Colin,
I often work with short single pulses, so I need to know what the scope is doing if I am to trust what I see on the screen. If I want to look at a repeating waveforms I have several analogue scopes I can use that are much faster to drive. The thing that DSOs do, that beats analogue scopes, is single shot storage of a waveform. There's not a lot of point talking about DSOs for repeating waveforms, that has already been solved. So for this discussion, I always refer to single shot operation.
All my comments above are valid for single shot, the buffer need to be filled no matter
if this is repeating or single event.
The “logic” behind the dots on screen is also valid for average and peak sampling modes,
the resulting count of dots might be different but that’s different story.
On my LeCroy there is an info screen, one button, anytime and you see all the channel settings, timebase settings, sample rate and memory acquired at a single glance. It is very clear to me what it is doing, and I know what pulses I can sensibly measure and what is likely to be aliased. By comparison, the Hantek/Tekway design seems to be hiding that information, so I wonder why. Also, they hide the change from Real Time operation to Equivalent Time operation, that is deeply annoying. LeCroy automatically switch to Equivalent Time (they call it RIS, or Repetitive Interleaved Sampling) but it is clearly visible at all times. The operator really <needs> to know when this is happening, otherwise the data on the screen might be totally undersampled and quite untrustworthy.
Right, but that’s the major diff between brand names and china manufacturers. Actually not everything will be displayed and if there is information you can’t really trust.
Why ? Well very simple, all china DSOs (except Rigol 6000) are sampling always with fixed and highest sample rate of the ADCs, the resulting information is just “converted” to kind of sample rate information on the display (if available). So actually you can’t undersample, the firmware is correcting automatically based on the horizontal settings. Of course they could do it different (meaning the sample rate setting will reflect changes to ADC clock), but then they would need better clock sources, an FPGA PLL is not good enough, some external clock logic would be necessary, probably with good clock jitter attenuators which cost money – so that’s why.
"actually it is very hard to say what is the real sample rate on these scopes." and "Probably we should ask the original developer"
I agree completely with these comments, but are we never likely to find who it was and ask them...
Actually in begin of 2007 a developer named
“rgj” started development of a DSO 1.0 platform.
Later, in April 2008 Tekway company was founded and started production of the first DSO models DST3000/DST4000/DST1000 – based on DSO 1.0 platform. End of 2008 this developer worked again on DSO development – named DSO2.0 platform – mid of 2009 Tekway started development and end of 2009 sales of this new (DSO 2.0 platform based) series - DST1000B/DST3000B/DST4000B, which was bought by Hantek shareholders June 2010. Since them developer teams from both companies merged together and started with firmware optimization – of course for both Tekway and Hantek models. (Btw, since them firmware issues started, probably the joined HanTekway team should ask “rgj” how to optimize the firmware …)
No idea if “rgj” was/is working for Tekway or only external resource (which I think),
the fact is this developer is not located in China.
Maybe he/she is reading this thread and will answer us some questions.
At same time Instek and Rigol started of development and later production of very similar DSO.
Coincident? no, I don’t think so.
You said:
"The DSO is using Altera Cyclone II, if you check the specs you will see the amount of internal SRAM is "big" enough for
only one channel 4k sampling"
OK, thanks for that information, I'm not experienced with FPGA so their specs are confusing to me. That would explain the default back to 4k with Autoset. It does change the way the scope must work, there are many compromises compared to my old LeCroy. Even though my 9310 is only 100MS/s, I can trust that what I have on the screen, since I know always what sample rate it being used.
The Cyclone III, EP3C5 have 414kbits memory, Tekway/Hantek is using 8 ADCs x 8 bit x 4000 point,
so 250kbits, no way to have both channels sampled at same time into internal memory. Rigol is using exact the same FPGA, and sampling 2 x 16k (or it was 2 x 8k? no matter), this can’t match even for single shot, so they already shifting the information to external SRAM during short-memory shot /which is slowing down the design). Instek is using EP3C16, having 515kbits, this is big enough for single 8k or 2 channels 4k. I didn’t found any information about the memory size in short mem setting, but it would be probably 4 or 8k.
I don't know what SoC/UI means.
The SoC is actually ARM9 Samsung application processor responsible for GUI. This baby is reading the sampled data and displaying on screen. It is doing additionally some math/FFT things and the communication to outside world.
In the LeCroy, the sample rate changes are completely logical and in keeping with the published limitations and specs. The Info screen always tells you precisely what is happening. In my LeCroy, the scope acquisition memory defaults to 10k, but you can select 50k. Using 50k increases the processing required to display the waveforms, so drops the update rate, but the choice is yours. As you increase the timebase setting, the scope it will automatically change to RIS (Repetitive Interleaved Sampling) or Equivalent Time automatically, but it clearly tells you this is the case, so you can't be easily fooled alias.
All china DSOs are doing is exact in same way, except the fact that ADCs are hard-clocked so everything in done in firmware itself.
You said:
"I said "finally" because in 8ns/4ns/2ns/div the diff is very small and hard to count exactly, but it seems to be
200ps, 100ps and 50ps."
Of course, this is what it would do for repeating waveforms. I was testing it for single shot only. Single shot waveform storage is the banner capability that separates DSOs from analogue scopes, so my focus is on single shot.
You said:
"This is of course with real time mode sampling selected. With equ mode selected and 20ns/div the diff between dots
is 100ps instead of 500ps like in real time mode."
Agreed, but I did all my tests on single shot acquisitions. I have analogue scopes and didn't buy a a digital scope to show easy repetitive waveforms.
Me too, there is no difference between single and repetitive run in the logic, except the fact that the memory will be purged and filled again and again and again.
