-
Nobody seems to be willing post anything from Rigol scopes, so I take it as the reason to assume that they are not that good.
I don't think Rigol scopes are bad, they have good reviews. Seems that people having Rigol scopes may be just busy and don't want to go into comparison with Hameg.But as I've try to point few times, I'm really interested in 10bit mode FFT screens, means that if You could turn in Your Hameg HMO scope-- HighPrecision mode ON and run FFT measurement would be very nice.
I'll try to find some time to do you an FFT. Note that you have no control over the number of bits used by the ADC. And as I say before, the ADC is probably just part of the system. For such low-level signal, you may have a preamp before the actual ADC.
Hi
Actually, I did not told that Rigol scopes are bad, but that probably(!!) not that good as Hameg HMO as far as low noise and FFT functionality goes..
..
Anyway bit pity that all Rigoll owners are so busy hire...
I'm keeping my fingers crossed about HighPrecision mode ON FFT measurement...
Please remember about including some marks in dB, I've write long post about it few days ago....
Rosendorfer -
Yes, very busied at the moment, no time to do additional setups. Devices not at work, used in my home lab.
As mentioned above, there's no big difference between high resolution on/off. And it's only a calculated higher resolution, not a physical one.
For your audio purposes any soundcard should be a better tool.
You can't get much better results using an ADC which is limited to 8 Bit physically, even not with a 10 Bit filter extension. If you want to do audio measurements
the span is also in issue, besides the time needed for calculation, which is not as fast as if using a PC with special software.
You should also consider that switching to high res mode does not affect the sensitivity of the scope, i.e., as with low res you can choose 1mV at min.
For audio measurements I've installed software and a M-Audio soundcard with a good SNR, easier to handle and in particular suited for the frequency domain up to 20KHz.
Are you from Germany? Then datatec might give you a try to test the HMO. Just call them, that's better than any screenshots. If there's time, I'll to post a screenshot, currently not, sorry for that.
Kind regards
Gunb -
Hi Rosendorfer,
Unsure what measurement you are after in FFT (dBm/dBV/Veff), bandwidth, on what signal. The Hameg does not seem to have a on/off setting for High Precision. Here's a couple of quick screenshot for a noisy 5mV 2kHz signal. The dBm/div can be turned down to 0.5dBm , but can only inspect the FFT at 5dBm / div
Hope that helps
-
Hi Rosendorfer,
Unsure what measurement you are after in FFT (dBm/dBV/Veff), bandwidth, on what signal. The Hameg does not seem to have a on/off setting for High Precision. Here's a couple of quick screenshot for a noisy 5mV 2kHz signal. The dBm/div can be turned down to 0.5dBm , but can only inspect the FFT at 5dBm / div
Hope that helps
Hi Bernie
First let me say THANK YOU...!!!!!
Finally after few weeks and tons of posts I have screens of Hameg HMO FFT...
And they are looking extremely promising.....!!!
As we already can see some from Your screens very interesting and surprising numbers... as Hameg HMO FFT is claiming it can see something!! down to 1.56uV or -115dBV ; uV to dBV seems to be calculated well. pls.check at: http://www.sengpielaudio.com/calculator-volt.htm
Well that looks very interesting as mater of fact there is even "small problem" these numbers are way too good even for 10bit system..
Looks more like 12 bit resolution... 
Well -115dBV ... sounds good enough to measure noise of most power supplies and Power Amps .
Seems that Modern scopes are not just simple 8 bit devices... Would be very interesting to have somebody comment on that... some reality check, pls...
Would be great if You could do test for noise floor of scope itself and see with no signal at all connected what FFT is telling us at 1 and 2 mV/div , test with probe disconnected also would be interesting.
Second test if You please, using some clean signal, with possibly low noise floor!!! for example sine wave from CD player or Sound Card, sounds like interesting candidate, worth to try ..
It does not have to be done at 1mV/div sensitivity ... Can be at 20, 50 or 100mV or so but is important to use full screen vertical space / 8 divisions / as I described in my previous posts and set up FFT markers with dBV and/or dBm scales so we could see the max signal value and minimum signal above noise floor.....
