Netbook Performance?

Netbook Performance?, a forum discussion on Cleverscope. Join us for more discussions on Netbook Performance? on our Questions forum.

Back to Forum Index : Back to Questions   RSS

17 Dec 2010
Posts:

Without naming names, I've tried several USB scope software applications--sometimes just the demo version, sometimes with the real hardware--and have found many consume quite a bit of PC resources and/or get really slow (as in waveform refresh rate) on older or netbook class (i.e. Intel Atom) laptops.

The manual for Cleverscope just says ""Pentium PC with at least 256 MB of memory"". But I also found where you're offering an older version (3309) of the software for Pentium II computers implying the latest version needs at least a Pentium III or better?

Obviously Netbooks are small, light, cheap, don't take up much space, etc. so in many ways they're ideal for running virtual instruments on. But, at least some of the manufactures have applications that are not very netbook (or old slow laptop) friendly. So I'm curious:

1 - What are the real-world recommendations for decent performance with the latest Cleverscope software?

2 - How indicative is the performance and CPU loading of the demo app versus the full app talking to real hardware?

3 - Is there much difference in CPU load between USB and ethernet? Will the older less resource intensive software (3309) support ethernet?

4 - Some have implied the entire Cleverscope app is a LabView application. Most of the application seems to be in a single large exe file with a few small DLLs. Can LabView apps be compiled into single DLL's without any sort of runtime environment? Or is Cleverscope written in C++ or some other language and perhaps just uses the LabView API as its interface at the driver level?

Thanks!

18 Dec 2010
Posts: 396

Hello Solargeek
As we have added capability, our memory use has increased, but only if you use the capability. For example if you do a large (1M point) spectrum, or do a long protocol decode, or full buffer Maths, memory usage means you will need at least 512M in your machine to get good performance. But memory is so cheap these days, even netbooks support 1G.

In terms of Netbook performance, why don't you ask another poster - BrianM - a few posts back, who used a netbook for a field strength survey. We ourselves have not yet acquired a Netbook to check out performance, so any posts from nebook users are welcome!

We no longer offer the 3309 software (which is much lighter, as you say), because our latest firmware is not compatible with it. We made the decision at 6430 to change some major data structures to allow for some of things that people were wanting. So new applications can support old firmware, but old applications cannot support new firmware.

Anyway, essentially our software only loads resources when it needs to. So to keep things light, don't use memory hungry functions.

To answer your questions:
1. We think you need a pentium III with at least 512M of memory to get reasonable performance.
2. The loading on the demo app is actually a little higher than on the real app, because it has to generate signals in real time. The real app gets it's signals in bursty packets via USB or Ethernet.
3. We don't see much difference between USB and Ethernet. The farme rate (20 fps) is the same. If you want to reduce CPU loading, set the frame rate lower (you can choose 20, 10 or 5 fps). Most of the loading is in things like spectra, signal information, protocol decoding and Maths. Don't have those windows open, and CPU loading reduces.

As an example loadings on a pentium IV, 2.66 GHz with 1 G memory (now an older machine are), all windows visible:
a. Scope graph only, 6 divisions wide, 20 fps, 3ms wide graph: 12%, 179M
b. Scope graph only, 5 fps, 3ms graph: 3%, 179M
c. Scope + Tracking Graph, 20 fps: 30%, 181M
d. Scope + Tracking + Information, 20 fps: 40%, 181M
e. Scope + Tracking + Information + Spectra, 20 fps: 54%, 181M
f. Scope + Tracking + Information + Spectra + Maths (a+b) : 64%, 182M
g. Full buffer use: Scope graph only, get 2M Samples - Memory = 214M.
h. Full buffer use: Scope graph + Spectrum graph, 1M point spectra - Memory = 453 Mbyte
(If you return back to lower spectra resolution, memory drops back to 214M)

4. The entire app is written in Labview. The labview Runtime is loaded as part of the install. The app is in one exe file. There are a small number of support DLL's, all loaded by the installer. For Labview 6, the entire environment was put into a single exe. However from Labview 7 onwards this is no longer possible, and it's an application exe + labview Runtime. We are currently still using Labview 7.1, but shortly we will be moving to Labview 8.6.1 because it supports Windows, Linux and MAc OS all in the one source file. This will allow us to compile Linux, Windows, and MacOS versions without having to maintain multiple source versions. The sad collateral damage to this is that the runtime has grown to about 35M. However, hard disks have got bigger - even on netbooks!

Just for your interest, looking at http://www.cpubenchmark.net/cpu_list.php, I see that a typical netbook processor, the Intel Atom 230 at 1.6 Ghz has a Passmark of 315, While a Pentium iV at 2.66 Ghz is 335. So there is not much difference between the two.

I hope this helps.

