Puzzle: one ON-OFF-ON switch and some caps

Hi,

I've been thinking about a topology with an ON-OFF-ON switch and some capacitors for switching the cutoff frequency of a filter.

The arrangement with the smallest resulting cap-value in the middle position is a no-brainer, but obviously one would like to have a smallest-middle-highest sequence for the values.

So a small puzzle: which topology could realize an increasing cap-value between two terminals ?

I'm actually not sure it has a solution, since closing a contact increases the total cap-value both for (1) adding another cap in parallel as well as for (2) shorting a series-cap. But please surprise me !


The Rules:

The switch is a single pole, three positions, middle position keeps the pole floating (ON-OFF-ON) and you can use any amount of caps.

Negative cap-values, switch driven relays, balanced cross-connections etc etc are out.


Bye,

Peter

I had the same problem with the lowcut of my channel strip. I thought of it day and night an ended up with two relays.
Maybe a mathematic guy can bring a final argument if it is even possible or not.

I readily admit that it's just a way of smuggling in negative capacitance, but the relatively easy way would be to simply switch among gains of an amplifier, for example, -1, 0, and +1. A capacitor of value C would be equivalently grounded by being tied to the low-Z output of the switched-gain amp. A high-Z buffer would look at the free end of the C and provide a buffered input signal to the switched-gain amp.

At -1 gain you double C, with 0 gain it's just C, and with +1 it's zero. Other gains could give a wider range. You put another cap in the circuit to common if zero or negative C is not usable.

I have seen ostensibly DPDT toggle switches that are actually configurable to realize a 1P3T function, if you must have a toggle. That would be the simplest of all.

Exactly what an impossibility proof would amount to is hard to say. I suppose you could do it a la the four-color theorem proof and just exhaust all of the topological possibilities. You have quite a few even with this simple a set of contacts, capacitors, and drive points. Once you add capacitors beyond a certain number it would become clear that more of them didn't change things qualitatively. If you exclude floating sources [EDIT: and transformers] (and it sounds as if you have) it would trim things down a bit.

Thanks both for chiming in :thumb:

Yes, the DPDT-route Brad suggested would be doing it and still keep things simple.... but it's just that the 'context' of this is a 2-pole LPF (or HPF) circuit with variable cutoff-frequency. So to be able to switch the caps for 3 different freq-ranges would require a 4PDT (with center-OFF), which does exist, but it less common.

I must admit that two ranges will be enough for that circuit (so a '2 position DPDT-switch' can do it) but then the curious side popped up. As in: I might not be really needing it, but if I did, how could it be done ?


I made a overdrive-circuit with switchable corner-frequencies for the following filter years ago. It was there & then that the issue popped up
and when running into it again this week I started wondering again about how it could be done...
... but without sneaky negative values or additional circuitry it doesn't seem possible...

Bye, thanks & enjoy,

Peter

Actually I should have said a way of sneaking in a cap multiplier, not necessarily negative capacitance. But for a substantial range change it would mean one position was yielding negative C.

[quote author="clintrubber"]Hi,

I've been thinking about a topology with an ON-OFF-ON switch and some capacitors for switching the cutoff frequency of a filter.

The arrangement with the smallest resulting cap-value in the middle position is a no-brainer, but obviously one would like to have a smallest-middle-highest sequence for the values.

So a small puzzle: which topology could realize an increasing cap-value between two terminals ?

I'm actually not sure it has a solution, since closing a contact increases the total cap-value both for (1) adding another cap in parallel as well as for (2) shorting a series-cap. But please surprise me !


The Rules:

The switch is a single pole, three positions, middle position keeps the pole floating (ON-OFF-ON) and you can use any amount of caps.

Negative cap-values, switch driven relays, balanced cross-connections etc etc are out.


Bye,

Peter[/quote]
A not very elegant solution would be to build 3 independent filter sections. Feed a buffer connected to the wiper from the intermediate filter through a compliance, perhaps a few Kohms. When the switch is in either extreme position it is hard connected to the low or high section shunting the mid.

One minor benefit is this is that it accommodates a wide range of filter topologies. In a real production situation the temptation is to tool the right switch for the application.

JR

Like bcarso mentioned, 3 position mini toggle switches are readily available that will do what you want. They are called on-on-on, and I think you can find them easily enough at Mouser et al. That would be the simplest solution.

> I'm actually not sure it has a solution

It doesn't and it does.

