I can't resist jumping back in at this point. These full bridge devices are mostly used as motor controllers. In such applications you just need to turn it on and have it supply an appropriate AC signal while the motor is running and then turn it off. There's never any need for fine control or signal modulation. Also, the full bridge design, on its own, doesn't lead directly to any solution for the problem of superimposing the Q pulses on the loading current.
Of course you're free to go your own way, but I think the motor controller approach may be more difficult than just trying to adopt Godes' design directly. If you look at the first figure http://3.bp.blogspot.com/-7NVukY_dlR0/UISB4e_LSAI/AAAAAAAAAW4/Rl9BROYHIHQ/s1600/Q-Pulses-1.PNGfrom here http://pdxlenr.blogspot.com/2012/10/thoughts-about-godes-brillouin-patent.htmlyou'll see two traces, the green one at top and the blue one at the bottom. The green spikes are the Q-pulses and the blue pulse train is the input from the microcontroller. The input pulse train from the microcontroller has a 50% duty cycle, but the Q pulses are narrow. In other words, the Q pulse width is not a function of the width of the input pulses. Instead, each positive-going edge on the input causes a narrow positive-going Q pulse, and each negative-going edge on the input signal causes a narrow negative-going Q pulse. The characteristics of each Q pulse are set by the choice of inductor and capacitor (labeled L1 and C2 in my circuit) and the load (R1 in my circuit), and not directly by any control signal. Note that my C2 is equivalent to Godes' C5 in figure 3C of the patent application. I apologize for not paying more attention to these labeling issues. Also note that my circuit includes an "ideal" voltage source V1 at upper left. A real circuit needs a discharge capacitor to simulate an ideal voltage source. This is shown at extreme upper left in Godes' figure 3C. Confusingly, the discharge capacitor is labeled "C2" in Godes' designations. Now, the distinctions between my partial circuit and Godes' complete one. First, in Godes' circuit you see a transformer, T8 (part number "F626-12") in place of my inductor L1. That transformer is playing two roles. These are (a) its primary winding acts as an inductor, playing the role of my L1. And (b), the Q pulses couple across to the secondary winding; but in the secondary, which shares no ground reference with the primary, Godes is free to establish any ground reference (or DC loading current +V) he likes. As you can see from figures 3C and also 3B and 3A, Godes uses the center tap of the transformer as "ground" (or +V) for the loading current. Now, since the transfomer-coupled Q pulses are swinging "end to end" across the secondary winding and the center tap of the secondary is the reference point, the Q pulses are swinging positive and negative relative to the reference point of the loading current. In other words they are AC. The reason I used the term "ground (or +V)" and "reference point" above is that it doesn't matter for the superimposed Q-pulses. It does matter for the loading current; you have to pick the loading current polarity so that the center tap of the transformer leads to the electrochemical anode. The ensures that the core will be the cathode, so it will evolve the H2 to load. You can more clearly see in figure 3A, where the core is labeled "15". Figure 3A also shows how the two ends of the secondary of T8 (which is not labeled, but there's only one transformer in figure 3A) are across the core; thus, as the Q pulses swing positive and negative, the polarity reverses across the core, which is the true meaning of "AC" in this case. In summary, you could probably generate interesting pulse trains with a variety of techniques. But I think the clever use of T8 is essential. I'm not going to try and explain why I think this, it's partly gut feel. I just wouldn't imagine trying to solve this problem in other ways when the Godes' circuit shows a way of doing it that I believe will work. Also, to summarize the parameters that Godes can vary from the microcontroller, they are: (1) the amplitude of the Q pulses, labeled "55a" in figures 3A, 3B, and 3C; (2) presumably the width of the input pulses, which control the spacing between positive- and negative-going Q pulses; (3) the timing of the input pulses, which controls the timing of Q-pulse pairs. But not the "shape" of the pulses, which is determined by the inductance of the primary of T8, the value of C5, the load on the secondary of T8 (i.e. the impedance of the wet cell) and the coupling characteristics of T8. It is this last bit that explains why I think I would need decent test equipment to get this circuit working - the AC characteristics are going to be weird and will need to be discovered bit by bit. For example changes in the AC impedance of the wet cell caused by ongoing electrolysis could cause the whole secondary circuit to begin oscillating under the drive of the Q pulses through T8, that sort of thing. Jeff On Fri, Nov 23, 2012 at 6:42 AM, Jack Cole <jcol...@gmail.com> wrote: > Arnaud, > > Looks like 240V max and 4A max was used by Godes in phase 1. The RMS > current is 12 mA. > > More recently, looks like his circuit has capacity up to 35A (doesn't > specify the voltage) and a minimum pulse width of 250 ns. > > I'd be happy just replicating the phase I for now. > > Looks like those transistors could get in the range of ~313 ns pulse width > from looking at the data sheet @ 300V and 9.5A. > > If I understand you correctly, are you saying you could use a second PWM > signal to turn all the gates off to get the dead time? > > Thanks, > Jack > > On Fri, Nov 23, 2012 at 4:12 AM, Arnaud Kodeck <arnaud.kod...@lakoco.be>wrote: > >> e of the power part of the circuit?**** >> >> ** ** >> >> Arnaud**** >> ------------------------------ >> >> *From:* Jack Cole [mailto:jcol...@gmail.com] >> >> **** >> >> Arnaud (or anyone who can answer),**** >> >> ** ** >> >> So if I understand correctly, you could use a PWM pulse with an H bridge >> to get AC from a PWM signal? I think I looked into this before, and the >> problem would be that you wouldn't have the "dead space" in the current. >> Let's say you have a 100 ns + current and when this is switched off, the H >> bridge allows the - current for the remainder of the duty cycle. This gets >> you closer, but is still not what is needed. If I understand correctly, >> you need a bipolar pulse (then no current in between the pulses).**** >> >> ** ** >> > >