Line mixer (active)

[JohnRoberts]

[quote author="NewYorkDave"]All right, I've been thinking about various ways to handle the bussing and mixing. I wanted to post a sketch of this idea before incorporating it into the updated schematic; let me know if you see any flaws that I may have overlooked.

This is for assignability to a main L/R buss and two subgroups. It could be increased to main L/R plus four subs by changing the 22K resistors to 33K.

The reason for the unusual assignment switching is twofold. First, the panpot wiper needs to be loaded with about 10K to give the correct "law." Second, for my own perverse reasons, I don't want the noise gain of the mix amp to change as channels are assigned to/removed from the buss.

I used a variation of the differential mixing setup shown in Fred's PDF, with a little tweak to minimize input offset voltage--which may not be anything to worry about, anyway, since the amp output is capacitively-coupled.[/quote]


Except for my personal objection to using "buss" for "bus", it looks straightforward.

I don't know if you have RF concerns in NYC :roll: but one trick you can play with 100 ohm in series with - input is to actually incorporate a 2 pole active LPF with C to ground at actual bus and feedback resistor now connected to bus side of 100 ohm resistor and C from opamp output to minus input. C to ground perhaps useful for antenna length bus runs. Not practical for audio bandpassing but high noise gain buses can actually rectify in high RF environments when using bipolar input opamps. I recall once correcting a RF problem in a small mixer by swapping in a BIFET opamp on the mix bus. Noise wasn't a problem in that case since it was only a 6 input mixer.

Running constant noise gain isn't perverse but not common in commercial products because customers don't understand residual noise floor with nothing assigned, despite it being inconsequential in real world use.

Side benefit of constant noise gain design is possible use of decompensated opamps or seriously decompensated discrete approaches. Note: you can't take advantage of decompensation if building in active LPF as that requires unity gain stability.

Does value of 1K on + input (bus side of cap) need to be raised to = 1k in parallel with 22k for proper forward ground reference ?


JR

 

[NewYorkDave]

John, thanks for weighing in... Word on the street is that you're no stranger to mixer design :grin:

Running constant noise gain isn't perverse but not common in commercial products because customers don't understand residual noise floor with nothing assigned, despite it being inconsequential in real world use.

Side benefit of constant noise gain design is possible use of decompensated opamps or seriously decompensated discrete approaches. Note: you can't take advantage of decompensation if building in active LPF as that requires unity gain stability.

Definitely a tangible advantage. My main reason was to avoid changing the feedback factor depending on the # of inputs assigned (which could, potentially, have a tangible effect on "the sound" of the mix amp). As for the noise floor not decreasing with fewer inputs assigned... we can call it "built-in dither" and it becomes a feature instead of a drawback! :wink:

In lieu of the active LPF, there's always the old trick of putting a small inductor in series with the inverting input (along with the aforementioned small resistor).

Does value of 1K on + input (bus side of cap) need to be raised to = 1k in parallel with 22k for proper forward ground reference ?

Technically, the correct value there would be 1.047K. The error seems small enough that it'd be swamped out by resistor tolerances.

 

[radiance]

The idea behind this bus assignment switching is that the "+" bus (witch goes to the inverted input of the opamp) always "sees" a 22K resistor wether the channel is assigned or not right?

Also, when the 22k resistors are increased to 33k, there can be 6 busses (1 master and 4 groups) in total. Does this mean that a channel can be assigned to all 6 busses? (or is it buses?)

 

[NewYorkDave]

Yes and yes.

The idea is for the noise gain of the mix amp and the load on the panpot to remain the same regardless of assignment.

I would not expand this beyond L/R + 4 subs because as the value of the mix resistors is increased, their thermal noise contribution increases as well. Buffers after the panpots would ease this limitation; but as I stated at the beginning, my overarching goal was to do the job with the fewest amplifiers.

 

[NewYorkDave]

[quote author="JohnRoberts"]...but high noise gain buses can actually rectify in high RF environments when using bipolar input opamps. I recall once correcting a RF problem in a small mixer by swapping in a BIFET opamp on the mix bus. Noise wasn't a problem in that case since it was only a 6 input mixer.[/quote]

Germane to the conversation is this excerpt from Steve Dove. (I grabbed it from Amazon's "preview" function and I think it's within the limits of Fair Use to quote it here, but someone please correct me if I'm wrong).

 

[JohnRoberts]

[quote author="NewYorkDave"]

Germane to the conversation is this excerpt from Steve Dove. (I grabbed it from Amazon's "preview" function and I think it's within the limits of Fair Use to quote it here, but someone please correct me if I'm wrong).

[/quote]

Yes, Steve Dove knows his way around inside a console and speaks from experience only gained by doing. The thing to bring into focus is a) that everything matters (especially with large numbers of channels), and b) there are often competing goals.

Bus combining impedance is one such example. I will add another anecdote to confuse the issue further. I had one console design where I dropped the impedance of the combining resistors lower than needed for optimal bus noise to reduce crosstalk between channels which scaled down with that resistance.

