How much noise floor and other things matter in oscilloscope usability

[mawyatt]

I was only able to find power devices.  Do you have a link?

Had some discrete GaN devices from Nitronix and Raytheon over 15 years ago, also had some from Cree but these were higher power types (~1A), recall the Nitronix & Raytheon were ~ 0.5A. These were utilized as a proof of concept for the GaNsistor and DD2A (Direct Digital to Antenna) which culminated in patents 7939857 and 7903016. We didn't have direct access to a GaN fab, so used Raytheon, Cree and Nitronix. A few researchers were delving into GaN MMW MIMICs like mixers, LNAs and such back then, so discrete devices were available to help characterize the GaN MIMIC processes which were plagued with memory issues, but don't know if any made it into the COTS world.

Recall these small signal GaN devices were extremely ESD sensitive, likely because of the very thin gate region & low capacitance. The power devices were more ESD hardened tho, and maybe why no discrete small signal GaN devices are readily COTS.

Best,

[Fungus]

Take a look at the screenshots in this thread? Where's the famous "thick traces"?

https://www.eevblog.com/forum/testgear/rigol-mso5000-artifacts/msg3947554/#msg3947554

[2N3055]

Take a look at the screenshots in this thread? Where's the famous "thick traces"?

https://www.eevblog.com/forum/testgear/rigol-mso5000-artifacts/msg3947554/#msg3947554

You weren't happy with the answer there or ?

[Performa01]

Well, if I were a Rigol fanboy and was to defend the MSO5000, I would certainly not point to a thread that strikingly demonstrates the excessive noise of this instrument and how this can ruin measurements even at low sensitivities like 50 mV/div (which would be 500 mV/div when using a x10 Probe).

If the trigger level is placed close to the peak amplitude, then it picks up the noise and triggers on the peak noise level, thus highlighting it.

Siglents can do this as well – to a far lesser degree, that is. And they need to be at a 1 mV/div sensitivity to show an effect that is still not as ugly (with a kink that cannot exist in reality, hence we have reconstruction errors despite the high sample rate) as on the Rigol at 50 mV/div. See first screenshot:

SDS2354X Plus_Sine_120MHz_4mVpp_T3.3Hz

Mind you, for this, we need to use the full bandwidth (580 MHz measured) and highest sensitivity (1 mV/div). The trigger event occurs about 3.3 times per second, so that’s 27.5 ppb of the rising edges in the input signal.

Things are almost back to normal again if we use just a little bit less sensitive range, like 5 mV/div:

SDS2354X Plus_Sine_120MHz_20mVpp_T2Hz

The peak value occurs only 2 times per second, that is 17 ppb of the total potential trigger events.

On an SDS2000X+, we also have a HiRes mode, that is actually working. So, if we can make do with just 100 MHz bandwidth, we don’t have a problem even at 1 mV/div:

SDS2354X Plus_10Bit_Sine_100MHz_4mVpp_T6.37Hz

The peak value occurs only 6.37 times per second, that is 63.7 ppb of the total potential trigger events.

[ozkarah]

For example the noise floor:
In what situation really matter to have an oscilloscope with a "low" noise floor?

When you're measuring very small signals.

I am going to use It for many different things, I am a "begginer" but I do digital stuff with embedded electronics and I also aim to learn more things as possible about analog electronics starting from working on an old valve radio that i would like to repair and experiment with.

It won't make any difference at all on your digital stuff.

For the radio? If it turns out to be a problem you can easily add a preamplifier and make it even better than a lower-noise oscilloscope.

So how much this matter in electronics? How this could preclude its usability?

It's not a showstopper. You can still do everything, just maybe not as easily for a few specific things.

The real questions are: How often do you do those things? How much would you have to spend to get a lower-noise 'scope with the same abilities as your Rigol? Is the extra money well-spent?

Not true. The amplifier designed here is working between only 10Hz-100KHz bandwidth. After 100 KHz there is a big gain roll-off.

