??? 04/17/06 10:52 Read: times |
#114371 - Be prepared to make a mistake Responding to: ???'s previous message |
Prahlad, this is obviously a commercial product and you as the engineer have to make the decisions. What would you do if there was no internet? You would have to either rely on the experience of someone senior or rely on your best judgement, and that may entail making a mistake. Some of our best learning comes from doing something wrong, realising that there is a problem and then solving it. I've put three products I've designed through the full CE testing and one failed due to the transzorb failing. I now know that I should use a MOV on the DC input.One of the products was mixed signal with a 24bit a/d converter. Initially, I used a split ground plane then moved to a single ground plane - the analog noise was slightly lower going to the single plane. However, every board is different. In doing all this, I've learnt a lot about the CE testing and what is required. I too had many questions over what I should do regarding the layout, but I did some research, tried to understand the concepts and took the plunge. I try not to rely on the internet doing the design for me.
As for your questions: Is the filtering adequate or too much? What are you trying to achieve? Best analog performance? CE (or other) emissions & susceptance?. You must have a fixed goal. Do you need to make changes? Perform some testing, that will guide you. The 7660 is a charge pump - I'd be keeping that away from low level analog stuff. Can it create trouble? If you suspect it can, then maybe you should look closer at it. For the -ve supply - what ripple on the supply can your application tolerate? What is the PSRR (power supply rejection ratio) of the op-amp? Then measure the ripple on the supply - that should answer that question. What can be done to keep noise problems to a minimum? Follow the basics...... I am targetting 10 bit accuracy - then you'd best be using a converter with better than 10bit resolution! Saying you want 10bit accuracy doesn't tell us much - at DC it may be quite simple, at 1GHz it might be bleeding edge. Qualify and quantify. Repeat after me, 'qualify and quantify' |