|
|
Product : AT3600Serial Number : VARIOUSFirmware : CURRENTSoftware : CURRENTUnit self test : PASSAccessories : N/A
Transformer / Inductor : Small toroid transformer.Fixture Details : Wilco test fixture, kelvin connectionsTests Used : Ls test at 25kHzQuestion : We have 13 AT3600's. Recently we ran an experiment where we took 10 parts and tested them on a Wayne Kerr three separate test sessions, compensating between each test session. We then took the same 10 parts to four different AT3600's and performed the same three test sessions of the 10 parts, compensating in between each session. We notice the repeatability of some Voltech's between each test session is much better than others. (Note we understand that there is an absolute accuracy offset from the Wayne Kerr due to frequency- it is the repeatability or precision we are concerned about ) We recalibrated the most variable unit (FN in attached chart) with no improvement. Can this be improved with Voltech service?
|
| |
|
|
|
 | |
|
|
Hi Chris,
Thanks for the comprehensive results. Calibration or other servicing will not improve the repeatability of results, but can I please clarify a few points with you so we can work out what may be happening?
- First of all, is the test frequency 250KHz and not 25KHz written above?
- What is the nominal inductance value? (Is the graph scale in uH?).
- What is the test voltage or current?
- What integration value do you have set in the test program? (Long, Medium or Short)
- Can you please send me a picture of the fixture - does it have leads that move or are they fixed?
- If you take one of the parts and test it 10 times without removing it from the fixture or compensating, what is the spread then?
With the information above we can determine whether the variability is due to the basic repeatability of the AT3600, what the specification is, and whether the fixturing could have anything to do with it. From there I will be able to tell you whether there's anything we can do to improve it.
I look forward to hearing from you soon.
Best regards,
Tim.
|
| |
|
|
| |
|
|
Tim
I thought replacing relays might have impact on repeatability. Anyway, the part is tested at 100mV, 250kHz, and the nominal Ls is about 62uH. We have moved to long integration, without seeing any improvement. I attached a photo of the fixture. Each set of 10 tests used this fixture, which we moved from Voltech to Voltech. If the fixture was an issue, we would expect repeatability of the median values to vary on all Voltech's. Regarding testing one part 10 times without removing, I'll try that, however I expect the readings to be very consistent. If you are familiar with boxplots (the box is 50% of the data, the line in the center is the median value, and the upper and lower winkers are each 25% of the data), you can see that total range or variation is about the same. The median shifting between each set of tests (compensation) is the issue, because we can apply an offset from the known true value when we can depend on approximately the same shift. Also note we understand the inherent error from the Wayne Kerr due to testing at high frequency.
|
| |
|
|
|
| |
|
|
Hi Chris,
Thanks for the information. If the AT measures consistently without compensating or moving the part as you suspect then that rules out any relays or other hardware issues with the AT3600. Self-test checks the relays as well, so assuming you have ran this and it passes, you can be sure they are OK. Are you performing both short circuit and open circuit compensation, and when doing so, are you fitting shorted and blank bobbins into the same fixture? If you open circuit compensate without any part fitted it may be possible for compensation to pass without the Kelvin contacts touching each other. This will effect the compensation readings. Other than this, I cannot imagine anything else happening with the fixture to cause the effect.
Best regards,
Tim.
|
| |
|
|
| |
|
|
Tim
Thank you very much for all this info. Embarrassingly, I've never thought about a "blank" open circuit compensation piece. I'll explore this further. I may tie in to the high frequency testing, as does the node contacts we explored in an earlier thread.
Regards
Chris
|
| |
|
|
| |
|
|
I have my fingers crossed that this resolves at least some of the issues. Let me know how you get on.
Regards,
Tim.
|
| |
|
|