EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: pascal_sweden on June 14, 2020, 10:44:36 am
-
Is it correct that typically the 10-bit, 12-bit and 14-bit modes are only supported for the lower bandwidth range of the oscilloscope?
What is the max. bandwidth range for these higher resolution modes?
Owon has offered an oscilloscope with 12-bit and 14-bit mode in the past.
Did the Owon oscilloscope come with a real 12-bit/14-bit ADC?
Siglent offers a 10-bit mode in the SDS2000X Plus and SDS5000X series.
Does the Siglent oscilloscope come with a real 10-bit ADC or is it emulated in software?
Rigol does not seem to offer a 10-bit mode in the MSO5000/MSO7000/MSO8000 series.
Does the new Rigol chipset on which they spent 5-10 years of development, really only support 8-bit mode?
-
Hello Pascal
The new Siglent models with 10 bit (8 bit plus hi rez) the 12 bit are genuine 12 bit even at 1Ghz currently China only units these are Lecroy hardware licences from an earlier generation HDO models I believe
Keysight s series and new MXR are genuine 10 bits plus high resolution modes in software
The Rigiol manages around 9 bits with the high resolution and precision option selected. The big R&S scopes are pure 8 bits with 8 bits in software
The HDO series and Wavepro Lecroy models are genuine 12 bits all the time even at big bandwidth
Tektronix 5 and 6 series the closest the 6 series gets to a genuine 12 bits is at 1Ghz at other frequencies no. Many occasions the 6 series would drop to 8 bits even at low frequencies no quite sure that pans out with claiming 12 bits though :-DD
What I would say is that correct implementation of whatever formula that company uses to produce a high resolution scope needs to be implemented correctly and the potential user selects according to their requirements.
-
Even a Rigol DS1054Z can pretend to have more than 8 bits if you put it in averaging or hi-res mode.
IMHO: A 'scope shouldn't be advertised as 10bits if doesn't have a true 10bit converter.
Does the Siglent oscilloscope come with a real 10-bit ADC
No.
-
IMHO: A 'scope shouldn't be advertised as 10bits if doesn't have a true 10bit converter'
Spot on :)
-
IMHO: A 'scope shouldn't be advertised as 10bits if doesn't have a true 10bit converter'
Spot on :)
Even a Rigol 1054Z can pretend to have more bits in averaging and hi-res modes. Nobody says they're 9bits or 10bits.
-
Siglent offers a 10-bit mode in the SDS2000X Plus
Correct, also with further enhancement with ERES if desired.
8 bits is default.
All websites list: 10-bit mode
and SDS5000X series
.
Nope, only 8 bit plus ERES enhancement.
Does the Siglent oscilloscope come with a real 10-bit ADC or is it emulated in software?
All are 8 bit at this time.
-
IMHO: A 'scope shouldn't be advertised as 10bits if doesn't have a true 10bit converter.
What makes a converter truly 10 bit? I think you are just trying to create a new fight. :)
-
What makes a converter truly 10 bit?
10 bits of output from the ADC? :-//
-
What makes a converter truly 10 bit? I think you are just trying to create a new fight. :)
10 bits of output from the ADC? :-//
Was that a joke? Its hard to tell on the internet.
-
What makes a converter truly 10 bit? I think you are just trying to create a new fight. :)
10 bits of output from the ADC? :-//
Was that a joke? Its hard to tell on the internet.
10 data lines coming out of the ADC. What's difficult to understand about that?
-
Siglent offers a 10-bit mode in the SDS2000X Plus
All websites list: 10-bit mode
And we already have one person who's confused about whether that means a real 10bit ADC or not.
-
What makes a converter truly 10 bit? I think you are just trying to create a new fight. :)
10 bits of output from the ADC? :-//
Was that a joke? Its hard to tell on the internet.
10 data lines coming out of the ADC. What's difficult to understand about that?
So, you just want more output bits? Those are easy to add. Most of us want meaningful bits, which are much harder. Most 8 bit ADCs in scopes seems to be more like 6 to 7 bits linear, so making them really give 8 bits of linearity would be a big improvement on what we have now.
-
10 data lines coming out of the ADC. What's difficult to understand about that?
So, you just want more output bits? Those are easy to add. Most of us want meaningful bits,
Do you think many people were confused by the phrase "10 bit ADC"?
