Frequency-Dependant Low Pass Filter

[Hides-His-Eyes]

I just can't see the point in reinventing the wheel, when it has been perfected over 30 years ago!!

a: It wasn't perfected. It's functional, but rather inelegant.
b: What if you could do the same thing with fewer parts...

What do you mean by inelegant exactly? Surely you realise that if EH could have cut the component count they would. Think rack not 1590 box.

"The microsynth is the size it is and EHX couldn't do a similar thing with fewer components, so therefore you couldn't do the same thing with fewer components...."

EHX aren't perfect, they had to work within cost constraints we don't and there have been lots of new ICs since then!

[StephenGiles]

Please tell us all about these new ICs and how they might be used to provide a lower component count for an EH Guitar Synth update - I'm all ears and ready to plug in my soldering iron

[liquids]

Please tell us all about these new ICs and how they might be used to provide a lower component count for an EH Guitar Synth update - I'm all ears and ready to plug in my soldering iron

In your own words "Or you could just ask Howard Davis nicely!" and from another member "Funnily enough, I did ask Howard Davis if he could explain it to me, but replied that it was a slightly clunky technique compared to what he does now with the pigtronix mothership."   http://electro-music.com/forum/post-304693.html

That being said, I am going to give it a go and see if there is any notable improvement in my tracking.  Satisfied?   

Thus far, the compression (based on the Hollis flatline) did help a little bit with certain notes having improved tracking and apparent sustain (into a 4046), but sadly, the electronotes circuit made the low octave skipping no better, and made high frequency note tracking less responsive.  To be continued...

[Hides-His-Eyes]

Please tell us all about these new ICs and how they might be used to provide a lower component count for an EH Guitar Synth update - I'm all ears and ready to plug in my soldering iron

so you're saying the technique is perfect and I'm saying the world's a different place and all of a sudden I'm designing it from scratch yeah?

ICs: single chip VCAs, signal to RMS voltage chips; in fact THAT will give you I think two VCAs and an RMS detector (and a couple of op-amps) all in one DIP. So that could probably be pretty useful.

CA3094s are now @#$%ing £7 each or some bulllshit like that. And you need six. So yeah, I think we can do something about that with modern ICs.

[StephenGiles]

I think you completely misunderstand what I am saying, but please don't get overheated - it's hot and humid enough here in the UK as it is! I think if you manage this it would be a fantastic achievement.

[amptramp]

Let's do a thought experiment here.  Suppose you had an unlimited supply of analog, passive and low-level logic devices at you disposal.  How would you approach this tracking filter project?  One way would be to use equalizer-type frequency filters and level detectors to determine the lowest (fundamental) frequency at any given time.  The problem is that as you make the filters narrower, they take longer to respond and give an unambiguous determination of what frequency is fundamental.  If you use a counter to determine the frequency via measuring the period between zero-axis crossings at each filter output, what time response are you looking for?

The question isn't whether you can make a tracking filter.  The question is whether its response will be fast enough to do what you want to do, which is transform the sound into something else via manipulation of the parameters of another filter that takes the signal input and produces the new sound.  This would be separate from the tracking filter.

[Hides-His-Eyes]

I think you completely misunderstand what I am saying, but please don't get overheated - it's hot and humid enough here in the UK as it is! I think if you manage this it would be a fantastic achievement.

sorry man, can you tell I spent all yesterday shovelling gravel in that sun?

sorry for snapping.

[StephenGiles]

I think you completely misunderstand what I am saying, but please don't get overheated - it's hot and humid enough here in the UK as it is! I think if you manage this it would be a fantastic achievement.

sorry man, can you tell I spent all yesterday shovelling gravel in that sun?

sorry for snapping.

No problem, I don't envy you doing that! Storm clouds are brewing now so should be some relief from the heat.

[Hides-His-Eyes]

That's the magic, man. The rain is going to wash all that gravel for me so I don't have to.

[HarryBBD]

Hiya all. A few comments on the whole thread to get caught up.

The Electronotes tracking filter is only useful if you can be sure that the upper and lower bandpass filters are above and below the desired fundamental... make that
range the width of possible guitar notes and those filters are far apart. If you can get on the WRONG skirt of the filter, you are hosed (I KNOW this because I built a
Deluxe Octave Multiplexer (DOM) tracking filter and mistakently made it a bandpass, not a lowpass :^). Another caveat is that a bandpass filter has resonance (it must) so when you hit it with
a transient (like a note starting) it will take (precious) time to start responding and even more to stop when the note ends,

The GR300 solution only has to track a 2:1 range, not bad if you have a hex pickup.

