How to accurately sim tone controls?

[GibsonGM]

Hi,

Here's what I think is an interesting question: in LT Spice, when I model various tone-shaping circuits, I come up with results that can vary WILDLY from 'known accurate' information (internet, books...).   This is, I believe, because I'm not properly defining what the output impedance of the previous stage is (loading effects), and also am missing the input impedance of the stage following.

I've tried adding 'the stage' (say, an op-amp buffer) before the circuit, and following with a 1M resistor to ground for 'next stage input'....and I'm still getting odd results.   How are the 'simulation gurus' getting accurate results when using sim software??   Is it as simple as something you can set up in your input signal voltage source?  Changing the voltage series resistance doesn't do much....

Thanks!

[artifus]

i'm not much of a simmer but check out some posts by gus and tca for clues and explanation.

*edit*  clues examples and explanations is what i meant. man - that woulda bin so much cooler. doh.

[induction]

I had a similar problem when comparing tone stacks in LTSpice with their equivalents in the Duncan Tone Stack Calculator.  I don't think impedances are accounted for in the Duncan, so they should match.

I believe the differences in this case might arrive from the definitions you use for logarithmic pots.  I found pot definitions at the Yahoo LTSpice group, and then modified them.  My results don't match Duncan perfectly, but they are pretty close.  The major differences I find are near the minimum settings of log pots.

[tca]

Let us be specific... What circuit and what parameters are you trying to simulate?

[GibsonGM]

Thanks, TCA - thought you might be able to shed some light on this )    I'm looking at a midcut snippet from Merlin Blencowe's book - the circuit looked nice.  Allowed you to 'dial in' a mid cut with the 1M pot at top....at 0% rotation, you get a flat response.      In my sims, I get all manner of weird mids...sort of Big Muffish (more tilt than an increasing mid cut).  Changing the load value doesn't do a lot, nor does altering series resistance of the 'generator'.   Maybe I'm missing something?
He suggests it works well with about 40K output impedance (like from a 12AX7); no clue if that is the problem.

[PRR]

Show results.

http://i.imgur.com/SzZ1Kgv.gif

My sim-pot is broke; I used a tapped string instead.

[ggedamed]

@ PRR: The "load" connection (R2-R3-R5) should move along the tapped string, isn't it? Looks like taking the output from the cursor of a tapped potentiometer for me.
Anyway, in Tina-TI is easy-peasy to do this (it's made for mouse-noobs like me ):





Man, the Flickr interface is so @#$%ed up!

Haha, I like the way censoring works here.

[GibsonGM]

Thanks, guys; I'm not getting results like yours, PRR - the graph you posted is close to what should be happening.  Mine doesn't 'recover'...the Q changes and it looks more like a tilt control.   I'll post more after work!

Yup, the 1Meg I threw in as a load should go to ground, actually....that was just messing around time to see if anything changed.  It didn't.   Odd - I must have something set up wrong.  It's a simple ckt, it's just not simming correctly....

[teemuk]

Spices are usually pretty much on point in basic math like RC filters. I'd have more doubts about the accuracy of other softwares.

Anyway...

I believe the differences in this case might arrive from the definitions you use for logarithmic pots.

Indeed.

Note that when a potentiometer is "logarithmic" or "audio" it simply means the taper isn't linear.

What is not so often mentioned is that logarithmic/audio tapers come in several different types which have several different "steepnesses" of the log function. Therefore the percentage of total resistance in center position of the dial can vary widely with values like 10% or 30% being the most common types. They are still all logarithmic or "audio" pots but you can understand how such variation might easily create huge differences in plotted responses of the pot dial.

This user error is, I think, the most common reason why simulations tend to generate differing results.

Correct to right logarithmic function and the results should be quite accurate.

[ggedamed]

Actually, using PRR's circuit my sim gives the same result as his:



And it's the best adjustable Bridged-T I've encountered. I'll try it for myself with a rotary switch.


LATER EDIT: actually it degrades under load. I think the involved impedances have to match better.

[ggedamed]

Man, am I stupid or what? To make a chart in Tina-TI, you have to define
1. Anaysis > Mode > Parameter stepping, then
2. Anaysis > Select Control Object, then
3. Anaysis > AC Analysis > AC Transfer Characteristic.

Thing is, at the step 2 you have to click on the parameter you wanted stepped (doh! ).

Here it is again, with a 1M load:



Then with a 10M load:



And, finally, buffered by op amps:



So the load has its importance, although even the 1M version is not too shabby.

[GibsonGM]

Nice graphs, guys! Yes, it looks like a really useful circuit!  Said to be 'optimized' for a 40k output impedance.  I think that means more that it will *work* with the output Z of a typical 12AX7 stage.

Turns out my issue was with my .STEP PARM commands...when I just used {x} to control the LT pot wiper value, it worked out!
Should've changed my chart background to white, sorry...

With a 10M load:

[PRR]

> 1Meg I threw in as a load should go to ground, actually

Just counting on thumbs: 1Meg loading on 1Meg pot will give DC response 0dB at left and -6dB at right. In audio terms: bass drops off at right end of pot.

This may or may not be what you want. IAC there's other ways to shave bass. I suspect (without peeking in the book) that you want NO-load.

The extra 1Meg plausibly goes to wiper, which indeed makes a difference overlooked in my hasty-sim. It probably drops the half-rotation response just a bit, perhaps not enough to matter; it also shifts the dip-center lower when deeper as seen in ggedamed's plot. 

40K is perhaps a reasonable upper-limit on source impedance before it just gets sucked-down. This is indeed the typical Zout of 12AX7 with 100K plate load (66K||100K).


> comparing tone stacks in LTSpice with ... Duncan Tone Stack Calculator. I don't think impedances are accounted for in the Duncan

TSC has "Zsrc", source impedance.

It lacks explicit load impedance, though most variants have an output resistor which may be modified to taste.

[GibsonGM]

Thanks, Paul, from another Mainer )  It was all gomy for a while, but we got it!    I figured 1M might be what I'd expect if I followed this circuit with an opamp...less for a BJT.   Just playing with some ideas I'm having for a preamp with more tonal features than the typical stompbox!
   
I think ggedamed's got it, actually sticking an opamp in there!  Nothing like using the actual device you'll be working with, ha.