Hi there,
you all take nice screenshots and press the Hires-button because Rosendorfer likes to see screenshots - no problem with that - but do you really know what you're measuring? What Hires-mode is really doing and what the fundamental basics are behind it?
Honestly I've got the feeling that you're not familiar with time discrete signal processing - no complaint, no accusation.
I mean, it's easy to switch on a scope and produce some nice colorful screenshots, but knowing how it works and what it shows is quite a different thing. And that's why I would argue that screenshots should no be the first chain link but the last.
Let me explain the mistake that I can see in all of your measurements and why you don't see any or only a small difference between normal mode and Hires-mode in most of the screenshots that you already have done.
As already explained to Rosendorfer in one of my previous comments the concept of Hires-mode is to take a few real samples and calculate one to be displayed. This is done at slower timebases where sample rate of the memory is limited and the ADC can provide more samples than can be stored with that speed.
Instead of throwing samples away as in normal mode the ADC runs with its fastest sample rate and takes a group of N samples to calculate 1 to be displayed in Hires-mode. The advantage is that more samples come into play, that the vertical resolution becomes higher depending on the ratio between the maximum sample rate of the ADC and the displayed sample rate and a lower noise floor because higher frequencies are filtered. The disadvantage is a slower signal refresh caused by averaging of the N samples per group and, most of all, that the bandwidth of the signals to be measured becomes lower with each extra bit of resolution!
This method is named "Boxcar Averaging" and the concept is: "don't throw away information, if you can use it".
Where's the limit?
-> It depends on the sample rate the scope is currently working with and that's depending on the selected timebase!
Example #1: you've got a scope with 4GSa/s ADC, the timebase is set to a value that the scope works 4GSa/s, too, that means, it works with its fastest sample rate -> pressing the HR-button has absolutely no effect because there are no samples left to average anything!
Example #2: same scope, but now working with 2GSa/s in normal mode. That means, that the ADC potentially throws every 2nd sample away. Pressing the HR-button now takes exactly 2 real sampled values and calculates exactly 1 displayed sample from it.
Example #3: only to make it crystal clear: My HMO2524 has got an ADC of 2.5GSa/ at max. When I choose a timebase where it samples with 625MSa/s, the ratio is 2.5GSa/s to 625MSa/s equals 4. It means that 3 of 4 samples are thrown away in normal mode. Pressing the Hires-button now takes all 4 samples and calculates the average "on the fly". The result - 1 sample - appears in screen then.
No you might ask: and what all of this has to do with HIGH RESOLUTION?
??
Well, as mentioned many times the algorithm behind that is the "Boxcar Average" that can be implemented as a simple FIR filter algorithm on DSPs, microcontrollers, FPGAs ....
It's a so called one pole filter that takes N samples for 1 average value.
One effect is that the gain of resolution follows the formula:
0.5*log (base 2) N - VERY IMPORTANT to understand!!!
This formula tells you how many extra bits you can win the more N samples are taken into account to calculate your average!
Example #4: If you sample with the maximum sample rate of your scope's ADC, exactly 1 sample is taken into the filter algorithm, i.e. N=1, that is: 0.5*log (base 2) 1 = 0!!! No gain of resolution!
Example #5 :If you sample with half of the ADCs maximum sample rate, i.e., N = 2, the formula says: 0.5*log (base 2) 2 = 0.5!!! You will win 0.5 extra bits of resolution, that is 8.5 bits instead of physical 8 bit
Example #6: If you sample with a quarter of the ADCs maximum sample rate, that leads to 4 samples per group, i.e., N=4, the result will be 0.5*log (base 2) 4 = 1!!! No we're at 9 bit resolution.
Do you understand now why you don't see any difference when you press the HR-button when sampling with the ADC's maximum sample rate?!?!? Or there's nearly no effort if using half of it?!
And that's not all!
One disadvantage of the Boxcar average - the Hires-mode - is that it works as a one pole filter where the -3dB bandwidth depends directly on the
displayed sample rate.
It follows the
3 db bandwidth limit:
0.44 x displayed sample rateSo, if you're scope has got a 2.5GSa/s ADC, the display shows a sample rate of 625MSa/s depending on the chosen timebase, the filter limits the maximum signal frequency to 275MHz with a resolution of 9 bit.
If you really want to see the difference, you should follow these guideline:
#1: calculate the
ratio between the maximum sample rate and the actual displayed sample rate and assign it to
N#2: set it into the formula:
0.5*log(base 2) N and you'll get the
extra bits of resolution#3: depending on the displayed sample rate calculate the
limiting bandwidth of
0.44 * sample rate and investigate if your signal frequency is below that
#4: for demonstration use signals where noise is interfering a small vertical detail to see the difference
#5: then take a screenshot to demonstrate the difference.
--> my explanations show:
- why all manufacturers speak about "up to x bits" extra resolution
- why it never can replace an ADC with real higher resolution
- why it absolutely makes no sense to post screenshots without understanding these fundamental basics, otherwise it's nice & colorful only.
- and most important, that High Resolution is not a constant for any chosen timebase. Now you can calculate yourself how to adjust the scope to reach it.Beside that, look into Rigol's datasheet, they tell more precise than Hameg at which timebase and sample rate the resolution is.
And: I've also the DSA815 but comparing FFT with a real analyzer can be tricky in case of non-periodic signals and knowing nothing about smearing effects and how windows have to be used. But that's quite a different story
Hope it's helpful for further measurements
Kind regards
Gunb