Acceptance criteria for a wireless network?

Hi. I am having a wireless network developed. The implementation
provides for API's to send and receive data from many nodes talking to
a PC over a "master" node (i.e. we are responsible for the application
protocol and the contractor is responsible for the PHY/MAC/NWK
layers). The deliverables are hardware "nodes" and a software
component that runs on the PC and gives the application access to the
nodes and other network information/functions.

My question is, what should be specified and included in the
Acceptance Criteria for such a system? What performance and other
indicators? (other than "it should work....")

Here is a list of what I could think of, not sure how great it is...
- Range test (i.e. performance at prescribed range)
- Throughput (i.e. minimal bitrate?)
- "Loading" (how many nodes, something like that)
- Uptime?
- Error rate?
- Stability?
- Hmmmm......

I would appreciate some pointers in the right direction... :)

Thanks

On Jun 25, 1:48 am, ElderUberGeek <aribl...@gmail.com> wrote:

> - Range test (i.e. performance at prescribed range)
> - Throughput (i.e. minimal bitrate?)
> - Error rate?

I would specify this as a conducted sensitivity test yielding a
specific BER at a specific signal strength, e.g. BER < 1E6 @ -103dBm.
Often there is a table for different signal strengths vs BER.

Specify antenna pattern as spherical within 3dB.

And if you like, specify minimum open-field range. But it can be very
difficult to measure, particularly if it's a long range (power > a few
tens of mW) and/or other frequent talkers share the frequency.

In article <1182750539.366882.164270@j4g2000prf.googlegroups.com>, 
aribloch@gmail.com says...
> 
> Hi. I am having a wireless network developed. The implementation
> provides for API's to send and receive data from many nodes talking to
> a PC over a "master" node (i.e. we are responsible for the application
> protocol and the contractor is responsible for the PHY/MAC/NWK
> layers). The deliverables are hardware "nodes" and a software
> component that runs on the PC and gives the application access to the
> nodes and other network information/functions.
> 
> My question is, what should be specified and included in the
> Acceptance Criteria for such a system? What performance and other
> indicators? (other than "it should work....")
> 
> Here is a list of what I could think of, not sure how great it is...
> - Range test (i.e. performance at prescribed range)
> - Throughput (i.e. minimal bitrate?)
> - "Loading" (how many nodes, something like that)
> - Uptime?
> - Error rate?
> - Stability?
> - Hmmmm......
> 
> I would appreciate some pointers in the right direction... :)
> 
> Thanks

Be sure that the HW node source code is a deliverable.
At a minimum, get the HW node source code placed in escrow.

You absolutely do not want to be at the mercy of the contractor for 
future updates to the nodes; you also want to be able to review it for 
errors, poor design, and back doors.

--Gene

ElderUberGeek wrote:

> Hi. I am having a wireless network developed. The implementation
> provides for API's to send and receive data from many nodes talking to
> a PC over a "master" node (i.e. we are responsible for the application
> protocol and the contractor is responsible for the PHY/MAC/NWK
> layers). The deliverables are hardware "nodes" and a software
> component that runs on the PC and gives the application access to the
> nodes and other network information/functions.
> 
> My question is, what should be specified and included in the
> Acceptance Criteria for such a system? What performance and other
> indicators? (other than "it should work....")

The performance criteria are relatively straightforward and easy to 
evaluate. The real pitfall for the wireless network could be the failure 
modes. You need to make sure that the network will not go crazy or lock 
up in any situation, including a random or malicious interference. 
Designing and testing the networking protocols for that is not a simple 
task.

Thus:

1. The failure modes state machines should be clearly defined.
2. The network should be tested for a long hours operation under a heavy 
load and in the presence of the random interferrence.


Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com