[sdiy] LC delay lines and scanner chorus/vibrato taps

Richie Burnett rburnett at richieburnett.co.uk
Sun Sep 8 20:31:38 CEST 2019

Hi Tom,

That cascade of LC pi-filters is what engineers call a lumped-element model 
of a transmission line.  You can delay a signal just by sending it down a 
length of coaxial cable, but it travels at the speed of light in the cable. 
So you need a looooong length of cable to get any significant delay at audio 
frequencies.  This is where the lumped-element approximation comes in handy. 
All those cascaded L's and C's model the behaviour of a much longer length 
of cable with taps every "30km" or so!  The catch is that modelling the 
cable like this in large "lengths" limits the effective bandwidth to only a 
few kHz.  Not a bad compromise given the intended application though.

In order to work properly the lumped element delay line (or a length of coax 
for that matter) needs to be driven from a source with the correct source 
impedance and terminated into a load with the correct load impedance.  If 
you don't do this you get reflections from the impedance discontinuities, 
and your signal bounces back from the ends of the delay line resulting in 
standing waves.  Think of the way that light reflects from the surface of a 
window or from the surface of water when it has to transition from one 
medium to another one.  (As an aside, this is why most signal generators and 
scopes designed for RF work have the option to use 50R source and input 
impedances so that everything can be matched throughout and you don't get 
any reflections.)

Now, if you apply any loading at all along the transmission line (or at the 
taps of the delay line) it will disturb the characteristic impedance of the 
line at that point and will cause reflections.  It doesn't have to be a 
resistive load, even adding some capacitive loading at a point along the 
line will cause reflections, and this modifies the frequency response.  So I 
would say that the high input-impedance buffers are essential to prevent 
loading the transmission line and modifying its operation.  Especially given 
that there are many of them spaced along the delay line.  Also every time 
you load the line resistively you are removing energy from it.  This loss 
adds to the "insertion loss" of the delay line caused by all of the 
inductors and capacitors, and results in a weaker signal coming out the end 
of the delay line.

This lumped element delay line is very amenable to simulation in something 
like LTspice if you have it.  So you could use that as a reference to debug 
your real circuit, and also to assess the amount by which loading at 
different tap points messes up the delay line operation.  If you simulate it 
in LTspice try to include the DC resistance of the inductors too, then you 
will get a better idea of the "insertion loss" through the whole delay line.

It's an interesting project, and I look forward to reading more about it as 
it progresses!


PS. As a practical note, make sure that the inductors you're using aren't 
magnetically coupling to each other.  If you're using toroidal inductors 
they keep most of the magnetic field inside.  But I don't know what type of 
indutors you are using.  If using ferrite stick inductors or gapped 
laminated iron-core inductors these tend to p!$$ out magnetic fields and 
couple effectively to nearby inductors that also have gaps in the magnetic 
core.  If all your inductors are gapped and right next to each other, your 
signal could potentially just couple magnetically from one inductor to the 
next to get to the output and "short cut" the delay line altogether!

-----Original Message----- 
From: Tom Wiltshire
Sent: Sunday, September 08, 2019 4:39 PM
Subject: [sdiy] LC delay lines and scanner chorus/vibrato taps

Hi all,

I’ve been playing with a modified version of Jürgen Haible’s scanner 
chorus/vibrato, described here:


Since JH has conveniently done the calculations for the delay line part and 
the 33mH inductors he used are still cheap and available, I’ve used that 
part without modification:


What I’ve changed is the control and crossfade circuit. My version has a 
pair of DG406 16-to-1 switches for selecting taps from the delay line, and 
then an AS3360 dual VCA for crossfading between those taps. Control signals 
for these are generated in firmware on a PIC 16F1778. This chip usefully has 
10-bit DAC channels that can be used to drive the VCAs, and by setting the 
DAC reference to the correct level, the output is perfectly scaled for the 
VCA. I’ve done a variable speed “LFO” in firmware that can scan the delay 
line at three depths (as per the Hammond organ, more or less), but you can 
also feed in a 0-5CV and use that to scan the delay line.

However, while I’m comfortable with all that part, I know very little about 
transmission lines, beyond what I’ve read here:


My question is “Are JH’s buffers on each tap necessary?”. The paper noted 
above suggests that a load on each tap should be at least 10x the impedance 
of the delay line (calculated by JH as 840 ohms). That’d mean resistors to 
take signal from each tap should be roughly 10K or greater. That would work 
perfectly into an inverting mixer or virtual ground node (such as I have on 
the VCAs) and I could simply use weighted resistors values to get all the 
taps at the same level. However, I can’t get this part to work, and looking 
at the signals in the transmission line doesn’t tell me much since I don’t 
know what I’m looking at or looking for.

So, do I need the buffers? Would a 10K to virtual ground be too much load 
for each tap? At this rate, I’m going to finish up putting simple op-amp 
buffers on each tap as an experiment.


       Electric Druid
Synth & Stompbox DIY

Synth-diy mailing list
Synth-diy at synth-diy.org

This email has been checked for viruses by AVG.

More information about the Synth-diy mailing list