As defined, cap-switching, the only answer is the negative cap.

Why are you switching caps?? Resistors affect frequency also. I'm too lazy to look at it now, but one way adds another cap to go down, the other way shunts a resistor with another resistor to go up.

Less parts than John's three filters, less brain-strain than bcarso's negative caps. And while on-on-on toggles exist, recently I couldn't find one to match my other switches.

while in this situation it is (probably) not that relevant ,
I prefer switching caps in filters and EQs cus
load on active elements remain constant
(unlike classic console SVF implementation of EQ
where on high end of sweep opamps outputs
are tormented by few kohms or less)

cheerz
urosh

[quote author="PRR"]Why are you switching caps?? Resistors affect frequency also. I'm too lazy to look at it now, but one way adds another cap to go down, the other way shunts a resistor with another resistor to go up. [/quote]
I've been thinking of that and for fixed frequency filters it should most likely be possible to make that working.
But the circuit at hand is for variable frequencies, with a topology as this one, so I'd like to keep using the same dual (rev)log pot (used as a rheostat).

As said, three ranges will not really be needed in practice, but I appreciate it a lot that you all responded to this little puzzle, be it out of curiousity, brain-exercise or whatever reason.

Thanks !

Peter

[quote author="JohnRoberts"]In a real production situation the temptation is to tool the right switch for the application.

JR[/quote]
Could understand that. W.r.t. the right switch, for DIY a 2P3T rotary could do the job (for a complete 2-pole filter).

Front panel space is limited so the toggle seemed attractive, so a DPDT for two ranges will be used in the actual circuit.

Regards,

Peter

[quote author="David Kulka"]Like bcarso mentioned, 3 position mini toggle switches are readily available that will do what you want. They are called on-on-on, and I think you can find them easily enough at Mouser et al. That would be the simplest solution.[/quote]
That's good news. In the center position, do these connect the two outer 'non-poles' or do these have an additional terminal ?

Thanks,

Peter

[quote author="recnsci"]while in this situation it is (probably) not that relevant ,
I prefer switching caps in filters and EQs cus
load on active elements remain constant
(unlike classic console SVF implementation of EQ
where on high end of sweep opamps outputs
are tormented by few kohms or less)

cheerz
urosh[/quote]
Valid concern indeed ! The circuits I've seen have a clear correlation between the small valued resistor (that's in series with the rheostat) and the used opamp. TL07X not below 2k2...2k7, and for the 5532 I saw 620 Ohms.

Bye,

Peter

[quote author="clintrubber"][quote author="David Kulka"]
Like bcarso mentioned, 3 position mini toggle switches are readily available that will do what you want. They are called on-on-on, and I think you can find them easily enough at Mouser et al. That would be the simplest solution.
[/quote]
That's good news. In the center position, do these connect the two outer 'non-poles' or do these have an additional terminal ? [/quote]

very interesting discussion - I wasn't aware that there are three position on-on-on switches. I have always wanted to have an addtional -10 dB position on all my -20 dB preamp pads, without having to use a rotary switch. Sorry for not being of any help...

cheers, Marten

[quote author="clintrubber"][quote author="PRR"]Why are you switching caps?? Resistors affect frequency also. I'm too lazy to look at it now, but one way adds another cap to go down, the other way shunts a resistor with another resistor to go up. [/quote]
I've been thinking of that and for fixed frequency filters it should most likely be possible to make that working.
But the circuit at hand is for variable frequencies, with a topology as this one, so I'd like to keep using the same dual (rev)log pot (used as a rheostat).

As said, three ranges will not really be needed in practice, but I appreciate it a lot that you all responded to this little puzzle, be it out of curiousity, brain-exercise or whatever reason.

Thanks !

Peter[/quote]

I've never attempted more than 2 ranges for switching variable EQ sections. As that often involved several poles. Pot tracking between gangs also becomes an issue over too wide of an adjustment per range as it can mess with Q.

JR

[quote author="clintrubber"][quote author="David Kulka"]Like bcarso mentioned, 3 position mini toggle switches are readily available that will do what you want. They are called on-on-on, and I think you can find them easily enough at Mouser et al. That would be the simplest solution.[/quote]
That's good news. In the center position, do these connect the two outer 'non-poles' or do these have an additional terminal ?

Thanks,

Peter[/quote]

They look like DPDT switches. You usually have to make an external connection for the wiper function IIRC.

