Chinese oscilloscopes with 12 bit AD converter in the analog front-end

[pascal_sweden]

Owon has launched an oscilloscope with a 12 bit AD converter:
http://www.owon.com.hk/products_info.asp?ParentID=57&SortID=87&ProID=182#sthash.EhjBvy98.dpbs

What about the other Chinese/Taiwanese manufacturers? (Rigol, Siglent, GW-Instek)
Does anybody know another Chinese or Taiwanese company besides Owon that produces an oscilloscope with a 12 bit ADC?

How come that most Chinese oscilloscopes only have an 8 bit ADC, while most AVR microcontrollers already have a 10 bit ADC? I understand the sample rate is lower on these microcontrollers, but still, why do they expect a higher vertical resolution in microcontroller applications?

Regarding High Resolution mode on the Rigol scopes:
How is it implemented? By averaging or by oversampling? Or any other techniques?

I found some related documents that are good background reading on this subject matter:

Silicon Labs - Improving ADC Resolution by Oversampling and Averaging:
https://www.silabs.com/Support%20Documents/TechnicalDocs/an118.pdf

Actel - Improving ADC Results Through Oversampling and Post-Processing of Data:
http://www.microsemi.com/document-portal/doc_view/131569-improving-adc-results-white-paper

Agilent - Evaluating High-Resolution Oscilloscopes:
http://cp.literature.agilent.com/litweb/pdf/5991-1617EN.pdf

Analog Devices -  Pushing the State of the Art with Multichannel A/D Converters:
http://www.analog.com/library/analogDialogue/archives/39-05/multichannel.pdf

[marshallh]

It's cheap and relatively easy to throw on a 12 or even 14bit SAR ADC, the flipside is they are very slow (usually not much faster than 1Megsample/second). Just takes 1 comparator.

Basically all DSOs use flash ADCs - a large resistor network and many comparators. They use quite a bit more power, take up more die space and so are more expensive. But you can clock them (most fast ones are heavily pipelined synchronous) into the Gbps range.

Internally many faster ADCs use 2 or 4 or more individual flash ADCs interleaved to give a faster apparent speed. However this introduces additional problems - either the ADC or you have to compensate for gain/offset mismatches and sampling phase delays (not all bits are coming out at the same time).


I think probably Owon has either bought up a large quantity of some major manufacturer's 12bit stock at a cheap price, or they hired a chinese fab to do a die clone of the chip.
It's also possible that its just typical Chinese marketing, where it's a 8bit converter but does averaging at lower speeds.

[Alfons]

"12 bits" is at first marketing screaming. If you want to make 12-bit really usable, the whole thing has to be constructed in accordance with quality around it. And that is very expensive. The units from China are not bad, as long as they move at a normal level. They always lag behind a bit and then bring equipment at very good prices. This can the Chinese perfectly. If they would not only look at the profit and a bit more would be innovative, they could really teach all fear.

[commie]

with 8 bit flash you need 256 comparators. 12 bit you need 4096 comparators and given each comparactor contains about 30 transistors that's 4096x30=122880 transistors approx. for 12 bits. Ooops, needs an encoder aswell.

[nctnico]

Owon has launched an oscilloscope with a 12 bit AD converter:
http://www.owon.com.hk/products_info.asp?ParentID=57&SortID=87&ProID=182#sthash.EhjBvy98.dpbs

What about the other Chinese/Taiwanese manufacturers? (Rigol, Siglent, GW-Instek)
Does anybody know another Chinese or Taiwanese company besides Owon that produces an oscilloscope with a 12 bit ADC?

How come that most Chinese oscilloscopes only have an 8 bit ADC, while most AVR microcontrollers already have a 10 bit ADC? I understand the sample rate is lower on these microcontrollers, but still, why do they expect a higher vertical resolution in microcontroller applications?

Just for fun calculate the jitter requirement to sample 12bit for a 200MHz bandwidth. I doubt the clock in the Owon oscilloscope is that good!

[rf-loop]

It's also possible that its just typical Chinese marketing, where it's a 8bit converter but does averaging at lower speeds.

quessing

Truth: (And well explained in other thread.)

Owon XDS3102A (A=12bit model) ADC is Analog Devices / Hittite Microwave Corporation HMCAD 1520

Inside scope:


So, why speculate this and that. 

[amyk]

How come that most Chinese oscilloscopes only have an 8 bit ADC, while most AVR microcontrollers already have a 10 bit ADC? I understand the sample rate is lower on these microcontrollers, but still, why do they expect a higher vertical resolution in microcontroller applications?

Those 10-bit ADCs are orders of magnitude slower too.