The DOM can be built with LM13600, no need to use the unobtanium CA3094. It works fine.

I have tried to improve the DOM by using a four pole filter (which takes more precious time), using stagger tuned parallel filters, adding S/H amps to try and keep the loop filter stable
when the input misbehaves, etc. I have been (so far) unable to add to the design and make it work better than the original. Its a very well thought out design that makes the trade-offs
in exactly the right places (imho) save the loop filter integrating cap, which should be a little larger.

P/V conversion need not have ripple if you use a ramp-hold approach rather than a tachometer method. You can get a CV witin a single (good) cycle. The best way (imho) is to use the guitar
waveform directly (like the GR-300 does, in effect).

I'm a "spare no expense" kind of guy.  I don't think you could pull many parts out of the DOM and have it work. You might improve the compressor a bit, get the noise down a bit, but I don't think
you can get the tracking filter much better.

RG: Your 'delay' idea has a bit of merit, but you might have to delay a lot longer than you'd like unless you do something like the AXON or Roland methods where you try and guess the length
of the string based on the echo of the attack reflecting back down the string (very simple explanation). The delay would probably be a half or full cycle of the lowest input frequency to be useful.

I once toyed with the idea (simulation) of using a frequency shifter to move the guitar fundamental to an ultrasonic frequency, P/V convert it there where you'd have very short times. Unfortunately
you need to get the sine/cosine to do that with what is known as a 'dome' filter. To allow a 80Hz low end, it would take about 18ms for the signal to get to the other end of the dome filter, negating the point of doing that in the first place :^)

There is no free lunch, infact there isn't even a free snack :^)

H^)

[liquids]

Harry - thank you very much for sharing your thoughts here!     

[earthtonesaudio]

Yes, thanks Harry.  Very useful to hear from someone who's tried 'em all.

[liquids]

Here's what I have so far for an LM13x00 version.

I doubt it's right as is; I also think the 'exact' simulation of the CA3094 may not be needed...but here's a start?

Do note that the Eh Guitar Synth schematics and Deluxe Octave Multiplexer schematics seem to show slightly different feedback arrangements, though values are quite similar, should be negligible...so long as you account for where the inversions take place & negative feedback...?

[DDD]

It seems that the subject is almost impossible to be figured out.
Guitar fundamental frequencies are approx. 80-1200 Hz, so how one can extract fundamental tone? Especially in the "bass" range of the guitar sound, where 2nd harmonic oftenly 2 times stronger than fundamental.

[StephenGiles]

Looking very good, just a thought at this stage - if you put dots where there are connections and gaps where lines simply cross it would be even better.

[StephenGiles]

It seems that the subject is almost impossible to be figured out.
Guitar fundamental frequencies are approx. 80-1200 Hz, so how one can extract fundamental tone? Especially in the "bass" range of the guitar sound, where 2nd harmonic oftenly 2 times stronger than fundamental.

Well this filter does just that

[liquids]

Looking very good, just a thought at this stage - if you put dots where there are connections and gaps where lines simply cross it would be even better.

There are no places where lines simply cross....but I'll clean it up eventually..it's mostly for me, but I suppose when I share, intelligibility helps.  

TAOE P 1056:

RULES

1. Wires connecting are indicated by a heavy black dot; wires crossing, but not connecting, have no dot (don't use a little half-circular "jog"; it went out in the 1950s).

2. Four wires must not connect at a point; i.e., wires must not cross and connect.

3. Always use the same symbol for the same device; e.g., don't draw flip-flops in two different ways (exception: assertion level logic symbols show each gate in two possible ways).

4. Wires and components are aligned horizontally or vertically, unless there's a good reason to do otherwise.

[R.G.]

Then there's the old Japanese version. Lines which end at another line at a T connect. Lines which cross at right angles do not connect. No dots anywhere. Then there's the "meet at angles" style, where two wires meeting at another wire take a 45 degree jog in opposite directions just before connecting. Lines crossing over do not connect. Sometimes the 45 degree jogs have dots, sometimes they don't.

[StephenGiles]

If the 13600 will not work, and I have no reason to suppose that it won't, in the interests of reducing component count the SSM 2164 quad VCA might be tried.

[merlinb]

If the 13600 will not work, and I have no reason to suppose that it won't, in the interests of reducing component count the SSM 2164 quad VCA might be tried.

Since the 2164 is also obsolete, that might not be much of a step forward!