[OverBugg]

Hi All,

I know this subject is grinded to death between the Siglent SDS2k+ and Rigol MSO5000 but I'd love if you gurus could answer a couple of questions to a beginner considering getting one of the two (SDS2104X Plus or Rigol MSO5104).
I'll use a scope mainly for testing power glitching and decoding high speed digital signals for known/unknown protocols and on occasion will do analog stuff just for learning so analog is a second priority, I've read endless reviews/posts and got all pros/cons for each, but since I'm a beginner I have no clue whether:
1. Will the noisy front end for the Rigol will prevent me from achieving my goals (currently I don't know what steps I'll use for glitching) if you have experience here I'd love your advice 
2. Will the higher sample rate of the Rigol 8Gb/s will be an advantage against the Siglent 2Gb/s all channels or will not make a difference for probing unknown high speed digital signals? (knowing the Rigol's sample rate is for one channel and splits for 2/4, its acceptable by my needs).
My two options currently are:
1. Rigol MSO5104 with Logic Probes
2. SDS2104+ with some generic 1Gb/s Logic analyzer (i.e. dslogic U3pro32 or if you have a better option..)
3. -=Add your dream setup here=-

Thank you so much...

[Fungus]

All you'll get is a repeat of people pointing out the large numbers of people happily doing analog work with their Rigols and people saying it's completely unusable for any sort of work whatsoever.

The only conclusion to be drawn is that the truth is somewhere in between so you will have to decide how much you'll be using the 'scope for analog work in the mV range and buy accordingly.

If they cost the same then the decision would be easy but they don't. As noted you can have a Rigol plus logic probes for the same (or less?) money as a Siglent without them.

[nctnico]

Hi All,

I know this subject is grinded to death between the Siglent SDS2k+ and Rigol MSO5000 but I'd love if you gurus could answer a couple of questions to a beginner considering getting one of the two (SDS2104X Plus or Rigol MSO5104).
I'll use a scope mainly for testing power glitching and decoding high speed digital signals for known/unknown protocols and on occasion will do analog stuff just for learning so analog is a second priority, I've read endless reviews/posts and got all pros/cons for each, but since I'm a beginner I have no clue whether:
1. Will the noisy front end for the Rigol will prevent me from achieving my goals (currently I don't know what steps I'll use for glitching) if you have experience here I'd love your advice 
2. Will the higher sample rate of the Rigol 8Gb/s will be an advantage against the Siglent 2Gb/s all channels or will not make a difference for probing unknown high speed digital signals? (knowing the Rigol's sample rate is for one channel and splits for 2/4, its acceptable by my needs).
My two options currently are:
1. Rigol MSO5104 with Logic Probes
2. SDS2104+ with some generic 1Gb/s Logic analyzer (i.e. dslogic U3pro32 or if you have a better option..)

3. -=Add your dream setup here=-R&S RTB2004   
 

[bdunham7]

1. Rigol MSO5104 with Logic Probes
2. SDS2104+ with some generic 1Gb/s Logic analyzer (i.e. dslogic U3pro32 or if you have a better option..)

I have the Siglent with the logic probes and I can look at 16 digital channels plus look at 4 of them with the analog channels just by connecting to the second pin on the grabbers.  That's the true utility of having both functions in one unit--you can trace a glitch in one domain with one in another.  So whichever you get, even if you do get a separate PC-based logic analyzer, the logic probes for the scope are worth the money.  So, if money is tight and you aren't overly concerned with low-level analog, perhaps the Rigol is for you.

To answer your specific questions, you won't know if the noise issue will be a problem until one day it is.  And perhaps that day may be never.  For me, it would be intolerable.  But I'm not doing what you're doing and I'm not sure what you mean by 'glitching'.  The sample rate will probably be a non-issue for 99.9% of use cases.  I think you'd have work pretty hard to demonstrate why the Rigol is better because of the sample rate.

3. -=Add your dream setup here=-

Fully loaded Tek MSO68B with 4 Iso-Vue fiber probes and 4 logic probe modules.

https://www.testequipmentdepot.com/tektronix/oscilloscope/8-flexchannel-mixed-signal-oscilloscope-mso68b6bw10000.htm?ref=gbase&gclid=CjwKCAjwo8-SBhAlEiwAopc9W8lJoVoqUW9D8yJ7PC7Z4WjSFinuHNkxyGz1nXVdtFSy93NMdIppRxoCHxcQAvD_BwE

https://www.tek.com/en/products/oscilloscopes/probes/isovu-isolated-probes

https://www.testequipmentdepot.com/tektronix/accessories/oscilloscope-probe/logic-probes/8-channel-flexchannel-general-purpose-logic-probe-tlp058.htm?ref=gbase&gclid=CjwKCAjwo8-SBhAlEiwAopc9W_bJBrnhzfK_PiGpB_1BhEzqroBQIiHfAjJfWQ2ADK5B1DITeFGxZhoCdoYQAvD_BwE

[Fungus]

To answer your specific questions, you won't know if the noise issue will be a problem until one day it is.

