Voltage Controlled Filter

[MAntunes]

Hello again.
Due to my growing interest in analog electronics, I decided to design an analog (or analog/digital, but mainly analog) synthesizer.
The first step is going to design a simple Voltage Controlled Filter (low pass).
I have a little bit of knowledge in analog filters, but not much.
Could you recomend me some books, articles, etc about analog filters and how to voltage control them?
Thank you in advance,

Miguel

[f5r5e5d]

https://www.google.com/#q=jfet+voltage+controlled+resistor&pws=0 is one common technique for varying RC circuit frequency response

https://www.google.com/#pws=0&q=voltage+controlled+amplifier can be used in conjunction with state variable filters where the frequency response depends on gains too

[IconicPCB]

A somewhat simpler approach might be switched capacitor filter topology.

[LvW]

The first step is going to design a simple Voltage Controlled Filter (low pass).

I suppose, you want to electronically control the 3dB frequency, correct?
In this case:
* Degree of the filter?
* Gain requirements?
* Which characteristics (Butterworth, Chebyshev,..)?

[Yansi]

Voltage controlled filters in analog? Look at application notes for CA3080 or LM13700 or any other transconductance amplifier.

[kripton2035]

if you want to go the digital way, take a look at cypress PSoc and their switched capacitors modules
there are tutorials on youtube.
you can make the whole system in a single chip.

[Yansi]

Nah, those damn psics... Get yourself a proper DSP.

But didn't op ask for an analog solution???
Then the OTA or switch cap filters is the way to go

[tggzzz]

Thanks for prompting a train of thought.

When I get time I'm going to reimplement a high-Q audio bandpass filter using modern components, to see how far I can push the technology now. I could make that voltage controlled, but I suspect the problem would be ensuring the centre frequency was sufficiently stable. No, it doesn't use OTAs and the like

[T3sl4co1l]

OTAs and switched caps are the basic building blocks, for simple variable filters and whatnot.

DSP?  GAH!

Tim

[kripton2035]

Nah, those damn psics... Get yourself a proper DSP.

But didn't op ask for an analog solution???
Then the OTA or switch cap filters is the way to go

the cypress psoc mix efficiently analog and digital in the same chip
it's not a dsp, it has analog switched capacitor blocs inside, and a way to program them the way you want
all in a single chip.
with high end psocs (3 and 5) you can even speak of a small dsp inside.

[MAntunes]

The first step is going to design a simple Voltage Controlled Filter (low pass).

I suppose, you want to electronically control the 3dB frequency, correct?
In this case:
* Degree of the filter?
* Gain requirements?
* Which characteristics (Butterworth, Chebyshev,..)?

Yes, I want to electronically control the cutoff frequency of the filter, sorry if I didn't express myself right. And I want it completely analog.
I wanted a first or second degree, nothing too complicated (second degree would be nicer).
Gain of 1 in the passband.
I'm more inclined to a Butterworth filter. As I say, this project is for me to learn more about analog eletronics, so if I could learn how to design both of them would be nice.

if you want to go the digital way, take a look at cypress PSoc and their switched capacitors modules
there are tutorials on youtube.
you can make the whole system in a single chip.

I have been interested in the PSoC, but not for this project though.

[LvW]

MAntunes, it is hard to find a second-order lowpass with the ability to change the frequency without disturbing the pole-Q (that means: mantaining the filter characteristics).
More than that - it is even harder to find such a circuit where the control element is grounded (because this is required, for example, for replacing a resistor by a FET).
For these reasons - and as indicated in previous answers - the only practicable solution seems to be on OTA basis (Operational Transconductance Amplifier). An OTA is an opearational amplifier with a current output (high output resistance).
As a modern alternative you could look for Current-Conveyor realizations (AD844).   

[MAntunes]

Thank you for your answer!
And for a first order? Would it be easier?

I found some filters that look like this:


But is there some books or articles that I could read about this?

[LvW]

Thank you for your answer!
And for a first order? Would it be easier?

Depending on the required tunung range there is a simple solution for a tunable 1st order filter:
Replace the first (input) resistor in an active inverting lowpass (R||C in the feedback path) by a voltage-controlled resistor which is realized by a FET (operated in the resistor region).
This is possible because of the virtual ground principle.

