Req: Paper on Modulation Compension Circuit

Hi all,

Have you heard about Modulation Compension Circuit (MCC)? In traditional wireless application because of use PLL (phase locked loop), when there is a long string of zero or one in NRZ application, it likes a very low frequency and this kind of low frequency cannot be modulated in PLL because PLL is a high-pass in nature with signal below loop bandwidth cannot be modified. So two kind of method, one is open-loop modulation and another two-point modulation.

But some chips use close-loop with employ MCC technique, any idea?

Thanks in advance!

Data scrambling

Scrambling of data is used to keep the long strings of 1 and 0 from happening. Another trick is to use a mixing of two oscillators, one of them fixed frequency and capable of being FMed. It is at a low enough frequency that it will not drift much.

Re: Data scrambling

flatulent said:

Scrambling of data is used to keep the long strings of 1 and 0 from happening. Another trick is to use a mixing of two oscillators, one of them fixed frequency and capable of being FMed. It is at a low enough frequency that it will not drift much.

Click to expand...

Yes, data scrambling can solve the problem, but in some case for example the data control and some ID cannot be scrambled due to some protocol nature. And your second scheme is same as the two-point modulation.

The reason I asked for MCC because some ATMEL transceiver as adopted some scheme to achieve the close-loop modulation for fast data rates.

I need some reference materials.

BR

Another idea

Sorry about reinventing the wheel. Here is another idea. You low pass filter the data stream in a side path at the same cutoff frequency as the high pass filter effect frequency of the PLL. You then subtract half the difference in voltage from the 1 to 0 level and add this DC value to the modulating voltage to the PLL. This way if there is a string of ones, the lpf output will rise above the half voltage level and bias the PLL towards the one level and maintain the ones frequency. If there is a string of zeros, the net output willl be a negative bias that will shift the PLL towards the zero frequency. These bias shifts will ramp at the same rate as the PLL corrections and cancel them.

The data stream should have a half voltage level when no data is being sent.

There is an old technique called DC-restoration that compensates (at the receiver side) for the loss of low-frequency contents in NRZ when the signal passes by a channel with a zero at 0 frequency. It is mentionned in the classical book by Lucky, Salz and Weldon (Principles of Data Communications).
The idea is a very simple type of fixed decision-feedback equalization. Supose that the channel has a transfer function H(s)=s/(s+1). After detection, pass the NRZ stream by a low-pass filter Hr(s)=1/(s+1) and add its output to the received baseband signal (before entering detector): this regenerated low-pass signal compensates the lost low-frequency components. The equivalent transfer function is H(s)+Hr(s)=1 .
It works very well if the the BER is not too high.
Regards

Z

out of band

The trouble that is being compensated for is that with a long string of ones or zeros the PLL goes towards center frequency and the when the next data bit that is the opposite type comes along the transmitter deviates out of the assigned chanel and possibly the receiver IF pass band.