Full or half wave rectifier in VLF

[pepe]

Hi

In most schemes known to me VLF detectors used half-wave synchronous rectifiers in input stage. On diagram Whites V3i I see full-wave rectifier. I'm sure that this solution increase signal two times. But I wonder if full-wave synchronous rectifiers can cancel some distortions ?

B.R
pepe

[golfnut]

Ive been thinking about this too recently, a lot.

If you consider x to Y channel phase then i think advantage in full wave.


If your simply looking at a positive recieved signal going over a threshold, then Im of the opnion that a single, unipolar mixer is best - FET type.

The TGSL mixer only conducts on half cycles - where the 4066 type conducts pos and neg signals thru.

A FET mixer does not output a signal for the Fe During Disc use - this helps reduce Iron masking.

A 4066 conducts + and - halves so an Iron signal comes thru as a -ve and the wanted + sig from a non Fe target - these opposite polarities cancel to zero - as Iron Masking.

Its only my opinion and I could be wrong

[Davor]

There is more to it, and I'll make some simulations on this soon. It is clear that conversion loss is greater in half wave detector, but given the fact that you are detecting a signal that is already pre-amplified, you are not losing anything. It is the harmonic behaviour (aliases), and offset that makes more difference.

[pepe]

I made a simplified simulation in LabVIEW. I generated 10 kHz sine wave with an amplitude of 1 and added a 50 Hz sine wave with an amplitude of 0.1. Then made the synchronous rectifier single / full and 100Hz low-pass filter. The output from the filter graph:

red - half-wave rectifier,
black - full-wave rectifier,

I'm not 100% sure for the accuracy of the simulation but the chart suggests itself.

[Tepco]

Taking all risk to be incorrect, or at least misinterpreted on this, my answer can be: Only really important difference between half and full wave demodulator is, half wave is like SSB (single side band) and full-wave is like DSB (double side band suppression) in normal communication technique. Only, here IF is centered around zero and followed by very narrow band high gain motion filter stages. As a result, half wave SSB suppressing type will leak part of low frequency component (usually ground proximity generated but may come from other sources too) within filter bandwidth, directly to next stage , full-wave will reject it. Nothing to do with ground balance, just demodulator output DC or LF level will change more or less. Full wave is better, will not leak, other considerations, like noise, interference, distortion etc are less noticeable and important.


Just try to simulate RX plus some component within few Hz effect to output after filter, you will see difference. Use full wave, not some dramatic change, but better.

[pepe]

The same simulation but with 2 Hz noised modulation.

[Overtheedge]

Consider the output voltage and ripple characteristics of half wave rectification versus full wave rectification of AC off the mains.
pepe's simulation pretty much shows 2X for full wave and substantially reduced ripple.
That should sum it up nicely.

eric

[ODM]

The case between half wave or full wave is not as difficult as it sounds. It's important to notice that we're actually talking about demodulation here. A full wave AM demodulator inverts the other half of the signal. For an undistorted signal there is not much difference.

If we "gate" the signal full time, instead of just one half of it, both halves of a distorted waveform are equally applied to the output signal, which helps in suppressing even order harmonics. Offset errors are also suppressed, which shows as a great improvement over DC and relatively rather low frequency rejection.

The "level" or losses of detection is not really that important when considering topology. Thus it's not as important to sample a peak value, if we have fast enough steered integrators, for example. Integrating the full waveform instead of its peaks pays off in additional noise suppression, which makes it attractive over quadrature sampling when fast enough integrators are available.

[Davor]

I can fully agree with most of the conclusions here, but still there is more to it than it seem.
I'll start with one often overlooked problem, and that is PWM due to the offset of the phase comparators. What happens is that we are phase shifting a Tx signal and supplying such signal to a comparator. Because comparators are not ideal, our Tx signal is not pure, and our topologies tend to "leak" some offset, the resulting square wave is not ideal, but slightly PWM modulated. When such signal is supplied to the synchronous detectors, the resulting signal is also shifted, and I have some results.

For offset of -20dBc:
- half wave detector: -31db(offset) or -51dBc
- full wave detector: -43dB(offset) or -63dBc (same for double balanced)

For offset of -40dBc:
- half wave detector: -32db(offset) or -72dBc
- full wave detector: -44dB(offset) or -84dBc (same for double balanced)

Both measurements are obtained in a simulated GB channel to minimise carrier problem and mimic the "air signal" injection. Regardless of the offset level, the resulting mixer output offset is consistently reduced by a constant factor. Full wave detector is some 12dB better.

Considering 80dB ratio between "air signal" and target signals it is clear that full wave is a better choice. It is also clear that -40dBc of offset at the phase comparators is enough to trigger falses, and is most probably responsible for most of the chatters, especially with half wave detectors.

It is important to notice that this injection happens as interplay between "air signal" and offset variation that results in PWM of the phase comparators output. Offset can also be induced by the Tx harmonic content and their phase variation, as it also results in PWM of the phasing signal. To my surprise, my simulation showed that 3rd harmonic is even more energetic in this sense than the 2nd, but it also depends upon the topology. In any case, as harmonics amplitude is far more dependent upon the coil loading, it is clear that they'll produce falses, and make discrimination more fuzzy.

In both cases full wave performs better, and surely will provide sharper discrimination.

I'll investigate harmonic performance off half wave, full wave (also double balanced), and QSD detectors. I'll keep the -80dB(2mV air signal) as a benchmark. It already seem quite interesting.

[Davor]

I played a bit with QSD against harmonics, and various timings other than typical T/4. It appears that QSD, however tempting from the layout simplicity point of view - is not optimal for harmonic behaviour. It is obvious that T/3 is a perfect choice for suppression of harmonics, as it attenuates 2nd, 3rd, and 4th, of which the 3rd is attenuated by at least 30dB. Fair enough. With all the filtering and at most -40dBc of 3rd harmonic in Tx signal, it passes a -80dB(2mV air signal) benchmark with flying colours.

I think I've got a winner It is a full wave switching mixer with T/3 duty cycle.