Simple voltage controlled low pass filter

[markusw]

Hey all,

for my PLL RingMod-Synth project (http://www.diystompboxes.com/smfforum/index.php?topic=55075.0) I would need a simple 1-pole voltage controlled low pass filter to reduce the 15th and 17th harmonics produced by the sine generator. Sine I plan to control it via the 4046's control voltage it should be linear.

Any ideas would be highly appreciated 

Thanks a lot in advance!

Markus

[R.G.]

You're almost ideally set up for a switched capacitor filter.

Almost.

Since your sine is generated by a frequency-locked high frequency clock, you could just use the high frequency clock as the input to a switched capacitor filter to do the filtering. The only reason this won't work well as is is that the clock that you generate to make the sine is not high enough frequency. If you could rearrange your PLL to make sure the lowest frequency it generates at the lowest frequency sine to be generated is, say, 44kHz, then you could run that directly into the switched cap filter.


The simple option is to use an OTA as a variable resistor in a one-R, one-C lowpass. You buffer the control voltage, feed it to the Iabc of the OTA, then use the OTA output to feed a cap to ground. The output is at the cap.

Since the only production OTAs are duals now, might as well go for two-pole and get better rejection. LM13700 works.

[DDD]

IMO single OTAs are in propduction now as well as the dual ones. In any case there's a lot of LM3080 (not CA3080, but LM3080) on the local market here in Kazakhstan. And they seem to be produced not so long ago.

[markusw]

Thanks a lot for your help R.G.! 

Since your sine is generated by a frequency-locked high frequency clock, you could just use the high frequency clock as the input to a switched capacitor filter to do the filtering. The only reason this won't work well as is is that the clock that you generate to make the sine is not high enough frequency. If you could rearrange your PLL to make sure the lowest frequency it generates at the lowest frequency sine to be generated is, say, 44kHz, then you could run that directly into the switched cap filter.

Actually, it would require just another divider in the PLL loop to get to around  50 or 100 times the cutoff freq of the filter. For the highest notes in the guitar range I'm currently running the PLL at about 200-300 kHz. So, to get to let's say 100-times the cutoff freq (with an adiitional divide by 8 ) the PLL would run at about 2 MHz. Would 50-times the Fc be sufficient?  Are there any switched cap filter ICs you could recommend?

The simple option is to use an OTA as a variable resistor in a one-R, one-C lowpass. You buffer the control voltage, feed it to the Iabc of the OTA, then use the OTA output to feed a cap to ground. The output is at the cap.

Since the only production OTAs are duals now, might as well go for two-pole and get better rejection. LM13700 works.

Another Q:
I found that 3 or 4 octaves down (see http://www.aronnelson.com/DIYFiles/up/PLL_AMS100_PD_NE555_current_sch_31-03-07.pdf ) gives a nice tremoloesque sound when used as a carrier for the RingMod. So I will need a 12-position rotary since a 8-postion isn't available. Since the 4046's CV doesn't change with the octave selector switch I had to use a different cap for each octave (with the OTA approch). Even if I had a 2P8T rotary I could switch just one cap in parallel to the octave selctor. Therefore, I'd prefer a single-pole low pass filter because it then would take only two rotary switches.


BTW, do you see any chance to switch the caps in parallel with the 4040 using only one 1P12T rotary. Maybe with two 4066 and some other parts? 

IMO single OTAs are in propduction now as well as the dual ones. In any case there's a lot of LM3080 (not CA3080, but LM3080) on the local market here in Kazakhstan. And they seem to be produced not so long ago.

In general, I'm not sure whether using a OTA is a really good choice noise-wise. As fas as I understood the audio signal has to be pretty low for OTA's. For the RingMod I want a pretty strong signal. Therefore, I fear that it would introduce quite a lot of noise. What do you think?


Hey DDD,

thanks for the tip!  I will have a look for them 


Regards,

Markus

[Mark Hammer]

Doesn't get a whole lot simpler than this: http://filters.muziq.be/files/schematics/paia_2720-3l.gif

[R.G.]

Actually, it would require just another divider in the PLL loop to get to around  50 or 100 times the cutoff freq of the filter. For the highest notes in the guitar range I'm currently running the PLL at about 200-300 kHz. So, to get to let's say 100-times the cutoff freq (with an adiitional divide by 8 ) the PLL would run at about 2 MHz. Would 50-times the Fc be sufficient?  Are there any switched cap filter ICs you could recommend?

