Twitter   Patreon WIGGLING 'LITE' IN GUEST MODE

 Goto page 1, 2  Next [all]
Author Pan CV with 2164 vs LM13700
cackland
 I'm laying out a circuit that has cv panning using 2164 ic and wondering if this looks correct? Ive seen a load of circuits using the standard LM13700 however wanted to use the 2164 instead. Following the data sheet I assume it is, however would like a second opinion. Appreciate it
jorg
 The control curve of the 2164 is expo; for LM13700 it's linear. This will cause major differences in behavior in this application. I don't think it will behave at all as you expect.
cygmu
 Yes, as jorg says, the exponential response is the big difference. Moreover, the 2164 can attenuate by 100dB but only provide +20dB gain so its not symmetrical either. One way to fix this is to use the Irwin linearisation circuit as in Fig 9 of the SSI2164 datasheet http://www.soundsemiconductor.com/downloads/ssi2164datasheet.pdf This uses an extra VCA per channel but you have four on the IC anyway. If you send x to one and 5-x to the other you'll get a linear pan response.
cackland
 jorg wrote: The control curve of the 2164 is expo; for LM13700 it's linear. This will cause major differences in behavior in this application. I don't think it will behave at all as you expect.

Good call. Totally overlooked that.

Thanks gygmu, would this suffice?

Added an led driver for left and right signals.

guest
 you will have to put the linearisation on each amplifier, after the inversion stage. otherwise you get a control signal of -ln(Cv/Vr), which, when put into an exponentiator makes e^-ln(CV/Vr) = Vr/CV and not -CV/Vr, which is what you want (CV = control voltage, Vr = reference voltage).
cackland
 guest wrote: you will have to put the linearisation on each amplifier, after the inversion stage.

Is this what you meant?

Graham Hinton
 cygmu wrote: If you send x to one and 5-x to the other you'll get a linear pan response.

Except that pan responses are not linear.

 jorg wrote: The control curve of the 2164 is expo; for LM13700 it's linear. This will cause major differences in behavior in this application. I don't think it will behave at all as you expect.

It won't and both are wrong at all points except the two extremes. The acid test is applying a triangle LFO and getting a smooth transition from side to side.

You could use either device as a control element, but the 2164 would be a better choice than an LM13700. Both would need different curve bending of the CVs to get something that pans correctly. It is easier if you make 0V = centre and deal with +ve and -ve CVs separately.

You can get PICs now with 10bit ADCs and DACs, the PIC16F1778 would be suitable for doing the whole job in one 28 pin package, or you could use two smaller ones with a single DAC output.
megaohm
 This THAT Design Note might be useful to you. It would need changes / additions to work with the 2164 IC. http://www.thatcorp.com/datashts/dn120.pdf
CLee
 megaohm wrote: This THAT Design Note might be useful to you. It would need changes / additions to work with the 2164 IC. http://www.thatcorp.com/datashts/dn120.pdf

I’ve built this design and it works well

http://clsound.com/vcpanner.html
CLee
 BTW, with the THAT design you can get exponential VC gain by using the unused (grounded) control ports.
cackland
 Thanks for the recommendations guys. Never considered THAT2180. The reason why I’m choosing the 2164 IC is because the # of VCAs per IC. With the recommendation from guest, does the implementation look correct?
cygmu
 I can see a few things that I think should change. Why is R118 1M? Do you really want a gain of 10 on the input there? You need resistors between the linearising VCA outputs and the control voltage inputs. Since you are using a -2.5V reference via a 49.9k resistor, if you are aiming for 0 - 5V control range, you want 100k resistors there. (Between IC19A pin 1 and U4C pin 12, and between IC10BV pin 7 and U6C pin 12). It looks like the control you are sending to the VCAs is equal and opposite but I think that what you really want is x and 5-x. This linearised VCA doesn't respond properly to negative control CV anyway. You can create your 5-x at IC10B by adding another input to the inverting terminal: -2.5V through 49.9k will do it. I might have made mistakes so proceed with caution! And Graham is right to point out that this will create a linear panning law in the end, which is not optimal -- you could look into equal power panning laws for example.
neil.johnson
 For equal power panning (-3dB) or any other law (e.g., -4.5dB) you need to do some control bending with diodes. Or just go with Graham's excellent suggestion and drive it from a cheap PIC - one ADC input for the CV and two DAC outs to drive the 2164s directly and a table to map the ADC input to the L and R channel DAC settings. Neil
Graham Hinton
 cygmu wrote: -- you could look into equal power panning laws for example.

