REVIEW - Owon SDS7102 - A look at the SDS series from Owon

[akis]

Hi guys, I have tried to read some of this thread but it is long and am not sure what is the most recent info. I want to buy a new scope, and was thinking of the OWON SDS7102 or the Hantek dso5202b - for example. Any advice would be welcomed.

Or maybe there are other, better scopes in the price range I could go for?

[marmad]

Hey guys, sorry - I've been travelling but am now settled down with some free time - and can hopefully finish updating this thread and re-posting my video review this week.

SH@RK & akis: I returned the Owon quite awhile ago - and was waiting either for Owon to significantly update their firmware - or for a better 'bang for the buck' scope to appear. I'm still waiting - and in the meantime, I have an 'oscilloscope piggy-bank' - where I throw spare change - so that I can afford (without seeming to spend more money) to jump to a higher-tier of scope.

What scope is best for you depends on a lot of factors - not just the price. Are you a hobbyist, semi-professional, or professional? What do/would you use the scope for most of the time: general purpose, audio work, high-speed digital circuitry, etc? Are there particular features which are critical for you: screen size, PC control and communication, battery-operation, etc?

I might still own the Rigol DS1052E if it had a larger screen (and a quieter fan would be good - though you can fix that yourself if desired). I might still own the Owon if it had either better firmware tools for large record lengths - or a faster waveform update rate with small record lengths. I might still own the Hantek if it had either better PC c/c or less buggy firmware. Each scope has attributes and faults which have to be balanced against your needs.

I will post a more thorough comparison and suggestion guide later this week, but in the meantime - in short - my opinion is this:

You are a hobbyist/semi-professional AND you do mainly audio work: you need Bode plotting for filter construction/debugging - consider the Velleman PCSGU250 - a great USB scope bargain (and the only with built-in Bode plots) at around € 130 in the EU.

You are a hobbyist/semi-professional AND glitch-hunting/debugging high-speed digital circuitry is NOT your primary scope use: you need deep memory OR battery-operation OR VGA output (for teaching or presentation)- AND PC ctrl/comm. is NOT important - get the Owon.

You are a hobbyist/semi-professional: you want large screen OR hackability - AND PC ctrl/comm via standardized SCPI commands is NOT important - get the Hantek.

You are a hobbyist/semi-professional: you need cheapest price OR PC ctrl/comm via SCPI - AND screen size is NOT important - get the Rigol.

You are semi-professional/professional: you need large screen AND fast update rates AND PC ctrl/comm - save some money and get something that costs 2 - 2.5x more - but is 5 - 10x better.

Edit: BTW, by hobbyist, semi-professional, and professional - I strictly mean (in this context) that you will make zero/a bit of/a reasonable amount of/ money from the use of the oscilloscope.
 

...but that said the update rate for most of the ranges beats the hantek, the hantek is appalling most of the time

Dave.S: What data are you basing your statements on? Did you have both scopes and run side-by-side measurements? Or did you just own the Hantek for awhile before the Owon? My own impressions, after having both scopes, was that the Hantek was significantly faster with 1k record lengths on many ranges.

[T4P]

Yes but only on 1K.

[0xdeadbeef]

While this again is a bit offtopic, I recently investigated into scopes in the 1k€-3k€ price range. And I'm kinda depressed right now, since all of them lack at least one of the features that I's like to have.
Stuff like measuring in the sample data (only for Hameg and LeCroy I'm really sure that they use the sample data), statistics (Tektronix DPO2000 simply has none, others have strange implementations), gating for measurements (no Agilent scope seems to have sensible gating as they obviously think the zoom window would be good enough, the Hamegs only support a kind of gating for semi-manual measurements), counting pulses/edges (surprisingly the Rigol DS4000 seems to miss this point completely). Also basic features are surprisingly badly designed. E.g. as the Owon, the Agilent DSX3000 has no button to switch between auto/normal trigger mode though this is something that you need all the time.

Generally, I'm underwhelmed by the displayed precision for cursor positions and measurements. With 1GSa/s and above, it should be somehow possible to measure if a pulse is 100µs or 100.01µs. That's 10ns and with 1GSa/s 10 points in the sample data. Still, scopes in the <8k€ range seem to cripple their measurements by either measuring in the display data or at least showing only three or four digits. The best result you can expect is something like 100.0µs, which reduces the accuracy to 100ns. Better than nothing you could say, but still not enough for certain purposes and I still think that a 1GSa/s scope should allow you to measure a period down to the sample period (1ns), even if there will be a certain jitter of course.

