| ??? 02/15/06 14:42 Read: times |
#110054 - You got it! Responding to: ???'s previous message |
Prahlad said:
These are MKP1837 type Vishay make 47nF precision metalised polypropelene capacitors with max manufacturing intial tolerance of 1% and aging drift <0.5% in 2 years. I know them, they are brilliant! I have used 330nF types in a three way active loudspeaker box with fourth order Linkwitz-Riley filters. You can improve your design, though, by using KP1836, if you need 47nF types. They show even better performance. Prahlad said:
But due to againg effect I will need to choose new master atleast twice a year. Am I right? This highly depends on flowing current and heat. So, if you have them in a box and store them in a cold and dry (!) room, then long term drift is even much lower. Take care, that no condensing water is in the box!!! High humidity can influence capacitance. Put a sicca bag into the box, to keep moisture out. Prahlad said:
And efforts are made to keep the process centered between UCL and LCL ie at 47nF thus the peak of bell distribution curve should ideally be at 47nF but processes rarely stay centered they tend to shift from center ie. peak of curve would deviate from 47nF in that case the statistical analysis from single lot may give slight error leading to error in value of reference cap. But again if samples from diffrent batches and diffrent sources are taken than this error will disappear. Am I right Kai? Totally right! If you take caps from same lot, you will notice, that errors between the individual caps is much smaller than +-1%. You will notice for instance, that all the caps don't differ by more than +-0.2%, but the average value of them being at let's say +0.4%. In a few hours same production line will show an error of average value of let's say -0.3%, but again with all the samples not differing by more than +-0.2%. That's the reason why you should not use components sitting on the same strap. (Is 'strap' the right word?) You can solve the problem also, if you buy these caps from same manufacturer but at different times. So, to be precisely, you do not need many caps, but many lots! Prahlad said:
Bur unfortunately I have only one supplier availbale so I will try to collect samples from 10 diffrent batches atleast 5 samples. Exactly! Prahlad said:
Great method and probably the simplest too. Remember, statistics has to do with infinite large numbers of samples. This nethode is only as good as statistically distributed your sampes are. So, try to get samples which fill the whole +-1% band. Kai |
