4th order delay as an alternative to equal-staged phasers!

[stm]

Hi!

In a previous thread I posted an allpass delay network suitable for TZF.  There was some interest there for a variable delay network, so I came up with the design presented below, where four identical resistors (i.e. 4 photocells) can be used to control the delay of a 4th order allpass network.

This presents a novel alternative to the typical 4-stage phasers that repeat four times a 1st order stage, whether with equal or staggered capacitor values.  Whether this new approach sounds good in its own right, better or worse than existing phasers is yet to be tried!  

This is the schematic:



And these are the performance curves:



The circuit's curves are in red.  The blue curves represent the delay of an equal capacitor four stage phaser, i.e. like a small stone.

As I said, this is an allpass, so amplitude is flat and fixed at 0 dB.  Notice also that the variable resistors are equal, but they are not referred to ground as in the other designs, so only the photocell approach would work.

The direct path and its summation with the delayed path is not included in this circuit.  Neither is the LFO.  This is just a building block.

Let me know what you think.  :wink:

[ExpAnonColin]

That's awesome.  I'd like to hear it.

Perhaps you could try using some H11F3's or ever just some FETs to make sure that the resistances will be matched.

-Colin

[Vsat]

Hi stm,
Built a 2nd order phaser (three  2nd order sections producing 3 notches) using matched LDR's (Mark H. has seen it in action). Called it the Box-O-Vibe since the original enclosure was a cardboard box with knobs. I got three matching pairs of LDRs from a bag of fifty unmatched units... they need to to match both under static and dynamic conditions. Sounds good... notch spacing is closer than with first order... sweep generator shows it behaves quite well.
Regards, Mike

[Peter Snowberg]

Another awesome filter from STM!

You're on a roll!

[stm]

Thanks, Peter.   I'm deeply involved with this filter stuff!

And there's still more to come...

[stm]

I simulated the amplitude curves in phaser mode for three different phasers, obtaining the following:

a) Blue: 4th order allpass phaser (stm)
b) Red: Four Equal stages (small clone)
c) Green: Four different stages (univibe)



It can be seen that the use of a true 4th order allpass produces the closest notches.
Also, if variable resistors (photocells) are not identical, some amplitude modulation will appear, which I consider a plus (!?). This amplitude variations will change dinamically according to the tracking of the photocells, modestly resembling what happens in a leslie cabinet!
I verified the circuit is quite tolerant to frequency resistor variations, where up to 20% difference between photocells is still acceptable. One natural option is to use two dual photocells. Also, I remark I don't see a way to use FETs on a circuit like this (for good or bad).

Finally, the following image shows a more practical implementation for a phaser. It requires only 4 opamps (a quad) for the audio section. The LFO and LED drive circuits are omitted, and can be copy-pasted from other good existing designs (i.e. easyvibe).



See you!

STM

[puretube]

information lost...

[stm]

Puretube, I just arrived at the office, so I will have to wait until evening to tackle the curves you are asking.  I think it will be an interesting exercise, and my sim is capable of large designs.

I think the real value of this circuit is that we are talking about something different, perhaps in-between a small clone and a phaser.  I'll see if I can try it soon to decide the actual usefulness of this sound.

As for the photocells, definitely a quad unit would be just great. Problem is matching, and as Mike (Vsat) said, he matched three pairs out from a bag of 50, so that's not very practical.  Maybe what I'm going to say should be mentioned on its own thread, but... I've been studying info available around photocells and LEDs, and found the following:

1) Light from an LED is quite proportional to its current, except when you aproach to 10 or 20 mA, where you get "diminishing returns". Anyway, I wouldn't expect powering an LED with more than 5 to 10mA on a battery pedal, and also, you can go with high efficiency devices, so in practice you should be running LEDs in their linear region.

2) There is much discussion about optimal color LED. From what I've seen, red and yellow would work reasonably good. Somewhere I read that yellow LEDs have more production spread, and thus red ones are preferrable for good matching (even though the photocell has its peak sensitivity around yellow).

3) Photocells are a complete world. Their resistance is inversely proportional to the light received, raised to a power near unity (something like 0.95,  or so, according to the device I saw).  This is nonlinear, and it is what makes matching difficult.

4) I've been thinking in using a jumbo LED (10 mm) to illuminate four 5mm photocells at a time.  This can make a quad.  Photocells can be mounted in a square on the PCB, while the LED must be supported above by some means.