[MAntunes]

Thanks a lot!
I'll search about FETs as variable resistors and then I'll simulate and see if it works.
Any advice you think would be important?

[sync]

Search for "Moog VCF". It's one of the classic analog synth VCF. You will find tons of information.

[LvW]

Thanks a lot!
I'll search about FETs as variable resistors and then I'll simulate and see if it works.
Any advice you think would be important?

Yes - the voltage VCE must not be larger than some tens of millivolt - otherwise you have to much distortion (due to FET non-linearities).
This range can be extended somewhat using feedback. Some schemes are available in relevant papers (keyword: FET as resistors).
The source node must be connected to the inverting input (virtual ground).

[MAntunes]

Search for "Moog VCF". It's one of the classic analog synth VCF. You will find tons of information.

Thank you, will do that.

Thanks a lot!
I'll search about FETs as variable resistors and then I'll simulate and see if it works.
Any advice you think would be important?

Yes - the voltage VCE must not be larger than some tens of millivolt - otherwise you have to much distortion (due to FET non-linearities).
This range can be extended somewhat using feedback. Some schemes are available in relevant papers (keyword: FET as resistors).
The source node must be connected to the inverting input (virtual ground).

Thank you.
http://www.vishay.com/docs/70598/70598.pdf - From here I could find that for the J111 JFET the RDS(on) varies between 20 and 60 Ohm.
The final circuit would be something like this, right?



With a capacitor of a few nF I could have a good range of frequencies. I'll investigate a bit more. The problem I'm seeing here is that I would have to feed a negative voltage.

[iampoor]

http://www.musicfromouterspace.com/index.php?MAINTAB=SYNTHDIY&PROJARG=STEREOMIXER2006/STEREOMIXER2006.php&VPW=1270&VPH=525

This website will be your new best friend. 

[MAntunes]

http://www.musicfromouterspace.com/index.php?MAINTAB=SYNTHDIY&PROJARG=STEREOMIXER2006/STEREOMIXER2006.php&VPW=1270&VPH=525

This website will be your new best friend. 

Yes, I gave a look at that website, but the designs seemed to difficult for me. But I'll give another look.

[LvW]

From here I could find that for the J111 JFET the RDS(on) varies between 20 and 60 Ohm.
The final circuit would be something like this, right?

With a capacitor of a few nF I could have a good range of frequencies. I'll investigate a bit more. The problem I'm seeing here is that I would have to feed a negative voltage.

Yes - that is the most simple alternative. Of course, you could - in addition - use another resistor in series or in parallel to the FET (for better tuning or to enlarge the allowed voltage range). I didn`t check the J111 part - however, I am sure the resistor range must be much broader (from some tens of ohms up to hundreds of kohms). However, the sensitivity to control voltage variations may be too high for very large resistive values.   

[Yansi]

My goodness why don't you choose the OTA? The jfet thingy is only a distortion generator. Nowhere near a high Q filter! Get yourself a proper OTA. Only that way you can achieve repeatable results with well defined behavior.

[LvW]

My goodness why don't you choose the OTA? The jfet thingy is only a distortion generator. Nowhere near a high Q filter! Get yourself a proper OTA. Only that way you can achieve repeatable results with well defined behavior.

I think, we don`t speak about a high-Q filter but on a first-order lowpass only.
Distortion generator? A well-designed combination of an ohmic resistor and a JFET (eventually with a distortion reducing feedback scheme) will give good results.
It is not a problem to limit the voltage between D and S at a level which allows sufficient linearity. 

I should add that it is also possible to reduce distortions if you use two FET`s in combination (parallel or series).

[Kleinstein]

Pure analog adjustable filters tend to be rather nonlinear and produce quite some distortions or noise if amplitude are reduced very much to stay in the small linear range. For a synthesizer some distortion may not be a big problem and amplitude are essentially fixed too.

Today I would consider digital pots to adjust an active filter, at least in the audio range.

[sync]

My goodness why don't you choose the OTA? The jfet thingy is only a distortion generator. Nowhere near a high Q filter! Get yourself a proper OTA. Only that way you can achieve repeatable results with well defined behavior.