If we follow that logic to the end we should all buy 4x4 pickup trucks just in case we ever need to take bales of hay up the hill to the horses in winter.

Some people do live on ranches and own horses so the logic holds for them. Other people live in the city and could buy things they actually need with the difference in price between a monster pickup and a Honda Civic.

[bdunham7]

If we follow that logic to the end we should all buy 4x4 pickup trucks just in case we ever need to take bales of hay up the hill to the horses in winter.

If you had only read a little bit further...

[OverBugg]

bdunham7 and all, Thank you for your elaborated response, that was extremely helpful (tending to the Siglent atm), to be more specific regarding the power analysis I want to perform testing on power like described in this paper: https://mdpi-res.com/d_attachment/cryptography/cryptography-04-00015/article_deploy/cryptography-04-00015-v3.pdf, my concern is that the noise will prevent me "seeing" uV/mV changes in this resolution, if the Rigol can do that then I prefer it over the Siglent since I'll use the 8Gs/s more often later on with other testing scenarios, but if not then the Siglent + Logic probes as you suggested is the next lineup.


Thanks again,

[David Hess]

bdunham7 and all, Thank you for your elaborated response, that was extremely helpful (tending to the Siglent atm), to be more specific regarding the power analysis I want to perform testing on power like described in this paper: https://mdpi-res.com/d_attachment/cryptography/cryptography-04-00015/article_deploy/cryptography-04-00015-v3.pdf, my concern is that the noise will prevent me "seeing" uV/mV changes in this resolution, if the Rigol can do that then I prefer it over the Siglent since I'll use the 8Gs/s more often later on with other testing scenarios, but if not then the Siglent + Logic probes as you suggested is the next lineup.

What you need is an amplifier which allows a "slideback" measurement, as they used to call it.  This is also known as a differential comparator.  Essentially it is a differential probe with a variable reference voltage applied to one input, so a precision DC offset can be subtracted from the signal to be measured.  The classic examples of this type of oscilloscope input amplifier are the Tektronix 7A13 and 7A22 but there were earlier ones.  The 7A22 supports 10 microvolt/div measurements with up to 1 volt of offset, but bandwidth is limited to below 1 MHz.  Modern DSOs implement this as an "offset" function but without as much performance as old instruments.

Even with the low noise 7A22, achieving 10 microvolt/division sensitivity is questionable at its full bandwidth of 1 MHz.  A modern implementation could do better because there are some better parts available finally, but not by a lot.

Since you are measuring a current, there are some better ways.  Use a transimpedance amplifier to make a virtual ground (or supply) for current measurement and produce a low noise fast voltage output proportional to supply current.  Then do a subtraction with an adjustable reference voltage and use the DSO of your choice.  Performance will be limited by the transimpedance amplifier implementation but that is a well researched topic.

[Fungus]

my concern is that the noise will prevent me "seeing" uV/mV changes in this resolution, if the Rigol can do that then I prefer it over the Siglent since I'll use the 8Gs/s more often later on with other testing scenarios, but if not then the Siglent + Logic probes as you suggested is the next lineup.

Why do you assume the Siglent can do it?

At some point you're going to need an amplifier to see tiny signals. The Siglent just delays that moment a bit longer than the Rigol.

[2N3055]

With side channel attack current analysis, probing solution is the most important one.
I recommend @OverBugg to do some research on that. There are many ways to do it, including non-contact probes (take a look at Little Bee magnetic probe, for instance..)

[2N3055]

my concern is that the noise will prevent me "seeing" uV/mV changes in this resolution, if the Rigol can do that then I prefer it over the Siglent since I'll use the 8Gs/s more often later on with other testing scenarios, but if not then the Siglent + Logic probes as you suggested is the next lineup.

Why do you assume the Siglent can do it?

At some point you're going to need an amplifier to see tiny signals. The Siglent just delays that moment a bit longer than the Rigol.

