How to use oscilloscope correctly?

[21KUZY073]

I used analog oscilloscopes before but now I bought my first digital one. I just believed before in what I saw on monitor. Now I have enough time and possibility to compare.
Which of signal is correct - same signal. I only changed input resistance and bits resolution on oscilloscope. I did not touch generator. (connection by RG58U cable)  SIGLENT2352 Plus

[bdunham7]

The image says you have your probe set to 1x mode.

He isn't using a probe, but rather a direct cable connection.

[bdunham7]

I used analog oscilloscopes before but now I bought my first digital one. I just believed before in what I saw on monitor. Now I have enough time and possibility to compare.
Which of signal is correct - same signal. I only changed input resistance and bits resolution on oscilloscope. I did not touch generator. (connection by RG58U cable)  SIGLENT2352 Plus

The most correct is the 50R 8-bit display.  At 50 MHz you do have to pay some attention to termination, the impedance of the input is causing the distortion and amplitude difference between the 1M and 50R pictures.  However, the difference between the 10-bit and 8-bit is due to the fact that the method that the scope uses to fake 10 bits reduces it's bandwidth to under 125MHz by my estimation (I could be off on that, it's just a quick guess) which means that the scope can't even properly display the third harmonic of the square wave you apparently are feeding it. 

[21KUZY073]

Thank you for answer bdunham7. Yes, there is written in manual that resolution "10bits" works up to "about" 100MHz. It looks like it is only fake resolution maybe for "zoom function" for LF signals. Now I understand problematic a bit more. Result when using probe is also different but it will be next question and I want to read about probes before I will ask stupid questions.

[vk6zgo]

I used analog oscilloscopes before but now I bought my first digital one. I just believed before in what I saw on monitor. Now I have enough time and possibility to compare.
Which of signal is correct - same signal. I only changed input resistance and bits resolution on oscilloscope. I did not touch generator. (connection by RG58U cable)  SIGLENT2352 Plus

The most correct is the 50R 8-bit display.  At 50 MHz you do have to pay some attention to termination, the impedance of the input is causing the distortion and amplitude difference between the 1M and 50R pictures.  However, the difference between the 10-bit and 8-bit is due to the fact that the method that the scope uses to fake 10 bits reduces it's bandwidth to under 125MHz by my estimation (I could be off on that, it's just a quick guess) which means that the scope can't even properly display the third harmonic of the square wave you apparently are feeding it.

As someone without a modern DSO, that looked a bit counter intuitive!

The few old ones I used (under protest) would have been "coughing their last" at 50MHz, so I always expect "the world" out of the current ones.

I suspected the funny results at 10 bit resolution were something to do with memory & sample rates (isn't it always?), & expected some clues on the display, but no!

Logically, a DSO capable of native 10 bit resolution would, by default, run in that mode, so is it always "fake", or does it just "run out of grunt" at higher frequencies?

A lot less "traps for young players" with my old Tek 7613!

That said, learning how to use the delayed timebase in those old, fairly complex analog 'scopes could be stressful-----------I sweated blood learning how to use it on a 545b, back in the day!

[srb1954]

As someone without a modern DSO, that looked a bit counter intuitive!

The few old ones I used (under protest) would have been "coughing their last" at 50MHz, so I always expect "the world" out of the current ones.

I suspected the funny results at 10 bit resolution were something to do with memory & sample rates (isn't it always?), & expected some clues on the display, but no!

Logically, a DSO capable of native 10 bit resolution would, by default, run in that mode, so is it always "fake", or does it just "run out of grunt" at higher frequencies?

A lot less "traps for young players" with my old Tek 7613!

That said, learning how to use the delayed timebase in those old, fairly complex analog 'scopes could be stressful-----------I sweated blood learning how to use it on a 545b, back in the day!

My opinion on the digital vs analog scope debate: analog scopes won't tell you the whole truth but digital scopes will tell you lies.

[tautech]

I suspected the funny results at 10 bit resolution were something to do with memory & sample rates (isn't it always?), & expected some clues on the display, but no!

Oh it's there alright if you know where to look. 
SDS2000X Plus indicate which mode they're whether it's 8 or 10 bit.

Bottom right of display just above the sampling/memory/timebase box.