Relay interested what will come out of that..
Well wonder how about HR mode on/off ..??
Anyway once more thank You for posting Hameg HMO FFT ..screens..!!!
Regards
Rosendorfer
-
Actually the dynamic range is determined by the ADC. Most scopes have an 8 bit ADC which could theoretically yield an FFT range of 48dB (6dB per bit -> 8*6=48). Since most ADCs are noisy, may be clocked by clocks with jitter the effective number of bits (ENOB) will be lower. In worst cases less than 6 bits which would translate in about 36dB. So called high resolution and other averaging methods won't get you more bits of resolution than the ADC has. That doesn't make FFT in a scope totally useless though. Its very handy to check filters and do a quick bandwidth check. Just don't expect a huge dynamic range like a spectrum analyser has.
-
Quote
So called high resolution and other averaging methods won't get you more bits of resolution than the ADC has.
Sure they will. The quantization noise of an ideal ADC is about 1/4 LSB spread over the entire nyquist interval. If you filter and decimate your data, it is just as good as starting with a lower speed, higher resolution ADC. For example, an ideal 8 bit, 1 GSPS ADC with 10 mV full scale with enough dither has a theoretical noise floor of 0.5 nV/sqrt(Hz). That is less than the johnson noise of a 50 ohm resistor. Obviously you can't actually get the implied performance from a real ADC: non-linearity, jitter, and clock feedthrough all limit your useful dynamic range.
-
Actually the dynamic range is determined by the ADC. Most scopes have an 8 bit ADC which could theoretically yield an FFT range of 48dB (6dB per bit -> 8*6=48). Since most ADCs are noisy, may be clocked by clocks with jitter the effective number of bits (ENOB) will be lower. In worst cases less than 6 bits which would translate in about 36dB. So called high resolution and other averaging methods won't get you more bits of resolution than the ADC has. That doesn't make FFT in a scope totally useless though. Its very handy to check filters and do a quick bandwidth check. Just don't expect a huge dynamic range like a spectrum analyser has.
Hi Nctnico
Well if I may ....
Seems that You are just telling what have been probably true Years ago with first wave of digitall scopes....
Please check for Yourself if TODAY even with simple cheap 8Bit dac Your claiming 36dB is still true, from what I have seen!!! it is not...
The reason of that tread is to find out if that what You are claiming is still true or it is just history,
and we all can learn something NEW
I keep my fingers crossed that guys with HMO scopes will have some time to do few few simple test and share the results, so we will just see!!!! what is possible TODAY...
Hope that hope for progress is not anything wrong...
Regards
Rosendorfer
-
Just imagine you have a comparator which can detect 2 levels and a DC level sitting somewhere. If you want to detect the DC level you'll have to add enough noise so the comparator output on average represents the DC level. This scheme has several sources of errors: the quality of the noise and the offset voltage of the comparator. A modern high speed ADC is a string of comparators each with errors which are barely small enough to make the ADC reach its specifications. You can average (oversample) what you want but the result is still loaded with errors.
-
It is important to note that one can always increase the gain and measure smaller and smaller signals. But for dynamic range, it is a bit different situation.
Signal with just one sinusoid is not a good test for dynamic range, better test would be to test with two sinusoids, one with large amplitude and one smaller, how small the smaller can be until it is not reliably identified (e.g. falls below distortion products or noise of larger one)?
Regards,
Janne -
Just imagine you have a comparator which can detect 2 levels and a DC level sitting somewhere. If you want to detect the DC level you'll have to add enough noise so the comparator output on average represents the DC level. This scheme has several sources of errors: the quality of the noise and the offset voltage of the comparator. A modern high speed ADC is a string of comparators each with errors which are barely small enough to make the ADC reach its specifications. You can average (oversample) what you want but the result is still loaded with errors.
Hi
I'm pretty "close to earth" person and like to use something more real than my imagination.