Beware of following red herrings dragged across the course by inexperienced graduates writing application notes for semiconductor manufacturers.

There is no mathematical law to describe panning, the laws used in most mixers derive from data published by the BBC in the 1930s which was established by experiment. They just said this is what we want and we're not buying anything that does not conform, without saying how to do it.
Anybody with a personal computer and some audio software has better equipment to repeat the experiment than the BBC R&D had 80 years ago. You just have to use your ears and what is holding them apart.

Think about it. The so called equal power law (cos^2 + sin^2 = 1) is describing a circle, but panning is intended to move across a linear "stage" between two speakers.
-3dB (half power) at the centre sums to make +3dB so it is not mono compatible (not that there is a huge requirement for AM radio these days, but there used to be).
Besides that, it just doesn't sound right and the proof of a pudding is in the eating.

The actual laws used have a small attenuation in 50% (-4 or -4.5dB) and a large one in the other 50% (<-80dB). The perceived position depends a lot on the shape of the latter curve and the balance between the two sides. A small error of 1dB can result in a 25% position error.
neil.johnson
 Graham Hinton wrote: There is no mathematical law to describe panning, the laws used in most mixers derive from data published by the BBC in the 1930s which was established by experiment. They just said this is what we want and we're not buying anything that does not conform, without saying how to do it. Anybody with a personal computer and some audio software has better equipment to repeat the experiment than the BBC R&D had 80 years ago. You just have to use your ears and what is holding them apart.

That's another advantage of using a PIC or other small micro with a lookup table - you can create whatever gain curves your ears tells you sounds "best".

Neil
cackland
 cygmu wrote: I can see a few things that I think should change. Why is R118 1M? Do you really want a gain of 10 on the input there? You need resistors between the linearising VCA outputs and the control voltage inputs. Since you are using a -2.5V reference via a 49.9k resistor, if you are aiming for 0 - 5V control range, you want 100k resistors there. (Between IC19A pin 1 and U4C pin 12, and between IC10BV pin 7 and U6C pin 12). It looks like the control you are sending to the VCAs is equal and opposite but I think that what you really want is x and 5-x. This linearised VCA doesn't respond properly to negative control CV anyway. You can create your 5-x at IC10B by adding another input to the inverting terminal: -2.5V through 49.9k will do it. I might have made mistakes so proceed with caution! And Graham is right to point out that this will create a linear panning law in the end, which is not optimal -- you could look into equal power panning laws for example.

Yes, that was a typo on my part. Changed the 1M feedback resistor to 100K and also added the 100K resistors as you suggested.

Would you suggest perhaps a direct -5v through a 100k resistor into the linear vca pin and removing the -2.5v through the 1M resistor?

I don't understand this 'x and 5-x'. Could you explain further or perhaps somewhere I could read up on this?

The parts for each channel is seeming to grow which makes me think whether I should go back to LM13700 if it is an easier implementation / less part count and can perform the same outcome?
cackland
 neil.johnson wrote: For equal power panning (-3dB) or any other law (e.g., -4.5dB) you need to do some control bending with diodes. Or just go with Graham's excellent suggestion and drive it from a cheap PIC - one ADC input for the CV and two DAC outs to drive the 2164s directly and a table to map the ADC input to the L and R channel DAC settings. Neil

With Grahams suggestion, using a PIC micro controller is not the path I was aiming for.

Would like to keep this circuit completely analog.
cygmu
 cackland wrote: I don't understand this 'x and 5-x'. Could you explain further or perhaps somewhere I could read up on this?

Each section with two parts of a 2164 is a linear response VCA set up so that 0V control voltage gives 0 gain and 5V control voltage gives unity gain.