Then again, the SDS lacks all of these features. It's nice for observing the shape of waveforms and you can do rough measurements with it, but for precise measurements, it's completely useless. I'm not convinced though, that any other scope in this price range is really much better in this regard.

[marmad]

While this again is a bit offtopic, I recently investigated into scopes in the 1k€-3k€ price range. And I'm kinda depressed right now, since all of them lack at least one of the features that I's like to have.

I think that DSO prices in general are rather depressing. They certainly seem overpriced compared to other electronic goods - and while you can argue that they are 'test equipment' and about development costs - to me it seems to have at least as much to do with history and captive market as anything else: the analog scopes cost X amount - and since these are similar (or better) in capabilities (though cheaper in terms of manufacture, materials and calibration), they should still cost X amount too.

[0xdeadbeef]

Yeah well, measurement equipment is a small market compared to consumer electronics - and low quantity usually means high prices.  Also the problem is of course that mainly companies buy scopes and e.g. 10k€ is not really much money for a company. That's about what an external engineer costs per month. That also explains the crazy prices for addons, plugins etc.
E.g. an SPI decoder for a LeCroy 6Zi costs around 1k€ (excluding VAT). That's simply insane for something so simple considering that measuring periods is already implemented. Then again, for a company that's petty cash. As a side note: surprisingly, Rigol seems to be the 2nd most expensive company regarding addons. They sell I2C, SPI and UART separately where each protocol costs as much as a 2-protocol bundle from Agilent or Tektronix. Let's see if they offer better prices for the DS2000 line at least.

Anyway, I still hope that the prices will come down eventually - also regarding arbitrary generators. But currently it doesn't really look like it. There is nearly no competition amongst dealers either. The prices of all the brand scopes I observe are rock stable, close to no current DSO is sold on eBay and if so, people pay nearly the full price...

[T4P]

OOPSOOPSOOPS

[akis]

While this again is a bit offtopic, I recently investigated into scopes in the 1k€-3k€ price range. And I'm kinda depressed right now, since all of them lack at least one of the features that I's like to have.
Stuff like measuring in the sample data (only for Hameg and LeCroy I'm really sure that they use the sample data), statistics (Tektronix DPO2000 simply has none, others have strange implementations), gating for measurements (no Agilent scope seems to have sensible gating as they obviously think the zoom window would be good enough, the Hamegs only support a kind of gating for semi-manual measurements), counting pulses/edges (surprisingly the Rigol DS4000 seems to miss this point completely). Also basic features are surprisingly badly designed. E.g. as the Owon, the Agilent DSX3000 has no button to switch between auto/normal trigger mode though this is something that you need all the time.

Generally, I'm underwhelmed by the displayed precision for cursor positions and measurements. With 1GSa/s and above, it should be somehow possible to measure if a pulse is 100µs or 100.01µs. That's 10ns and with 1GSa/s 10 points in the sample data. Still, scopes in the <8k€ range seem to cripple their measurements by either measuring in the display data or at least showing only three or four digits. The best result you can expect is something like 100.0µs, which reduces the accuracy to 100ns. Better than nothing you could say, but still not enough for certain purposes and I still think that a 1GSa/s scope should allow you to measure a period down to the sample period (1ns), even if there will be a certain jitter of course.

Then again, the SDS lacks all of these features. It's nice for observing the shape of waveforms and you can do rough measurements with it, but for precise measurements, it's completely useless. I'm not convinced though, that any other scope in this price range is really much better in this regard.

I am very, very new to the idea of digital scopes and sampling used for displaying a trace but:

Some rough calcs  : with 1 billion samples per second rate, I'd want say 20 samples per period to be able to show a decent curve, and that would mean a max input frequency of 50 MHz. If the curve is more complex I'd want even more samples per period meaning less max frequency capability. In that respect if the scope is being sold as a two channel 50 MHz scope I'd think it's about right. At 100MHz that would leave 10 samples per period on one trace and 5 samples per period for dual trace. Can you really show a trace with 5 or 10 samples over its period? I leave that to the experts to answer

Then it is also the sampling depth , or whatever else we call this : "bits per sample" - that means if the screen has 8 vertical divisions, how many bits do we allocate per division? I'd hope we would allocate at least 3 bits per division (0 to 7, ie seven distinct levels within each verticule) making this 3 * 8 = 24 bits per sample. Modern VGA cards work at 32 bits per pixel and at GHz levels so I do not see a problem here. But something tells me that all these digital scopes do not have 24 bit sampling depth - am I right?