I started to whip in a quick reply, then I went off and looked for chips.

Bottom line - you want the TI  TLC04/MF4A-50. You want it in the plastic DIP package, but Digikey only stocks it in the surface mount SO package. Here's why you want that one.
1. You can get it. The other candidates are the National LMF-100, Maxim 29x series, Maxim 74xx series, Linear Technology universal active filter building blocks, a few others. They are quite hard to find.
2. You can afford it. The available Maxim, LT, and other filters are usually over $10. There are LT's for as low as $6 a pop. The TI is $2.25 in ones. The LMF-100 is $6 and change.

I'd say find a place that stocks the plastic DIP of the TLC04.

You want the -50 version, which makes the cutoff at 1/50th the clock frequency. Don't put the cutoff right at your sine. Put it at 2x or 4x your sine frequency. There's a long way from there to the 15th harmonic, so you have room. So use the 50:1 part and run it at a clock of 100x the sine you're wanting out. Make sure the lowest clock you use, lowest sine frequency, is over 44kHz to keep clock gook out of the signal lines.

Another Q:
I found that 3 or 4 octaves down (see http://www.aronnelson.com/DIYFiles/up/PLL_AMS100_PD_NE555_current_sch_31-03-07.pdf ) gives a nice tremoloesque sound when used as a carrier for the RingMod. So I will need a 12-position rotary since a 8-postion isn't available. Since the 4046's CV doesn't change with the octave selector switch I had to use a different cap for each octave (with the OTA approch). Even if I had a 2P8T rotary I could switch just one cap in parallel to the octave selctor. Therefore, I'd prefer a single-pole low pass filter because it then would take only two rotary switches.
BTW, do you see any chance to switch the caps in parallel with the 4040 using only one 1P12T rotary. Maybe with two 4066 and some other parts?

Welcome to the mixed signal world. You have available to you 1:8 analog switches in the CD4051. You can feed two 4051's the same control bits and have a 2P8T. Or a 5p8T. Just keep adding 4051's. You can switch as many things as you like. 1P12T rotary switches are available, and they usually have adjustable stops from 2 to 12 positions. Or you could use a CMOS up/down counter IC with up/down inputs and have an up-down counter. Use another 4051 to select one of 8 LEDs or one of eight LED brightnesses, or a CMOS encode/decode/7segment driver to get it to a single digit display. The problem is not how, it's which one of the many ways.

In general, I'm not sure whether using a OTA is a really good choice noise-wise. As fas as I understood the audio signal has to be pretty low for OTA's. For the RingMod I want a pretty strong signal. Therefore, I fear that it would introduce quite a lot of noise. What do you think?

"Quite a lot" is subjective. Maybe. Maybe not. There will be some. Whether it's too much is up for question.

[markusw]

Doesn't get a whole lot simpler than this: http://filters.muziq.be/files/schematics/paia_2720-3l.gif

Thanks for the tip Mark! 
I already had a look at this one but it seems that you can't control the cutoff frequency with the CV but only the gain above the cutoff.
At least according to Spice sims  It's also mentioned at http://filters.muziq.be/model/paia/2720/3l
"this VCF has a fixed frequency, the controls can only modify how much of the frequencies about the cut-off frequency (500Hz) is removed."

Thanks again for your help R.G. ! 

At a first glance the cheapest switched cap filter chip I could find here in Austria is the MF10 for about 4 Euro (should be about 4,50 US$).
Will check if I can get the TLC04.....

You want the -50 version, which makes the cutoff at 1/50th the clock frequency. Don't put the cutoff right at your sine. Put it at 2x or 4x your sine frequency. There's a long way from there to the 15th harmonic, so you have room. So use the 50:1 part and run it at a clock of 100x the sine you're wanting out. Make sure the lowest clock you use, lowest sine frequency, is over 44kHz to keep clock gook out of the signal lines.

Thanks for the tips! There is one thing I don't get.    Assuming 4 octaves down, the low E will be about 5 Hz. So the ratio is about 1:10000 for 44 kHz. But I need a ratio of 1:100??
For the high guitar notes above 1000 Hz the clock frequency  would be around 10 MHz.