18 Dec 2010
Posts:

Yes, that helps a lot. Thank you. I'm running the full version on a 2 year old netbook with a 1.2Ghz single core Atom and 1GB of RAM under XP. And, using the simulated generator, it's running 20% - 45% CPU loading. So if the real hardware is similar, that should work just fine. And that's substantially better than some competing products.

I didn't notice the LabView runtime installation. On my installation it's about 25MB worth in the National Instruments folder but I didn't install the VISA ethernet driver. Runtime or not, apparently your approach works better than most of the competition. So it seems to be worth it. National Instruments is one of those rare companies that really tries to do things right. They have a reputation for retaining really smart employees. So that's probably a good foundation to build on.

I have to say reading through the forums here, your PDF documents, videos, etc., I'm really impressed. None of the competition, overall, seems to offer what you do.

Now the only problem is your sole dealer in the USA only has the more expensive 8MB versions in stock and they don't seem to know when they'll be getting more of the regular 4MB models.

18 Dec 2010
Posts: 396

Thanks for the feedback. In fact we are shipping to them on our Monday (your Sunday), so if you want, you can tell us what you want, and we can make sure that we include a suitable unit with the shipment. It should be in the US on your Wednesday. After Monday things get a bit sticky as New Zealand closes down until 10 Jan for Christmas and our Summer Holidays.

You will need to install the NI-Visa driver to use a real system. Just let the installer that comes with the unit install what it wants. Ni-Visa is a universal conduit to USB, Ethernet (and GPIB and Serial) we use to abstract the communications. We do find the NI stuff reliable.

Thanks for considering Cleverscope.

18 Dec 2010
Posts:

What I want is a basic CS328AE. The 12 bit upgrade is somewhat tempting, but I'm not convinced it will offer any useful real-world benefit over averaging the 10 bit ADC. So a basic 10 bit ethernet unit with the logic analyzer digital inputs should be fine.

Saelig has some friendly folks, but they seem to often be out of stock on whatever people most want. You might want to consider adding a US distributor with deeper pockets that's not afraid to have more than one or two on the shelf (i.e. Mouser, DigiKey, etc.). DigiKey, for example, sells the TechTools products and at the moment have 20 DV3100's and 9 DV3400's in stock. That's about $24,000 worth of product for just 2 SKUs--a dollar amount that would probably send Saelig into cardiac arrest. They also have online inventory, same day shipping, and a vastly more professional and full featured website.

I know FTDI tried to use Saelig as their exclusive US distributor for several years and finally added some of the bigger guys and watched their sales go up by a factor of 10. So it's worth considering.

Some of us need a solution *right now*. And if Saelig is out of stock, many will likely buy a different product somewhere else even if it's inferior. I almost did that today. So availability is important.

19 Dec 2010
Posts: 396

Hi Solargeek,
We will put a CS320AE in with the shipment. Thanks for the comments on stocking.

10 bits is , as you say, sufficient for most real world uses. Where 12 or 14 bits come into play is where you need real ENOB's of that level. The LTC2280 we use for 10 bits has a DNL of +/-0.1 LSB. Thus you can average to about 3 bits better before revealing the DNL non-linearities. The LTC2282 (12 bit) has a DNL of +/-0.2 LSB which gives you about 2 bits of improvement with averaging to 14 bits. So you get 6dB better averaging performance, and 6 dB better spectral response - useful for some folks. (Just as a comparison some of our competitors, using 8 bit converters (such as the Natsemi ADC08D500) with DNL's of +/-0.15 bits, should only be able to claim 2 bits of improvement to 10 bits. However they talk about ENOBS of 12 and 16 bits after averaging. That's hard to understand).

20 Dec 2010
Posts:

Thanks. I'm looking for a CS328AE (not CS320AE) with the digital inputs. There are several things that set Cleverscope apart, but the mixed signal capability is a big one. There just are not many PC-based devices that do it, let alone do it with decent hardware and software.

And yes, some of the competition get ""creative"" in their ENOB calculations. One has to wonder if some of them even have the necessary equipment to properly measure their product's true performance? Linear Tech makes excellent well characterized ADCs but the implementation (pcb layout, power supplies, EMI, etc.) is also critical. I think some companies just slap an ADC on their board and base their specs on the raw ADC numbers. But, as you point out, sometimes even that doesn't justify their specs.

I have a 2 GSa/s Agilent 6000 series scope that I'm very happy with. I also have a handful of USB-based instruments that I'm less happy with. Perhaps one of these days I'll get around to quantifying their relative performance in various ways and publish my results on the web.

A person can go online and find countless reviews of say a particular digital camera that measure lens distortion, true resolution, noise performance, frames per second, battery life and lots of other parameters. But there's almost no independent objective data published for test equipment. And, to further confuse things, the big 3 (Agilent, Tek and Lecroy) have taken to private labeling Asian-designed products that, in some cases, look like their higher-end models but are nothing like them once you start using them.

