PaiA motion filter

[antonis]

I'd generously reduce all Megohms resistors values and 500k pots ones..

[DUCKFACE]

I'd generously reduce all Megohms resistors values and 500k pots ones..

even the input buffer ?
i really like the idea of frequency control. it will be great on bass

[antonis]

even the input buffer ?

Why not..??
A pair of 10k resistors and a 470k bias resistor should do the job..

P.S.
I'd be more focused on transistor bias configuration..

[Mark Hammer]

The Motion Filter uses a half-wave rectifier to provide an envelope voltage.  It has come up many times over the years that, when a circuit uses a half-wave rectifier, someone says "I don't know why, but it sounds like it is distorting".

Half-wave rectifiers take a waveform that goes up and down, and chops half off so that the result only goes up, back to the midpoint, then up again.  A full-wave rectifier doubles the number of upswings, as in the picture shown.  Every time the envelope voltage goes up and back down again, in response to very brief changes in the envelope voltage (what is called "ripple") it moves the filter just a bit.  That very fast millisecond movement of the filter CAN sound like distortion; especially during the decay phase, when the string is producing all sorts of artifacts.

There are three cures for this.  One is to replace a half-wave rectifier with a full-wave one, but that will require a more complicated circuit than maybe you are prepared for. 

A second strategy is to make the filter sweep so fast that you don't get to hear the ripple for very long.  Try a 100k resistor in parallel with the 1uf cap to ground.

A third strategy is to smooth out the envelope by using a larger-value cap than 1uf, such as 4u7 or even 10uf.  IN some circuits, that cap is referred to as an "averaging capacitor" since it responds more slowly and "averages" out the envelope over a longer period of time, like blurring a charcoal drawing with your thumb.  The attack AND decay time will be slower but the envelope ripple will be reduced.  If you do this, the decay time can be shortened by placing a "bleed" resistor in parallel with the cap to quickly drain it off.  With a 10uf cap, using a 47k fixed resistor in series with a 1M pot will be able to get a range of decay times fro very short to fairly long.

[Fancy Lime]

I have something with the same type of filter on the breadboard. Some important points:

The ratio between the two 1M resistors at IC1c sets the low frequency gain (to 1 in this case). The absolute value of the resistor in the negative feedback loop (between pins 8 and 10) sets the cutoff frequency and Q factor of the filter, together with you the two 10n caps and the transistor. 1M gives a lot of Q, leading to clipping of the filter. 470k worked well with 10n caps for me. I would replace both these 1M with 470k or even 330k.

Another point is that the Q gets higher, the lower the resistance of the transistor get and the higher the filter frequency gets. This will also lead to clipping on loud signals. Putting an additional 47p cap between pins 8 and 10 balances the Q a bit better and also kills a lot of noise.

The 47k resistor at the base of the transistor needs to be adjusted to the beta of the transistor to get a good response curve. The higher the gain of the transistor, the higher the resistor value should be. Try putting a 100k or 220k or 470k here to see what sounds best.

There is a better way of setting the start frequency. Scrap the 500k pot and the 1M resistor attached to it and instead add a 100k resistor to ground right after D1 (bleed resistor to empty the cap, 100k should be ok when using a 1u cap here, I used something else, see last point below). Then add a 100k log pot wired as a variable resistor to ground from point "A".

Instead of normal Si diodes, use Schottkys. The lower forward voltage drop gives you more sensitivity and a better feel of the envelope. 1N5817 have among the lowest forward voltage drop and work nicely on my breadboard at the moment.

Change the 3M9 at IC1b to something less ridiculous. 1M or 470k are probably much more useful.

Last but not least: what Mark said! Envelope ripple can sound like distortion. The strategy I used to get rid of that was strategy Nr. 4: After D1, there are two 1k resistors in series and after each resistor there is a 4u7 cap to ground. This gives a nice slowish attack time and kills all the rippling. I then used a 10k fixed resistor to ground after that for quick decay.

HTH,
Andy

[Mark Hammer]

#4 is a good strategy. One doesn't have to use a resistance value as high as 1k, but the idea is to provide what amounts to 2-pole lowpass filtering of the envelope.  Perhaps not as helpful as full-wave rectification, but a big step in the right direction.

[m4268588]

Do not turn up the "sensitivity" knob too high. If you turn it up too high, you will also lose bypass sound.

The drawing of this section is questionable.

[PRR]

> The drawing of this section is questionable.

The note bottom right explains why the trim-pot.