And World War II didn't happen because you don't like Siglent...

His presumption comes for the fact that one piece of equipment has 10X more sensitive input amplifiers and more than 4-5X less noise...
I would say that delaying that moment 10 times is VERY relevant.
If you need that kind of analog performance, Siglent is outperforming Rigol MSO5000 by a very large margin.
Siglent SDS2000X+, literally HAVE 10X amplifier built in , compared to MSO5000.
Full range, propper oscilloscope front end amplifier...

If you don't need very best analog performance, then it doesn't matter.

[Fungus]

Why do you assume the Siglent can do it?

His presumption comes for the fact that one piece of equipment has 10X more sensitive input amplifiers and more than 4-5X less noise...
I would say that delaying that moment 10 times is VERY relevant.

If somebody comes in and says they want to look at uV signals then it seems like the question should be asked. Signals don't magically stop where Siglents say they do.

[bdunham7]

bdunham7 and all, Thank you for your elaborated response, that was extremely helpful (tending to the Siglent atm), to be more specific regarding the power analysis I want to perform testing on power like described in this paper: https://mdpi-res.com/d_attachment/cryptography/cryptography-04-00015/article_deploy/cryptography-04-00015-v3.pdf, my concern is that the noise will prevent me "seeing" uV/mV changes in this resolution, if the Rigol can do that then I prefer it over the Siglent since I'll use the 8Gs/s more often later on with other testing scenarios, but if not then the Siglent + Logic probes as you suggested is the next lineup.

OK, interesting.  I think the Siglent will be an order of magnitude better at this particular job, but I don't have the Rigol to test directly so perhaps someone else can, using the signal I explain below?  David Hess has pointed out that for the general case, you need an amplifier or probe that can provide or accept a DC offset--some differential and single-ended FET probes do have this feature but you'll have a hard time finding one that is 1X, let alone amplified.  So you might need to solve that issue if the built-in capabilities of the scope aren't enough.  You also just might be able to use AC coupling in this sort of setup, but that would take more getting into the specifics.

So to be specific, I looked at  Fig. 2 on p. 33 of your link, this is showing a few-millivolt signal on a 600-ish millivolt offset with some detail.  I sort of replicated this with an AWG set to put out a 5mV_p-p 1 kHz 8-step staircase signal with a 650mV offset.  Then I set up the SDS2354X+ at 1ms/div, 1mV/div and a -650mV offset (which takes a bit of doing) and was pleasantly surprised to find that in this case, the scope was perfectly capable of making the reading directly.  Eventually you run out of offset, but that's in the manual.  I noted a little less than a 1mV apparent error in the offset, or ~0.15%, which seems very good to me.  I captured these two screenshots--one in 10-bit with 20MHz BW and one with ERES @ 3.0 bits.  As the document you linked stated, further methods like averaging can be used in some cases.  As you can see, the signal appears to be fairly clear, but you certainly wouldn't want more noise. 

[bdunham7]

If somebody comes in and says they want to look at uV signals then it seems like the question should be asked. Signals don't magically stop where Siglents say they do.

What question?  If the signal is lower than can be readily observed directly on the Siglent, than you need to consider amplification.  But AFAIK there's no readily available amplification product that will render the differences between the two scopes moot.  You will always be much better off starting with the quieter, more sensitive scope and using less amplification.  I think you'd have a serious and expensive challenge just making a preamp that would get the small-signal performance of the Rigol to merely equal what the Siglent already has.

[OverBugg]

@fungus I don't assume anything because in reality I don't actually know the level of precision needed for a specific use case, I have worked mostly with logic analyzers in the digital domain in the past (reverse engineering several devices) but never in the "analog" side of things which I would like to explore more and learn, my use case is very simple: get a microcontroller with encryption, pull one leg with 1k/10/100k resistor and measure voltage while its doing its computation to extract the key, I have seen and read cases using oscilloscopes such as the Rigol DSO2072A and Keysight DSO-X 2004A, my thought was getting an MSO with a bit more "pawa.." to be able to learn analog stuff as well as my primary usage which is mostly digital (decoding/capturing), that is it
so the only question remains, will one of the two(SDS2k+/MSO5000) will do the trick in regards to precision/sample rate/etc., I am also open to get a DSO if you can recommend one that will be a better option for these things.