[T4P]

While this again is a bit offtopic, I recently investigated into scopes in the 1k€-3k€ price range. And I'm kinda depressed right now, since all of them lack at least one of the features that I's like to have.
Stuff like measuring in the sample data (only for Hameg and LeCroy I'm really sure that they use the sample data), statistics (Tektronix DPO2000 simply has none, others have strange implementations), gating for measurements (no Agilent scope seems to have sensible gating as they obviously think the zoom window would be good enough, the Hamegs only support a kind of gating for semi-manual measurements), counting pulses/edges (surprisingly the Rigol DS4000 seems to miss this point completely). Also basic features are surprisingly badly designed. E.g. as the Owon, the Agilent DSX3000 has no button to switch between auto/normal trigger mode though this is something that you need all the time.

Generally, I'm underwhelmed by the displayed precision for cursor positions and measurements. With 1GSa/s and above, it should be somehow possible to measure if a pulse is 100µs or 100.01µs. That's 10ns and with 1GSa/s 10 points in the sample data. Still, scopes in the <8k€ range seem to cripple their measurements by either measuring in the display data or at least showing only three or four digits. The best result you can expect is something like 100.0µs, which reduces the accuracy to 100ns. Better than nothing you could say, but still not enough for certain purposes and I still think that a 1GSa/s scope should allow you to measure a period down to the sample period (1ns), even if there will be a certain jitter of course.

Then again, the SDS lacks all of these features. It's nice for observing the shape of waveforms and you can do rough measurements with it, but for precise measurements, it's completely useless. I'm not convinced though, that any other scope in this price range is really much better in this regard.

I am very, very new to the idea of digital scopes and sampling used for displaying a trace but:

Some rough calcs  : with 1 billion samples per second rate, I'd want say 20 samples per period to be able to show a decent curve, and that would mean a max input frequency of 50 MHz. If the curve is more complex I'd want even more samples per period meaning less max frequency capability. In that respect if the scope is being sold as a two channel 50 MHz scope I'd think it's about right. At 100MHz that would leave 10 samples per period on one trace and 5 samples per period for dual trace. Can you really show a trace with 5 or 10 samples over its period? I leave that to the experts to answer

Then it is also the sampling depth , or whatever else we call this : "bits per sample" - that means if the screen has 8 vertical divisions, how many bits do we allocate per division? I'd hope we would allocate at least 3 bits per division (0 to 7, ie seven distinct levels within each verticule) making this 3 * 8 = 24 bits per sample. Modern VGA cards work at 32 bits per pixel and at GHz levels so I do not see a problem here. But something tells me that all these digital scopes do not have 24 bit sampling depth - am I right?

Yep. Modern VGA cards working at 32bits @ 1680x1050 as i'm typing away ... i'm not too sure if they run at GHz levels OR not , at least i know the pixel clock on mine is at 550MHz but i agree it's GHz levels because of the 2.2 Gigapixels/sec pixel fill rate ( It's a Mobility HD 5430 anyway )

[krenzo]

I REALLY wish there was competition , i mean like okay 1Gigsamples ... What's the speed of a standard ADC converter in a off the shelf notebook you find now?
They have 24 bits 5gigsamples ADC's in there.

Really?  Please elaborate on where I can find such a notebook and why it would need such a powerful chip.

[alm]

I REALLY wish there was competition , i mean like okay 1Gigsamples ... What's the speed of a standard ADC converter in a off the shelf notebook you find now?
They have 24 bits 5gigsamples ADC's in there.

Where? The only analog input signals that come to mind are audio and wifi, and neither has anywhere near 5Gb/s bandwidth.

AND quad-core 1.4GHz 35W chips ... i mean what's stopping agilent to make cheaper stuff ? They have HP by their side and a fab of their own ( they make the megazoom asics so i assume they have a fab of theirs )

As was already answered by 0xdeadbeef, scale. Look at the number of CPU's Intel sells compared to the number of scopes Agilent sells. I can't be bothered to do the research now, but I expect many orders of magnitude difference.

I think that DSO prices in general are rather depressing. They certainly seem overpriced compared to other electronic goods - and while you can argue that they are 'test equipment' and about development costs - to me it seems to have at least as much to do with history and captive market as anything else: the analog scopes cost X amount - and since these are similar (or better) in capabilities (though cheaper in terms of manufacture, materials and calibration), they should still cost X amount too.

Analog scopes were much more expensive in their heydays. In 1959, the 30 MHz Tektronix 545A cost $1550 without the essential plug-in amplifiers, about $12,000 in 2011 money. In 1977, a 100 MHz Tektronix 465 cost $2145, which would be something like $8000 in 2011 money. In 1993, when DSOs were already available, the 100 MHz Tektronix 2245A (cheaper construction, less parts) cost $2595, about $4000 in 2011 money.

