4066 buffered or unbuffered?

[bioroids]

Hi, how are you?

I've been bulding a few switches using the 4066 CMOS switch (buffered version). I was wondering should I be using the unbuffered version? The ones I got sound great but it's hard to remove the pop when switching.

Should I be using 4016? I think the 4016 can switch slowly (depending on the control signal) and the 4066 only hard switches? Am I wrong?

Any tip or info is welcomed

Thanks ahead

Miguel

[bioroids]

Anyone?

Thanks

Miguel

[Peter Snowberg]

Very interesting .

I don't know if one switches more "digitally" than the other, but from the Philips datasheet you can see the internal schematic. It shows the substrate of the N channel MOSFET being switched between one of the inputs (the one connected to the drain) and ground. This is the secret of the lowered on resistance I'm sure.

I would guess from this that the 4016 will work more like a pot, but I really don't know.

Here are the Philips data sheets in any case:
http://www.philipslogic.com/products/hef/pdf/hef4016b.pdf
http://www.philipslogic.com/products/hef/pdf/hef4066b.pdf

I would try it out and see what happens. The chip is not made to be used this way and the operation will probably be different between brands.

If you are getting pops, do you have DC on the switch?

I think if your inputs and outputs are capacitor coupled, and you bias the pins of the chip to 1/2 of Vdd, you should get a minimal amount of pop.


Good luck,
-Peter

[The Tone God]

Hi, how are you?

Tired but fine otherwise.

I've been bulding a few switches using the 4066 CMOS switch (buffered version). I was wondering should I be using the unbuffered version? The ones I got sound great but it's hard to remove the pop when switching. Should I be using 4016? I think the 4016 can switch slowly (depending on the control signal) and the 4066 only hard switches? Am I wrong?

Check out the "4016 vs. 4066" section in the Wicked Switches article at the site to answer these questions.

http://www.geocities.com/thetonegod/

Andrew

[bioroids]

Hi there

Nice info, i'm gonna breadboard a little test circuit to see what happens, now that I bought a 4016.

Let's hope the audio does not suffer much on the 4016, maybe the fidelity is good enough for guitar signals.

I'll post any results later.

Good luck

Miguel

[gez]

Let's hope the audio does not suffer much on the 4016, maybe the fidelity is good enough for guitar signals.

I recently messed around with both chips and had a similar problem (but that was my own fault, I was trying to be clever!).  I can't say I noticed any deterioration in quality when using the 4016.

Here's some stuff that I've collected in my 'switching' file over the past few months.  I'm pretty sure the author for all of it (various posts) was RG:

"There are two things that cause pops with CMOS switches that are relatively unavoidable, and some things that can be sidestepped.

1) Control signal feedthrough; the very fact that you're putting a fast-rising edge into the chip lets that edge get coupled through the tiny capacitances inside to the audio path. Some chips (the 4016) have one of the two biphase drivers available and you can slow that one down with an RC. The 4066 and other swtiches have all internal buffers and inverters on the control signal and you can't slow this down. Newer designed chips are better.

In most of these chips, if you have the DC level of the signal sitting at half the DC power supply level to the chip, the control signal feedthrough from the Pchannel and Nchannel sides semi-cancel. So always DC bias the inputs and outputs of the switch chip to half it's logic voltage. That means NOT DC coupling the input signals. Usually it's OK to use 1uF caps and 100K resistors to the bias voltage on each in/out.

2) Fractional signal waves. When you flip the control signal, the switch changes in about 100nS. This is instantaneous compared to audio, so if the signal happens to be near the peak of a wave, the wave is sliced in half time-wise. That sudden change from an instantaneous level to none is audible as a click under most circumstances. This is unavoidable with hard-switching. Even mechanical switches do this. Only a variable gain block fading the signal up and down avoids this.

3)Switching DC levels; this is entirely avoidable, as in 1).

See the article at GEO on bypassing with the CD4053 for more discussion of CMOS switching.

CD4053:

I've used that circuit in a number of things. It does work. I can't take credit for originating it - it's what bypasses most of the Dano Burned Toast/Cheezburger/Yellow Biscuit run of pedals, as well as having been used in some earlier pedals.

The circuit it bypasses can make a difference. Some circuits need you to reduce the half-meg or 1M bias resistors to minimize pop. Going as low as 100K has straightened out even hard cases for me.

The thing you're (maybe) fighting is charge injection. Whenever you move the gate of a MOSFET around, the voltage change on the gate is transferred to the conducting channel by the gate-channel capacitances. The industry calls this charge injection, as charge is deposited or removed from the channel. N-channel and P-channel devices tend to cancel one another out, and this cancellation is best if they are both biased at exactly half the chip's supply voltage. Hence the bias to half the battery.

There is some residual charge, and this has to leak out through the bias resistors. Different brands of chips have slightly different results, but most of them work with no particularly noticeable pop.  

CMOS switches cannot be completely prevented from popping.

Neither can any other non-optically coupled electronic switch. What you can do is make the switching transient inaudible. That is quite possible.

Whenever you switch a CMOS device (or a JFET, or a bipolar, or a relay for that matter) the unwanted small capacitances in the device couple the switching voltage into the audio-conducting electrodes. Whether this is a huge pop or an inaudible one depends on how well you do the design.

In CMOS particularly, there are two devices being switched at the same time. The control voltages are equal and opposite, so if the DC condition at the audio channel is in the middle of the power supply, the transients tend to cancel each other out. Hence the advice to bias them in the middle of the power supply.

Switching a high impedance audio line makes it worse; if you have a 1M input to the amp and couple in 25 picocoulombs ( a not-unreasonable figure) you'll get a pop. Shunt the amp with 10K and you get no audible pop. Impedances matter. 100K and biasing in the center has worked for every brand of CD4053 I've tried, although they do vary and some are substantially pop free up to 1M.

The business of slowing down the switching works with all of the technologies. The idea is that if the voltage transition on the control element is slow enough, the tiny coupling capacitances will not be able to couple the slow moving voltage. Works great on JFETs, which are high impedance voltage mode devices. Works so so on CMOS, particularly Craig Anderton's 4016 circuit, where you have the control line go directly to the N-channel device. Unfortunately, the P channel device is driven by an internal inverter, so no matter how slowly you drive it, the P-channel device can still inject charge.

On relays, slowing down the coil voltage change works GREAT. I have several circuits at GEO to show how to do this.

Opto coupled things are in theory free of control line coupling. LED/LDRs work (but are expensive and not very good switches). LED/photo FETs like the H11F1 are good, but don't stop in the middle of the swing for a variable resistor - you'll get distortion on guitar level signals. LED-MOSFET photo-relays hold great promise, but you have to use two MOSFETs to get around the integral diode in MOSFETs and they're quite expensive right now.

It's really all a matter of degree."

[bioroids]

Thats great info gez, I'll read it at home and make some experiments.

Sadly I bougth the wrong IC, something called MC4016P and it has 16 pins (not 14) so I dont know what the heck is that. I'll have to trip again to the electronics store ...

Good luck

Miguel