[Performa01]

I used analog oscilloscopes before but now I bought my first digital one. I just believed before in what I saw on monitor. Now I have enough time and possibility to compare.
Which of signal is correct - same signal. I only changed input resistance and bits resolution on oscilloscope. I did not touch generator. (connection by RG58U cable)  SIGLENT2352 Plus

I repeat my answer from the other thread:

1.   The 10 bit mode captures cannot be correct, because as we all should know, 100 MHz bandwidth is way too low for the reproduction of a 50 MHz squarewave.
2.   Basic transmission line theory rules out the 1M "measurements" as well, because we cannot expect any signal fidelity/accuracy at frequencies that high with a severe mismatch like this.

You wanted to say ns, not ms, right?
You did not deliberately attach screenshots that don't have enough resolution to clearly see the parameters, did you?
You have had a look into the manual, haven't you? Still you've obviously missed the part where it says that 10 bit mode is always limited to 100 MHz bandwidth

- Yes, it is 1ns not 1ms - I already corrected it.
- I can read details on picture I added without problems. I can even read all scope settings on my pictures.

It’s fine that you can “even read your settings”. If you want help from others, you might want to make sure that settings are actually readable, not just guessable. Because it’s no fun to try to decipher that pixel-mush.

My fault, sorry for that.

- I read manual, this is why tested signal is 50MHz not more. Isn't 50MHz signal in 100 MHz bandwidth? Or is it something else?

Pardon? Are you kidding me? We are talking about square waves here. 100 MHz bandwidth isn’t sufficient to reproduce _any_ squarewave at 50 MHz, no matter what the transition times are, because even the 3rd harmonic is already outside the bandwidth then.
These are absolute basics, and it’s the very same on analog scopes…

I thought the higher impedance the lower source loading and in result more clear/readable/low deformed measured signal.

This is so at DC and in the audio range of frequencies - and can even be stretched up to a few MHz in many circumstances. But as soon as the length of a link exceeds about 1/20 of the wavelength of the signal that is to be transferred, we enter serious territory, known as HF. Now (among other things) transmission line theory becomes increasingly important and there is no room for mismatched connections anymore.

For 50 MHz fundamental frequency, the wavelength is about 6 m. This means that any link length greater than 30 cm is critical already. Take the NVP (Nominal Velocity of Propagation) of RG58 coax cable into consideration (0.66), then it’s only 20 cm. And this is only for the fundamental frequency! If you want a reasonably undistorted signal, all relevant harmonics must be correctly transferred too, so you can easily divide the number by e.g. 7 (to ensure proper transfer up to the 7th harmonic), and the link must not have a physical length of more than 3 cm - as long as it is not properly terminated, that is. There is a reason why better (= high bandwidth) scopes have a switchable 50 ohms input impedance (and the really high bandwidth ones might have only 50 ohms and nothing else)!

Yes, there is written in manual that resolution "10bits" works up to "about" 100MHz. It looks like it is only fake resolution maybe for "zoom function" for LF signals. Now I understand problematic a bit more.

Are you sure? 😉
So you call a 100 MHz bandwidth “LF signals” now. An interesting statement for someone who doesn’t seem to even know transmission line theory (or the bandwidth requirements of a square wave).

Apart from R&S (10 bit) and LeCroy (up to 12 bit), who have scopes with true high resolution converters, most of the better known DSO manufacturers (Keysight, Tek and also R&S) offer “fake” resolutions, even up to 16 bits sometimes. It works well, it trades in resolution for bandwidth, it’s common and, most importantly, it’s an option. Take it or leave it.

I suspected the funny results at 10 bit resolution were something to do with memory & sample rates (isn't it always?), & expected some clues on the display, but no!

The results are not “funny”, but that’s what you get when you feed a 50 MHz square into a system with only 100 MHz bandwidth.
I have attached a screenshot that is actually readable. Have a look at it!

Logically, a DSO capable of native 10 bit resolution would, by default, run in that mode, so is it always "fake", or does it just "run out of grunt" at higher frequencies?

As explained earlier, this is an 8 bit scope and nowhere is it advertised as 10 bit. It happens to have a 10 bit HiRes mode, it's there if someone wants to trade in the higher resolution against bandwidth. In this forum there are already a number of postings (with screenshots) that explain and demonstrate the value of that mode.

A lot less "traps for young players" with my old Tek 7613!

I know. It was so much better in the good old days. 

[bdunham7]

It’s fine that you can “even read your settings”. If you want help from others, you might want to make sure that settings are actually readable, not just guessable. Because it’s no fun to try to decipher that pixel-mush.

Not sure what the issue is, but his photos are clear and sharp on my monitor. 

As explained earlier, this is an 8 bit scope and nowhere is it advertised as 10 bit. It happens to have a 10 bit HiRes mode, it's there if someone wants to trade in the higher resolution against bandwidth. In this forum there are already a number of postings (with screenshots) that explain and demonstrate the value of that mode.