[pascal_sweden]

So the scope really has a 12 bit AD converter?
What about the jitter on the clock?

Does anybody on this forum have such an actual scope from Owon, and if so, are you willing to do some measurements on it, to check how it performs.

Or if Owon is reading this, can you ship the 12 bit AD converter model with 200 MHz BW to Dave Jones, for a full review video? =)

[rf-loop]

So the scope really has a 12 bit AD converter?
What about the jitter on the clock?

Does anybody on this forum have such an actual scope from Owon, and if so, are you willing to do some measurements on it, to check how it performs.

Or if Owon is reading this, can you ship the 12 bit AD converter model with 200 MHz BW to Dave Jones, for a full review video? =)

After ADC datasheet still question, why?

So the scope really has a 12 bit AD converter?

Around 2 months ago   there was available models XDS3102 (8bit 100MHz) and XDS3102A (8/12bit 100MHz).
I do not have more new information.

Yes, I have one XDS3102A with options Wi-Fi, TOU, and VGA.

There is other thread with some tiny tests just after I get it. After then I have been out from work nearly 2 months. So not yet made more tests .  Also it need note that when I get scope it have very early FW. (1.0.0.0)

[robert_]

Would be interesting to see how good it really performs. I sort of would have expected some other major manufacturer would be the second one to go 12bit, not a 3rd tier chinese... Interesting nonetheless.
I did have the opportunity to get my hands on a real 12bit scope (HDO6k series), and it really IS a huge improvement. 10x vertical zoom doesnt result in excessive noise (<1div) at full 500Mhz and with no filtering needed...

[Kleinstein]

From the ADC's datasheet, the ADC is considerably (e.g. 20 times) slower in 12 Bit mode than in 8 Bit mode. This is still better than oversampling. So noise can be lower, but only about a factor of 4 at best - not a factor of 16 as you might expect from 4 extra bits.

Noise also depends on the input amplifier - usually at the highest amplification this is setting the noise limit, not the ADC.

[pascal_sweden]

That's probably why they have provided the menu option to switch between 12 bit mode and 8 bit mode.

But does the vertical resolution affect the memory depth?

My understanding was that the memory depth only relates to the time period which you can record.

Or do they store two 12 bit samples as three 8 bit samples internally in memory?

[Neganur]

It's definitely not 12-bit at 2GSa/s. That ADC can only do 500 MSa/s at 12-bit.
It's a lot of "maximum....", "up to....." put together in a summary.

EDIT: I like the ability to select higher bit rates at lower speeds tho, wonder why it doesn't take full advantage of the 14-bit for some 250 MSa/s stuff.

[rf-loop]

It's definitely not 12-bit at 2GSa/s. That ADC can only do 500 MSa/s at 12-bit.
It's a lot of "maximum....", "up to....." put together in a summary.

EDIT: I like the ability to select higher bit rates at lower speeds tho, wonder why it doesn't take full advantage of the 14-bit for some 250 MSa/s stuff.

Yes, it is of course not 2GSa/s. Models XDS3102 and XDS3102A are max 1GSa/s 100MHz models.
Without A it is 8bit  max 1GSa/s.  Model A is max 1GSa/s 8bit  and max 500MSa/s 12bit.

In these preliminary tables can see 8bit and 12bit  first 1 channel in use and next table 2 channels in use.
(note: these are tested with very first 1.0.0.0 FW)
wfm/s speeds are average speeds, not short time peak values what are perhaps used in advertisements.






 

[pascal_sweden]

@rf-loop: In your profile it says that your home country is China. But in the footer of your postings it says that you are reseller in Finland. Can you provide more details? How come you have all these detailed tables from Owon? =)

[rf-loop]

@rf-loop: In your profile it says that your home country is China. But in the footer of your postings it says that you are reseller in Finland. Can you provide more details?...

I have two home. One in China and one in Finland. At this time I'm home in China. My main home country is Finland and I'm Finnish people. My family is Finnnish/Chinese.
Soon I'm in Finland (and there I'm normally mostly)
My profile show current situation. Forum can not show double country and flag.

... How come you have all these detailed tables from Owon?

I have made these. (data in tables are from my own tests, not from manufacturer or some other sources)

[rf-design]

I have made these. (data in tables are from my own tests, not from manufacturer or some other sources)

I am interessted in your experience with the HMCAD1520. The datasheet actual shows only typ values for the spurious. Are you one of the pilot customer of ADI as these adcs with quad/dual/single modes are specific for scope markets? Do you got more informations than a sample, datasheet and evalboard?

