There are two main ways to accomplish the sampling in scopes of this
caliber. First, it is important to point out that DSOs have two
different specs. Sample rate and bandwidth. They are related, but
separate. High bandwidth is usually accomplished by sampling a waveform
multiple times at slight offsets. For example, if I have a 1GS/s ADC,
Nyquist says 500MHz is my theoretical limit and as was pointed out, you
really need way more than 2 samples per period to get a recognizable
waveform. Let's say, for the sake of math that we want 10 (This is what
Tek uses). That limits us to 100MHz. To get 1GHz bandwidth, you adjust
your sample timing. Instead of sampling every 10ns you sample every
11ns. This increases your effective resolution on a repeating waveform
but a factor of 10, giving you an effective 1GHz bandwidth. Note this
only works for repetitive wave forms. For non repeating wave forms, you
are limited to 100MHz. Though for single event triggers you have a
resolution of 1ns.
To increase actual bandwidth, you have to have multiple ADCs each
sampling a a slight time offset. To get the necessary ten samples per
period for a 1GHz waveform you need ten 1GS/s ADCs each sampling 0.1ns
apart.
This is all correct. I didn't realize that Tek wants people to sample that high. In any case I agree it takes a whole stack of lower speed ADCs to get enough samples. The scope OEMs get by this by having custom ICs that do much of the work. Building it from discretes is going to be a pain without a foundry. Also the currently available ADCs are only have 8 - 6 bits of resolution. In addition MAX and Rockwell have export and useage restrictions on those parts as they can be used in radar systems. Also much of the back magic is in the design of the sampling heads. There is just so much twitchy technology with it.
Max makes a 600Msps 16 bit ADC. So 17 of those at 35$ each about 600$. 400$ for DSP, board, sampling head, etc. Should be doable for 1000$ in quantity. Kind of sucks to not have some of it integrated into an ASIC though. I think if I were to try to attempt it I would do something like this: http://www.lecroy.com/tm/Library/WhitePapers/PDF/DBI_Explained.pdf
The basic method there is to use a diplexer and down converter to break a high frequency signal into two lower frequency components. Then do the needed sampling on them. They also have a FPGA that does processing and lookups to compensate for the systems distortion. It seems like a better way to go about things.
> A 1 Ghz scope might be doable. But I can see it being several hundred
> in parts alone.
If you could put together a 1GHz DSO for $1k in parts you could probably
sell more than you could ever make. A low end Tek DSO (TDS2014) is
going to set you back around $2kUS and that buys you a 4 channel 100Mhz
- 1GS/s DSO. A 4 channel 1Ghz - 5GS/s Tek DPO4104 has a list price of
$14kUS.
Patrick M
_______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
