FET switching vs mechanical on bus assigns

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

dogears

Well-known member
Joined
Nov 15, 2017
Messages
548
Location
Texas
In putting together a scheme for channel assign to bus, I was thinking there might be a benefit to using a JFET vs a mechanical assign, with the aim of making switching quieter.

I am looking specifically at the solo/mute arrangement from the Studer 169 console.
MHXfVor.png


Comparing the approach to channel assign with the switch is the resistor always connected to bus, and the other end switched between bus and ground with the alternate of using a series and shunt JFET mute arrangement (showm simplified) this seems effectively the same.

miQI4Zh.png


You have the advantage of multiple ways to activate the control voltage if desired. I am thinking this should be quieter. Possible disadvantage would be distortion. Are there any big pitfalls to doing the JFET approach?

Thanks


 
Actually it occurs to me you could also reverse the arrangement of the two FETs to accomplish Self's "optimized" system with better offness and reduced noise gain when you reduce the number of assigned channels:

Rpwd69S.png
 
dogears said:
Comparing the approach to channel assign with the switch is the resistor always connected to bus
That's the first concern I have. When the resistor is switched away from the bus, multitrack bus have the big advantage of operating with low noise gain, since there is usually quite a small number of channels routed.

and the other end switched between bus and ground with the alternate of using a series and shunt JFET mute arrangement (showm simplified) this seems effectively the same.
On the attached schemo, the FET's must be of opposite polarity (or receive control voltages of opposite polarity). Anyway it wouldn't really work as is. You need to provide adequate timing to the FET's, making sure the series one closes after the shunt one has opened. And you want to apply half-voltage to the gate in order to reduce distortion.

You have the advantage of multiple ways to activate the control voltage if desired. I am thinking this should be quieter.
Is switching noise such an important issue? I have never encountered a situation where I'd need to assign a channel to a bus while recording.
In order to make switching quieter, you need to ramp the control voltages, and that is not an easy task. I pulled some hair over the issue many years ago.
Don't neglect the actual performance of CMOS switches. They are often criticized for being inadequate for switching voltage, having to resort to current switching. I think it's a sequel of the past, when only the CD series existed.
I use extensively the ADG419 and have no issues with performance.
 
Hi Abbey - thanks. The little sketch is simplified, I put the full schematic of how the JFETs are connected in the first image. Q10 acts as an inverter, which flip-flops the two FETs. One is always on, the other off, and vice-versa. I was going to use the J112.

Re: noise. I don't know. A friend specifically asked about being able to assign a bus on the fly while recording, but I didn't really ask why.

This does have the advantage of simplifying switch requirements by being able to control as many as I like from a CV, much like a relay.  For example, if I wanted to switch on a balanced stereo bus I'd need a 4PDT switch, more if you wanted an LED indication. Here you can reduce any number of "switches" to an SPDT (or DPDT with indication). Cheaper and significant power saving over relay(s) too.

 
Driven from an op amp output, to a high impedance op amp input I’m with you. But in the case of on/off a bus you’re going through a bus resistor onto a virtual earth. No concerns about distortion there?
 
dogears said:
Driven from an op amp output, to a high impedance op amp input I’m with you. But in the case of on/off a bus you’re going through a bus resistor onto a virtual earth. No concerns about distortion there?
Depends on your expectations. Linearity graphs show that the distortion level for a +20dBu signal is about 0.02% when driving a 10k load.
 
The high voltage analog gates are a great option. But there are ones that are a lot lower Ron with better Ron flatness accordingly. You just have to dig through the parts. And of course they're a little expensive but for a one-off project I wouldn't care.

I have also wondered if it would be worth exploring using mosfets and an inverter IC to make a discrete gates that would obviously support potentially very high voltage with Ron in the mohms range. Although I also have to wonder why such parts don't already exist so maybe I'm missing something. Maybe the body diodes would be a problem?
 
I am skeptical about the complexity being worth the utility unless you wish to automate the channel assigns for digital control.

