak will appear the same intensity as a constant signal. An analog scope has an intensity more or less proportional to how frequent something happens, so uncorrelated noise will appear fairly faint.

[rigol_owner]

I know that shorting the probe tip with the ground clip is not the best way to measure scope noise.
I also know that noise looks different on an analog scope.
But what the Rigol is showing is excessive. I also have an analog Philips PM3055 scope here that doesn't show a hint of RF pickup in the same setup.
I also use Agilent scopes at work that have never shown this behaviour.

In my opinion the noise could be explained if the BNC plug is not properly (low impedance at all frequencies) grounded to the ground plane of the scope's front end.
This way, any common-mode noise picked up by the probe lead and connected circuits are directly coupled into the front-end amplifier.
Looking at the photo's from the ds1052e teardown (eevblog #37it "seems" like the BNC plug is grounded through a small PCB track instead of soldered firmly.
However, I cannot see it clearly. I haven't checked on my scope as my scope is still under warranty.

Btw, I love the Rigol DS1052e, even with the noise problem! I use it at home for my hobby stuff (currently designing/building an integrated tube amplifier) and it's better than any scope I've ever had. Offcourse, the Agilents at work are better but for the money, it can't be beat imho.

Regards,

Rigol_owner

[saturation]

Yes, hellene has found that some of the newer production units are noiser than the older models; see photos of my older Rigol near the start of the thread.  The spectra I've found of the noise using FFT is similar to your first post, but at lower amplitudes.

..Btw, I love the Rigol DS1052e, even with the noise problem! I use it at home for my hobby stuff (currently designing/building an integrated tube amplifier) and it's better than any scope I've ever had. Offcourse, the Agilents at work are better but for the money, it can't be beat imho.

Regards,

Rigol_owner

[A Hellene]

Hello, Rigol_owner,

Thank you for joining in to share your observations and opinions.

I also like the DS1052E design and behaviour, and this was the reason why I got myself the same device after having given away the first unit I purchased. Though I acquired both the units within a couple of months, the first one returned decent results; I could probe current shunt resistors (of a few milliohms value, being electrically equivalent to shorting the probes) of a switching PSU I was working on and the device returned nice waveforms. But this is not the case for my second unit. Though the noise floor of both the units was almost the same, when grounding the probes of the second one its display was literally filled with very high levels of an apparent VHF sinusoidal-like noise, of a frequency that was suspiciously identical to the 100.0MHz internal timebase of the device. See the "Noise 8" and the "10. Ch1 probe" screenshots, I posted in the first page:


Noise 8: Ch1 probe on the probe calibration ground (sampled at 500MSPS)


10. Ch1 probe on the probe calibration ground, in dot mode (sampled at 1GSPS)

In both the waveforms above, the signal period width seems to be very close, if not equal, to 10.0ns. Is it a coincidence? I think not, since there is a 10.0ns period crystal oscillator timebase on board (based on the 'AHC04 hex inverter) that feeds with a 100.0MHz clock directly the CPLD (that clocks the DSP, the SRAM, the FLASH and the LCD controller), the FPGA (that clocks the ADCs) and the optional logic analyser Logic Head header.

Furthermore, there seems to be a small pause of the 10.0ns signal in the "Noise 8" figure, interpreted and depicted as a glitch at the time point -0.8 .. -0.2 div, where the signal phase is somehow inverted by 180.0°; this might be the missing pulse that prevents the (hardware) frequency counter from reporting a 100.0MHz signal and shows a ~95MHz signal only.

Is this a poor decoupling related problem? It might be; but this is a 6-layer PCB and it is next to impossible to verify the ground and the ground plane vias continuity. The very noisy internal SMPS certainly contributes to the overall noise of the device; but its switching frequency (~200KHz) is nowhere near the 100MHz region...

Does it also seem to be the result of an overloading I/O stage? Some sort of excessive subthreshold drain currents or crowbar power dissipation? Probably, since the noise problem seems to be symmetrical for both the input channels, something that suggests that the problem exists in the common sections beyond the analog front ends...

Anyway, I will agree with you that I have never seen these levels of VHF noise when grounding the probes in any other oscilloscopes. I guess that we have both been, amongst others, the lucky winners of this specific (non-)QC market...


-George

[firewalker]

I hadn't read this thread for a while.

Go go George!!!

Kala, milame, ta SPAS!!!

Alexander.

[A Hellene]

Thank you, firewalker!

Special thanks, also, for your encouraging comment!