Actually if you look a the data sheet, second page, it discusses the 10-bit acquisition mode twice without any footnote or mention of the bandwidth limitation.  The scope also continues to display "2GSa/s" in the 10-bit mode, but I don't think that's really true.  You can say 'RTFM' but that doesn't change the fact that what is said on the screen and on the datasheet is not 100% correct.

[Performa01]

Not sure what the issue is, but his photos are clear and sharp on my monitor. 

You are right. The issue was on my side. Sorry for the rant.

Actually if you look a the data sheet, second page, it discusses the 10-bit acquisition mode twice without any footnote or mention of the bandwidth limitation.  The scope also continues to display "2GSa/s" in the 10-bit mode, but I don't think that's really true.  You can say 'RTFM' but that doesn't change the fact that what is said on the screen and on the datasheet is not 100% correct.

Here's the current datasheet:

https://siglentna.com/wp-content/uploads/dlm_uploads/2019/12/SDS2000X-Plus_Datasheet_DS0102XP_E01A-2.pdf

The only mention of 10 bits I can find on page 2 is this one: "10-bit mode (with typical 100 MHz bandwidth)"

As for the sample rate, you are technically correct of course. The initial sample rate is 1 or 2 GSa/s, but the averaging used for enhanced resolution is of course a form of decimation, so the effective sample rate will be four times lower.
But the same would be true for ERES or HiRes (no matter if they are implemented as genuine acquisition modes or math functions), these modes exist for a long time and no one has ever complained that the sample rate display is not adapted accordingly. In fact, there are (or at least have been) even scopes that don't tell you the current effective sample rate at all.

As can be seen in my screenshot, the 10 bit mode is clearly indicated (and users should know what this means, also in terms of effective sample rate and bandwidth) as well as even the bandwidth limit is stated in the channel tab, which I think is a pretty neat and not very common feature.

[vk6zgo]

I suspected the funny results at 10 bit resolution were something to do with memory & sample rates (isn't it always?), & expected some clues on the display, but no!

Oh it's there alright if you know where to look. 
SDS2000X Plus indicate which mode they're whether it's 8 or 10 bit.

Bottom right of display just above the sampling/memory/timebase box.

I see it now! ------it's the bottom left, but OK.

[vk6zgo]

I used analog oscilloscopes before but now I bought my first digital one. I just believed before in what I saw on monitor. Now I have enough time and possibility to compare.
Which of signal is correct - same signal. I only changed input resistance and bits resolution on oscilloscope. I did not touch generator. (connection by RG58U cable)  SIGLENT2352 Plus

I repeat my answer from the other thread:

All good, but it doesn't help when, like me, someone hasn't read the other one, & thinks this is a standalone thread.

1.   The 10 bit mode captures cannot be correct, because as we all should know, 100 MHz bandwidth is way too low for the reproduction of a 50 MHz squarewave.

As above.

2.   Basic transmission line theory rules out the 1M "measurements" as well, because we cannot expect any signal fidelity/accuracy at frequencies that high with a severe mismatch like this.

You wanted to say ns, not ms, right?
You did not deliberately attach screenshots that don't have enough resolution to clearly see the parameters, did you?
You have had a look into the manual, haven't you? Still you've obviously missed the part where it says that 10 bit mode is always limited to 100 MHz bandwidth

- Yes, it is 1ns not 1ms - I already corrected it.
- I can read details on picture I added without problems. I can even read all scope settings on my pictures.

It’s fine that you can “even read your settings”. If you want help from others, you might want to make sure that settings are actually readable, not just guessable. Because it’s no fun to try to decipher that pixel-mush.

My fault, sorry for that.

- I read manual, this is why tested signal is 50MHz not more. Isn't 50MHz signal in 100 MHz bandwidth? Or is it something else?

Pardon? Are you kidding me? We are talking about square waves here. 100 MHz bandwidth isn’t sufficient to reproduce _any_ squarewave at 50 MHz, no matter what the transition times are, because even the 3rd harmonic is already outside the bandwidth then.
These are absolute basics, and it’s the very same on analog scopes…

I thought the higher impedance the lower source loading and in result more clear/readable/low deformed measured signal.

This is so at DC and in the audio range of frequencies - and can even be stretched up to a few MHz in many circumstances. But as soon as the length of a link exceeds about 1/20 of the wavelength of the signal that is to be transferred, we enter serious territory, known as HF. Now (among other things) transmission line theory becomes increasingly important and there is no room for mismatched connections anymore.

