Mid Side M-S Matrix Uses No Precision Resistors

[Mastertone]

Switches is cool, just got another bunch of russian mil spec rotarys, enormus torque on theese bastards, but they should survive a nuclear blast.


Amazing work Roger and Wayne!

[mediatechnology]

Let me expand on Roger's question: What gain trim amounts and number of steps would be required?

+/- 3 dB 6 dB?

6 steps, 12?

[Mastertone]

I have a hard time seeing me use this matrix with more that 3db cut/boost, i would say its enough, and my vote goes for 1/2 (0.6) steps. But a range switch would be cool!

[mediatechnology]

Thanks for the input everyone.

Another option is, of course, a knob with a pushbutton to switch between "fixed" and variable. With only +/- 3dB range the knob would have fair repeatability (and less chunkiness compared to a switch) based on pointer position. The switch or relay would only need to be SPDT since mid or side would be controlled individually.

[Ruairi]

Hi all,

thanks for the invite to give input Roger. Apologies that it's at the end of another 14 hour day so it may not be worth a whole lot. I reread the thread to try to get a handle on the technical constraints presented by this circuit. Rather than write a long essay I made up a small diagram of my ideal layout.

http://i58.photobucket.com/albums/g266/ ... MSplan.jpg broken link

A couple of notes
-If we put the "pot" after the line receiver and buffer it with an opamp before the MS decoder we should be ok right? The 1200 series can drive down to 2K ohms and the buffer will present a near zero output impedance to our MS decoder network?

-I'd rather build this with relays to allow freedom WRT to front panel positioning of switches etc (but it barely matters), if switches make life much easier then great

-If we have a gain control on both the M and S channels we can get a maximum gain difference of 6dB while still having 0.5dB accuracy. I mastered a cd today and used MS extensively as the vocals were too low and muffled throughout the album. In many cases 1.5 dB was enough gain for the M channel (along with eq but 2 dB was too much. It may sound nit picky but in a mastering context small changes make a big difference.

-I would love if the "Hard Bypass" of the whole circuit could be relay based for to be as clean as possible.

Hopefully I'm not muddying the waters.

All the best,
Ruairi

[Gold]

Thanks for the input everyone.

Another option is, of course, a knob with a pushbutton to switch between "fixed" and variable. With only +/- 3dB range the knob would have fair repeatability (and less chunkiness compared to a switch) based on pointer position. The switch or relay would only need to be SPDT since mid or side would be controlled individually.

I thought the control had to be center tapped? In any case I like the 12 position switch idea. +/- 2.5 dB is enough for me but .6dB steps are fine too. Center tapped pots are unobtainium in small quantities. Unless Wayne feels like stocking them I think it's a no go for a pot.

[mediatechnology]

Paul - Having a 2 pos switch in (addition to the pot) which has "fixed" (calibrated) and variable positions wouldn't require a center tap on the pot. (There would be one switch each for M and S.) The center position on the pot wouldn't have the 0.05 dB/0.005% accuracy. But that accuracy wouldn't be needed if the gain trim stage was bypassed by the switch for the "precision" mode where the gain error of the trim stage equals that of wire. This would be a pretty simple circuit.

It would require 4 op amp stages, 2 pots and 2 DPDT relays for both M and S. This is what Roger originally did without the "fixed" mode switching. (Think of it like an oscilloscope input where you have precise deflection in V/div (1-2-5 sequence) and then a variable switch and pot to tweak the vertical amplifer gain for nulling applications such as A+B and A-B. In our case of course our scale is dB not V.

Even if a stepped switch were used there still would need to be a fixed/var switch if 0 dB is in the middle to offer both gain and attenuation. To eliminate the fixed/var switch altogether, a second pole would be required with a unique bypass connection at 0 dB. O dB has to be exactly 0 dB. Any of the other steps can have tolerance. This needs to be drawn by me.