5) Going back to matching, at least you can do the following to match 4 photocells in an arrangement as above:

  a. Define minimum and maximum LED current, perhaps in a 40:1 ratio.  For this example, let's consider LED current ranges from 200 uA to 8 mA (we are talking about high efficiency LEDs here!).

  b. We'll do matching at two points near (but not at) the extremes, so overall tracking should be more evenly distributed.

  c. Let's measure all photocells resistances at 4 mA (half the maximum current). Now take the value of the highest of the four photocells as a reference, and place in series with the other three photocells some resistors adequate for equaling said reference value.

  d. Let's measure all photocells resistances at 400 uA (twice the minimum current). Take the value of the lowest of them all, and use it as a reference. Now add resistors in parallel to the other three photocells so you match the value of the lowest.

  e. With the above method you need six resistors to do some matching at two points along the operating region. At least it should be better than running randomly selected photocells.

6) A more scientific method would consist in measuring each photocell at three or more currents, thus characterizing its curve. Then, some algorithm (or brute-force approach) can be used to determine optimum series and parallel resistors to add to each photocell so they have best possible matching within a region of interest.  Just to clarify, series resistors should be quite small, while parallel ones should be quite large.

7) One last point is that you could filter some light entering into one or more photocells in case its sensitivity is greater than its neighbours. This could even be done painting a small area on the LED quadrant that illumintes the particular photocell. Also, you can mount the photocells at slightly different heights from the PCB so they get more light.  Of course this is rather difficult to do because at this point there is some lack of well defined methodology for this adjustment.  Anyway, I'm working on this...

That's all for now.

STM

[puretube]

information lost...
:?:

[Vsat]

My apologies in advance for this off-topic post:

How does one include a gif (eg. for a schematic/graph...) in a reply to the list?
Regards, Mike

[Peter Snowberg]

Mike, take a peek at the end of this thread.

http://www.diystompboxes.com/sboxforum/viewtopic.php?t=21

[StephenGiles]

STM - that's quite something - una parva distinta de pescados!
Stephen

[puretube]

information lost...

[Vsat]

stm and list,
Have you seen a schematic for a 4th order allpass stage using a single op amp? This would be very interesting (but again would require precision components). The Eventide Instant Flanger uses a single LM301A op amp as a 4-pole lowpass antialiasing filter.

If 4-pole allpass could be done with a single op amp, then a 4-stage phaser (producing two notches) could be built with one dual op amp - one half for the allpass, the other half for the LFO (driving an LED to illuminate the LDRs), with resistive summing on  output. A Univibe built with a single 4558, so to speak.
Regards, Mike

[puretube]

haven`t seen the schem lately, but iirc, it`s not equal-R...

[StephenGiles]

See here on the input.

http://img78.photobucket.com/albums/v317/StephenGiles/EVph_comb.gif
Stephen

[Vsat]

Steve,
Yup, that's the LM301A-based lowpassv (nice re-draw BTW). Now if an allpass  could be built in somewhat similar fashion  with a single op amp, four capacitors (not necessarily same value) and four (preferably equal value) resistors, some very simple but neat things could be built!

Here's a gif comparing responses of a first-order 6-stage phaser (six identical first order sections) with a 2nd-order 3-stage phaser (three equal 2nd-order sections in series):

http://img.photobucket.com/albums/v484/icmax/boxovibe-4.gif

(watch the word wrap). Both phasers generate three notches - the central notch is at the same freq in both graphs.
Mike

[Mark Hammer]

As Mike noted, he demoed his Box-o-vibe for me (the one illustrated in the graph above) and it has a really nice sound.  The closer spacing of notches (especially with some regen) has a more focussed sound, almost like a synth-ey VCF, and like some of the best instances of the Mutron Bi-Phase you've heard on record.

Closer vs farther spacing is not really a better/worse distinction, but more a matter of application appropriateness.  Farther spacing is better suited to rhythm work and other set-and-forget instances; the broader spacing and shallower notches of Uni-Vibes even moreso.  In contrast, tighter notch spacing is more of a single note oriented solo tone.  Both musically valid, but the one is suited for where you don't want to pay attention to the sweep quite so much.

[Peter Snowberg]

Of course this gives rise to the question.... what about using both circuits and then adding or differencing the outputs?

[puretube]

information lost...