I think, we don`t speak about a high-Q filter but on a first-order lowpass only.
Distortion generator? A well-designed combination of an ohmic resistor and a JFET (eventually with a distortion reducing feedback scheme) will give good results.
It is not a problem to limit the voltage between D and S at a level which allows sufficient linearity. 

I should add that it is also possible to reduce distortions if you use two FET`s in combination (parallel or series).

The Q factor of an analog synthesizer VCF is (voltage) controllable. Up to self oscillation. They are usually 12dB/octave (2nd order) or better 24dB/octave (4th order).

Low distortion is not that important. A perfect clean and low distortion synth sounds sterile and boring.

[MAntunes]

As I said, I'm doing this project to learn. I like to explore all the alternatives.
In my electronics classes I learned about FET's, so I was more familiarized with them, so I went with them first.
The OTA seems like a very nice device after I gave it a look. I read the LM13700 datasheet, it has a lot of circuit examples, including a VCF (2 pole). But I know nothing about this device, is there any basic tutorial, explaining how they work?

Another thing, for simulating circuits, what simulator do you use? I am currently using Multisim 10, but it doesn't have a lot of parts and I don't like it very much.. Are there any better alternatives?

[T3sl4co1l]

OTAs fundamentally work because of Ebers-Moll equation:
Ic = Is * exp(Vbe/Vth)

The derivative is the transconductance, and the derivative of an exponential is the exponential again, so
dIc/dVbe == gm = Ic/Vth

The circuit is arranged as a differential pair, which makes it look like an operational amplifier.

Hence, Operational Transconductance Amplifier.

The current into the control pin is literally the collector current of the input differential pair, so the small-signal gain is proportional to that current.

For larger signals, one must account for the physics of the differential pair: it has a tanh(x) function, which is only linear for small values (perhaps 10 or 20mV).  By adding an input dropping resistor and biased diode pair (included in the LM13700), this is somewhat pre-compensated with a complementary current through the diodes.  For the same linearity, input range is extended to about 60mV this way.

Tim

[LvW]

MAntunes - in case you are considering OTA`s it may be of interest that you have two basic options:
(1) OTA replaces opamp: First-order OTA-C lowpass (also called gm-C lowpass). In this case, the bias current Iabc controls the transconductance gm and, hence, the corner frequency. But be aware that you need - in addition - an opamp for buffering purposes because the OTA-C lowpass has no low-resistive filter output 

(2) OTA as resistor: In a classical active first-order lowpass (with opamp) you can realize the first resistor with a FET or you can use an OTA.
This is because an OTA with 100% feeedback (peace of wire from output to inv. input) simulates at the inv. input a grounded resistor R=1/gm (to be controlled with Iabc). 

[MAntunes]

Thank you all for your help.

Another thing, for simulating circuits, what simulator do you use? I am currently using Multisim 10, but it doesn't have a lot of parts and I don't like it very much.. Are there any better alternatives?

And about this, any opinions? Is there any good simulator with lots of parts or that allow the use of libraries of components?

[T3sl4co1l]

Thank you all for your help.

Another thing, for simulating circuits, what simulator do you use? I am currently using Multisim 10, but it doesn't have a lot of parts and I don't like it very much.. Are there any better alternatives?

And about this, any opinions? Is there any good simulator with lots of parts or that allow the use of libraries of components?

You mentioned Multisim, surely it has extensive libraries?

I don't think I've seen any other EDA package with libraries as vast, and this with almost full simulation data (if not always the best models) as well as schematic and footprint data (if paired with Ultiboard).  In my opinion, Multisim is quite good at simulation, considering its general quirkiness and price range.

Tim

[MAntunes]

I had an older version of Multisim, but now I updated it and it has a lot of new features and a lot of new components.
I hope it works well, thank you for your help!

[MAntunes]

Hello again,
after searching a bit more I came across this paper: https://camilotejeiro.files.wordpress.com/2012/12/voltage-controlled-low-pass-filter-analog-design.pdf and some other books.
I came across this 2nd order filter :



With this filter I can vary the cutoff frequency without messing with its characteristics. I already have the transfer function for this one calculated.
Then I added resistor dividers to all the imputs of the OTA to maintain Vd < 20mV, for linearity reasons.