Yep. Modern VGA cards working at 32bits @ 1680x1050 as i'm typing away ... i'm not too sure if they run at GHz levels OR not , at least i know the pixel clock on mine is at 550MHz but i agree it's GHz levels because of the 2.2 Gigapixels/sec pixel fill rate ( It's a Mobility HD 5430 anyway )

You're confusing a lot of specs. The pixel clock of a 1680x1050 VGA signal is most likely something like 150 MHz, and is produced by a DAC, not an ADC. Of those 32-bits per pixel, 8 bits are alpha channel and never transmitted to the monitor, and the other 24 bits are split across three different colors. Also note that VGA connectors on computers are usually outputs, not inputs. Unless you have a fairly old monitor, chances are that you're using a digital (DVI/HDMI/DisplayPort) connection, which does not involve any ADC/DAC. 550 MHz is most likely the clock rate of the (digital) GPU. 2.2 Gigapixels (texels?) / sec refers to the number of pixels the GPU/graphics memory can process per second.

[0xdeadbeef]

Some rough calcs  : with 1 billion samples per second rate, I'd want say 20 samples per period to be able to show a decent curve, and that would mean a max input frequency of 50 MHz. If the curve is more complex I'd want even more samples per period meaning less max frequency capability. In that respect if the scope is being sold as a two channel 50 MHz scope I'd think it's about right. At 100MHz that would leave 10 samples per period on one trace and 5 samples per period for dual trace. Can you really show a trace with 5 or 10 samples over its period? I leave that to the experts to answer

I'm not sure if or how this could relate to my posting. Does it at all? If not, why quoting it?
Anyway, to reconstruct a sine, you need a sample rate that's at least twice the sine frequency (Shannon and friends). 10 samples per period are not that much, so a non-sinewave might look a bit ugly depending on the reconstruction algorithm applied, but that this shoudln't influence period measurement. I was talking about measuring a 100µs pulse period +/-10ns. So in the sample buffer there are ten more values with the same high voltage level and even if the automatic measurement fails and the displayed edge looks a bit SINCinsh, I want to be able to set the cursors at the samples that I consider to be the edges and get an accurate delta time displayed (sample_number/sample_frequency) with a good precision.
A decent scope can do this and I don't see why implementing this should be difficult.

Then it is also the sampling depth , or whatever else we call this : "bits per sample" - that means if the screen has 8 vertical divisions, how many bits do we allocate per division? I'd hope we would allocate at least 3 bits per division (0 to 7, ie seven distinct levels within each verticule) making this 3 * 8 = 24 bits per sample. Modern VGA cards work at 32 bits per pixel and at GHz levels so I do not see a problem here. But something tells me that all these digital scopes do not have 24 bit sampling depth - am I right?

Now you lost me completely. Even scopes in the multi k€ price range only have an 8bit sample resolution (and "hires" modes are usually nothing but oversampling). But this has absolutely nothing to do with the screen resolution or the display bandwidth or whatever. Honestly, I don't see your point at all.
Anyway, when implemented correctly, the display resolution (letting aside the bandwidth) has nothing to do with measurement accuracy. Again, with a decent scope, I can set both cursors in a maximum zoom mode on the exact sample and get the correct distance between those two cursors, no matter what the current zoom level is. As said before: measuring in the display data is a design flaw and I don't get why e.g. Agilent engineers and customers see this differently.

[T4P]

I REALLY wish there was competition , i mean like okay 1Gigsamples ... What's the speed of a standard ADC converter in a off the shelf notebook you find now?
They have 24 bits 5gigsamples ADC's in there.

Where? The only analog input signals that come to mind are audio and wifi, and neither has anywhere near 5Gb/s bandwidth.

AND quad-core 1.4GHz 35W chips ... i mean what's stopping agilent to make cheaper stuff ? They have HP by their side and a fab of their own ( they make the megazoom asics so i assume they have a fab of theirs )

As was already answered by 0xdeadbeef, scale. Look at the number of CPU's Intel sells compared to the number of scopes Agilent sells. I can't be bothered to do the research now, but I expect many orders of magnitude difference.

I think that DSO prices in general are rather depressing. They certainly seem overpriced compared to other electronic goods - and while you can argue that they are 'test equipment' and about development costs - to me it seems to have at least as much to do with history and captive market as anything else: the analog scopes cost X amount - and since these are similar (or better) in capabilities (though cheaper in terms of manufacture, materials and calibration), they should still cost X amount too.

