Rigol DS1052 vs Siglent SDS2202X oscilloscope spectrum analyzer mode (FFT)

[@tautech8196]

Some comments if I may as a Siglent distributor. FFT results are dependant on timebase and memory depth as with a much larger memory and twice the sampling rate than the Rigol it takes time for the processor to compute all the data points into a FFT trace. Unlike the Rigol, the Siglent will give better results with more waveform cycles on the display and any discerning latency can be addressed by setting a lesser memory depth if required. Detailed inspection of very small signals can be accomplished by changing the reference level and the vertical scale. Also if the noise floor is unstable try selecting Max Hold. Reducing the FFT points is not the way to use Siglents FFT for best results. Oh and BTW, you are using the incorrect Siglent model number, what you have is SDS2202X-E, a 200 MHz DSO with faster processing, better sensitivity and FFT than the older 2000X series. As these are just $619 they are also much cheaper than an equivalent 2000X model. Have fun learning your new scope.

[@TestEquipmentAnonymous]

Thanks for jumping into the thread, tautech.

[@and1424]

Which one is newer / better: Sds2202x-e or sds2202x? Thanks!

[@tautech8196]

@@and1424 Certainly for FFT SDS2202X-E as the newer faster processor they use allows for the 1 Mpts FFT. Although the display size is smaller and they have shared controls their total feature set is better than the older X models. If you're seriously considering a new DSO have a look at the mighty new SDS2000X Plus models.

[@cassvirgillo3395]

Hello John, Glad your doing well. Thank you for the demo and explaining the scopes. What a great channel for chip amp builders and audiophiles in general. Wishing you and yours all the best, C.

[@z0lt4n]

Amazingly helpful video, with many insights for a hobbyist as I'm. Thank you, John, for devoting your time to create such knowledge sharing content! And also, what a surprising outcome (for me) that the Rigol seems to fit better to this kind of measurement.

[@tautech8196]

It might seem that way however I can assure you it is not when you know best how to drive the Siglent. All their later model DSO's FFT is excellent in their respective classes and cost.

[@EngineeringEssentials]

GW instek, rigol 2000series scopes only have usable spectrum analysis mode, But for audio anything should work. John but for you, an accurate spectrum analyser is essential. Because you do hi-fi audio amplifier projects. I hope someday you will receive such an equipment ♥️

[@alberti.2334]

thanks John for showing this information. nice informative video and full of details. I hope one day you will have access to more advanced tools.

[@deanc6515]

Thanks John, it makes sense when you explain it.

[@tangerinq]

You could use an instrumentation amplifier to null the signal (sinusoidal wave) and then amplify the distortion so that it could be more easily analyzed. Just an idea for an interesting and useful project.

[@zdravkomosnak5970]

I use similar Siglent mostly for FFT. Not sure about Rigol, but Siglent can do much better than seen here, but you have to know how to use it. There are a lot of things to adjust, but then you get very good audio and RF spectrum analyzer. I manged to get easy 60-70dB of dynamic range when measured harmonics of RF transmitter and got same result with real spectrum analyzer.

[@MrDunk66]

Thanks John. My experience with a lesser spec SDS1102CML has been very similar to yours when testing my projects using your method. Please let us know if you can get better performance as there might be other frustrated users watching. I see there are some suggestions in the comments.

[@migalito1955]

Revised for clarity on 5/4/20 Hi John, I have a question for you regarding analog to digital conversion. I don't have a problem with the concept of only two samples are needed to fully determine a signal's frequency, but I do have a problem with only two samples are needed to have determined both the signals frequency and amplitude. .. I came back and edited my question by providing a visual example to make my question a bit more clear. I am sure that a change from a negative value to a positive value is only needed to tell us that a sinusoidal waveform has gone from its first half cycle to its second half cycle along its full interval or period and from this given the timing of our sample rate we can determine precisely the frequency of a signal with as few as 2 samples. However, think of a 20Khz tone as the signal. Now allow one's imagination to zoom in to just one interval or complete period of the tone along the time domain. What one is visualizing is a complete cycle of the tone along an axis of time. Suppose t=0 is the time at the beginning of this cycle, t=e is the time at the end of the cycle and t=m is the time where the cycle transitions from negative to positive by having a function value or amplitude of zero. In my model where are the peak amplitudes? Well, they are at (t=0+t=m)/2 & at (t=m+t=e)/2. Where along the axis of time were the two samples taken along this complete cycle of 20Khz? If the two samples were not taken at a point on the time axis that coincides with the peak amplitude of the tone in my example then what? Well, then we have an approximation for the amplitude and seemingly only precisely have determined frequency, right? So, my question is essentially are people reading more than what is real into the concept that as little as two samples fully determines an analog signal. Otherwise, should they be saying, restricted to frequency and given a sampling rate, as little as two samples will within limits fully determine the frequency of the signal and provide an approximation for the signal's amplitude that becomes less precise with respect to amplitude as the frequency increases? Aside from my question I hope your brother is doing as well as possible given the additional burden he was given to carry. Thanks, Mike

[@KissAnalog]

Hi Mike, that’s a great question that I think actually can take a full video to explain well, but in short; it is the Nyqiest Theory that says that you can reproduce a wave shape in 2 samples per cycle. But, I think most scope manufacturers shoot for 2.5x to 5x samples per cycle. Great topic!