[DUCKFACE]

after all iv read above here is the schematic.



im confused about the frequency part ...

[PRR]

....im confused about the frequency part ...

It's a Bridged Tee Filter.


The transistor is R1. There's a third cap which may not do a lot.

[Fancy Lime]

Nice! Some things to consider, though:

1) The sensitivity range is too high, probably. Right now the minimum gain of the side chain is 47x, the maximum is effectively open loop gain of the opamp. You likely want much less to have a useful sensitivity knob. I would change R26 to 100k and the sensitivity pot also to 100k logarithmic, with a 1k resistor in series.

2) The Volume pot R30 has a range of 1x to 3x, so it can only boost. That is likely not ideal. Change R29 to 2.2k for a more useful range ot 0.5x to 2.5x.

3) C18 probably has no discernible effect and can be left out.

4) 100k for R28 gives a very long decay. This may or may not be what you want. I prefer a very quick decay and used 10k in that position. 22k or 47k sound ok to me too but do not encourage quick playing as much. Using a 10k with a 100k pot in series here may be considered worth the extra pot, at least by some people. Me, I want fast decay only, but that's me.

5) R34 probably doesn't need to be a pot in the final design. Once you found the approximately right value for the transistor that you're using, you could swap in the nearest resistor value at hand. Or make it a trim pot but probably not a "user settable pot".

6) The Freq pot: in these bridged-T filters, the transistor provides a variable resistance between point "A" and ground, which is adjusted by the envelope detection circuit (aka the side chain). If the resistance between"A" and ground is high, the filter frequency is low. If the resistance is low, the frequency is high. Now: the Freq pot is supposed to provide an adjustable resistance parallel to the transistor, thus setting the maximum resistance between "A" and ground, when the transistor is turned all the way off. Thereby, you can set the lower frequency from which the filter starts when tracking the envelope. To do that, the pot needs to be wired as a variable resistor (ports 1 and 2 of the pot are direct connected together) and point "A" needs to be connected to the junction between the pot and the transistor. Also, for a useful range, I would suggest a 100k logarithmic pot, possibly with a 2.2k or 4.7k resistor in series.

7) the filter frequency is going to be very low. Using 10n for C20 and C21 will make it much more useful.

@PRR: the 47p C19 does nothing until the filter goes to its highest frequencies. There it starts shaving the top off right when the Q of the filter would normally get extremely high. The result is a much more even Q over most of the frequency range. However, for that to really work, there is indeed something missing, namely a 100 Ohm or so resistor in series with either C20 or C21 which I neglected to mention earlier. Alternatively, C19 can be left out and the new resistor be made 1k for much the same effect of throttling the Q.

HTH,
Andy

[antonis]

A cap on R4/R24 junction should do no harm..
(even if it isn't used for AC ground..)

[DUCKFACE]

Nice! Some things to consider, though:
......

1) The sensitivity range if its 100k starts to react at 80% of the pot.

2) The Volume pot is set to 10k in series with 2.2k resistor. Its really a pleasure to control the volume

3) C18 is out of the game

4) Decay correction with 100k pot is just fine. For powerchords sweeps out the sound on long decay

5) R34 "user settable pot".

6) The Freq pot 100k logarithmic pot, possibly with 4.7k resistor in series cuts off the low end. E string on bass is untouched until 1-st octave. This is something to think about. This setup has wide range on mids. You can correct the high freq at the end of the pot. Very fast transfer from mids to high at the end of the pot.

7) the filter frequency with 10n for C20 and C21 cut of the lows. probably ill try it with 47n or combinations 47n and 10n.

There are still little distordions when you reach the desired freq. Still didnt use the mod with 100 Ohm resistor in series with either C20 or C21.

how actualy i have to place 1k resistor? between c20 and c21 ?

Thanks Andy

[DUCKFACE]

probably a stupid question but ... how can i use 500k pots instead of 100k? iv tryed the pot cheater https://www.diystompboxes.com/analogalchemy/emh/emh.html but the value of 500k is too big for the calculator.

[antonis]

Code   1 2 3 4
----    -------
0   - - - -
1   - - - X
2   - - X -
3   - - X X
4   - X - -
5   - X - X
6   - X X -
7   - X X X
8   X - - -
9   X - - X
10   X - X -
11   X - X X
12   X X - -
13   X X - X
14   X X X -
15   X X X X

Trimpot Value =   500000
                R1 =   270000
                R2 =   120000
                R3 =   68000
                R4 =   33000