Analog scopes were much more expensive in their heydays. In 1959, the 30 MHz Tektronix 545A cost $1550 without the essential plug-in amplifiers, about $12,000 in 2011 money. In 1977, a 100 MHz Tektronix 465 cost $2145, which would be something like $8000 in 2011 money. In 1993, when DSOs were already available, the 100 MHz Tektronix 2245A (cheaper construction, less parts) cost $2595, about $4000 in 2011 money.

Yep. Modern VGA cards working at 32bits @ 1680x1050 as i'm typing away ... i'm not too sure if they run at GHz levels OR not , at least i know the pixel clock on mine is at 550MHz but i agree it's GHz levels because of the 2.2 Gigapixels/sec pixel fill rate ( It's a Mobility HD 5430 anyway )

You're confusing a lot of specs. The pixel clock of a 1680x1050 VGA signal is most likely something like 150 MHz, and is produced by a DAC, not an ADC. Of those 32-bits per pixel, 8 bits are alpha channel and never transmitted to the monitor, and the other 24 bits are split across three different colors. Also note that VGA connectors on computers are usually outputs, not inputs. Unless you have a fairly old monitor, chances are that you're using a digital (DVI/HDMI/DisplayPort) connection, which does not involve any ADC/DAC. 550 MHz is most likely the clock rate of the (digital) GPU. 2.2 Gigapixels (texels?) / sec refers to the number of pixels the GPU/graphics memory can process per second.

Oops, but yeah i am aware the actual speed is 150MHz, STILL faster then a low end scope though.
I correct whatever i said above, just ignore it. 2.2gigpixels is a different thing to texels it has like 8.8 Gigasamples/sec samples fill rate and 4.4gigtexels fill rate
ANYWAY, i do know 550MHz is the core clock, yes i know the stuff i work on as i design graphical display outputs
But still 32bits is 32bits

[tinhead]

no, i'm not protecting HanTekway in any case, they still have, let me think, 10+ bug for sure,
but let's compare a bit to Owon to see the MAJOR hadrwae difference (and this was already said multiple times here
and in Tekway thread).

Yes but only on 1K.

no, it's not like that, you have to understand the hardware first.
Generally I recommend www2.rohde-schwarz.com/file/1ER02_1e.pdf


HanTekway when sampling with 1GSs and 4k buffer is able to capture 2500wfrm/s (4us capture time).
These 2500wfrm/s are back calculated relative to 10DIV, which is in real live 2100 when 16DIV visible and
1900wfrm/s when 20DIV visible.

While capturing with 40k buffer the sample rate is 400MSs or 500MSs (which depends on hw model version),
when you take the latest model this give us 80us capture time.
The relation between active acquisition time and fixed(blind) time is not fixed to screen refresh rate (which is the case for Owon),
it is like on every other DSO an dynamic value. As long we do not change the setup (e.g. by adding an extra FFT measure or what
so ever affecting postprocessing - even timebase!) while sampling with different sample depth the ratio change proportional to active
acquisiton time. This mean when sampling witk 40k buffer and 500MSs you can assume that the 80us need to be divided by
known value 4us (from which we know its equal to 2500wfrm/s). This give us factor 20 ... now divide the 2500 by 20 and you
will get 125 wfrm/s per 10DIV (real live for full 20DIV screen 95 wfrm/s) which is the value for 20ns/DIV and 40k depth single channel.

When you do the same with 1Mpoint, this is 2ms capture time and 500 times slower than when with 4k, give us 5 wfrm/s
per 10DIV, real live worst case value is then 3.8wfrm/s. This seems to be what visually observed by others on these DSOs.

So what is wrong with Owon? Well, they decided to made the fixed (blind)time fixed to somewhat 25ms - no matter what memory
depth or sampling rate. When the depth is set to 1k the buffer will be of course postprocessed very fast (smallest capture time is 1us)
- but then the loop takes still 25ms - that's 39.99 wfrm/s. Now when you take 10Mpoint, that's 10ms capture time giving total loop
time of 35ms - that's 28.57 wfrm/s.

We might wonder where this 25ms is coming from. My guess: this is one thread system, therefore you need to have some kind
of fixed post-processing turn around on which then display buffer, sample buffer, i/o and all other systems are (relative) based.
The 25ms seems to be good value, that's 40 display buffers per second, or 40 x 10Mpoint (so 400M/s for buffer memory)
and for sure fast enought to capture i/o (this is why you don't see any front panel lag when in 10M in compare to 1k on Owon).
This calculation might be of course wrong, only Owon can answer it in detail, however for my hardware understaing it makes
sense why it works like that.