Welcome to the mixed signal world. You have available to you 1:8 analog switches in the CD4051. You can feed two 4051's the same control bits and have a 2P8T. Or a 5p8T. Just keep adding 4051's. You can switch as many things as you like. 1P12T rotary switches are available, and they usually have adjustable stops from 2 to 12 positions. Or you could use a CMOS up/down counter IC with up/down inputs and have an up-down counter. Use another 4051 to select one of 8 LEDs or one of eight LED brightnesses, or a CMOS encode/decode/7segment driver to get it to a single digit display. The problem is not how, it's which one of the many ways.

Thanks. Now the problem isn't anymore "HOW" for me 

"Quite a lot" is subjective. Maybe. Maybe not. There will be some. Whether it's too much is up for question.

I can easily check this on breadboard....

Regards,

Markus

[R.G.]

At a first glance the cheapest switched cap filter chip I could find here in Austria is the MF10 for about 4 Euro (should be about 4,50 US$).
Will check if I can get the TLC04.....

If you can get an MF-10 for E4,00, that's a good choice too. I didn't find a source for the MF-10.

There is one thing I don't get.   Assuming 4 octaves down, the low E will be about 5 Hz. So the ratio is about 1:10000 for 44 kHz. But I need a ratio of 1:100?? For the high guitar notes above 1000 Hz the clock frequency  would be around 10 MHz.

Actually, it was something I didn't get. If you're making sines for the four octaves starting at 5 Hz, you're going to want to use a 100:1 filter. I somehow got it into my head that you were tracking the basic guitar frequency.

I was advising on the clock multiple based on not wanting to have to have the 4046 running in the multi-MHz on the high end as that gets layout-critical, and not wanting to dip under twice audio to keep any clock junk out of the audio band. But you can't do that and still get to 5 Hz.

So you're going to have to use the switched capacitor to clean up the low sine, but an analog filter to clean up any clock remnants that still happen in the audio path. For instance: if you are producing a 5Hz sine, the clock is running at 500Hz, and it's running the full power supply at 500Hz in a harmonic-rich square wave. Actually, you'd want the low cutoff down at 2.5Hz to avoid starting to lose the sine you're trying to pass, so the clock would would be running at 250Hz, and that will be critical to keep out of your signal path.

[Paul Perry (Frostwave)]

In general, I'm not sure whether using a OTA is a really good choice noise-wise. As fas as I understood the audio signal has to be pretty low for OTA's. For the RingMod I want a pretty strong signal. Therefore, I fear that it would introduce quite a lot of noise. What do you think?

I don't know exactly how pure a sine wave needs to be for a ring modulator. I use a triangle that has been turned into a sine by overdriving an OTA (this gets you to within 2% or so accuracy). Considering what is likely to be on the otehr input to the ring modulator, I think that would be plenty good enough.
The main problem with a ring modulator is getting it balanced accurately so that when there is no audio in, there is no carrier leaking through. That is a much tougher job than the purity of the sine, and more important too, because it is so much easier to notice.
I think the inherent noise in the multiplier core is likely to be of the same order as that of an OTA (same technology, really). It is vital to get the inputs to the multiplier as high as possible without clipping for best signal to noise.
Something else worth watching is that if there is any mains hum on the audio input line, then that generates sidebands around the carrier frequency, and when the audio is quiet, the carrier + and - 50 or 60 Hz is much more intrusive than just the hum would have been.

[markusw]

So you're going to have to use the switched capacitor to clean up the low sine, but an analog filter to clean up any clock remnants that still happen in the audio path. For instance: if you are producing a 5Hz sine, the clock is running at 500Hz, and it's running the full power supply at 500Hz in a harmonic-rich square wave. Actually, you'd want the low cutoff down at 2.5Hz to avoid starting to lose the sine you're trying to pass, so the clock would would be running at 250Hz, and that will be critical to keep out of your signal path.

Thanks again R.G. !
So it would probably be easier to use an OTA and risk the noise (if any)........I will give the OTA a try and see if it's OK..

Hey Paul,

thanks for your help!

I don't know exactly how pure a sine wave needs to be for a ring modulator. I use a triangle that has been turned into a sine by overdriving an OTA (this gets you to within 2% or so accuracy). Considering what is likely to be on the otehr input to the ring modulator, I think that would be plenty good enough.