For 50 MHz fundamental frequency, the wavelength is about 6 m. This means that any link length greater than 30 cm is critical already. Take the NVP (Nominal Velocity of Propagation) of RG58 coax cable into consideration (0.66), then it’s only 20 cm. And this is only for the fundamental frequency! If you want a reasonably undistorted signal, all relevant harmonics must be correctly transferred too, so you can easily divide the number by e.g. 7 (to ensure proper transfer up to the 7th harmonic), and the link must not have a physical length of more than 3 cm - as long as it is not properly terminated, that is. There is a reason why better (= high bandwidth) scopes have a switchable 50 ohms input impedance (and the really high bandwidth ones might have only 50 ohms and nothing else)!

Yes, there is written in manual that resolution "10bits" works up to "about" 100MHz. It looks like it is only fake resolution maybe for "zoom function" for LF signals. Now I understand problematic a bit more.

Are you sure? 😉
So you call a 100 MHz bandwidth “LF signals” now. An interesting statement for someone who doesn’t seem to even know transmission line theory (or the bandwidth requirements of a square wave).

Apart from R&S (10 bit) and LeCroy (up to 12 bit), who have scopes with true high resolution converters, most of the better known DSO manufacturers (Keysight, Tek and also R&S) offer “fake” resolutions, even up to 16 bits sometimes. It works well, it trades in resolution for bandwidth, it’s common and, most importantly, it’s an option. Take it or leave it.

I suspected the funny results at 10 bit resolution were something to do with memory & sample rates (isn't it always?), & expected some clues on the display, but no!

The results are not “funny”, but that’s what you get when you feed a 50 MHz square into a system with only 100 MHz bandwidth.
I have attached a screenshot that is actually readable. Have a look at it!

Well, now having partially read the "original thread" it is fairly obvious that the hardware is quite happy to
500MHz, & the 100MHz limit is just the result of dropping the sample rate.

The same old syndrome present in the ancient DSOs, just in an unfamiliar form!
The OP's screenshot is quite readable , even on my cruddy iPad!

Logically, a DSO capable of native 10 bit resolution would, by default, run in that mode, so is it always "fake", or does it just "run out of grunt" at higher frequencies?

As explained earlier, this is an 8 bit scope and nowhere is it advertised as 10 bit. It happens to have a 10 bit HiRes mode, it's there if someone wants to trade in the higher resolution against bandwidth. In this forum there are already a number of postings (with screenshots) that explain and demonstrate the value of that mode.

A lot less "traps for young players" with my old Tek 7613!

I know. It was so much better in the good old days. 

You didn't include my qualification to that comment in your quote.
(How hard it was for noobs to get a "handle on" delayed timebase mode in the old Teks, HPs, etc)

[Performa01]

Well, now having partially read the "original thread" it is fairly obvious that the hardware is quite happy to
500MHz, & the 100MHz limit is just the result of dropping the sample rate.

The same old syndrome present in the ancient DSOs, just in an unfamiliar form!

Well, I don't think it is comparable at all.

• First of all, everyone would have been quite happy if the ancient DSOs had provided a realtime bandwidth of 100 MHz. Even by today's standards, 100 MHz is quite usable and there are many scopes that have this or an even lower (e.g. 50, 70 MHz) bandwidth.
  
• Secondly, even if they did, the ancient DSOs could not maintain the full samplerate up to long timebases, due to the lack of deep memory. This means, the realtime bandwidth dropped pretty quickly at slower timebases, which rules out some important use cases for a DSO.
  
• Thirdly, these limitations are inescapable on an ancient DSO, whereas they are just the price to pay for the option of having an enhanced vertical resolution on the modern SDS2000X+. The majority of today's DSOs are still 8 bit - because we just don't need more for the majority of tasks.
  

Finally this is an optional feature, coming for free in a cheap upper-entry level DSO - no one is forced to use that feature and unlike other instruments, this one keeps the user informed about all important acquisition parameters, like sample rate, record length and 10 bit mode (if active) at the bottom right in the timebase tab, as well as input coupling, input impedance, attenuation factor and detailed bandwidth limit information in the channel tabs (starting from the bottom left).

[bdunham7]

The only mention of 10 bits I can find on page 2 is this one: "10-bit mode (with typical 100 MHz bandwidth)"

Yes, there it is, but page 2 is the third page, I said the second page which is page 1.    