At the moment it looks like your schematic is set up so that when the control voltage sent to one section is 1V, for example, the other section gets -1V. The second VCA will not respond properly to -1V.

I was suggesting that, in this linear arrangement, you would want to send e.g. 1V to one and 4V to the other; and when one gets 2.5V the other would get 2.5V also. Certainly when one gets 5V the other should get 0V (hard pan left or right.)

That is what I meant by "x and 5-x" -- if the CV to one channel is x, the other gets 5-x, so that as one channel goes from 0 to unity gain, the other drops from unity gain to 0.

In fact as Graham and Neil have pointed out, proper panning circuits don't do this. The extremes are right but as you pan across you want some appropriate curve between the two VCAs. What is certain is that x and -x as currently set up will not work.
cackland
 Right ok, I understand. Reading over this: http://www.cs.cmu.edu/~music/icm-online/readings/panlaws/ I also found this diagram to help make sense of it. Currently the circuit is a linear circuit (Left graph) and what I want is the right graph, or a resemblance of it.
dksynth
 this THAT app note is a decent read on the subject: http://www.thatcorp.com/datashts/dn120.pdf
Graham Hinton
 cackland wrote: Would like to keep this circuit completely analog.

You want analogue? Here's how I did it 35 years ago which was good enough for WDR:

The input CVs are +/-2.5V and complimentary. It drives two pairs of dbx2150s (the forerunner of the 2180). When in the centre the outputs are both 0V = 0dB and the centre gain is set by the VCA input and output resistors.

Four years later I did it for the BBC and used a Z80 with much better results. PICs did not exist then and 1% resistors were expensive. The drawback of analogue shaping is the difficulty of getting the steep curve going down to virtually off at the extremes which happens in less than 10% of the CV range. Then you have to find a curve that sounds right on both headphones and loudspeakers. Good luck with that...
cackland
 Thanks graham. Appreciate you sharing. After reviewing the THAT 2180 pan circuit, I would be eager to try it, although I'm put off by the fact that there is only 1 vca per 2180. Needing 8 for four channels takes up a lot of space on the pcb. Im still eager to learn how to use diodes (as suggested) to bend / control for the 2164 ic. Does anyone has any suggestions for where I can read up on that? I done a big search on using the 2164 in a pan circuit (using non linear panning) and haven't had any luck. Assume everyone has either used the THAT 2180 or LM13700. After reading a big long thread about vca comparison: https://www.muffwiggler.com/forum/viewtopic.php?t=191492&postdays=0&po storder=asc&start=150 https://www.muffwiggler.com/forum/viewtopic.php?t=193783&start=0&postd ays=0&postorder=asc&highlight= it appears 2164 is more advantageous to use in comparison to the LM13700.
clorax hurd
 cackland wrote: although I'm put off by the fact that there is only 1 vca per 2180

I believe it should be quite easy to convert the 2180 circuits to 2164 as both are exponential and the negative CV pin (which 2164 VCA blocks doesn't have) is not used in the 2180 schematics anyway.... So perhaps just scaling the control voltage and changing values of the 2 resistors and 1 capacitor connected to VCA's audio input pin should do the trick.
Read both (2180 & 2164) datasheets to see how exactly...
Graham Hinton
 cackland wrote: After reviewing the THAT 2180 pan circuit, I would be eager to try it, although I'm put off by the fact that there is only 1 vca per 2180.

There's a dual smd version: THAT 2162.

 Quote: Im still eager to learn how to use diodes (as suggested) to bend / control for the 2164 ic. Does anyone has any suggestions for where I can read up on that?

Nonlinear Circuits Handbook Chapter 5

 clorax hurd wrote: I believe it should be quite easy to convert the 2180 circuits to 2164 as both are exponential and the negative CV pin (which 2164 VCA blocks doesn't have) is not used in the 2180 schematics anyway....

It is, but all you have to do is reverse all the diodes and reference voltages.
cackland
 Thanks Graham. I'll definitely have a look and read. Glad there is a dual smd version. Much better.
 Goto page 1, 2  Next [all] Page 1 of 2