Your absolutely right! For the ring modulator sine purity is not that critical. A bit of low pass filtering (even with pretty high cutoff frequency) astonishingly is already OK.
Alternatively, I want to use the approximated sine signal with an envelope derived from the audio signal. At the moment this is done with a LED/LDR combo which controlled by the rectified audio signal (derived from the NE571). Unfortunately, for this purpose the 15th and 17th harmonics are much more noticeable than for the Ring modulator.
That's the main reason I want to have a pure sine. The high harmonics are a bit annoying with the envelope...

I think the inherent noise in the multiplier core is likely to be of the same order as that of an OTA (same technology, really). It is vital to get the inputs to the multiplier as high as possible without clipping for best signal to noise.
Something else worth watching is that if there is any mains hum on the audio input line, then that generates sidebands around the carrier frequency, and when the audio is quiet, the carrier + and - 50 or 60 Hz is much more intrusive than just the hum would have been.

Currently. I'm using a 2-xfmr diode ring modulator (I know you prefer the balanced modulator chips  ) with a schottky diode array instead of the dicrete diodes  and carrier bleed through is not noticeable. Will have to check at higher volumes though.
Anyway, I think I will give the OTA VCF a try on breadboard.  I have a LM13700 at home, so there is nothing to loose....
Will also take care about mains hum. Thanks for the tip!! 

Regards,

Markus

[DDD]

As for me, the 3080 OTAs are not too noisy at all.

[R.G.]

You know, a single LM13700 would make for a nice voltage controlled low pass filter, especially if you did it as a state variable filter and used the lowpass output. I believe that the frequency control is linear with Iabc, not exponential, so it would fit with the control voltage on the PLL pretty well.

[markusw]

As for me, the 3080 OTAs are not too noisy at all.

Good to know!  So the 13700 should be fine too....

You know, a single LM13700 would make for a nice voltage controlled low pass filter, especially if you did it as a state variable filter and used the lowpass output. I believe that the frequency control is linear with Iabc, not exponential, so it would fit with the control voltage on the PLL pretty well.

I did a simulation for a VCA using a LM13700 and response was linear. So I suppose cutoff frequency should be too..
The state variable filter would be cool. Will have a look at the datasheet. IIRC there was a schem.
Alternatively, I could use the second half of the 13700 instead of the LED/LDR combo for gain control via the envelope.
Some more breadboarding 

Markus

[markusw]

One more Q:  does the CV from the 4046 need to be buffered before fed into the LM13700? The 137000 shouldn't draw that much current anyway....

[grapefruit]

From my experience with the LM13600/13700 you're best off using a voltage controlled currrent source for Iabc. It was a while ago but from what I remember you can get more range if you use that method. Not sure if it's in the datasheet, but the circuit used for the PAIA Fatman works well, and is linear.

Cheers,
Stew.

[markusw]

From my experience with the LM13600/13700 you're best off using a voltage controlled currrent source for Iabc. It was a while ago but from what I remember you can get more range if you use that method. Not sure if it's in the datasheet, but the circuit used for the PAIA Fatman works well, and is linear.

Cheers,
Stew.

Thanks a lot for your help Stew! Could you target me to a schem? 

Markus

[grapefruit]

http://www.paia.com/manuals/docs/9308-fatman-manual.pdf

On page 6 IC13 and Q8, Q9 form the VCCS for the LM13600 (IC13).

Cheers,
Stew.

[markusw]

http://www.paia.com/manuals/docs/9308-fatman-manual.pdf

On page 6 IC13 and Q8, Q9 form the VCCS for the LM13600 (IC13).

Cheers,
Stew.

Thanks a lot!!!  Will give it a try on breadboard for sure...

Regards,

Markus

[markusw]

An update:
Yesterday eve I added the LM13700 voltage controlled low pass as shown in the data sheet on breadboard. Amplifier bias voltage is taken from the output of the LF398. Compared to the sims the cap needs to be a bit larger.
Response is not perfectly linear judged from scoping since at lower control voltage there is a small drop in signal amplitude for about one octave (maybe 20%) but then it stays constant. Anyway, in my applications it's not noticeable . 
THD for the filtered sine is about 0.5%. Noise from the LM13700 is definitely not an issue!

Don't know yet what to do with the second half of the LM13700. Two pole low pass is definitely not necessary. Maybe a VCA......?

Thanks for your help!!!

Will try do do an update for the schem within the next days....

Regards,

Markus

[StephenGiles]

Space Drum sweep filter perhaps, my circuit is here somewhere......