I'm not really asserting that the scope or the feature is bad, just that the marketing is sometimes a bit careless about disclosing these things--and that is a problem when they advertise a 'feature' that appears to match a competitor's feature but is really not the same.  I fail to see how this '10-bit' is actually any better than ERES at all.  Perhaps I'm missing something.

[Performa01]

The only mention of 10 bits I can find on page 2 is this one: "10-bit mode (with typical 100 MHz bandwidth)"

Yes, there it is, but page 2 is the third page, I said the second page which is page 1.    

Page one is just a brief overview, without any details that otherwise only appear in the fine print. This sort of introductional talk is quite common amongst all manufacturers and I honestly don't think that every single mention of "10 bit mode" requires yet another note regarding the bandwidth limit. Page two on the other hand contains the "Key Specifications", so this is now part of the actual datasheet. Here important details like bandwidth limits should be clearly visible - and they are.

I'm not really asserting that the scope or the feature is bad, just that the marketing is sometimes a bit careless about disclosing these things--and that is a problem when they advertise a 'feature' that appears to match a competitor's feature but is really not the same.  I fail to see how this '10-bit' is actually any better than ERES at all.  Perhaps I'm missing something.

What do you mean "to match a competitor's freature"?

Competitors are mostly blatant about their HR features. Rightfully so in case of true HR converters like R&S entry level and lower midrange (RTB, RTM...), but also for the "fake" resolution enhancements, which are usuallly limited to the higher end products anyway. I would have been the first to complain loudly if Siglent marketing had called the SDS2000X+ anything like HR or 10 bit.

As for the advantages compared to ERES: 10 bit mode has a constant bandwidth of >100 MHz, whereas ERES bandwidth changes with the effective sample rate and can only be calculated using a table from the manual. This is prone to error; any altered setting, that leads to a different effective sample rate (like changing the timebase with short memory selected for instance) would alter the bandwidth and this could be bad if it goes unnoticed. In any case there is no bandwidth display for this.

Apart from that, the SDS2000X+ still has ERES as well, it's just not a genuine acquisition mode like in other Siglent DSOs, but a math function here. So you can have whatever you like better - or both together.

Further explanations and a bunch of links to demonstrations how the 10 bit mode and ERES can be combined to even 16 bit resolution can be found here in reply #2347

https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg3398420/#msg3398420

[vk6zgo]

Well, now having partially read the "original thread" it is fairly obvious that the hardware is quite happy to
500MHz, & the 100MHz limit is just the result of dropping the sample rate.

The same old syndrome present in the ancient DSOs, just in an unfamiliar form!

Well, I don't think it is comparable at all.

• First of all, everyone would have been quite happy if the ancient DSOs had provided a realtime bandwidth of 100 MHz. Even by today's standards, 100 MHz is quite usable and there are many scopes that have this or an even lower (e.g. 50, 70 MHz) bandwidth.
  

Indeed!, & this is something I have pointed out previously on this forum.
Many of the fairly early Tek & HP DSOs did offer a sufficient sample rate to reproduce 100MHz sinewave signals, but needed to use very short time/div settings to do so.
This leads into your second point:-

• Secondly, even if they did, the ancient DSOs could not maintain the full samplerate up to long timebases, due to the lack of deep memory. This means, the realtime bandwidth dropped pretty quickly at slower timebases, which rules out some important use cases for a DSO.
  

Yes, the first generation of these these supposed "100MHz" 'scopes, if used to observe analog video, reduced the sample rate so far, that, even at TV "line rate", the higher frequency components (around 5MHz) of such signals just disappeared into a mass of aliasing.
The next lot could (just) do line rate, but trying to see anything useful at field rate (50 Hz) was impossible.

• Thirdly, these limitations are inescapable on an ancient DSO, whereas they are just the price to pay for the option of having an enhanced vertical resolution on the modern SDS2000X+. The majority of today's DSOs are still 8 bit - because we just don't need more for the majority of tasks.
  

I suppose there are use cases, (other than looking for low frequency noise in a RF "noisy environment") where we would deliberately reduce the bandwidth of an instrument, but it is a bit hard to relate to this when I don't have a modern, or indeed, any DSO.

[/list]
Finally this is an optional feature, coming for free in a cheap upper-entry level DSO - no one is forced to use that feature and unlike other instruments, this one keeps the user informed about all important acquisition parameters, like sample rate, record length and 10 bit mode (if active) at the bottom right in the timebase tab, as well as input coupling, input impedance, attenuation factor and detailed bandwidth limit information in the channel tabs (starting from the bottom left).