Anyway, thanks for the added info on resolution. Any time I need more than 10 bits it's likely I can oversample and average. And I can live without the extra 6db but can see how it would be very useful for some.

20 Dec 2010
Posts: 6

Solargeek,
I have a small company that is a specialist in an obscure field. We started off applying simple ""chip level"" digital electronics to our products and then moved on to more complex systems. We have no true electronic engineers on staff, and certainly not in the field, so I am viewing my association with Cleverscope from the point of view of a practical electronics user of relatively simple systems that have to be installed, tested and work in tough environments.

My comments:
An electronics expert could certainly work through Digi-key etc and successfully purchase an ""in-the-box"" system and put it into use. Personally I prefer working through the folks at Saelig so that I can have someone to hold my hand. I thought they were great, they supplied my system at short notice including adding the 8 meg memory upgrade and changing a box they had in stock from 10 bit to 12 bit ADC. Sealig provides old-fashioned support for which I am perfectly willing to pay extra and wait longer if I have to do so.

Although our use of the Cleverscope from an electronic point of view is probably not too demanding its ease of use and reliability in the field is critical. The ease of use requires battery operation, ease of transportation and ease of set-up. Our first test involved using our CS328A with an Acer Aspire One netbook running XP-Pro. For our purposes, apart from screen size, the system operation was indistinguishable from using the Cscope with our desktop. Using individual batteries for the several components in the system was a pain so I will be looking at building our complete test kit, including storage for test leads etc, in a rolling-suitcase type idea with a substantial battery pack in the lower section with voltage regulators for the various power requirements.

Pluses:
Battery operation, light weight, adequate performance, no analog controls to get contaminated with dust, ability to simply record data to disc, ability to manipulate data with the computer to provide, for example, graphical displays. The size of the Cscope and netbook combined is much smaller than the volume of my Tek 4 channel scope and I do not need to carry and deploy a mains power cable.
Minuses:
Computer screen can be difficult to view in bright conditions (same with the Tek), potential dust contamination of computer, potential heat build up in components, Csope lacks a front-panel on/off switch which would be very useful.









c

20 Dec 2010
Posts:

Brian M,

Thanks for your added comments. I wasn't suggesting dropping Saelig, just perhaps adding an alternative that can offer things Saelig seemingly cannot--like deeper inventory levels, online inventory status, same day shipping even for orders placed late in the day, etc. Saelig is a great resource for inexpensive hobbiest products. The bigger distributors are more oriented to professionals. I think Cleverscope bridges both markets and could benefit by having both levels of distribution. DigiKey and Mouser, by the way, each have far more dedicated application engineers than Saelig. And it's usually easy to get one on the phone.

It would be nice if the Cleverscope was USB powered. There are some comparable devices that are. And it would also be nice if the USB interface was isolated as some of the competition has done. But ethernet solves the isolation issue and offers many other benefits. And I can live with needing a dedicated power supply especially with the wide input voltage range.

The Cleverscope isn't perfect but it seems to have the best combination of features and performance for the price given my needs. And it's encouraging it's being steadily improved and enhanced (primarily with software and firmware upgrades). There are similar products out there that haven't had even a software revision in 2+ years. And others have let their software get bloated, buggy and resource heavy while still lacking much of what Cleverscope offers and having an inferior user interface.

20 Dec 2010
Posts: 396

Hello Solargeek and BrianM,
Thanks for the suggestions. First we will include a CS328A (sorry about that). While we are putting it in, do you want a CS700A sig gen included? It's useful if you want to do Transfer Functions, or stimulate stuff. A new firmware upgrade coming out after Christmas will improve the stimulus to include low frequency ARB (1 MSPS) digital pattern generation (3 bits) and serial, spi and I2C output.

The other points - a power switch on the front panel - we have been asked about this occasionally, and it makes sense in terms of a battery powered unit. We have been thinking about doing a bit more in the power management area, which will reduce power when not being used (we already do a bit of this, but not as much as we could). We will contemplate if the next model should have a power switch.

USB powered - this is primarily a safety issue. We have had around 12 people connect the scope ground clip to the mains, and who did not have a ground fault interrupter in their mains system. There is a sacrificial earth track between the BNC earth terminal and the main system earth which vaporizes when this happens. Sometimes the USB module suffers damage as well, but the rest of the system is fine, and easy to repair (even by the user). However the main point is that the fault current flows through our system ground to the mains earth, and the user is safe.

If we made the scope USB powered, and the user attached the ground clip to the mains, the scope and PC would become live. In non-US countries this is 230V, and the user could die when they next handle the PC. We cannot live with this risk, so we don't do it.

We do offer an isolated option - Ethernet - as you say, which necessitates an external power supply. But with batteries, the scope is isolated, and the user can make measurements without ground loop issues.

