EMP (Evaluating the Measurement Process) Studies for Bias Effect

[Raquel]

I need some help cause I am really a begginer and have some issues trying to understand what the limits on the Bias Effect Charts means on the Evaluating the Measurement Process. I am using as reference Wheeler and Lyday Literature...

I know that they are calculated taken into account the test /retest errors and the grand average of the measurements so what means to have a chart were the limits are very tight and all the data is out of them?

Does this mean the instrument I am using to test it is not adecquate?

Thanks in advance for any help regarding this.

 

[Bev D]

Re: EMP Studies for Bias Effect

... some issues trying to understand what the limits on the Bias Effect Charts means on the Evaluating the Measurement Process. I am using as reference Wheeler and Lyday Literature...

I know that they are calculated taken into account the test /retest errors and the grand average of the measurements so what means to have a chart were the limits are very tight and all the data is out of them?

The limits are calculated using the variation between the two test - retest measurements. The points that you have plotted are the averages for the different fixtures. A guage or system that has very little measurement error for test - retest will give very tight limits. If the different guages - or in your case fixtures - are biased from one another (they have different center points) and that difference is larger than the measurement error, the plotted averages for each fixture will be outside the limits. This means that the fixtures are different from each other.

Does this mean the instrument I am using to test it is not adecquate?

without understanding the characteristic and it's specification limit it is not possible to understand if the bias is large enough to matter; we can only know that it is statistically larger than the repeatability measurement error...what are the spec limits? are you using the instrument for SPC or acceptance testing? what is the instrument and what is the fixture?

 

[Raquel]

Thanks for your prompt answer Bev D...

REgarding your questions...

What are the spec limits? are you using the instrument for SPC or acceptance testing? what is the instrument and what is the fixture?

Well in this case the characteristic we are meausring is the Minimun Insertion Loss for a Passband Filter, the specification we have already in place for this part is 5.99 dB Maximun and depending on the measurement we got we pass or failed the PN for shipment so it is an acceptance testing.

The fixture is what we use to test the part in the machine. It is a PCB that permit us to simulate the electrical behavior of the filter. We measured the part using a Network Analyzer, however since the PCBs has not an ideal performance we need to compare the differences between PCBs and connectors to check if some connector or PCB has damaged and we used the EMP to check if the Fixtures are statistically different.

Thank you

 

[Miner]

Bev D gave you a good explanation that I will expound upon. The test tells you that there is definitely a statistical difference between the fixtures. You must then decide whether this difference is of practical importance to you.

Factors that you should weigh when making this decision include:

• Importance of the characteristic measured,
• Cost of shipping bad product,
• Cost of rejecting good product,
• Problems arising when manufacturing discovers that product rejected by one fixture will pass when tested on another fixture, such as loss of credibility of test, using the fixture that will give the desired results, etc.

 

[Raquel]

Thanks for your Answer Miner. This is what exactly is giving us a headache cause we have used this tool for several and almost none of the fixtures looks similar and as you mentioned some parts passed in one but fail in another so I have been using it just for reference but used to do some other test and try to used the fixtures that gave me values near to the gold standard values.

However the question for my mate is about my criteria to decide what fixture I must use? I mean for her they must use the ones that gave better yield, but I used to use the ones that gave me poor yield which ensures that the customer will receive actually passing parts not marginal ones...this is because they already lost the reference for a specific PN.

I prefer a Happy Customer than a Happy Manager

What do you think?

 

[Hershal]

A side question as a bandpass filter is in discussion, what frequency range is the filter for? What connectors are used? What cables are used and how long are they? What power level is generated by the source?

Those all may affect insertion loss to a different degree.

This ia a bit early to mention, but depending on the frequency the rho and gamma of both source and the pc board you are using may be a contributing factor in the decision also.

 

[Raquel]

Thanks Hershal...

But, in order to try to avoid part of the problem we perform a calibration that is suppose compensates cables lenghts, power level source and some other loss that are out of the fixture... we have an "ideal" system.

So ideally when we measured a part what we are seen is the connectors and traces of the PCB summed to the loss of the DUT by itself.

In this case we have a RLL 201 MHz used for GSM and EDGE, but this same problem is seen at higher frequencies...2 GHz...