-
10 data lines coming out of the ADC. What's difficult to understand about that?
So, you just want more output bits? Those are easy to add. Most of us want meaningful bits,
Do you think many people were confused by the phrase "10 bit ADC"?
I find they usually are. You seem to be using the term as though it has a hard meaning, when the number of effective bits coming out of an ADC is a complex question of linearity, noise and other factors. What might be effectively 16 bits for one application, that does a lot of noise averaging over many samples, might be effectively 12 bits for an application where each sample is treated as an accurate measure of a signal, if the ADC is pretty linear, but has a poor SNR. ADC performance is an entire engineering discipline.
-
ADC performance is an entire engineering discipline.
Sure... but I don't think anybody's confused about how to count the number of wires coming out of an ADC.
-
Most of the scope are only using 8bit ADC chips and using software to enhance the resolution, often called hi-res mode (depends on the company how they called it)
in return of processing overhead.
I think many of the scope using 8bits ADC chips and has a feature to enhance the resolution similar to 10bits equivalent with software processing.
Please see the page 26 of R&S RTE1000 since it elaborates in the chart which has only 8bits ADC and using software to enhance up to 16bits.
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/RTE_bro_en_3606-9033-12_v1501_web.pdf
The models which are using 10 or 12bits ADC chips, such as SIGLENT, OWEN, Pico, R&S, Tek and others, will be select-able between 8bits and 10bits or 12 bits but normally it will impact the sampling rate and/or bandwidth.
The description on Tek 4 Series data sheet was easy to understand.
Page 12:
https://download.tek.com/datasheet/4-Series-MSO-MSO44-MSO46-Datasheet-48W615585-new.pdf
-
10 data lines coming out of the ADC. What's difficult to understand about that?
So, you just want more output bits? Those are easy to add. Most of us want meaningful bits,
Do you think many people were confused by the phrase "10 bit ADC"?
I find they usually are. You seem to be using the term as though it has a hard meaning, when the number of effective bits coming out of an ADC is a complex question of linearity, noise and other factors. What might be effectively 16 bits for one application, that does a lot of noise averaging over many samples, might be effectively 12 bits for an application where each sample is treated as an accurate measure of a signal, if the ADC is pretty linear, but has a poor SNR. ADC performance is an entire engineering discipline.
Though effective bits may be a more accurate measure of the capabilities of the underlying instrument, as consumers we need some sort of objective measure of the capabilities of the said instrument. 10-bit ADC should mean an ADC that outputs data that is sampled and is resolved at 10-bits that are coming out of the ADC. The fact that marketing obfuscates the hardware capabilities of higher resolution ADCs with its software-processed "effective" resolution modes or whatever is misleading at its best, and downright lying at its worst.
On the same note, I say publish more data. Another categories to compare for our test instruments, I wouldn't having another column in the comparison tables for effective bits. I bet it would become one of the most popular comparison measure for the sales and marketing professions no doubt!
-
There are several bit number for the ADC:
One is the actual number of data lines coming out. This does not necessary mean the lower bits are accurate and free. The noise and distortion free part is called ENOB. This is less than the physical lines, often something like 1 bit down.
With averaging one can increase the resolution seen by the user, but this reduces the speed or needs multiple waveform sampling with a periodic signal. For the user it makes not real difference if the higher resolution is from averaging or direct support from the HW. However oversampling and averaging mainly improved noise, but does only little to the linearity. To get 2 more bits it needs 16 samples, so the speed goes down quite a bit. This averaging mode over multiple triggers (kind of box-car integration) can still be very useful, as it also reduces the amplifier and input signal noise, not just ADC noise.
So while averaging can not really replace a higher resolution ADC, a higher resolution ADC can not replace a good high res mode (box-car integration).
Beside the ADC also the sampling memory / display part has to support the higher resolution. Higher resolution may cut the memory in half.
-
There are several bit number for the ADC:
One is the actual number of data lines coming out. This does not necessary mean the lower bits are accurate and free. The noise and distortion free part is called ENOB. This is less than the physical lines, often something like 1 bit down.
With averaging one can increase the resolution seen by the user, but this reduces the speed or needs multiple waveform sampling with a periodic signal. For the user it makes not real difference if the higher resolution is from averaging or direct support from the HW.
