TDA1541 DEM reclocking

[grundigger]

I agree that in 4xOS mode you would need to run the DEM at 705.6kHz to get a "correct" current for each individual oversample period. But the oversampling adds no extra information about the original wave sampled at 44100 Hz. Although each oversample period uses only one DEM state with fDEM at 176.4kHz, each sequence of 4 uses all 4 DEM states, so the average current at the DAC output for each original 44100 Hz sample period is proportional to the CD sample value, and the wave is reconstructed correctly.

8 of the DEM current switches operate at fDEM, and 6 at fDEM/2. Caps at pins 7, 8, 10, 12, 19, 21, 23, 24 filter switch ripple of frequency fDEM. Caps at pins 9, 11, 13, 18, 20, 22 filter switch ripple of frequency fDEM/2. The DEM also produces a very small ripple at the output of frequency fDEM/4. The Philips document says that the DEM frequency must be greater than 4 times the highest audio frequency so that the fDEM/4 ripple does not fall within the audio pass band.

When the DEM is not clocked, and the capacitor is used, Philips recommend a maximum 300kHz for fDEM, because at faster speeds distortion may increase due to poorer duty cycle accuracy of the free-running multivibrator. I do not know the maximum when the DEM is clocked, but good results are reported at up to 2.8MHz in this thread.

[oshifis]

I just measured the signal at each DEM pins after removing the capacitors. The results are different than you advised. There is a 4-step signal at each pins. That well corresponds to the 4-stage shift register driven by the DEM oscillator. The step levels show a great variability (withing about 20mV, I could not measure exactly), and they are different at each pin. But the repetition frequency is always fDEM/4.