See the bottom page 18 of this handy guide from Eaton:

http://www.eatonelectrical.com/unsecure/cms1/SWITCHTRAININGMANUAL.PDF

I had one once when I didn't know they existed. I didn't know what it was, and tried to use it as a DPDT with center off, which was what it looked like. Boy was that a puzzle for a while!

EDIT: And see the bottom of page 4 of this

http://www.nkkswitches.com/pdf/MtogglesBushing.pdf

which I accessed out of Mouser online. (EDIT again---they don't seem to be stocked). Note that with the one on the right you get your desired 2P3T function for two-pole-pair band switching.

Thanks Brad, these pdfs explain it well :thumb:

Didn't knew these existed either. The idea of a DPDT with one pole already going while the other pole stays where it was (as on that page 4 of http://www.nkkswitches.com/pdf/MtogglesBushing.pdf ) was new to me.

Regards,

Peter

[quote author="JohnRoberts"]I've never attempted more than 2 ranges for switching variable EQ sections. As that often involved several poles. [/quote]

You're right, more than 2 would be overdoing.

Pot tracking between gangs also becomes an issue over too wide of an adjustment per range as it can mess with Q.

JR

Click to expand...

How do you mean ?
Q determined by the (fixed) opamp feedback-resistors. But I could understand that too big a mismatch for the R,R,C&C-values will start to mess with the shape of the filter-transfer.

As seen for some circuits (Calrec), a slight adjustment to one of the caps could be done.
So I guess the mismatch of the dual pot could be compensated for but it might be triggering new hassle: the small series resistance in series with the rheostat needs to be scaled then as well.
In the end it may all be overkill or still not getting you there. For now I want to avoid this hassle; the pots I have match quite well.

Regards,

Peter

[quote author="clintrubber"]

Pot tracking between gangs also becomes an issue over too wide of an adjustment per range as it can mess with Q.

JR

Click to expand...

How do you mean ?
Q determined by the (fixed) opamp feedback-resistors. But I could understand that too big a mismatch for the R,R,C&C-values will start to mess with the shape of the filter-transfer.

As seen for some circuits (Calrec), a slight adjustment to one of the caps could be done.
So I guess the mismatch of the dual pot could be compensated for but it might be triggering new hassle: the small series resistance in series with the rheostat needs to be scaled then as well.
In the end it may all be overkill or still not getting you there. For now I want to avoid this hassle; the pots I have match quite well.

Regards,

Peter[/quote]
My worst experience was with a 4 ganged pot in an adjustable Linkwitz-Riley crossover. Pots are typically 20% tolerance for bulk resistance and somewhat better than that for rotational tracking but still bad enough to cause issues with ganged filter sections.

One helpful approach when using them in something like a state variable is to configure them so their end points of adjustment are independent of bulk resistance. So instead of a 10K pot with a 1k end limit resistor to deliver a 10:1 adjustment range use no end limit resistor so wiper goes all the way to ground and parallel the wiper resistor with something like a 9X R from the top of the pot. Now the endpoints are the 5% resistor tolerance or better not 20%. I still had a technician measure the pots at 50% rotation and add a shunt resistor as needed to make the 4 sections match at mid rotation as well as both ends.

Yet another reason why DSP rocks....

JR

[quote author="JohnRoberts"]
One helpful approach when using them in something like a state variable is to configure them so their end points of adjustment are independent of bulk resistance. So instead of a 10K pot with a 1k end limit resistor to deliver a 10:1 adjustment range use no end limit resistor so wiper goes all the way to ground and parallel the wiper resistor with something like a 9X R from the top of the pot. Now the endpoints are the 5% resistor tolerance or better not 20%.

JR[/quote]

I can't quite visualize that. Are you talking about a Kerwin-Huelsman SVF with two inverting integrators?

For that topology I used this arrangement: have a fixed series R, call it R1, to the inverting opamp input with feedback C. Hang the pot from the inverting amp input to common/noninverting opamp input. Have resistor R2 from the pot wiper to the driven end of R1.

Now with pot at max ccw you have exactly R1 as your integrator R. With pot at max cw you have the parallel value of R1 and R2.

Slight disadvantages: the noise gain is higher a bit because the pot end-to-end R is hanging there all the time. Also your previous stage is driving a heavier load at the low frequency endpoint than it would with a simple series R plus variable R.