The latter is not quite accurate. First of all getting more bits from an ADC needs a certain amount of white noise. If that noise isn't there then the extra bits won't appear. Secondly the ADC needs to be linear enough to get a useful result from oversampling. In reality 8 bit ADCs aren't designed to produce more than 8 bits and noise may not be there. I have done some experiments on a Tektronix TDS700 series oscilloscope a while back and in some cases I could get very 'interesting' signals in high-res mode which had nothing to do with reality. All in all IMHO an oscilloscope which says 10bit on the badge should have a physical 10 bit ADC under the hood.
For older DSOs the ENOB was mentioned in the datasheet. That somehow got lost in datasheets for modern DSOs.
-
Just want to add that, for the sake of correctness, the resolution of an ADC shouldn't be determined by "the number of lines" since you can have serial output.
-
...or differential pairs... ;)
Just have a look a the ADC08D1000 (https://www.ti.com/lit/gpn/adc08d1000) dual channel 8 bit converter that is interfaced to the sampling engine via a whopping 64 lines for the data only, housekeeping stuff not included.
-
Just want to add that, for the sake of correctness, the resolution of an ADC shouldn't be determined by "the number of lines" since you can have serial output.
There's ADC chips with serial interfaces, yes, but somewhere inside the chip there'll be a countable number of wires.
-
There's ADC chips with serial interfaces, yes, but somewhere inside the chip there'll be a countable number of wires.
Sure, I assume these chips aren't unifilar... ;D Let us not go that path...
-
Most of the scope are only using 8bit ADC chips and using software to enhance the resolution, often called hi-res mode (depends on the company how they called it) in return of processing overhead.
I think many of the scope using 8bits ADC chips and has a feature to enhance the resolution similar to 10bits equivalent with software processing.
Please see the page 26 of R&S RTE1000 since it elaborates in the chart which has only 8bits ADC and using software to enhance up to 16bits.
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/RTE_bro_en_3606-9033-12_v1501_web.pdf
The models which are using 10 or 12bits ADC chips, such as SIGLENT, ..............
Again, Siglent do not use 10 or 12 bit ADC's, at least in DSO's outside the China marketplace.
-
I believe Owon is using the HMCAD1520 for their 12 and 14 bit scopes.
The Owon XDS-2102A does real 12 bit sampling to a depth of 20 Mpts. It's pretty much unusable as a DSO, but it *will* acquire 20 Mpts at 500 MSa/s. The aesthetics are as bad as the software, but it does work. However, it does a zero phase sinc(t) interpolation both in dot mode and in vector mode. So you get a precursor on a fast step. Instek does the same zero phase interpolator, but at least they don't do it in dot mode, so you can get an accurate picture of the waveform using persistence and dot mode.
Have fun!
Reg
-
Is it correct that typically the 10-bit, 12-bit and 14-bit modes are only supported for the lower bandwidth range of the oscilloscope?
What is the max. bandwidth range for these higher resolution modes?
Owon has offered an oscilloscope with 12-bit and 14-bit mode in the past.
Did the Owon oscilloscope come with a real 12-bit/14-bit ADC?
Siglent offers a 10-bit mode in the SDS2000X Plus and SDS5000X series.
Does the Siglent oscilloscope come with a real 10-bit ADC or is it emulated in software?
Rigol does not seem to offer a 10-bit mode in the MSO5000/MSO7000/MSO8000 series.
Does the new Rigol chipset on which they spent 5-10 years of development, really only support 8-bit mode?
The discussion misses out some important issues:
- ADC are characterized not only by speed and the number of bits, but also by *noise*
- noise gets more disturbing as frequency is increased (and approaches the Nyquist frequency).
- you may trade in some speed by oversampling and boxcar averaging. This reduces noise, but
limits the maximum signal frequency.
- If your signal is too fast so no oversampling/averaging is possible, the resolution left is ADC bits minus noise.
- The important number is called ENOB (Effective number of bits). Approaching the bandwidth limit, this usually drops by a few
bits from the low frequency value. Example: Keysight S-Series with 10Bit ADCs: LF ENOB ca. 8, dropping to 6.X at some GHz.
Extreme example: Keysight UXR 110GHz 10Bit ADC: ENOB@110GHz ca. 5.
