The AD6644 can work in 65MHz, today's experiment, I use 40MHz crystal. Since the AD6644 conversion clock needs to use a differential method, the original is that my design is a conversion chip using TTL to ECL. Since the chip does not buy the chip, you will decide yourself to wind high-frequency transformers.
Wrap a high-frequency transformer requires a few conditions: the first need to use high frequency porcelain ring, before I didn't find a low frequency porcelain ring because there was no high frequency porcelain ring, the high frequency ingredient in the result signal was The loss is lost, and it is eaten by the porcelain ring in the form of eddy currents. When the original coil is 7 turns, the sub-coil is 14 turns, the input signal is 0.1U capacitance and 100 ohm resistance, and the signal obtained in the secondary coil is about 800 mV, and the rise time of the signal is significantly longer. (Because my error is not recorded at the time of rise)
Today I went to Cui teacher to find 4 high frequency porcelain rings. Cui teacher is very good, but I am very happy to take care of the porcelain ring, I have forgotten a happy holiday.
Got the high frequency porcelain ring, I immediately use the 0.37 mm diameter paint package line to wind, the original coil 7, the sub-coil 14, the result of the secondary coil input voltage, I need 800mV, then I took the original coil Changed to 5 turns, the sub-coil changed to 11 turns, and the results did not change significantly, and the secondary coil voltage varied. When I continue to reduce the number of turns, the number of the number of the secondary coil is reduced, the input amplitude to the coil is improved, at around 4.2V, there is no way to follow the ADI evaluation board, in parallel two The judged diode, the result voltage is 900mV. As for why, I have to experiment, I think it may be related to the capacitance between the coils.
I also welded the Xilinx XCR3128's CPLD today. The program-controlled end of the AD8370 is no longer completely suspended, because the CPLD is not in the program, the IO port is input), so each board is small gain, input The signal buffers the AD8138 to the AD6644, I saw the data output of the AD6644, the quality of the quality is very good, because the signal is small, the high data line is 0, when I use the tweezers to hit the input of the AD8138 The high data line also appears. Using tweezers, the CNC LATCH side of the AD8370 is triggered, and the amplification gain is happened, so that the output signal of the AD8370 can be observed by an oscilloscope, and it is found that the high position of the AD6644 begins with signals, and the signal varies with the frequency amplitude of the input signal. Obviously, the AD6644 is working fine! Just because there is no heat radiator, this 1W more guy, there are 50 to 60 degrees.
Plates need to be improved: The input interface has a BNC to SMA / SMB. Increase the filter capacitor at the power supply, change the TTL / ECL circuit to a high frequency transformer, and the package package is increased.
In addition, when using high-frequency transformers, it should also be noted that there is no DC component in the input signal, which requires an electricity to be compromised, which is mainly to prevent magnetic saturation.