Right now we do have screens from Bernie Hameg HMO FFT's. And while that screens alone are not enough to judge on the subject, there are some very interesting numbers to be seen and as for me that numbers are pretty hard to agree with thing You are telling us hire...
If things are like You are saying.... how Hameg dares to even be able to splash that numbers on its screen
I'm not saying that it alone proves everything but it is there... and looks pretty interesting, well definitely not in Your estimated 36dB range....
I say I like to see more... please.
Hope that Guys with HMO scopes will share more FFT's ...
Well maybe even somebody with access to Rigoll DS4000 would post some.
Regards
Rosendorfer
-
My Tek TDS744A also has a so called hi-res mode (10 bit). I set my signal generator to generate a triangle wave and made it fill the screen of the scope (first picture). Then I fed the same signal through a 40dB (100*) attenuator and set the scope to 100* magnification (second picture). Now check the differences. When the signal is magnified 100* you are looking at about 1.33 bits of the ADC. Now try to FFT that. A sine wave won't be so bad but a more complex waveform like a triangle is perfectly suitable to make imperfections visible.
Note that I used the trigger output of the generator to trigger the scope on the 2nd channel so the trigger is the same for both situations.
-
As someone else mentioned, I was surprised at the size of the Hameg, but the screen is very clear. For what I need, and the price point the Hameg is very good. I think it's a great scope and happy with the interface and features. The Rigol looks very good, but equivalent options would bump up the price compared to the Hameg. Would be interesting to see the Rigol screenshots/feedback.
I thought this link might be useful http://www.analog.com/static/imported-files/tutorials/MT-001.pdf
SNR = 6.02N + 1.76dB , for 10Bit would be 62dB + FFT noise floor 10log10(65536/2) = 62+45 = 107dB
I must admit, I hadn't noticed the HiRes option, it was stuck on the second page of the Acquire menu. It defaults to off, and the other setting is Auto! The FFT floor seems good to about 80-100dB, but I didn't see a great benefit for the FFT in HiRes. It all depends what the signal is and what you want to measure - eg a periodic signal benefits from averaging in the FFT, SNR, bandwidth, samples, window, freq resolution, include LP filtering etc etc.
Hope the attached screenshots help
1) No signal attached (SCR00)
40KHz AM modulated with 3KHz, 500mV pk-pk
2) Waveform (SCR72) , averaged FFT 8Bit (SCR77 / 79), averaged FFT 10Bit HiRes (SCR78 / 80)
3) Some 200mV audio out from an iPhone at 8Bit (96/97/98) - anyone guess the tune
Bernie
-
a first video of my HMO1022, simple start (sorry that it isn't sharp)
http://youtu.be/S3CBZLbjtqo
I measured the voltage AC on a 10V DC power supply on 5V/div and then viewed it on 20mv/div.
It seems the step size was 50mV...
What was the total range 8*5V??
This would mean 40/0.05=800 --> 9.6 bits (10bit <-- range = 10.24div*5V)
8bit, 255*0.05 = 12.75V or 2.55 div
or am I missing something?
(it is just showing the dots, lines can be confusing)
Edit:
Changing the timebase (after measuring 5V/div)
You get 50mv step at 0.5GSa and below (shows "HR" ), 10bit
You get 100mV step at 1Gsa (shows "HR" ), 9bit
You get 200mv step at 2Gsa (hides "HR" ), 8bit
(strange this is 2 bit )
If you disable HiRes you get 200mV all the time, 8bit
edit: added bit depth
-
Did you adjust the volt/div or magnification (zoom)?
-
Did you adjust the volt/div or magnification (zoom)?
Volt/div, maybe it isn't clear if you didn't watch the video, I took a single measurement set. -
If you change the volt/div setting the input signal is scaled (divided or multiplied) so the full range of the ADC is used to allow for a signal of 8 divisions in height.