Consoles are the toughest simple circuits to design, multiple simultaneous tradeoffs.

As I have surely posted before, bus noise in any competent design is usually well below one mic channel at nominal gain so IMO a non-issue in use, but of interest to consumers when doing their WFO tire kicking.

JR

(edit spelling)

 

[JohnRoberts]

[quote author="radiance"]..... all 6 busses? (or is it buses?)[/quote]

Both spellings are considered acceptable for the plural of bus.

JR

 

[NewYorkDave]

I expanded the fragment posted earlier to include the output section. Comments invited...

PDF

 

[JohnRoberts]

[quote author="NewYorkDave"]I expanded the fragment posted earlier to include the output section. Comments invited...

PDF[/quote]

I don't see the immediate benefit of feedback topology. Perhaps it's related to opamp choice? Any DC bias voltage or current could cause wiper noise in level pot.

JR

 

[NewYorkDave]

Well, I'll summarize my train of thought here...

Signal level leaving each channel is at -6dB. In order to preserve headroom, I set the mix amp for unity signal gain instead of trying to make up the loss there. So, between that and the 10dB of loss in the master fader (when set to "0" position), the output amp needs 16dB of gain.

I want to use 10K faders throughout since that's the most common value and easy to get. The input resistor to the output amp must be high enough to not have a major effect on the law of the fader, hence 33K. With a conventional feedback topology, the feedback resistor would need to be ~208K to give 16dB of gain.

The T-network using smaller resistors contributes less thermal noise, and is also convenient because all three values (4.7K, 22K and 33K) are already used elsewhere in the mixer. Plus, as noted on the drawing, the shunt leg of the T can be AC-coupled to keep the DC gain of the stage very low.

As you pointed out, some input bias current would be drawn from the pot wiper; this could be AC-coupled if necessary although I really wanted to keep coupling caps to a minimum.

 

[JohnRoberts]

[quote author="NewYorkDave"]Well, I'll summarize my train of thought here...

Signal level leaving each channel is at -6dB. In order to preserve headroom, I set the mix amp for unity signal gain instead of trying to make up the loss there. So, between that and the 10dB of loss in the master fader (when set to "0" position), the output amp needs 16dB of gain.

I want to use 10K faders throughout since that's the most common value and easy to get. The input resistor to the output amp must be high enough to not have a major effect on the law of the fader, hence 33K. With a conventional feedback topology, the feedback resistor would need to be ~208K to give 16dB of gain.

The T-network using smaller resistors contributes less thermal noise, and is also convenient because all three values (4.7K, 22K and 33K) are already used elsewhere in the mixer. Plus, as noted on the drawing, the shunt leg of the T can be AC-coupled to keep the DC gain of the stage very low.

As you pointed out, some input bias current would be drawn from the pot wiper; this could be AC-coupled if necessary although I really wanted to keep coupling caps to a minimum.[/quote]

Keeping to a small number of standard values is very much a factor in manufacturing, not so much for DIY but appreciated when buying parts.

Regarding thermal noise, the feedback resistor is in parallel with the input resistor for that analysis, so it's not as high as first appears, but indeed it will be somewhat lower with the T network values. I suspect the effective impedance for opamp input noise current analysis will be similar.

JR

 

[capnspoony]

Where did everyone go!

 

[radiance]

Well, time was not on my side lately so I'm still at the vero board stage....
Eventually I will finish this project though.
Any updates on the master section?

:guinness:

 

[capnspoony]

I used what dave had originally posted the first time for the master section. Only for my "bus" I made an identical master section. So channels 1-8 get sum'd/ have outputs and channels 1-4 (since those are the only channels that have aux sends L&R on my board) get sum'd/have outputs. I was thinking about having a return for my 1-4 bus and just having a rotary that would toggle between add/bypass to the master section.

Radiance do you not have what NYD had posted the first time?

best,

Richie

 

[radiance]

What I build on vero board is the latest schematic that NYD posted (BTW, it looks different than the pic I posted previously...)

I will have a look though... I guess I have all versions of this NYD mixer.

 

[capnspoony]

lets see some pics!

 

[radiance]

Ok, here goes...

All those IC sockets are for caps and resistors, so I can fiddle around with different values (resistors are taken out at the moment btw). The opamp can be a DOA or a dip 8 version.

Here with a DOA ....

It passes signal but I still have to figure out the exact values for the resistors since I'm not having a pan circuit like in NYD's schematic...

 

[capnspoony]

ah so you're doing stereo channels?

Certainly looks more organized than mine.

Yeah, I changed a few things around also by trial and error.

How many channels are you going to make?

 

[radiance]

Well, I think it will be something like 10 stereo channels.
Since this will be a "once in a lifetime" project for me I'll take my time. I will make pcb's for it (if I ever get to know Eagle that is...or Rimu ).

 

[capnspoony]

Having a work surface, although simple, is quite rewarding. I hate everything about digital mixing and this has really been a lot of fun to have around. I just added a headphones out and monitor control pots last night. Now I don't have to bend over and press mute on my protools rig after every vocal take.