-George

[rigol_owner]

First of all George,

You have done a impressive job investigating this problem. Removing, reballing and resoldering an FPGA is not something I would have dared 
Also , I'm sure I can live with the problem as the digital filter function of the DS1052E does a real good job of filtering out the noise. After all this scop is just a tool.
If you now it's weaknesses and it's strengths and you know what you're doing you can do anything with it.

As for the noise however, I'm still not convinced it is noise picked up from the internals of the scope.
Take a look at the screenshots i've attached.

The first one shows the result of a large loop; a 1:1 probe with the tip connected to the ground test point.
You can see the FM broadcast spectrum with a large peak around 106.6 MHz or so.

The second one is with a small loop; the probe lead is loosely twisted together. Notice the main peak is almost 10dB lower.

The third one is interesting. It is a large loop with a small transistor radio nearby tuned to 106.6MHz. You can clearly see the radio's LO 10.7MHz apart from the 106.6MHz peak!
When turning the tuning knob I can move this LO peak from left to right.

To me, all this indicates (at least in my scope) that the noise picked up is external. However, why is the Rigol so sensitive to this kind of noise? I still suspect it has something to do with grounding or with the layout of the front end.

Regards,

rigol_owner

PS This scope is really wonderful, even with it's 50MHz bandwidth it can be used as a spectrum analyzer to diagnose an FM radio 

[tinhead]

buy IPA, open both input stages and cleanup the PCB, this looks for me as self self oscilations from varicap or AD8370.

[scrat]

Interesting. Did something similar already happen to you?

[tinhead]

the varicap can easily self oscilate and pickup every crap, AD8370 as well. Especially when the manufacturer
didn't cleaned up the PCB properly or when the gridding (on AD8370 or LMH6552) was too deep.
And yes, it happens already on my DSOs - twice - first time AD8370 gridding too deep + dirty PCB (the scope
was self oscilating even without power, that was funny) second time (on a different scope) PCB around varicap dirty.

[scrat]

So they are probably not joking when in the AD8370 datasheet (p. 16), they state:

"Do not use silkscreen over the signal line because it alters the line impedance."

Wow, so what should be the accuracy of pcb manufacturing?

[A Hellene]

It's been a while, I know my fellow EEVBloggers, but I was busy during the last couple of months.

In an attempt to redeem myself I am posting yet another schematic sheet I have recently drawn!
This time it will be the Keypad & Indicators PCB.

In the circuit it is clear what the I/O bitstream should look like for any control asserted or any indicator activated. Now, regarding the Keypad PCB connectivity, the LEDs serial data comes from the SPORT0 primary channel of the DSP and the keypad/encoder serial data are fed to the SPORT0 secondary channel.
______________________________


As I wave written in a similar thread, when I resoldered the FPGA back on the mainboard the noise issue was neither improved nor deteriorated. After a few days, while discovering a few mistakes in the PCB wiring diagrams I drew, I removed the FPGA once again to correct them and resoldered it back; but it seems that the chip was not soldered properly aligned this time and mangled the calibration data... I am not even sure if I have damaged the FPGA by the accidental bridging. I guess I should be looking for a ZIF socket to put in there!
Fortunately, the rest of the device seems to be unaffected and the FPGA is a low cost component (~$15) that can be replaced directly.

Anyway, this was the excuse I needed in order to begin studying the Blackfin processor, since I do not have any prior experience with the ADSP family, in order to find a way to be accessing the boot FLASH memory without the need to be physically removing the chip; you see, EEPROM cells gravely hate to be heated...

Thanks to Krater, who published an IDA plug-in for the BlackFin processor he wrote, I am attempting to reverse engineer the Rigol firmware. Actually, it is not as difficult as it sounds to be: I have quickly spotted four ELF headers that boot the FLASH memory contents into the SDRAM system memory space, using the internal processor Boot-ROM. Funny thing is that even though C-type identifiers can be spotted within the firmware, some small portions of it seem to have been written in pure assembly! Of course, I could be wrong in that because the compiler could be using ready-made assembly routines in the background, since the .ASM black arts seem to be too difficult for the newer breeds of "engineers," who do not feel the need to know the inner workings of things, to be exercised...


-George

[amspire]

George,

Please keep us informed of your discoveries.  It is really interesting for us Rigol owners.

Wouldn't it be great if one of these companies just made the hardware and let the software be a fully open-source project?

Just like Lynksys did with the WRT54G routers, they would probably have a huge leap in sales.

Richard.