RTA plots from A/DA flanger clone at varying regeneration levels

[12Bass]

Hey Everyone,

For those interested in flangers, I've made up some static RTA plots using my modded A/DA flanger clone.  For the test, pink noise was used to graphically demonstrate the comb filtering effect.  In another thread there were some questions regarding the impact of adding regeneration (feedback/enhance) to the signal, so I've included some plots which show varying regeneration levels.  Note that this particular flanger has extended high frequency response compared to most analog flangers, which would roll off starting around 5 - 10 kHz.

Pink noise (effect off):



No regeneration, positive flange:



Low regeneration, positive flange:



Moderate regeneration, positive flange:



High regeneration, positive flange:



No regeneration, negative flange, same delay as above:



High regeneration, negative flange:



And, finally,

No regeneration, positive, but centered at 500 Hz, showing the comb filtering up past 10 kHz:



A few things are apparent in the plots.  The overall gain goes up as regeneration is added, while the comb shape becomes more severe and pointed.  Also, note that negative flanging cancels more of the low frequencies, and makes the sound more hollow.

As mentioned above, these are all static plots.  To help visualize what the flanging effect looks like in action, basically what happens when the LFO sweep is engaged is that the comb shape moves up and down the audio spectrum, or left and right horizontally on the graph.

Hope this helps anyone who might be interested in seeing pictures which serve to illustrate what they are hearing.

[Mark Hammer]

Lovely pictures!  Thank you.

Folks interested in this should read up on "theta processing".  This was first introduced by Bernie Hutchins in Electronotes in the late 70's, as far as I know, and the topic has come up here on several occasions.  The basic concept is that multiple allpass (phase-shift) stages are added to the wet signal, coming from the BBD, prior to mixing with dry.  These stages add phase-shift to low-frequency content, rather than high (so, cap goes to ground, rather than between last stage and non-inverting input, as we're used to seeing).  The result is that the notches at the low end are spaced a little further apart and the result is apparently more musical sounding.

Hutchins' experiments involved adding as much as 24 or more fixed stages, though as few as 4 can apparently do the trick, as you can see in Jurgen Haible's Stormtide Flanger - http://www.jhaible.de/jh_storm_tide_2.pdf.  Keep in mind Jurgen is not shy about adding as much circuitry as it takes to get the job done, so if he's happy with 4 stages, then I see no reason to add more.

You will note that the impact of adding regeneration is greatly diminished above 4khz or so, which is where the 2-pole lowpass filter in the regen loop kind of kicks in.

[12Bass]

Thanks Mark!

Still curious about theta... not sure if/when I'll give it a try.

As for the lowpass filter on the regeneration path, I've relaxed it as well, so it is pretty subtle on my build.  After building the A/DA close to the original values, I experimented with the various LPF sections until I got things sounding the way I wanted.  For the most part, the audible range is unaffected.  The resultant wider bandwidth of the BBD path allows deeper cancellations at higher frequencies and more dramatic flanging in the upper range of the sweep.  As a downside, a bit of clock and aliasing noise seeps in when using the lowest delay range.  But that's a worthwhile tradeoff, IMO, as I don't tend to like the super-low end of the sweep for flanging anyway. 

Here are a couple more plots to illustrate:

High regeneration, positive flange, first notch at 1 kHz:



High regeneration, negative flange, same settings as above:



As set above, the flanger is really screaming up toward the top of the sweep.  Note the huge hump in the bass as the comb filter moves upward.  It's actually a bit deceiving, as the ear tends to focus on the pitch of the pronounced 2 kHz and above peaks and doesn't notice the big bass increase.  The negative flange at the same settings sees the bass seriously reduced, which explains why the fundamentals practically disappear as a negative flanger nears the zero point. 

In its current tuning, my A/DA will go roughly an octave above the previous plots, though the BBD loses gain as the clock frequency goes up, so the comb filtering is less severe.  From the samples I've heard, digital flangers like the TC vortex have a more linear response as they sweep up toward the zero point. 

[Mark Hammer]

Experimenting with theta processing is pretty dang easy.  Trivial matter to perf 4 fixed allpass stages with a quad op-amp, and have it as a switchable add-on between the mixing node and the last BBD filter stage before the mixing node.

I can send you the Electronotes appnote (#117) on it if you're interested.

[frequencycentral]

Trivial matter to perf 4 fixed allpass stages with a quad op-amp....

[Mark Hammer]

.....or you could etch.

You can use that PCB layout, but do note that the cap and resistor swap places for this application.  Normally, we'd have a cap and a resistor coming from the previous stage.  The cap would go to the non-inverting pin, and there'd be a resistance of some sort from the non-inverting pin to ground, whether a fixed resistor, LDR, JFET, or switched resistor.

That cap and resistance to ground form a highpass filter that applies phase shift above a given frequency set by the RC joint value.  Flipping the R and C, so that the C goes to ground forms a lowpass filter such that phase shift increases the lower the frequency.  Theta processing introduces additional phase shift for the low end to stretch the spacing of those notches so they aren't as close together.