Cello Palette Style EQ Design (was High End Tone Control)...

If you where to build a Cello Palette functional clone, what technology do you want?


  • Total voters
    48
  • Poll closed .
Because I'd already set my model up for the 25kHz band, I put the resistor values in from Charles' superb data. To reveal the accuracy, I had to add 1.5 ohms in series with the 24dB position to set the maximum boost to precisely 24.00dB.

That reveals that the accuracy of the 25kHz boost/cut to be within +/-0.01dB at all positions. Whether those values were figured out by Burwen or Cello is beside the point - they give (at least with completely accurate resistor values) absolutely astonishing accuracy.
 
If only there were a way to mathematically calculate the Vout/Vin characteristic of a circuit in the frequency domain, then people could perform these calculations using different component values, and choose the assemblage which gave the most desirable result.

And indeed, if such a calculation were possible, then people could program a computer to do the work, saving time and drudgery. Luckily the FORTRAN-II programming language, which appeared in 1960, included a COMPLEX datatype, which is the perfect vehicle for Laplace domain calculations (AC analysis in electrical engineering). Additional breakthroughs in computer software during the late 60s and early 70s made such calculations even easier, faster, and more accurate.
 
Well that is interesting. I've put all the switched data that Charles kindly supplied into my model. The critical design attribute is that the center frequencies (120Hz, 500Hz, 2kHz and 5kHz) have a grounded center tap. That makes those bands constant Q. The outer bands (15Hz and 25kHz) do not have a center tap and are variable Q.

If you remove the grounded tap on the middle four filters, the step size goes totally whacky. So the design resistor values were worked out with a center tap.

Just to show what I'm talking about is the response of the 2kHz filter, attached. Again just boost (cut is identical response, just subtractive).

Craig
 

Attachments

  • 2kHz.jpg
    2kHz.jpg
    349.4 KB · Views: 89
Very interesting. So at the frequency extremes, does the Q decrease (implying greater bandwidth) as the boost/cut levels increase? If so, I'd opine that this is a "good" idea sonically. Now, does this / can this also apply to the National Handbook circuit (all paralleled filters)?

Thanks
 
OK - the only National/Cello Pre control with a 10k pot is the 20kHz one. So I grafted in one of the Audio Palette 10k switched resistors to the Palette Pre model.

With the center tap grounded, the frequency sweeps as you adjust the control. Which is useless. So that is an idea that does not work with that circuit.

Leaving off the tap ground the control does what is expected, but the steps are wrong - so a switched resistor would have to be designed for the impedances in the Cello Preamp, and with the correct step size.
 
Very interesting. So at the frequency extremes, does the Q decrease (implying greater bandwidth) as the boost/cut levels increase? If so, I'd opine that this is a "good" idea sonically. Now, does this / can this also apply to the National Handbook circuit (all paralleled filters)?

Thanks

For the Audio Palette the 15Hz/25kHz the initial few steps have very low Q (about 0.3). The Q increases as you advance the control to about 2 at maximum boost/cut. So the bandwidth stays roughly constant

Everything in-between operates with constant Q (or very close to).
 
If only there were a way to mathematically calculate the Vout/Vin characteristic of a circuit in the frequency domain, then people could perform these calculations using different component values, and choose the assemblage which gave the most desirable result.

And indeed, if such a calculation were possible, then people could program a computer to do the work, saving time and drudgery. Luckily the FORTRAN-II programming language, which appeared in 1960, included a COMPLEX datatype, which is the perfect vehicle for Laplace domain calculations (AC analysis in electrical engineering). Additional breakthroughs in computer software during the late 60s and early 70s made such calculations even easier, faster, and more accurate.

You don't need anything as sophisticated as Fortran. Excel has everything you need to do these sums. The trick is choosing the right E96 resistors given the idea sequence. Watch this space...
 
  • Like
Reactions: Rick Miller
Distortion of OTA-1 is also fairly sensitive to load resistance. Decent opamps will cheerfully work into 600 ohms without any increase in distortion or decrease in output voltage. That is not the case with OTA-1; of course it was designed with a particular circuit application in mind, with a multiple feedback bandpass filter wrapped around it. And it has a large output impedance, being a transconductance amplifier. Most if not all of Coleangelo's designs were transconductance with the output taken from the collectors rather than voltage amplifiers with the output taken from the emitters.
 
OTA ... yes.
I was thinking whether you could change the output section to emitter follower keeping the front end and ending with a (more normal) opamp with out the normal differential input pair (which will kill/reduce even harmonics).
I like the non symmetrical design, as it will ensure more even than uneven distortion ...
But this is of course sidetracking a bit ... sorry 😉