passive mastering console with PICS

Igor said:

Can share some stuff on ms, if this can be usefull....
The MS circuit from Wayne is not best crosstalk-wise, about -70db @10kHz practiacally.

Click to expand...

Yes please, share some stuff on ms.

Here a test wayne did on his matrix showing a -80dB to -84dB crosstalk.
http://www.ka-electronics.com/Images/html/M-S_Matrix_Encode_Decode.htm

How do you came to your -70dB figure?

Here is something I am working on here... It is a generic MS or MONO or Buffer or 9.54dB gain stage depending on how you throw the switches....  It is also the same circuit for the encoder or the decoder...

Resistors are all precision to stave off trimmers with their aforementioned TCR... The better the precision (and TCR), the better off ye be...  Of course, the more precision, the less selection of resistor values you may have, hence the single value of Resistors...  but for uber precision it may be better to substitute the proper equivalent values with high precision values...

Don't overlook capacitor precision too and ESR.... 

Opamps in this schematic are generic, but you can place your own favorite opamps... 

Don't forget the protection circuitry and RF EMI prevention and load driving capabilities for long cables...

This circuit is a proverbial "Swiss Army Knife" and covers many possibilities of audio operations (it is essentially a basic analog computer), but it may be beneficial for uber precision to hard wire some of the multi-purpose to more specific purpose configurations, which would get rid of some switches at the expense of functionality...  This circuit as it stands can supply M-S with Side on the "left" channel and Mid on the "right" channel OR vice versa (which may not be necessary so one could hardwire phase switches to simplify the circuit and hardwire functionality).

The way it works is denoted in the switches... The alternate channel is the opposite channel to be summed (in phase or out of phase) with the main channel... For example on a "left" channel PCB, the alternate channel would be a feed from the "right" channel PCB and the main channel would be the "left" channel proper...  The opposite is true for a "right" channel PCB...

The MS / MONO mode switch is also the CUT for the alternate channel which also renders the circuit into a "normal" buffer circuit; however, the main channel gain in this non-MS/MONO mode (normal mode) will have a nifty side effect of +9.54dB (or 0dB as in being a buffer where K5A separately shorts R4 and R15 to each virtual ground also through K1A)... This may be handy for some make-up gain after a compressor...  You can of course still change the main channel phase..

In MS / MONO mode, the main channel gain changes to either +6dB or -6dB as denoted... The alternate channel is now active and being summed as well with either +6dB or -6dB and of course, the alternate channel phase is available to be changed as desired to be either a true L-R  and on the other channel PCB, a true MONO sum (both main and alternate at -6dB).

With the main channel CUT switch (not shown; which could also be part of a bypass for any adaptation you wish) it can also allow a swap/flip of L & R if you add another switch to short across R9 and separately (on another pole of the DPDT switch) short out R11 (when put into MS / MONO mode with the main channel cut)... This would not be passive of course....

Cheers,
-chris

Wow, way to go, dude Those relays are pretty intense...

livingnote said:

Wow, way to go, dude Those relays are pretty intense...

Click to expand...

Danke!  The relays do perform "double duty" now and again... It took a while to figure out how to get the main channel to perform 0dB, -6dB and +6dB in the 2 different modes (MS or Normal)... And then the side effect of +9.54dB for the main channel being available when not in MS mode... wow... cool...

hmm.. that looks interesting indeed! more relays.  ;D

This morning i did some rmaa tests to see how much crosstalk i have. (forgot to save it though..)
The unit itself is clean, something like -117db with eq and comp inserted that are hardbypassed so basically just a lot of cables and relays inserted.
The crosstalk in stereo is the same as my dynamic range so, to me, no crosstalk at all.
But when i switch to MS the crosstalk is something like -76db which i think is ok (useable) but it could indeed be a bit better. THD is a bit more then in stereo but nothing shocking.

When i switch from stereo to ms with a bypassed unit inserted (so just en- >decoding) i don't HEAR any difference so the same as what Radiance describes and to me 'how it sounds', or in this case, 'how it doesn't sound' is more important to me then what the figures tell me.
But still... it would be nice to see a bit less crosstalk.

Great stuff!

