memomemomemo wrote on Feb 19th, 2010, 3:30pm:What is the minimum SNR level that the BERT can really work
because it needs a minimum ( BER < 10-2 ? ) to acquire lock?
It depends on bit rate.
I assume your concern is typical "Baseband Communication System Model".
About categories of "Communication System Model", see the following.
http://edocs.soco.agilent.com/display/ads2009/About+SinksUnder ideal and optimized sampling conditions, required Es/No(=Eb/No in this case) is about 5dB for BER < 1e-2.
See
http://edocs.soco.agilent.com/display/ads2009/Bit-Error-Rate+Measurement+Tutoria...About "Relationship between SNR, Es/No and Eb/No", again see the following. Here bit rate and System ENBW(Equivalent Noise Bandwidth) are required to evaluate SNR.
http://edocs.soco.agilent.com/display/ads2009/About+Sinksmemomemomemo wrote on Feb 19th, 2010, 3:30pm:What is the circuit mechanism that makes the lock in very low SNR? ( such as minus 30dB)
Generally commercial BERT try to test partial pattern matching during 32bits, 64bits, 128bits, etc. with some judgement for sync or unsync,
so BERT often give message of "Sync_Loss=Unsync" for low Eb/No data.
I don't use commercial BERT with auto synchronization for BER measurement of very low Eb/No data.
Generally such BERT is not useful for burst bit errors or relative long successive bit errors.
The following is my cheap "heppoko" BERT.
http://translate.google.co.jp/translate?u=http%3A%2F%2Fwww.kikusui.co.jp%2Fcatal...I use logic analyzer having long memory such as 64Mwords for evaluation of BER by full pattern matching.
http://cp.literature.agilent.com/litweb/pdf/5989-0129EN.pdfhttp://cp.literature.agilent.com/litweb/pdf/5989-0024EN.pdf