Started to Build the MP

[6666]

I might have a revision of a GB add-in soon

Davor any progress on your GB ?

[Davor]

All my projects are on hold right now. As for GB, there is almost identical scheme to what I already suggested, except that it goes with 1:3 or 1:4 timing instead of 1:2, in order to achieve less attenuation.

[6666]

Hi Davor, thanks, is the schematic something that you can post here ?, or its not ready yet ?

[Davor]

It is a mere extension of what I posted before. I'll keep the tuning mechanism because it gives relative tuning that remains valid for wide range of timings, and gives me opportunity to toggle timing and thus avoid holes. Moreover, if toggling happens fast enough, the very toggling sound may serve as means of discrimination.
The mechanism is not a secret. There are 3 things... first sample, second anti-sample, and a third EF sample. The first sample starts after a pause and its duration is roughly same as the pause, or 1:1. The second anti-sample is supplied to the integrator with equal gain as the first sample, so in order for it to work its duration must be equal to the time that lapsed from the flyback.
It goes like this... If time is quantised and pause is 1, first sample is 1, together they are 2. My initial scheme said that after these 2 comes anti-sample in duration of 2, but although it would do GB it results in somewhat higher attenuation. The EF in this scheme is antisample minus sample = 1 EF sample at some practical delay as obtainable by a single 4017. This scheme can be called 1:2
To alleviate attenuation one can use 1:3 or 1:4 schemes.
1:3 would work as follows: 1 delay, 1 sample, 1 delay (altogether 3 by now), 3 antisample, long delay, 2 EF
1:4 would work as follows: 1 delay, 1 sample, 2 delay (altogether 4 by now), 4 antisample, long delay, 3 EF
Both 1:3 and 1:4 schemes can be obtained by simply cascading an additional 4017 to the original scheme, and you only need to put a few more diodes.

I'd stop at 1:4 because delays become ridiculously long, and you get better off with integrators that support different gains in alternating channels, but my only point is that this scheme can be applied to MPP or surf or any other simple PI rig out there that uses a differential integrator.

[6666]

Thanks Davor.

[Davor]

There is a quirk related to all PI GB approaches - a polarity reversal of long tau targets. Also there is a hole at a tau in between. But now I wish to announce that I'm tweaking an absolute value circuit that will fix this for MPP and provide true static response. It is a simple add-on that replaces a transistor prior to audio NE555.
More elaborate devices have the absolute value circuit designed with op amps, but this one is much simpler than those.
I caught cold, so don't hold your breath. I'll fix it at my slow pace.

[6666]

All good, Get over your cold Davor.

[Davor]

I feel awful, but here it goes...
It could perhaps undergo a tweak or two. Say, at this point the depth of a notch is tweaked by R6. What you get is shown in a result window ... current that is supplied to a timing capacitor of NE555 is shown in green, and fun part is that it is actually current value for both voltage extremes and it is the same
Trace in red is a logarithm of that same current and it spreads over ~5 decades.
It is nearly symmetric and should work fine for the intended use.
With reference to Mpp Rev B, V3 stands for output of U9, and except for the V- connection (ground in my schematic ... sorry) replaces R35, R36, and Q6.