This is just result of how they using sin/x interpolation and the way how the DSO is designed."
Clearly the maker must work within the limits of physics and with difficult cost restraints. So, I am not being critical of the makers. But, there is no reason for not letting the user know what the scope is doing. There is still something weird happening with the sample rate getting stuck at 400MS/s for so long. I cannot understand why this should happen. At higher time base settings, the scope appears to use 500MS/s and 1000MS/s sample rate, so it's not like the digitiser can't do it. It's not because of a 4k fast RAM limitation, since at 200nS/div, it samples 1.28k at 400MS/s. It just makes no sense to me. I can't understand why they break the 1-2-5 sequence of sample rate at lower timbase settings and go to 1-2-4. I'm looking to explain why it is like this. I can accept quirks of design, but I like to know <why> it is like that.
The 1-2-4 sequence is probably because of clock rates for FPGA/SRAM/ADC and amount of ADCs. (100MHz clocked, 200Mhz clocked, 8ADCs)
They could probably do 1-2-5 but it will cost some FPGA time slowing down waveforms refresh.
I don’t like it, but can live with that, one of the reason why I bought it was the 2500wfrm/s, the other choise would be at that time Rigol with 1-2-5 seq. but 800wfrm/s.
You said:
"This is just result of how they using sin/x interpolation and the way how the DSO is designed. "
What? Surely this must be a mistake. I saw nothing in the manual about the scope using sinx/x interpolation between actual data points. I detest sinx/x interpolation. I used to sell LeCroy DSOs many years ago, and I have seen many many comparisons of different digital scopes. I finished that work 8 years ago, so I have not seen the latest HP DSOs. Apparently, they have improved dramatically, but in my past experience, by far the worst alias problems my customers encountered were cases where sinx/x was operating. Linear interpolation is the safest method. As soon as you see the very characteristic alpine looking sharp peaks, you know you have undersampling and need to take care to ensure you aren't looking at alias. Since Hantek/Tekway under the Display menu allows selection of Dots or Vectors, I assumed Vectors meant linear interpolation. We must find some way to be sure about this, it really matters to me.
Well, that’s what you can find in the user manual – waveform interpolation : (sin x)/x
therefore you will see aliasing on all china DSOs.
Generally spoken all china DSO have a different understanding of sin x as for example Tektronix has,
however it is hard to guess what they exactly doing without source code.
You said:
"Therefore don't count the dots, they not one per sample and don't compare to other DSOs"
If the dots are not actual sample points, what else could they be? Don't forget, I have always used single shot mode and ensured Real Time mode was operating.
Physically these dots have to be interpolated dots and not actual sampled points.
Please don't think I am full of criticism, I'm just trying to understand how this thing works, and whether I can trust its single shot screens. I deal with photomultiplier and gas radiation detectors and fast one-ff pulses matter to me.
I can’t answer this question, these DSOs are comparable to Tektronix TDS2012, is such Tek is good for you then you fine, if not you will have to change it to better model, maybe Agilent DSOX will fit your need.
As general DSO philosophy, the vital thing is to understand the limitations of the technology so you are not fooled by alias. Aliasing happens all the time, we just don't often realise it. Trusting a DSO screen without questioning is a good way to fool yourself.
Another interesting factor is screen alias. The screen only has a limited number of dots across it, if you squeeze memory down to show the pixels, then data is not being shown and alias can occur. To my best knowledge, Hantek/Tekway at least make an effort to capture spikes that are present in memory and display something on the screen that grabs your attention and makes you zoom for more info. LeCroy have always done this. I believe GW Instek take the trouble also. Many little modern scopes simply ignore the spike in memory and if luck has it fall into the data that is decimated for screen display, then bad luck, you never see it. If you ask a Tek salesman about this, they get excited and say the data isn't lost, it's still in memory. But so what, why would you scroll through anything up to 1 meg of waveform data <every single time> to find something that you don't know is there!!
Actually Hantek/Tekway are very similar to Instek, and to be very honest good for spikes detection.
The only thing I’m missing is search function, sure with the double window (F7) you can switch both normal and zoomed view and actually you will see these spikes in normal view, you can set some markes but there is room for improvement.
I have no trouble with instrument and technology limitations. I have no trouble with compromise made for better cost. But if you can't find out what those imitations are, how can you ever know when to be alert for alias? Even a car manual tells you what speeds to use for each gear...
Well, no idea why these guys (now talking about all chinese DSO manufacturers) can’t just write everything down into the manual. Sure, even brand manufacturers are doing this, hiding some information’s, but you can always find such hidden information if you look on the competitor website …
What the do is to write down highlights, sure everything what good need to be known, everything else hidden.
I guess they doing this just because they worry nobody would buy their products. But actually we all know that you will have to pay a lot of money to get really good product, everything else is always compromise.
Please keep up the good work, I have very much appreciated your work in informing us all about these scopes. I'm enjoying tinkering with the HAntek/Tekway scope, the screen readability for my older eyes is very nice, I just don't trust it yet and may never fully, unless I get some answers. I have emailed this issue to Hantek, who have sent a polite "gone to engineering evaluation" reply, so I wait with interest. In the mean time, I thought it would be interesting to ask your thoughts on the matter. *smile*
Sure you welcome and i hope my answer will help a bit to understand some things about chinese DSOs.
I remember how hard was to get any answer from Tekway at the beginning of my Tekway-adventure, especially because I did found some
“strange things” and asked directly. Today the communication is better, especially after Hantek shareholder bought Tekway,
Hantek’s international experience is good for us.
Home
Help
Search
Login
Register