Does ADI have plans to bring up an AFE which is on par with a 12bit class ADC? To me it seems using a high resolution ADC with a 3% DC accuracy AFE little ugly. Also presenting 12bit w/o mention important specs as gain drift, offset drift, droop, overdrive recovery and so on is unfair for the uneducated user. So he wount know that the 12bit promise could not be delivered.

Do you build the AFE discrete? Or is it a mixture of OpAmps and discretes?

[Wuerstchenhund]

I have made these. (data in tables are from my own tests, not from manufacturer or some other sources)

I am interessted in your experience with the HMCAD1520. The datasheet actual shows only typ values for the spurious. Are you one of the pilot customer of ADI as these adcs with quad/dual/single modes are specific for scope markets? Do you got more informations than a sample, datasheet and evalboard?

Does ADI have plans to bring up an AFE which is on par with a 12bit class ADC? To me it seems using a high resolution ADC with a 3% DC accuracy AFE little ugly. Also presenting 12bit w/o mention important specs as gain drift, offset drift, droop, overdrive recovery and so on is unfair for the uneducated user. So he wount know that the 12bit promise could not be delivered.

Do you build the AFE discrete? Or is it a mixture of OpAmps and discretes?

My guess is that rf-loop simply has access to one of these OWON 12bit scopes, not because he's developing a product with that ADC or because he has some eval boards.

[coppice]

Basically all DSOs use flash ADCs.

Name one DSO made in the last 20 years that uses a flash ADC?

[tszaboo]

From the ADC's datasheet, the ADC is considerably (e.g. 20 times) slower in 12 Bit mode than in 8 Bit mode. This is still better than oversampling. So noise can be lower, but only about a factor of 4 at best - not a factor of 16 as you might expect from 4 extra bits.

Noise also depends on the input amplifier - usually at the highest amplification this is setting the noise limit, not the ADC.

maybe I'm ignorant, but what is the point in 12 bits? you would need a screen that resolution to have meaningful representation, and voltige measurements on a scope are affected by the poor DC parameter opamps in the signal chain.
Do you want 12 bits for FFT, or is there any practical reason?

[coppice]

From the ADC's datasheet, the ADC is considerably (e.g. 20 times) slower in 12 Bit mode than in 8 Bit mode. This is still better than oversampling. So noise can be lower, but only about a factor of 4 at best - not a factor of 16 as you might expect from 4 extra bits.

Noise also depends on the input amplifier - usually at the highest amplification this is setting the noise limit, not the ADC.

maybe I'm ignorant, but what is the point in 12 bits? you would need a screen that resolution to have meaningful representation, and voltige measurements on a scope are affected by the poor DC parameter opamps in the signal chain.
Do you want 12 bits for FFT, or is there any practical reason?

Having high res data for post processing, like FFT processing, would be the main practical use. As you said, it doesn't add a lot to a waveform viewed on the screen.

[C]

Think of the one 12 bit data stream as being more scope channels.
Channel one the full signal.
You could then have more channels showing an expanded view of the signal.
Channel two could show the expanded view of the positive peaks of signal.
Channel three could show the expanded view of the negative peaks of signal.
Channel four could show the expanded view of the zero cross of signal.

[tszaboo]

Think of the one 12 bit data stream as being more scope channels.
Channel one the full signal.
You could then have more channels showing an expanded view of the signal.
Channel two could show the expanded view of the positive peaks of signal.
Channel three could show the expanded view of the negative peaks of signal.
Channel four could show the expanded view of the zero cross of signal.

Oh actually it can be viewed like the Megazoom on Agilent scopes.
That makes more samples, so if you zoom in on the x axis, there is still data.
With this 12 bit, if you zoom in on the Y axis, there is still some extra resolution.
Makes sense!

[Neganur]

Name one DSO made in the last 20 years that uses a flash ADC?

Uh, all R&S scopes (and Hameg) unless I'm mistaken.

[coppice]

Name one DSO made in the last 20 years that uses a flash ADC?

Uh, all R&S scopes (and Hameg) unless I'm mistaken.

I can find material on the web which says R&S use converters of their own designs, and I found some performance information about them. I couldn't find any info on the type of converter, though. Do you have any? Its quite possible some people still use flash converters, but its not the mainstream technique these days. Its not the technique the TI, ADI and Hittite (yes I know that is also ADI now) converters in most low end scopes use.

People in forums keep saying that scopes use flash ADCs, but its mostly because they have never bothered to look and see what is really in a scope. Its hard to make a flash converter run fast. It takes too much power, and the die is too big. Unless you need the extremely low latency a flash converter offers it just makes no sense. Pipelined converters run faster, are more accurate, and take far less power.