I also have the transfer function for this one calculated.
When I was transitioning for the final schematic the problems started to come. The LT13700 has output buffers and I want to use them. This is the circuit for the filter in the LT13700 datasheet:



In the paper mentioned above, the final circuit was this one:



I don't understand how to to make this this final transition. First, the capacitors were connected to ground, but now they are not.
Thank you for your help until now and I hope you can help me with this one too.

[dentaku]

This document is quite useful and has a bunch of handy OTA circuits.
I've built a VCA with an LM13700 but not a VCF yet.
http://www.nutsvolts.com/uploads/magazine_downloads/11/May%202003%20Ray%20Marston%20-%20Understanding%20And%20Using%20OTA%20OP-Amps.pdf

You should check out the Synth DIY Facebook group.

[iampoor]

Another great one...

http://www.schmitzbits.de/ms20.html

[Yansi]

...

 This is the circuit for the filter in the LT13700 datasheet:



...

I don't understand how to to make this this final transition. First, the capacitors were connected to ground, but now they are not.
Thank you for your help until now and I hope you can help me with this one too.

Because it is a 2nd order filter, equivalent to this Sallen-Key filter: https://upload.wikimedia.org/wikipedia/commons/thumb/3/3f/Sallen-Key_Lowpass_General.svg/400px-Sallen-Key_Lowpass_General.svg.png

The first one with both caps to ground is "just two 1st order filters one after another".

I hope I have understood you question right.

[crash2140]

I agree with iampoor. MFOS has tons of info about building analog synths. Even though the circuits might seem complex at first glance, the complexity is necessary. If you're planning to build a synth (which I am as well), and not a single VCF, it needs to have good tracking, thermal stability, nice sound, etc.
As to simulation, I use LTSpice and it works well for me. Multisim pisses me off with its "convergence errors".

 PS: Nice to know you are from Portugal! As it seems I am not the only one here interested in synths

[magetoo]

Might be interesting to look at the lectures of the "Electronics for music synthesis" course if you want a bit of theory and not just cookbook circuits with no explanation.  (Video links on the course page are broken, use the first link.)

[MAntunes]

Because it is a 2nd order filter, equivalent to this Sallen-Key filter: https://upload.wikimedia.org/wikipedia/commons/thumb/3/3f/Sallen-Key_Lowpass_General.svg/400px-Sallen-Key_Lowpass_General.svg.png

The first one with both caps to ground is "just two 1st order filters one after another".

I hope I have understood you question right.

Thank you! Will give a look to that and try to do the analysis of the final circuit!

I agree with iampoor. MFOS has tons of info about building analog synths. Even though the circuits might seem complex at first glance, the complexity is necessary. If you're planning to build a synth (which I am as well), and not a single VCF, it needs to have good tracking, thermal stability, nice sound, etc.
As to simulation, I use LTSpice and it works well for me. Multisim pisses me off with its "convergence errors".

 PS: Nice to know you are from Portugal! As it seems I am not the only one here interested in synths

Yes, I am trying to build a synth, something simple as I have to start somewhere. I think it is a great project for one who likes analog electronics as it involves oscillators, filters, amplifiers, etc. Right now, I only know the basic building blocks of electronics I learned in classes so I will tackle the problems as they come up!
I'll give a look to LTSpice, does it have a nice library of components?
Nice to know you are Portuguese too, will send you a message some day!

Might be interesting to look at the lectures of the "Electronics for music synthesis" course if you want a bit of theory and not just cookbook circuits with no explanation.  (Video links on the course page are broken, use the first link.)

Thank you! I will give a look to those links! They seem very interesting!

[crash2140]

I'll give a look to LTSpice, does it have a nice library of components?

It has a fair amount of components , but it you can always add SPICE models of other components. Something it lacks is potentiometers. In Multisim you can add potentiometers and change their rotation in real time, which is really handy, while in LTspice you have to change the value of the resistance and simulate again. There's a library that adds them, but it was kinda confusing for me to use and I never got around to try it again.

Edit: and LTspice is free, by the way.