KVR Audio

zamrate · Post by **zamrate** » Mon Apr 10, 2017 6:30 am

Yesterday evening, more for fun than anything else, I tried simulating a Moog Ladder filter in LTSpice. The reference schematic I used is: http://www.experimentalistsanonymous.co ... CF%202.gif
I thought this would be an interesting test since LTSpice allows WAV files input.
Sadly, although I get an output from the filter, it doesn't seem to be filtered much. Also, although R8 (which controls feedback amount) is set to 1 Ohm, I don't get much resonance.
I'm new to LTSpice, I practically never used it before, so if anybody more familiar with LTSpice could tell me what's wrong, I'd be thankful.. Note that for most transistors I used generic PNPs/NPNs, since I couldn't find the TIS97 used in the original Moog ladder filter schematic which I used for reference.

For the simulation, unzip saw.wav into c:\samples

Capture.PNG

moog.zip

aciddose · Post by **aciddose** » Mon Apr 10, 2017 7:08 am

Here:

http://xhip.net/temp/other/moog.net

Adding the circuit for the exponential current source while interesting is completely a waste of time.

A nice neat net is more useful than a schematic a lot of the time although I'm sure some would appreciate having the schematic to select nets to probe visually.

You should find once you get the net operating correctly that spice is not accurate enough to provide realistic performance from audio-frequency circuits. Unfortunately you'll probably realize this on your own over time, but I can assure you that you are wasting your time if you are not motivated to do this because you are working with the same electronic circuits on a bench.

zamrate · Post by **zamrate** » Mon Apr 10, 2017 11:34 am

Thanks for the .net file. I just tried it and it works. Still don't understand why my schematic doesn't work though...
When you say SPICE is not accurate, you probably mean that a real Moog filter sounds better than the SPICE simulation of the same filter? Unfortunately I don't have a real Moog filter here to compare..

mystran · Post by **mystran** » Mon Apr 10, 2017 3:43 pm

I see that you already got a working netlist from aciddose, but I want to make an observation about the LTSpice transistors: the generic NPN/PNP transistors in LTSpice seem to be really simple models, and you should probably always substitute some actual transistor model even if you can't find an exact model or substitute.

If I'm not mistaken, the TIS97 in particular is an entirely uninteresting generic high-beta small signal NPN and everyone's favourite substitute for those is 2n3904 (which is common, cheap high-beta small signal NPN that usually does the job and happens to come as a model in LTSpice). This is what you have in aciddose's netlist too.

In general 2n3904 will work just fine in most audio circuits that require generic small signal NPNs (and it's complement 2n3906 for PNPs) and usually the particular transistor that the manufacturer happened to get a cheap bulk-deal for is far less important than getting Spice to use a more complex simulation model.

aciddose · Post by **aciddose** » Mon Apr 10, 2017 11:49 pm

If you actually wanted to bench-test this circuit however you might look at various IC options containing multiple matched NPN pairs. The performance of the active elements is really not very important and can be tuned easily by adjusting the bias resistors and current source. This is what made the Moog circuit practical in a time period where transistor performance was to be honest awful considering the cost of devices.

Today you can waste a lot of time matching and gluing 2n3904s for that "authentically awful" touch, or you can use widely available matched transistor ICs to get the circuit as near perfect as is reasonably possible. (Still awful compared to modern OTA-based designs.)

Unfortunately the available ICs change quite frequently and you're likely looking at smaller SMD footprints these days.

Urs · Post by **Urs** » Tue Apr 11, 2017 6:35 am

Uhm, I haven't compared the schematics to anything, but it looks like you're shorting the differential pair at the bottom, as if feeding the resonance to both inputs rather than just the inverting one. That looks odd to me.

mystran · Post by **mystran** » Tue Apr 11, 2017 9:12 am

aciddose wrote: Today you can waste a lot of time matching and gluing 2n3904s for that "authentically awful" touch

That's blasphemy, you're obviously supposed to source some vintage TIS97 and then match & glue those! Remember to use vintage caps too, modern caps are way too clean and besides it's necessary for caps to age properly before they will sound truly analog!

Urs wrote: Uhm, I haven't compared the schematics to anything, but it looks like you're shorting the differential pair at the bottom, as if feeding the resonance to both inputs rather than just the inverting one. That looks odd to me.

Good catch. That's definitely wrong (eg. cancels all the differential signal) whatever the circuit (although this looks like a Minimoog). Anything coming out of the circuit then is essentially common mode pass-through that should theoretically get cancelled out (so there's probably some other problem too, maybe with IO-levels, maybe we something else).

zamrate · Post by **zamrate** » Fri Mar 01, 2019 2:06 pm

Urs wrote: ↑Tue Apr 11, 2017 6:35 am Uhm, I haven't compared the schematics to anything, but it looks like you're shorting the differential pair at the bottom, as if feeding the resonance to both inputs rather than just the inverting one. That looks odd to me.

Indeed, that was it, thanks.

stratum · Post by **stratum** » Sat Mar 02, 2019 1:02 am

It took two years to figure that out?

Moog Ladder filter simulation in LTSpice