Browse Prior Art Database

Reading protocol for point to multipoint testing

IP.com Disclosure Number: IPCOM000130312D
Original Publication Date: 2005-Nov-25
Included in the Prior Art Database: 2005-Nov-25
Document File: 3 page(s) / 259K

Publishing Venue

Siemens

Related People

Juergen Carstens: CONTACT

Abstract

The considered scenario is a point to multipoint microcontrollers testing system where a PC Test application tests a large number (hundreds) of devices almost in the same time. In this particular case testing operation is performed downloading in parallel to all devices under test a micro-code with the test flow that should be executed in each single device. After a prefixed time the PC Test application reads in sequence the test results saved inside each device. This last operation wastes a lot of time and could be strongly shorted with the mechanism proposed. Up to now, reading operation is composed by a "Request" from PC Test to device and an "Answer" from device to PC Test. Such operation is repeated step by step for each device under test. Typically, the considered testing system structure contains complex communication architecture. Most of the time of a reading operation is lost during "round trip time" (RTT), from sending "Request" to receiving "Answer". Then, in this scenario, reading results as "killer operations"are mainly due to RTT. Since test time is proportional to production costs, this procedure should be optimized.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 3

S

Reading protocol for point to multipoint testing

Idea: Angelo De Poli, IT-Padua; Alessandro Scanferla, IT-Padua

The considered scenario is a point to multipoint microcontrollers testing system where a PC Test application tests a large number (hundreds) of devices almost in the same time. In this particular case testing operation is performed downloading in parallel to all devices under test a micro-code with the test flow that should be executed in each single device. After a prefixed time the PC Test application reads in sequence the test results saved inside each device. This last operation wastes a lot of time and could be strongly shorted with the mechanism proposed.

Up to now, reading operation is composed by a "Request" from PC Test to device and an "Answer" from device to PC Test. Such operation is repeated step by step for each device under test. Typically, the considered testing system structure contains complex communication architecture. Most of the time of a reading operation is lost during "round trip time" (RTT), from sending "Request" to receiving "Answer". Then, in this scenario, reading results as "killer operations"are mainly due to RTT. Since test time is proportional to production costs, this procedure should be optimized.

The main idea is to apply a CDMA (Code Division Multiple Access) concept to a particular point to multipoint testing machine to reduce drastically the reading operation time, avoiding wasting time in many single request transmissions. A simple visual description of the mechanism is shown in figure 1. Reducing the reading operation time to small fraction just packing the set of answers of all devices in a single answer can do this. In other terms all the answers are sent in parallel in a single shot answering to a single request. In such a way wasting time due to RTT is limited just only to a single RTT. The mechanism that allows doing that could be thought similar to a CDMA. To each device is assigned a unique code that converts each bit in a sequence of bits. The length of this sequence, in terms of bits, is strictly close to the number of devices. All the "Answers" are combined together then the new sequence, of the same length of the previous ones, is sent in one shot to the PC Test. From this sequence it is possible to decode all the single "Answers" just using the unique codes already known.

Prior art can broadcasting write operations, considering as write operation an operation that does not need a feedback from device, but reading operations, as operations that must get a feedback, are performed device by device in sequence as described in figure 1. As shown in the figure the RTT time is the most costly time.

Just to have an idea of the weight...