We don't offer isolated USB, because it is very expensive and difficult to isolate the 480 Mbit/sec USB 2.0 data stream (max skew is only 200ps between the D+ and D- channels). In addition the interface operates single ended for some parts of the USB protocol, making transformer isolation very hard. All the isolated USB options we have seen out there offer isolated USB at the 12 Mbit/sec 1.1 rate. We think the 100 Mbit/sec Ethernet option is both faster and more flexible.

Talking about the rest of the front end circuitry you are completely right - we had to agonize hard over the tradeoffs in bandwidth, noise and distortion in choosing the front end topology. And it is a simple truth that any circuitry ahead of the ADC will increase noise and distortion. As you say, averaging can be used to improve noise, so distortion becomes one the most important parameters to control - especially if you have a spectrum analyser built in that displays any imperfections in all their glory. So just for your interest here is our front end chain, with distortion at 1 MHz and 2V p-p signal output:

TI OPA355 (200MHz BW, HD2: -81 dBc, HD3: -93 dBc) - AD8337 (BW - 280 Mhz, HD2: -72 dBc, HD3: -66 dBc), AD8138 (BW - 320Mhz, HD2: -94 dBc, HD3: -114 dBc), LTC2280 (BW - 575Mhz, HD2, HD3 better than -85 dBc). The VGA distortion (AD8337) is the biggest limitation. Happily it reduces with signal level, and reduced gain, and by having two of them in series, we can partition the gain to get at least a 10 dB improvement in distortion. You will find that these devices are also pretty much the lowest noise devices available at the time we designed the CS328A.

Anyway, thanks again for your comments and suggestions. Have a very Happy Christmas!

20 Dec 2010
Posts:

Thanks for the reply Bart.

The CS700A seems to be a nice cost-effective option for those needing it but I don't need one. I have a Tek AFG3000 series 14 bit ARB and a low distortion analog generator. I use a TotalPhase device for SPI/I2C but that's a great feature to add.

If you are doing true 480 Mbit USB 2.0 then I understand the isolation difficulty. I figured, like most of your competition, it was only 12 Mbit USB 2.0 which is fairly cheap to isolate.

I understand the isolation safety issue but that only goes so far. An ungrounded laptop won't vaporize that earth trace in the Cleverscope and would still present a safety risk. And a grounded PC will remain at ground potential even if you use USB power. I do agree the damage could be more extensive using non-isolated USB power. But that really shouldn't be your problem if users are that clueless. You just get to sell more scopes when they blow theirs up.

The USB power, in theory, could be isolated if the power consumption was around 2.1 watts or less of the 2.5 watt USB budget. That allows for the losses of an isolated DC-DC converter. That adds some cost but the bigger trick is probably keeping the power consumption low enough. FPGA's, high speed ADC's, high bandwidth analog front ends, etc. can be fairly power hungry. But at least one competitor does 100 Mhz and 250 MSa/s with USB power (or 4 channels at 80 MSa/s). If only they had digital inputs, isolation, and better software.

If I can dream, the best of all worlds would be a dual interface device that can interface via isolated USB including isolated USB power. Or it can use ethernet with a simple USB 5 volt power power supply connected to the USB jack. That would also let you use one of the many cheap Li-Ion USB battery packs (used typically to extend the run time of iPods, iPhones, etc.). That way you could use the a direct ethernet connection to your laptop but have the scope be externally battery powered with a plug-and-play solution to prolong the laptop's battery life (especially a netbook which have about half the battery watt hours).

And to really dream, there are modular WiFi solutions that could be configured via a USB or Ethernet connection and then scope could run wirelessly from the external battery pack. Internal WiFi could be a plug-in option.

But the reality is, most anything anyone is going to want to connect the scope to will generally be line powered or be able to located near line power. Off-grid solar installations are a notable (but rare I'm sure) exception. It's just more hassle to travel with yet another AC adapter and increase the odds of delays at airport security from having a bit too much suspious looking stuff show up on x-ray. But at least with the wide input range of the Cleverscope it might be able to share an adapter with something else.

As for performance, Tek/Lecroy/Agilent have the advantage of being able to justify using full custom ICs in their higher end products. But those products are typically at least 5 times the price of the Cleverscope. Having done some analog design, I'm in awe of the really high-end scopes with 20+ Ghz of analog bandwidth and 80 GSa/s. It's probably easier to make pigs fly. But they also have six figure price tags so they can justify millions of dollars worth of R&D and one-time costs even if they only sell 100 of them.

Again, the above are just comments and suggestions. I think you've done a great job from everything I can tell so far. I'm looking forward to having a CS328AE in front of me.
Back to Forum Index : Back to Questions   RSS
You must be logged in to post a reply



You need to Register or Log In before posting on these forums.

×

Your shopping cart is empty.