[@migalito1955]

Edited for mis types and Amended to include a summary of what my review of Shannon's theorems actually imply attached to the end of the comment. Yes, Kiss Analog, a video on the topic would be fabulous especially if you can visually show that aliasing is occurring such as by perhaps inputting a 22 khz tone with a precisely measured amplitude into a ADC- DAC where sampling was at 44 K and show that the output is an aliasing artifact by not being an exact match of the input. Ok, I found my answer by finding Shannon's theorems related to digital signal reproduction. In particular Aliasing. Having no real electrical engineering background did not keep what I have a graduate degree in which is mathematics from recognizing something is amiss with the notion that as little as 2 samples will guarantee a faithful representation of the signal. It will have a problem with amplitude and precise frequency if the cycle being sampled has only one sample in each half of the full cycle. For example aliasing or misrepresentation will likely happen when the 44.1k sampling scheme samples 22khz. The data stream for an interval of 22khz sound that has been recorded might look like this: (1,-1),(1,-1),(1,-1),(1,-1)...ad infinitum. The trouble is that while frequency can be determined roughly from this stream even were it precisely identified there still are an infinite number of amplitudes having a fixed frequency that will match (1,-1) or any other pair of numbers indicating a similar type of wave. I did not know this was called aliasing until today, but that is the term for it. Apparently people like to filter out the troublesome high frequencies that lead to aliasing when using 44.1k sampling. I would prefer going to 88k sampling which eliminates the chance of aliasing at 22khz and similarly issue prone frequencies and keeps the recording faithful. The other thing of note is. The Nyquist-Shannon theorem people often refer to is actually several theorems grouped together that pertain to digital recording. Aliasing was the topic of one of the theorems and the rest pertain to other topics pertaining to digital reproduction. Footnote or summary: After reading through all of Shannon's theorems the general interpretation of his work that is fitting is: If 2 samples having the same sign are provided then and only then the amplitude and frequency of the full sound wave being sampled is fully and uniquely determined else aliasing occurs. Simply stated, 3 samples where 2 of the samples must have the same sign of (plus or minus) is required to avoid aliasing also known as misrepresentation. The common inference that only 2 samples are required to fully determine a complete sound wave is inferring too much by not recognizing the pair of samples must be in the same half cycle thus in reality 3 are needed to avoid aliasing. I have calculated aliasing or misrepresentation can occur at above 14.8 KHz when 44.1 K sampling is used. There is also a chance no aliasing in the human range of hearing occurs when 44.1 K sampling is used, but that is extremely unlikely unless the problem frequencies have been filtered out at input to the ADC. So, yes C D quality and Hi Fi quality are indeed different and having my druthers...

[@panosd5693]

Use a higher time base and you will get a better resolution

[@raymondheath7668]

I still struggle with my ancient 10mhz dual channel Phillips anolog scope. Good info

[@stephanc7192]

Very interesting! Great video

[@TestEquipmentAnonymous]

Interesting to see the difference in the FFT between these scopes, John. I'm pinging a friend who may be able to offer some tips on getting better results on the SDS.

[@pa3eka]

John can you make a filter for 1 khz signal for let’s say 20dB- 30dB then you can see 20 - 30 dB deeper Is that a solution for your 0.4 % problem

[@marksplace]

Nice videos, very informative.

[@uK8cvPAq]

Ever updated the firmware on these things?

[@jeffm2787]

Only RIgol scope I have is the DS2302A. The FFT is not too bad. I like that I can have the FFT and Trace on the screen at the same time. Really you should be using a PC with a voltage divider and 24 bit audio, or spend the thousands on dedicated gear for distortion analysis.

[@twobob]

The Signal path "Agilent 8445B Auto Preselector" YT that came out recently had a novel approach to squashing some of those issues around oscope testing. Dunno if you eyeballed that.

[@copernicofelinis]

HOW DARE YOU disturb His Majesty when He is sleeping on the throne!

[@gaborpetroci]

This measurements can mutch better with PC based scope or a better quality soundcard! SIR Audio , and Spectralissime way above ahead of those!

[@davidluther3408]

I ALSO QUESTION THE ACCURACY OF FFTS ON O SCOPES

[@davidluther3408]

I HAD TO HAVE MY CAT PUT AWAY SOME TIME AGO AND,I STILL MISS HIM.

[@pda49184]

Interesting stuff John .. Just shows that not all scopes are equal... Horses for courses eh !... Hope you both abstain from "19"..

[@BruceNitroxpro]

At least you didn't try and bull your way with the features you didn't understand. Good for what you did say, however.

[@misterjorous]

When working with oscilloscope fft, I like to use the avarages to reduce the noise floor... I have a Rigols aswell, and boy, that is a slowwwww scope...I'm thinking about adding the sds2000x 4 channel siglent scope for the lab. Not shure about that yet.

[@stmounts]

What is the price of the two DSOs?.

[@martinda7446]

Ha ha mute -19