Based on that I agree with marmad, Owon need to change the "base" in the firmware to speed-up to wfrm/s rate. The sample buffer
when smaller than 10M need to be postprocessed multiple times per one cycle, but then we would have a DPO-like and not
DSO-like device. Sure there are other non-DPO-like devices on the market having higher wfrm/s rate than Owon, this is not
because they better implemented but only because they do have variable cycle time for data processing and fixed for display
which means the data will be shown asynch to acquisition (this could be a good implementation for Owon too).
However in such systems you need to have two "clocks", which is not the case in one thready Owon firmware.
As Owon never ever managed to develop such implementation (SDS is the latest hardware and it does have the same gaps
as older Owon models - they also based on Samsung SoC and one thread firmware) i doubt they will be able to fix it now.

However what they can do is to implement all other missing things, then only the wfrm/s would be worse than on competitor models,
and that would be already very good thing for all of us using Owon hardware.

You are a hobbyist/semi-professional: you want large screen OR hackability - AND PC ctrl/comm is NOT important - get the Hantek.

PC ctrl/comm is luckily not a dark hole anymore, we have all the information how to control these DSOs and how to
understand the data. There is already great custom PC software (for Mac/Linux/Win) allowing to remote control,
to export buffer to csv (few standards), to capture screenshots or ot capture data to PC screen, etc.
What missing if only FFT (on the to do) and access to raw data buffer (not on to do as still some company confidential information
unknown - but i'm working on that).

Check screenshots on
http://www.dreisiebner.at/dso-usb-tool/
http://www.dreisiebner.at/dso-usb-tool/screenshots.htm

Nevertheless LabView support is still missing, however it can be easily build based on the PC protocol
informations posted in here
http://elinux.org/Das_Oszi_Protocol

and here

http://www.mikrocontroller.net/articles/Hantek_Protokoll

[EU1]

Yep. Modern VGA cards working at 32bits @ 1680x1050 as i'm typing away ... i'm not too sure if they run at GHz levels OR not , at least i know the pixel clock on mine is at 550MHz but i agree it's GHz levels because of the 2.2 Gigapixels/sec pixel fill rate ( It's a Mobility HD 5430 anyway )

1680*1050*100Hz = 176.4Mhz with only 8-bit per channel.

[EU1]

Then it is also the sampling depth , or whatever else we call this : "bits per sample" - that means if the screen has 8 vertical divisions, how many bits do we allocate per division? I'd hope we would allocate at least 3 bits per division (0 to 7, ie seven distinct levels within each verticule) making this 3 * 8 = 24 bits per sample.

You need 3+log2(8 ) = 6 bits per sample.

Modern VGA cards work at 32 bits per pixel and at GHz levels so I do not see a problem here. But something tells me that all these digital scopes do not have 24 bit sampling depth - am I right?

First of all, VGA cards have DAC's, but not ADC, thus it is not a legal comparison.
And, of course, there are 3 8-bit DAC's in a VGA card (one DAC per a color channel). 3 x 8-bit DAC is not the same as 1 x 24-bit DAC 

I REALLY wish there was competition , i mean like okay 1Gigsamples ... What's the speed of a standard ADC converter in a off the shelf notebook you find now?
They have 24 bits 5gigsamples ADC's in there.

24-bit @ 5GSPS ADC - it is just impossible. Moreover, notebooks have no fast ADC at all.

[marmad]

PC ctrl/comm is luckily not a dark hole anymore, we have all the information how to control these DSOs and how to understand the data.

Tinhead, I did think about this when I wrote the previous comment - and the HanTekway has certainly passed the Owon in this regard now.  It's just not as good/easy as the Rigol with it's standardized SCPI command set. But I went back and modified the post a bit to better reflect this distinction.

One thing I wanted to mention here in regards to the future of DSOs: USB 3.0. Many people (Dave included) have commented on the problems and limitations with USB scopes - and many of those were due to the bandwidth speeds of USB 2.0. But if you've used any USB 3.0 devices (like an external drive) you can see that it's a brave new world when it comes to communication speeds between the PC and external devices. Of course, USB scopes will never have the portablity of standalone DSOs - but, if you have a lab set up with work computers driving equipment like I do - that's not such a big issue. Try downloading the software for a PicoScope and running it with 4 different waveform windows running on a computer with a 1900x1200 screen - there ain't nothing like it in the standalone world for clarity, precision, and wealth of simultaneous information.

I believe a new collection of USB 3.0 scopes are on the horizon - and might very well make USB scopes, for many people, the bang-for-buck match of - if not the better deal than - standalone DSOs. Hopefully it will also help drive further down standalone DSO prices.

