sverk,
I have long thought a "PC Scope" a cool idea, and they have come a long way in the last decade or so....that being said, I have never used one. When faced with buying a new scope I keep ending up getting "free standing" scopes. There are 2ch import offerings starting @ ~ $400, and some really nice offerings from "name brand" companies starting @ not a lot more. Last year I ordered a Lecroy WaveJet 314A and I have to say, "WOW, what a difference." Of course @ 6 to 10 times the price of an "entry level, 2ch DSO", it should have some "whistles and bells", lol. @ the end of the day, I have just never been able to convince myself that the PC scopes would "satisfy" me, or that the difference in cost from a "stand alone unit" was enough to justify the disparity in specs. Following is a list of "features" that I feel are important in a DSO:
1) Memory
2) ADC Speed
One of the most powerful features of a DSO is the ability to capture and "freeze" an event in time for careful inspection. Assuming a screen width of 1024 pixels, one might mistakenly assume that 1kpts/ch is sufficient @ any given acquisition rate, but this is absolutely not true. Assuming even "basic" firmware/software features, a DSO can capture all of the events on a signal line for a "fixed" period of time before//after a trigger event and then allow you to "zoom" in. Let's take something simple like RS232-TTL level communications @ 115.2kbps. A 10 byte packet would have 100 bits of data spread over ~9mS. 100 bits of data is WAY too much data to make sense of on a screen that is only 1024 bits wide, but if you "zoom in" so that 10 bits are "visible" at any given time you can then "scroll" through the entire packet and verify the information quickly and easily. For this process to be reliable you need the ADC to sample @ a much higher rate (in this case a minimum of 10 times the rate one might expect to use for a 10 bytes @ 115.2khz signal), and to capture the signal @ this higher resolution you will need more memory (in this case at least 10 times the memory)....so, in this very simple example 10kb of memory would be the absolute minimum.
It might appear from my example that the DSO-2820 with it's 40k sample length might fit the bill nicely, but in my experience you would likely find the task of analyzing even this simple signal more than frustrating. I have an Instek GDS-1062A (in addition to my Lecroy) that would appear to have considerably better specs than the DSO-2820 you referenced (2MPts memory and a 1Gsa/s to start with) and I can promise you that analyzing a 10 byte packet of 115.2kbps RS232-TTL comm would be a real chore using it. (I paid ~$400 for the GDS1062A several years ago, that model has since been replaced, and currently the recommended replacement is the GDS1072-U at ~$500...NOT saying that is a good or bad deal, just a quick look).
All DSOs I am aware of in the 'accessible' price range "share" a single ADC and, it would appear, struggle to find a way to cram the data it spits out into memory and then onto the screen. Even most of the low-end DSOs have 1Gsa/s acquisition rates and I have a hard time wrapping my head around how they actually handle the data, but I can tell you one of my primary concerns involving PC-DSOs has always been the "lag time" between signal acquisition and "screen display". Even my Lecroy in "continuous" mode struggles to dump the data to the screen....by this I mean that if I am sampling something like a PWM signal in "automatic trigger mode" there is ~1-2 seconds of delay between samples and I wonder how long that delay might be if it had to be processed for and fed through the USB interface and then processed by a winblows application and only then processed for display and sent to the windows kernal to again process the signal to actually refresh the screen....? I don't know the answer, it is just a 'fear'.
As far as a PC scope being suitable for your needs....It may well suit your needs fine; I think a lot will depend on what those needs are and how important the up-front cost is. If $309 is all your wallet can stand then it really doesn't matter what $400 to $40,000 will buy....If the up-front cost is not the most important consideration, then I would take some time and look closely at what you plan to use the scope for and determine what the best choices are. If you have more time than money, you might consider playing with one of the Arduino projects...I have not looked at any of them in quite a while, but I feel 100% confident there are several open-source logic analyzer // DSO sketches out there that would let you get a feel for "what the specs mean" for very little $$$. I know you aren't going to be able to do much on the limited AVR platform, but for << $20 you could play with a low bandwidth PC scope and explore some of the limitations and difficulties w/o spending any real money.
At the end of the day, no matter how cool an idea I think PC scopes are, I have always ended up getting stand-alone units after carefully researching the actual capabilities and cost-wise comparing. My guess is that this is still true (but as stated, I have not looked in a while.)
Good luck!
Fish