The whole trim stage would require it's own small board because including it on the main board would make it too big and expensive for people wanting just a simple encoder/decoder or not wanting trim. The cost of boards grows geometrically based on HxW and this would add about the same area as the entire encoder/decoder. But, in mastering applications I could see how M/S ratio trims would be a requirement. So we have an option for that.

Ruairi- Thanks for your contributions here. The block diagram looks spot-on. The only change in the flow from what you have and what Roger has laid out is the way it's bypassed. We opted for internal bypass to loop from the single-ended M and S encoder outs to decoder ins. We discussed the reasons for that further up the thread but the main reason was for gain-accurate A/B comparisons. Even if the external processing was out of circuit (I think you labeled them L and R though you may have meant M and S) and a patchcord jumped from the balanced encoder out to balanced decoder in, there would be enough gain error in the 1646 to blow the separation to the point it wouldn't sound exactly like the source at the encoder L and R in. A full hardwire bypass would be nice but it wouldn't necessarily be gain accurate and takes a ton of relay poles. So we opted for the simple solution that was the most gain transparent that wouldn't bias A/B comparisons with minute level differences. We figure if you're not using the M-S encoder and don't want all those extra encode/decode stages along with the line driver/recievers in-circuit one could just pull the patches to achive hardwire bypass.

To answer your question, the gain trims could be put where you proposed, (decode in post-line receiver pre-decode matrix in) or at the encode, or in the L-R domain. The connections would be double-buffered: At the input of the inverting stage and at the output of the trim stage. Just as long as the connections to it are SE. Roger has put inserts for it all along the chain for the various possibilities.

Paul/Ruairi - If the +/- 3dB trim were on a pot would it be precise enough from a repeatability perspective? If not, one could build it as a stepped switch. It sounds like to me that you are leaning towards switches. Maybe we should make the board accept both types of controls.

[mediatechnology]

Roger -

I'm happy with the board pic you posted yesterday which has the one DPDT. You may want to change the labeling and pad orientation but I'd keep that one soft-bypass relay. No one has to stuff it and I think we're going to find a lot of people will use it. The size was OK too.

With that board you can build it into a Pico Compressor SE and do M-S compression with A/B pre/post efx monitoring. Makeup gain is adjustbale with the Pico Comp SE makeup gain and the mute pads are there. It's the only thing you'd need to make a Pico SE an M/S Pico SE processor.

We all need to keep in mind some of these encoder/decoders may be running off dual 9V batteries built into field recording preamps. This works with the small/simple approach you outline.

As to the gain stage I'll get you a drawing for one using a pot. Conversion to a switch will be easy and it can be adapted to relays.

Wayne

[mediatechnology]

Roger -

Having the relay bypass become pads for optional switches works for me. I don't have a problem using a good switch and since it's only DPDT at that location it's do-able.

My thinking was that for a field-recording battery-powered rig, the relays would be off for normal operation using the NC contacts or that the relay would not be stuffed and the NC contacts linked. But. with them off-board the point is "mute." (sic)

The total current draw for 8X line receivers and 4X line drivers are approximately 16 mA + 20 mA +signal currents (2mA/line-rx 5mA/line-tx). This is about 36 mA Iq typically and a lot of current for 9V batteries anyway. We may be in wall-wart country at that current for a portable rig.

Glad taking out the relay makes the routes easier.

Wayne

[mediatechnology]

OK, back to plan A. I love it. From just quickly looking at it you can see the symmetry. And it's simple with everything clearly labeled.

Looking at the layout you have a good low inductance path from the 1646 pin 3 to ground. That whole plane, to the left of the 1646 gets shunted back into Vcc by the 1240's bypass capacitors to their left. (Not to mention the 1646 Vcc/Vee bypass on the right-hand side.) That's a very good thing. I'm learning that pin 3 really likes to have it's "ground" inductance (and were talking less than a few nH) shunted to rails. Bypass caps seem to help pin 3 as much as they do Vcc and Vee.