Login Sign Up   Invite   Feedback  
Circuit Sage Forums | My Discussions | Start a Discussion
Circuit Sage Forum
avatar
Posted by Jesal Mehta on November 1, 2008, 7:47 pm
Does anyone know what are different ways to get rid of DC offset in Zero-IF receiver?

Replies to This Discussion

avatar
Reply by ryan on November 5, 2008, 6:21 pm
Best way is to have a 2 dacs at the output of the I and Q mixers. These dacs are calibrated during startup or during some frame sync time and periodically if temperature changes. Also a continuous Costas Loop can be used in certain modulation schemes.
avatar
Reply by Jesal Mehta on November 5, 2008, 9:02 pm
ThanksBut what if the system does not allow sync time in the burst. How will the offset change with gain/AGC as well as LO frequency in this system?
avatar
Reply by ryan on November 5, 2008, 11:33 pm
Hi Jesal,In that case u will have to put a error amplifer and dc offset cancellation loop where u measure the offset before the adc (after gain) and feed it to cancellation circuit at the output of the mixer. Just remember to make the bandwidth of the loop lower than the lowest frequency of interest.
avatar
Reply by Jesal Mehta on November 6, 2008, 12:42 am
Ryan,Have you done any calculation of settling time of dc cancellation loop vs change in the forward gain? I know that another worry is that if the dc offset at input/gain changes it will take some time for it to cancel. If it happens during the data it causes serve degradation.By the way do you like this site?
avatar
Reply by ryan on November 6, 2008, 3:37 pm
Jesal,The stepresponse of the loop will be commensurate with the order of the loop filter, most practically it will be first order. most de-mods also have a built in DC cancellation loop that has a bandwidth of a few HZ so what this outer loop has to do is remove the offset to a point that the ADC's dynamic range is not affected. So if the loop takes time to settle,, as long as it does not ring (which is why u want 1st order), it is going to be fine.
avatar
Reply by bernie katz on November 30, 2008, 10:44 pm
With a decent filter design program, a higher order filter can be designed with low damping value so it will not ring, such that the settling time for a given cutoff frequency will be faster. Mulitsim has a filter designer tool for this purpose. Your best response is probably Bessel. There are quite a few free filter design programs on the web. Bernie Katz. www.technicaldevelop.com <a href="http://www.technicaldevelop.com">Technical Develop</a>
avatar
Reply by John Wetherell on December 12, 2008, 9:36 am
I'm with Ryan on this. Power up DC with current injected into mixer output, then let the digital use a very low frequency highpass corner to remove the rest. If you have a higher order baseband filter with a programmble gain to optimize dynamic range as the signal changes, you may have problems with offset changing with gain changes. I prefer to have low order baseband filters and high SNDR ADCs with no baseband analog gain changes, with all of the channel selection filtering done digitally, so DC offsets don't change.

back to top