Thanks for your time helping a noob

[OverBugg]

AMAZING bdunham7.... thank you for taking the time for this. I'll look into your results, Beers on me when we meet..

[SpacedCowboy]

That's the true utility of having both functions in one unit--you can trace a glitch in one domain with one in another.  So whichever you get, even if you do get a separate PC-based logic analyzer, the logic probes for the scope are worth the money.  So, if money is tight and you aren't overly concerned with low-level analog, perhaps the Rigol is for you.

I’m somewhat interested in getting one of the two scopes under discussion (my current scope is a Rigol 1102D, so due an upgrade ) - but somewhat torn on the options available. They both seem like pretty good scopes, but while a low noise floor seems like a great property for a scope to have, modding the license code seems a lot more difficult on modern Siglent scopes than Rigol, and I’d hate to be stuck with the low-end model, however hard that is to ethically justify [grin]

In reference to the above quote, though, I *do* own a Salaea Logic Pro 16 analyzer (bought when it was about half its current price) and for digital work it’s pretty awesome - and it will (more slowly) do analogue sampling as well as digital. You don’t necessarily need a scope for mixed work, the lines are blurred in both directions these days - plus you can write your own protocol analyzers relatively trivially with the Salaea - I did one for PECI a year or so back. For digital signals, I’m not even sure the scope is the better option…

[ozkarah]

I’ve been at the same place with you. My options were MSO5104, SDS2104X+, and R&S RTB2004. Thanks to contributors here I read a lot, and I learned a lot.  Let me share my experience with you.

I don’t have a large budget. Price/Performance was important for me. First I eliminated RTB2004. Because it is more expensive (considered with options and upgrades) and not possible to hack it in the future. Then I had 2 options left:

RIGOL MSO5104:
Pros:
+ 8 GSa/sec sample rate (it is great if you need high freq) (4 Gsa/sec with 2 channels enabled, 1GSa/sec with 4 channels enabled)  (Siglent:  2 Gsa/sec with 1/2 channels enabled, 1GSa/sec with 3/4 channels enabled)
+ Easy to upgrade(or hack) for all options (including 500MHz bandwith)
+ Comes with 4 x 350 MHz probes (Siglent comes with 200 MHz probes)
+ 1 GSa/sec digital channels sampling rate. (Siglent has 500 MSa/sec)
+ 2 Channel AWG (option) (but up to 25 MHz. Siglent comes with 1 Ch 50 MHz option)
+ Better price. (and some very attractive promotions with bundle options)
+ HDMI output (but 1024x600 not scaled to fullscreen)
+ Web viewer.
+ 4 Math functions at a time (Siglent has 2)


Cons:
- The most popular one is the noisy front end. Bad for analog signals. I considered adding an amplifier. The most effective types of signal amplifiers are active/diffferantial probes. However, they are too expensive, even more expensive than the scope itself. There are also very cheap low-noise signal amplifiers. However, their problem is nonlinearity.  Their gain cannot stay at the same level as the frequency changes. So you can see the shape of the signal, but cannot trust to the measurements. It depends on your requirements. If just seeing the waveform is enough no problem.
- Buggy UI and firmware. Rigol used its own ASIC chip on this model. And users report that software bugs are less now but still there.
- 100 mega points (analog) standard memory. 100 mp is a good value but less compared to Siglent. (Siglent has 200 Mp out of the box. 200 Mp/ch when 2 ch enabled. 100 mp/ch when 4 channels are enabled. Rigol has 100 Mp/ch when a single channel is enabled. 25 Mp/ch when 4 channels are enabled. However, Rigol has a 400 USD upgrade option to 200 mp/ch, MSO5000-2RL, which doubles the numbers in the previous sentence. Still half of the Siglent, even with the deep memory option purchased.)
- 25 mega points per digital channel memory (Siglent has 50 Mp/ch)
- You have to buy separately the 16 channel logic probes and to use them and it is not cheap (probes: ~400 USD). But no sw license is needed. 
- You have to pay  (or hack) for all protocol decoding options (sometimes there are free bundle promotions. at Siglent basic protocols are included for free)
- There is no 50 Ohm input setting. You have to use 50 ohm terminator when your source impedance is 50 Ohms. 
- 1 mega points FFT