Result:
- Cheap scopes are no accurate digitizers whatever they claim. ENOB is not even specified in a lot of cases.
- If you need real 12-14Bit at a few MHz, you need a 16Bit 100MHz instrument plus some pocket money
(more a digitizer than a scope)
-
The classic use of an oscilloscope is to get an understanding of the waveform. It is a picture on a screen and somewhere not far beyond 8 bits the picture doesn't improve. Digital oscilloscopes have blurred the lines between an oscilloscope and a data acquisition system. If you are using the scope as a DAS there are many more questions than how many bits it has that must be answered.
-
IMHO: A 'scope shouldn't be advertised as 10bits if doesn't have a true 10bit converter.
What makes a converter truly 10 bit? I think you are just trying to create a new fight. :)
Most (all?) practical GS/s scope ADCs are based on a massive array of slower and less accurate ADCs that are then calibrated to produce a more accurate composite ADC. Where those corrections are applied isn't always in the ADC so much more data can be coming out of the ADC than the "specification" of the converter. There is a trade-off between bit-depth/ENOB and the sample rate (oversampling), most scopes don't expose that to simplify the calibration but some do. So a "true" 10 bit converter is like a "true" Scotsman, open to varying interpretations.
-
Most of the scope are only using 8bit ADC chips and using software to enhance the resolution, often called hi-res mode (depends on the company how they called it) in return of processing overhead.
I think many of the scope using 8bits ADC chips and has a feature to enhance the resolution similar to 10bits equivalent with software processing.
Please see the page 26 of R&S RTE1000 since it elaborates in the chart which has only 8bits ADC and using software to enhance up to 16bits.
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/RTE_bro_en_3606-9033-12_v1501_web.pdf
The models which are using 10 or 12bits ADC chips, such as SIGLENT, ..............
Again, Siglent do not use 10 or 12 bit ADC's, at least in DSO's outside the China marketplace.
Oh, I see. Thanks, I will fix the SDS2000X-Plus column in the oscilloscope charts. I meant to be only hardware ADC bits in that column.
I will review the other models as well.
-
Most of the scope are only using 8bit ADC chips and using software to enhance the resolution, often called hi-res mode (depends on the company how they called it) in return of processing overhead.
I think many of the scope using 8bits ADC chips and has a feature to enhance the resolution similar to 10bits equivalent with software processing.
Please see the page 26 of R&S RTE1000 since it elaborates in the chart which has only 8bits ADC and using software to enhance up to 16bits.
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/RTE_bro_en_3606-9033-12_v1501_web.pdf
The models which are using 10 or 12bits ADC chips, such as SIGLENT, ..............
Again, Siglent do not use 10 or 12 bit ADC's, at least in DSO's outside the China marketplace.
Oh, I see. Thanks, I will fix the SDS2000X-Plus column in the oscilloscope charts. I meant to be only hardware ADC bits in that column.
I will review the other models as well.
... maybe an ENOB column is helpful.
-
The classic use of an oscilloscope is to get an understanding of the waveform. It is a picture on a screen and somewhere not far beyond 8 bits the picture doesn't improve.
That's true but if the oscilloscope offers math functions like FFT then more bits makes a huge difference.
-
The classic use of an oscilloscope is to get an understanding of the waveform. It is a picture on a screen and somewhere not far beyond 8 bits the picture doesn't improve. Digital oscilloscopes have blurred the lines between an oscilloscope and a data acquisition system. If you are using the scope as a DAS there are many more questions than how many bits it has that must be answered.
I disagree. When you look at the new Keysight MXR series, ... the future is mixed spectral/time domain analysis. You could come from two ways: real time spectrum analysis OR real time scope with FFT capabilities. The faster things get (5G, ....) the more advantages the time domain approaches have. And then, you need more than just 8 bits otherwise the dynamic range of the FFTs become too small.
-
If I understand correctly, there are 2 approaches to increase the resolution: averaging and oversampling.
Averaging works only for repetitive waveforms.
It does not reduce the bandwidth.
It reduces the waveform update rate.
Oversampling works for both repetitive waveforms and non-repetitive waveforms.
It reduces the bandwidth.
It does not reduce the waveform update rate.
Are these assumptions correct?