[alm]

I don't think USB 2.0 is what limits the current USB scopes. USB 3.0 is not fast enough for real-time sampling, so you still need the sampling memory. It's not like the refresh rate is what's limiting the current USB scopes. The problem is that nobody designed an affordable scopes with an analog frond-end that does not suck and ADC and memory depth competitive with the often similarly priced bench scopes. Things have gotten slightly better since Dave did the comparison, but it's still very hard, if not impossible, to find a USB scope competitive with the Rigol DS1102E (100 MHz BW, 1 GS/s sampling rate, 1 Mpoints memory, 2 channels, though not all specs at the same time) at a similar/lower price. The Picoscope 3200 series comes close, but costs something like $1000 for 100 MHz BW.

[marmad]

I don't think USB 2.0 is what limits the current USB scopes. USB 3.0 is not fast enough for real-time sampling, so you still need the sampling memory.

You'll always need buffer memory - but memory is cheap, so what? And it depends what you think 'fast enough' is. USB 3.0 could support real-time sampling rates of 300 MS/s (8-bit) uncompressed - and with aggregation you could likely push it over 1 GS/s. Of course, it depends on clever design and good drivers and software.

The Picoscope 3200 series comes close, but costs something like $1000 for 100 MHz BW.

Yes, but to be fair, you should be comparing it to something like the Agilent DSOX2002A - and not the Rigol. Don't just look at it's sampling rate - look at it's collection of features, triggers, software, etc. It's miles above the Rigol in my opinion.

Anyway, there was a wave of USB scopes (and logic analyzers, etc) that came out after the dust settled on the 2.0 spec. - I believe the same thing will happen again with 3.0.

[EU1]

You'll always need buffer memory - but memory is cheap, so what? And it depends what you think 'fast enough' is. USB 3.0 could support real-time sampling rates of 300 MS/s (8-bit) uncompressed - and with aggregation you could likely push it over 1 GS/s. Of course, it depends on clever design and good drivers and software.

We need to separate DRAM and SRAM. In SRAM each bit is a trigger which consists of several transistors, and in DRAM a bit  is a single p-n junction. Therefore SRAM costs much more than DRAM.
But SRAM is easier to use for real-time sampling at high sample rate, therefore it is widely used in oscilloscopes.
Owon's SDS series is rather an exception - for sample rates up to 1GSPS they use cost-effective DDR2 DRAM which allows them to have 10MB of memory per channel.

[akis]

Has anyone so far got an opinion of which of these scopes comes with the best probes (I presume I mean smallest capacitance) / or with differential probes out of the box / or with the best ability to use the pseudo differential mode with two channels ?

I ask because I have been working with a "low" frequency device recently, 200 KHz, and am finding all sorts of things about my ageing scope and my probes that I had not thought of before.

[voidptr]

One bad pixel?
Not meaning to sound rude, but I am curious why you are even wasting you time trying to chase that down?
.....and you have two good probes now.
What more do you expect?
A new replacement scope?  If so, just come out and tell the vendor so, and pack it all back up for a refund.  Why drag it out?

ZAP  Why beat around the bush about it with post-after-post, documenting with very detailed pictures of the perpetrator-dot.   Get a new scope and be done with it.

dear pullin-gs,

i am so sorry you had to read my message 

i really don't know why you took the time to respond to it tho if it wasn't good enough for you ! 
oh i see you might be "the best message content estimator of this site"  ... 

for others people 
i repost cleaner pictures because i received 4 private messages asking about them.
it seems that at least some people care about that kind of silly problems...
i routinely buy lcd 1920x1200 without ANY bad pixel,  and i don't see why they sent me 2 bads probes on 4, maybe quality management fail somewhere.

so end of story,
i gave neutral review with comments to my ebay seller  (who is a member here i think)
i didn't get any response back to my 2nd message from Owon,
for now scope is ok for my need and i will buy something better when i will have more money.

 

[T4P]

One bad pixel?
Not meaning to sound rude, but I am curious why you are even wasting you time trying to chase that down?
.....and you have two good probes now.
What more do you expect?
A new replacement scope?  If so, just come out and tell the vendor so, and pack it all back up for a refund.  Why drag it out?

ZAP  Why beat around the bush about it with post-after-post, documenting with very detailed pictures of the perpetrator-dot.   Get a new scope and be done with it.

dear pullin-gs,

i am so sorry you had to read my message 

i really don't know why you took the time to respond to it tho if it wasn't good enough for you ! 
oh i see you might be "the best message content estimator of this site"  ... 

for others people 
i repost cleaner pictures because i received 4 private messages asking about them.
it seems that at least some people care about that kind of silly problems...
i routinely buy lcd 1920x1200 without ANY bad pixel,  and i don't see why they sent me 2 bads probes on 4, maybe quality management fail somewhere.

so end of story,
i gave neutral review with comments to my ebay seller  (who is a member here i think)
i didn't get any response back to my 2nd message from Owon,
for now scope is ok for my need and i will buy something better when i will have more money.