-----------------------------------------------------



SIGLENT SDS2104X
Pros:
+ Very low noise floor (It is around 70 microvolts which is really great fun when working with low amplitude signals) 
+ 500 microvolt/div range setting (which is not pixel doubling as it is in Rigol)
+ Easy to upgrade(or hack) for all options (including 500MHz bandwith)
+ Astonishing 10.1 display with 256 level intensity grading (Rigol has 9”). Touchscreen and gestures work smoothly (I saw some videos in which Rigol has some problems with that).
+ 50 MHz signal generator (optional) (but 1 channel)
+ 10-bit mode is really working fine (also you can add ERES filter to increase the effective bits on top of it)
+ If you have a signal generator from Siglent (SDG2042X I have) scope can communicate and command it flawlessly (through the local area network) to calculate the Bode plot.
+ DIY digital logic probe schematics are available in the forum.  (not sure if Rigol also has)
+ Decoding for basic (UART, I2C, SPI, …  ) protocols is included for free. You have to pay for some additional protocols (I2S, Manchester, CAN FD, FlexRay, MIL-STD 1553B, SENT)
+ Stable UI/UX.
+ Web control with full screen viewing option. (There is a VNC server running on the device. From any client you can connect and view scope UI)
+ 50 Ohm input setting is there.
+ 2 mega points FFT

Cons:
- In high frequencies, 2 GSa/sec per channel might be limiting (It is 1GSa/ch when 3 or4 channel is used). 
- When upgraded to 500 MHz bandwidth it is up to 2 channels at a time. If you need 3 or 4 channels simultaneously, bandwidth limit is 350 MHz.
- 200 MHz probes come with 100MHz and 200 MHz models.
- Price level is higher than Rigol. (Sometimes, also very attractive bundle promotions are there but usually still higher)
- You have to buy separately the 16 channel logic probes and a sw licence to use them and it is not cheap (probes: ~380 USD sw:~175 USD).
- No HDMI/VGA output. (You have to use web viewer or VNC)
+ Only 2 Math functions at a time (Rigol has 4)


-----------------------------------------------------

SUMMARY:
•   Both scopes are really good. You shouldn’t expect the best scope of the world at this price level.
•   Both scopes are open to liberation (hack). R&S RTB2000 is not as I know.
•   If you will mostly work with digital signals which don’t need more resolutions but higher sample rates Rigol has a 2 times sample rate compared to the Siglent. (2 GSa/sec when 4 channels are used compared to 1 GSa/sec). Also when you need 1 channel only 8GSa/sec should be very great at high frequencies. So with digital signals at high freq Rigol looks great. 
•   Having a low noise floor, effective 10-bit mode, and ERES filter Siglent is much better at analog signals (or when probing digital signals with analog probes).
•   Siglent has 4 times of memory of Rigol’s in most cases. (2 times when Rigol’s deep memory option is purchased)
•   Siglent is more expensive.
•   Siglent has a less buggy and better UI/UX (my opinion).
•   If you need BODE PLOT analysis and already have a signal generator from Siglent or Rigol you would want to consider the integration benefits of having a scope from the same brand.

Which one we should choose?

It really depends. You should consider your requirements and check the pro and con conditions above. If you only work with digital signals Rigol beats with a high sample rate and attractive price. If analog is also important and especially you need a good resolution then Siglent is better.

I usually work with digital signals but analog signal probing is also a must for me, and I didn’t wanna struggle with the noise problems when going analog. So, I went with Siglent. I bought it under a promotion price; bundled with AWG, 16 channel logic probes, and a logic probing software license. Sometimes I wonder how fun it would be if it had an 8 GSa/sec sample rate, but in common I am very happy with the Siglent. 

[Fungus]

SUMMARY:
•   Siglent has 4 times of memory of Rigol’s in most cases. (2 times when Rigol’s deep memory option is purchased)

Good summary.

Just a nitpick: Both scopes are probably going to be hacked so the Rigol's deep memory is "standard".

[ozkarah]

SUMMARY:
•   Siglent has 4 times of memory of Rigol’s in most cases. (2 times when Rigol’s deep memory option is purchased)

Good summary.

Just a nitpick: Both scopes are probably going to be hacked so the Rigol's deep memory is "standard".

Thanks. Till you liberate it is still an option. However, you are right too. Most probably it is not a big issue if you are not afraid of warranty validity risks.