If you set your scope to 20mV/div it means that the 8 bit resolution is used for a range of 160mV (625uV/bit). If you set the scope to 1V/div then the ADC output represents 31.25mV/bit. -
The measurement does point to 10.2 div instead of 8div. 0.2V*255/5V
edit:
I just did a test (I will make a video later) that shows that it is clipped at just 5.1div on the top and the bottom.
Making it 10.2 div for 8 bit. --> Resolution = 4.0% of V/div (25 steps/div) and in full HiRes 1.0% (100 steps/div) -
But that still doesn't uncover how much of those extra 2 bits are real information or something that is dreamed up by the scope. Using excess digits doesn't make a measurement more accurate. If a multimeter with 10% accuracy shows 1000mV then the real value can be anything between 900mW and 1100mV.
-
But that still doesn't uncover how much of those extra 2 bits are real information or something that is dreamed up by the scope.
there are many detailed information available describing how ADC resolution can be increaed
and what are the traps, e.g.:
http://www.actel.com/documents/Improve_ADC_WP.pdf
http://www.atmel.com/Images/doc8003.pdf
-
But that still doesn't uncover how much of those extra 2 bits are real information or something that is dreamed up by the scope. Using excess digits doesn't make a measurement more accurate. If a multimeter with 10% accuracy shows 1000mV then the real value can be anything between 900mW and 1100mV.
This should help you, the black dots are 9bit the blue ones 10bit (I did moved them up one pixel so they wouldn't cover each other)
Maybe you can check which algorithm fits to calculate the blue ones based on the black ones.
(you will have to zoom in)
Edit attached xls based on csv files
(this is changing 8 to 10 bit after the measurement, so it may work differently on Run)
Edit2:
ReSample2.xls (0.5MB) is with HiRes on and off (8bit vs 10 bit) on the same data both 50MSa.
-
Hi
Can see that KedasProbe is having fun with his new scope...
Having seen HMO FFT's screens I have to say it looks very nice.....
As per KedasProbe request for Rigol screens and to check about nctncio triangle shots,
I have spend some time with my old Rigol DS1052E and run some tests,
so KedasProbe and other HMO owners can see how nice FFT they have...
I hope it will add some practical, real life info to this discussion.
Would be perfect seeing something similar run on HMO scopes.
So There are some Triangles...My Tek TDS744A also has a so called hi-res mode (10 bit). I set my signal generator to generate a triangle wave and made it fill the screen of the scope (first picture). Then I fed the same signal through a 40dB (100*) attenuator and set the scope to 100* magnification (second picture). Now check the differences. When the signal is magnified 100* you are looking at about 1.33 bits of the ADC. Now try to FFT that. A sine wave won't be so bad but a more complex waveform like a triangle is perfectly suitable to make imperfections visible.
Note that I used the trigger output of the generator to trigger the scope on the 2nd channel so the trigger is the same for both situations.
So there are Triangles @10kHz run from Siglent SDG1025..
Screens are with Rigol FFT option set for full screen and markers @dBVrms .
First "full screen" Triangle @ 200mV/div that goes about 1,5V p-p.
then just FFT screen at 10kHz/div , so You can see just FFT more clearly.
Then 100 times attenuation that is 15mV @ 2mV/div
And again just FFT screen @ 10kHz/div and second FFT @ 25kHz/div so You can see that there is no distortion or any problems till about 300kHz. Everything looks nice and smooth as expected.
Wonder what is at nctnico screens.?
I went even bit further..
to.. 5mV p-p triangle..and even to..1mV p-p.
That are special shots for guys that are complaining at DS1052E scope noise...
Still all looks nice and smooth ...
So hope it is pretty safe to say that SFDR or dynamic range at THE CHIPEST RIGOL scope seems to be at list about 60dB.
And second is that FFT noise floor at THE CHIPEST RIGOL scope is about 108-110dBVrms.
And even with THE CHIPEST RIGOL scope some measurment can be done..
Hope that it will help a bit.
And again and again and again .... would be perfect seeing something similar run on HMO scopes.
Or actually any other scope.
Rosendorfer
Bit more in second post.
-
Hi
I have made few more shots I like to show, hope You will like them.