 

I will certainly do the same and get this as my starter scope then get a DSOX3000 down the road

[alm]

You'll always need buffer memory - but memory is cheap, so what? And it depends what you think 'fast enough' is. USB 3.0 could support real-time sampling rates of 300 MS/s (8-bit) uncompressed - and with aggregation you could likely push it over 1 GS/s. Of course, it depends on clever design and good drivers and software.

It may be usable on lower sampling rate, just like the long memory option on Rigol scopes, but at that point you're saving the relatively cheap DRAM, not the expensive RAM that can run at full speed. Not sure how much BOM costs this saves, especially if you factor in the premium for USB 3.0 transceivers.

Yes, but to be fair, you should be comparing it to something like the Agilent DSOX2002A - and not the Rigol. Don't just look at it's sampling rate - look at it's collection of features, triggers, software, etc. It's miles above the Rigol in my opinion.

Sure, it has more features, though I'm not sure the Agilent is a fair comparison, since its banner spec is the fast update rate. My point was that this is the cheapest scope made by Picoscope (and I can't think of too many cheaper competitors) that exceeds the entry level scope which has been dominated by the Rigol DS1052E/1102E for the past five or so years. And you pay $600 extra for the lack of screen and controls.

Has anyone so far got an opinion of which of these scopes comes with the best probes (I presume I mean smallest capacitance) / or with differential probes out of the box / or with the best ability to use the pseudo differential mode with two channels ?

I ask because I have been working with a "low" frequency device recently, 200 KHz, and am finding all sorts of things about my ageing scope and my probes that I had not thought of before.

If you're talking about Owon/Hantek/Rigol, then I believe they're more or less the same: they all suck in this regard. None of them have invested significantly in probe development as far as I know. Rigol sells an active probe, but I don't think it works with the DS1000E series. There are some generic differential probes by companies like Testec, which will work with any scope. The big vendors: Agilent, Lecroy and Tektronix each have a huge number of probes, including low capacitance, current and active probes, but the active probes usually use proprietary power connectors, so you either have to use their scopes or provide a DIY solution (it's usually just some DC voltages). They also make the best passive probes, although you would probably be hard pressed to notice this at 200 kHz (assuming we're talking 200 kHz bandwidth). Don't be surprised to pay more than for an entry level scope for an active probe. Some of the Tek/HP/Agilent branded probes are occasionally available on eBay for slightly less outrages prices.

Poor-man's differential probes don't work well on any scopes (scopes with proper differential inputs like the analog Tek 7000 series with 7A13 amplifier excluded). The old Tektronix TDS-200 series had a spec of 100:1 CMRR at 60 Hz, down to 20:1 at 50 MHz. Modern scopes often don't even spec CMRR anymore, so don't expect it to be much better.

[marmad]

It may be usable on lower sampling rate, just like the long memory option on Rigol scopes, but at that point you're saving the relatively cheap DRAM, not the expensive RAM that can run at full speed. Not sure how much BOM costs this saves, especially if you factor in the premium for USB 3.0 transceivers.

Well, it seems pointless to argue this. You don't seem to feel that USB 3.0 scopes will make any difference in the market - I disagree. We'll just have to wait and see.

Sure, it has more features, though I'm not sure the Agilent is a fair comparison, since its banner spec is the fast update rate. My point was that this is the cheapest scope made by Picoscope (and I can't think of too many cheaper competitors) that exceeds the entry level scope which has been dominated by the Rigol DS1052E/1102E for the past five or so years. And you pay $600 extra for the lack of screen and controls.

I'm not saying that USB scopes couldn't be priced lower (I think Velleman has made inroads in this direction), but I think you're a bit out of touch with the market. The Picoscope 2207 - which has the 100MHz bw and 1GS/s rate - plus a built-in AWG and function generator - lists for $740 on the Picoscope website.

Also, I don't personally feel the lack of a tiny, old-fashioned 320x240 screen and some cheap controls matters if you get a more powerful tool with greater capabilities. It seems that at least some of the disdain for USB devices on this blog comes from 'traditionalist' engineers (and the like) who seem to believe that any piece of test equipment that isn't a standalone device is not 'serious' equipment. This whole attitude strikes me as rather hilarious. I no longer own any standalone media devices (television, dvr, stereo, etc) - everything in my home being driven from HTPCs. I expect my lab to keep moving in that direction as well.