I have shared the circuits here before, back in the early 80s I performed bank switching in a large split console to use one switch/control voltage to toggle between tracking and mixdown.

I used CMOS transfer gates which have their known linearity concerns (not perfect). I was able to finesse the TG linearity by pulling it inside the NF loop, connecting the feedback resistor to the input side of the TG. This technique got the path distortion below the measurement threshold of professional bench test equipment back then (Sound Technology).

Unless you desire some kind of programmable recall, this is not worth the complexity (and distortion) IMO, but my opinion is not what is most important here.

JR 
 
I value y'all's opinions a lot - which is why I asked.

I'm considering either relays or JFETs. The consideration is that I can have make JFET arrangement SMT and cheaper+easier to make than the relays, and reduce power needs.

This is the arrangement for bus switching.
zNvsVO6.png


This sims well, but I'd love a cold eye review.
tJb9uqM.png


7OtVFud.png


 
john12ax7 said:
Do you specifically need remote switching? Or could you use a regular mechanical switch at the needed location?
One reason for it is to combine the mute and destructive solo functions.

Cheers

Ian
 
Yes, and for me I can get PCBs assembled with the JFET muting arrangement in SMT for the same or cheaper than I can purchase a relay and have to solder it myself. Some element of laziness there.

And yes - the ability to combine switches and functions like Ian pointed out is nice.

The other thing is a mechanical switch - to switch a balanced stereo bus needs a 4PDT plus another pole for indication. The switch alone is more expensive than either of the other two options!  :eek:
 
Works fine on a breadboard. Somehow I am still concerned there may be something I’m missing. Sensitivity to stray capacitance on the bus? Nothing?  :-\
 
dogears said:
Works fine on a breadboard. Somehow I am still concerned there may be something I’m missing. Sensitivity to stray capacitance on the bus? Nothing?  :-\

bus capacitance is an actual issue when you have a physical bus several feet long. Adequate feedback capacitance across the combining amp is generally adequate to maintain stability.

JR
 
dogears said:
Works fine on a breadboard. Somehow I am still concerned there may be something I’m missing. Sensitivity to stray capacitance on the bus? Nothing?  :-\
Is it for a large mixer? You opted for about 4.5dB attenuation (28k/47k). Is it necessary?
In terms of noise OPA134 is not the best choice IMO, although I love it very much in many positions. But you may have used it in the sim for proving the concept...
I question the choice of 47k resistors to the summing bus.
The main reason for this choice in API mixers is that it reduced the load imposed on the post-fader amp, since there were many stems, including the post-fader aux sends.
The choice of 47k was a compromise between this and the noise performance of the 2520, which has an OSI of about 5kohm.
The main reason why 10k-ish is preferred by many designers is that it improves by 15dB the offness of mechanical switches. In the case of FET switching it is very likely to present the same sensitivity to stray capacitances. Another reason is it results in much lower bus impedance, that can take advantage of VLN arrangements with large-geometry or multiple transistors. It may not be relevant to you if you're sold to the 2520.
 
Hey Abbey, thank you for the input.

Not a huge mixer, and the you are correct about the OP134 to prove out in sim - but I did intend to use a socketed DOA, so the same limitations of the 2520 apply.
 
dogears said:
Hey Abbey, thank you for the input.

Not a huge mixer, and the you are correct about the OP134 to prove out in sim - but I did intend to use a socketed DOA, so the same limitations of the 2520 apply.
all good info but don't stress out too much over selecting summing resistor values. All else equal lower values will exhibit less Johnson (thermal) noise, and less noise voltage due to bus amp input noise current, but these will rarely ever be the dominant noise floor source.

JR 
 
dogears said:
Hey Abbey, thank you for the input.
Not a huge mixer, and the you are correct about the OP134 to prove out in sim
So why don't you use the basic opamp in LTspice?
I wish we had access to such a part: zero noise, infinite headroom, tunable GBW and doesn't even need supply rails!  ;D
 

Latest posts

Back
Top