-
If I understand correctly, there are 2 approaches to increase the resolution: averaging and oversampling.
Oversampling works for both repetitive waveforms and non-repetitive waveforms.
It reduces the bandwidth.
It does not reduce the waveform update rate.
Averaging works only for repetitive waveforms.
It does not reduce the bandwidth.
It reduces the waveform update rate.
Are these assumptions correct?
Not quite...
At a previous company we referred to these modes as 'fast time averaging' and 'slow time averaging'. In this context, 'fast time' means the time between successive ADC measurements, at the 'raw' sample rate, e.g. 5GHz or so for an oscilloscope. 'Slow time' means the time between successive trigger events, which may be of the order of milliseconds.
You refer to oversampling - in my experience this generally refers to the difference between the rate at which an ADC samples its input voltage and the rate at which it generates (statistically independent) output samples. For example, a Delta-Sigma type ADC may sample its input 128 times for each output sample, and the input conversion may be only 1 bit wide, while the output may be 16 bits or more. This all happens inside the ADC.
Fast time averaging works as you have described for oversampling. In the oscilloscope world, this is generally referred to as Hi-Res mode, or ERES in some cases (LeCroy?) where something other than an equal weighting is applied to the input samples being averaged. This can, in theory, slightly improve the output SNR of the fast time averaged data. Fast time averaging in oscilloscopes is also generally used to reduce (decimate) the ADC sample rate, in order to store a long enough time record in the available acquisition memory. It doesn't have to work this way though, the average can be recalculated for every incoming ADC sample. This doesn't change the bandwidth reduction.
Slow time averaging does reduce the bandwidth (that's how it gets rid of excess noise) but instead of a passband running from DC to some high frequency limit, it creates a series of separate passbands, centred on the harmonics of the trigger repetition rate (plus DC) and filters out noise in the multiple stopbands ('gaps') between them. The highest frequency limit of the highest stopband will still be unchanged from its non-averaged value, though.
The waveform update rate may or may not be reduced. Most oscilloscopes implement what we used to call a 'slow-update' averaging scheme. As each new trigger event is added to the average, the output is updated. So the screen display will change with each trigger event received - unless the overhead introduced by making the averaging calculations increases the 'dead time' enough that one or more trigger events are missed.
-
Most of the scope are only using 8bit ADC chips and using software to enhance the resolution, often called hi-res mode (depends on the company how they called it) in return of processing overhead.
I think many of the scope using 8bits ADC chips and has a feature to enhance the resolution similar to 10bits equivalent with software processing.
Please see the page 26 of R&S RTE1000 since it elaborates in the chart which has only 8bits ADC and using software to enhance up to 16bits.
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/RTE_bro_en_3606-9033-12_v1501_web.pdf
The models which are using 10 or 12bits ADC chips, such as SIGLENT, ..............
Again, Siglent do not use 10 or 12 bit ADC's, at least in DSO's outside the China marketplace.
Oh, I see. Thanks, I will fix the SDS2000X-Plus column in the oscilloscope charts. I meant to be only hardware ADC bits in that column.
I will review the other models as well.
... maybe an ENOB column is helpful.
I just added the ENOB column.
https://docs.google.com/spreadsheets/d/1rtPqMAkNw2bSkqocuc7zcYkl57fsdPVlxzyqcJmbKIc
Appreciate anyone who adds the ENOB as a comments to it. :)
-
Is there a good book about ADC converters covering in details the oversampling and averaging techniques?
-
It's not about ADC convertors, but the best book about the mathematics is:
Random Data
Bendat & Piersol
4th ed
It's a comprehensive treatment of time series analysis in the tradition initiated by Norbert Wiener in 1940 with "The Extrapolation, Interpolation and smoothing of Stationary Time Series" which was declassified and published in 1949.
Averaging to increase the number of bits depends upon an assumption of Gaussian distributed random noise. However, there is more to it as it is possible to acquire single bit data and by cross correlation produce any arbitrary number of bits of resolution. Bendat & Piersoll give a brief summary of that also under the rubric of "hard clipping". Sam Allen acquired 1024 channel Vibroseis data using single bit geophones in the 60's before ADCs were capable of handling that many channels. The requirement was the signal had to be *less* than the noise, i.e. SNR < 1.0.
Have Fun!
Reg