Again from SDG1025 ;
So there is:
Sine wave Full scale 15mVp-p @ 2mV and just FFT's..
I like to show that as hire You can clearly see the noise floor of FFT.
And also low level sine 2mV p-p @2mV.
And...
Even more lowel level as it just triggers bit better:
Square wave at about 800uV p-p ... Yeas pretty clean 800uV p-p on DS1052E scope...
Now I hope on some more nice FFT's from You Guys.
Rosendorfer
-
Hi
You really need a book on how accuracy affects your measurement results
Can see that KedasProbe is having fun with his new scope...
Having seen HMO FFT's screens I have to say it looks very nice.....
As per KedasProbe request for Rigol screens and to check about nctncio triangle shots,
I have spend some time with my old Rigol DS1052E and run some tests,
so KedasProbe and other HMO owners can see how nice FFT they have...
I hope it will add some practical, real life info to this discussion.
Would be perfect seeing something similar run on HMO scopes.
So There are some Triangles...My Tek TDS744A also has a so called hi-res mode (10 bit). I set my signal generator to generate a triangle wave and made it fill the screen of the scope (first picture). Then I fed the same signal through a 40dB (100*) attenuator and set the scope to 100* magnification (second picture). Now check the differences. When the signal is magnified 100* you are looking at about 1.33 bits of the ADC. Now try to FFT that. A sine wave won't be so bad but a more complex waveform like a triangle is perfectly suitable to make imperfections visible.
Note that I used the trigger output of the generator to trigger the scope on the 2nd channel so the trigger is the same for both situations.
So there are Triangles @10kHz run from Siglent SDG1025..
Screens are with Rigol FFT option set for full screen and markers @dBVrms .
First "full screen" Triangle @ 200mV/div that goes about 1,5V p-p.
then just FFT screen at 10kHz/div , so You can see just FFT more clearly.
Then 100 times attenuation that is 15mV @ 2mV/div
And again just FFT screen @ 10kHz/div and second FFT @ 25kHz/div so You can see that there is no distortion or any problems till about 300kHz. Everything looks nice and smooth as expected.
Wonder what is at nctnico screens.?
I went even bit further..
to.. 5mV p-p triangle..and even to..1mV p-p.
That are special shots for guys that are complaining at DS1052E scope noise...
Still all looks nice and smooth ...
So hope it is pretty safe to say that SFDR or dynamic range at THE CHIPEST RIGOL scope seems to be at list about 60dB.
And second is that FFT noise floor at THE CHIPEST RIGOL scope is about 108-110dBVrms. The scope screen starts at approx -45dbV and ends with approx -110dBV with a noise floor at approx -100dB. The effective range is about 55dBV. You should check whether the harmonics have the expected values.
-
You really need a book on how accuracy affects your measurement results The scope screen starts at approx -45dbV and ends with approx -110dBV with a noise floor at approx -100dB. The effective range is about 55dBV. You should check whether the harmonics have the expected values.
Let's see...
So we are not talking about at list 20dB extra from Your 36dB dynamic range of 8 bit ADC but now according to Your "judgment" there are "missing" 5dB.....
But...I have proposition, maybe instead arguing about Rigol DS1052E decibels...
You can actually come back to Your Tek and run some nice FFT's at 8 and 10 bit mode and share results.....
Do not tell, You are going to leave us with impression that Tek TDS744A is that lousy as on Your screens..??
Rosendorfer
-
Maybe it is wise to compare apples with apples. dB itself is rather meaningless.
You are talking dBV, that is 20log and most times using the 1M input ( that, on a Rigol 1192 is dropping like a brick while frequency increases, so harmonics will not be true measurements anymore.
But spectrum analysers ( and my HMO ) can use dBm and 50 ohm. And the 50 ohm input is flat accros the bandwidth.
I have several secteum analysers and I seldom use the FFT of my HMO, not that it is bad but it is light years away in usability compared to a real spectrum analyser.
Rigol DS1102 DC imput impedance