People seem to be a bit obsessed about the X-talk. (It IS good to be obsessed to a certain level)
I would be delighted to know what is the X-talk you can actually hear (in a blind test). Very fast and accurate to test in DAW. I understand that having several pieces of gear in chain raises the performance requirements for one unit, but now we are talking about X-talk of only one piece, MS coder-decoder (and every gear in that MS-format) in a mastering setup and that is certainly gonna have much worse X-talk than any piece of gear in stereo format so we can forget them in my opinion.

I'm just wondering if comments about "good" or "acceptable" X-talk are based on listening experience or some solid knowledge, or are they just pulled out of the air or compared to specifications of brand X stuff. I can not hear better than -40dB with my untrained ears and the mastering engineer I work with a lot is perfectly happy with 0.25dB steps in his width control and he has pretty much perfect setup and good ears.

-Jonte

Thats a nifty device Chris!

Yes... Uhg.... more relays! the main and alt channel number of relays is balanced (with the main CUT not shown but being external to the schematic for my purposes here) to keep the paths the same number of relays going through the precision resistors... It is possible to move the alt phase relay to the output and hardwire the alt channel phase to always inverting (out of phase)... Then if you need to do MONO, flip the main phase and flip the output phase...  But this would unbalance the number of relays per channel...

Anyway.. With relays comprising 0.05 ohm contact resistance the insertion characteristics are not too bad.. drops a perfect (theoretical) -80dB CMRR to -78dB or so with the 3 relays, but the theoretical tolerance for TCR will drop it to almost -70dB for 40C rise (-73 for 20C) using +/- 5ppm precision 0.01% resistors... This is a partial worst case theoretical by just doing the math by simply assuming that the TCR changes the tolerance of the 0.01% to something else during a temperature rise and then 20 * Log (resistor match tolerance)... This assumes that one resistor drifts compared to the other, so these numbers could actually be worse ...

So I am curious to see MS circuits spec'ed over temperature rise inside the box....

The obsession question is valid... I can spec out 0.001% (yes, 0.001%) with a TCR of 0.05ppm and get into the -90dB and upper -80 dB range...  TCR drift would almost be negligible...  The price would not be... but it would keep it in the -90 dB range...  Law of diminishing returns?

Vishay makes these resistors but the lead time is probably 8 weeks and who knows the cost without getting a quote..  I know that the lower TCR resistors are like $9 and $17 each...  I don't know what these uber TCR resistors are.. yikes...

There is also "specsmanship" with the commercial folks too to get the better spec than the competition, perhaps establishing the vibe of what to expect...

'Tis true that mastering gear will probably have X-talk on that order unless one obsesses over this spec and builds all their own gear to be "better"... I do know of one mastering engineer (my "neighbor") down the street from me who built most of his own chain in an old backplane chassis... In that case he has control over input and output configurations between processors, and common but locally regulated power supplies... The reduced count of DC blocking parts on *both* input and output between pieces of gear is probably a help in the equipment "stacking" game...  He also uses unbalanced throughout so to avoid the matching between a balanced stage full of matched and high tolerance parts and to avoid multiple conversions between balanced and unbalanced internal to the processors..

Ultimately we are usually going into an A/D after analog processing and what is the match tolerance on the input... Datasheets show 0.1% recommended for the TI ADCs attenuator / driver....  This is something slightly different than the MS discussion going on of course, but speaking of precision...

Nice work, Dagoose.

It strikes me that crosstalk is part of what one wants to mess with ITFP when employing M/S processing.  One wouldn't insert it and do nothing with it so worry over a crosstalk spec seems of practical worthlessness, unless it's truly horrible.  

For those curious about the specifics of Wayne's final design, I'd start here, on page 18 of the design thread:

http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=112&start=170#p2714

block diagram here:

http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=112&start=180#p2917

I'm reading better than -90 dB for stereo crosstalk, which is within 6 dB of the absolute noise floor of a CD.  I recall that being roughly 50 dB better than the average noise floor of vinyl, and 35 dB better than the noise floor of tape sans NR.   All at a very low price, and with nearly no hassles.   Approval from Bob Katz and Paul Gold, to boot.  One can easily note the signal path of the M/S encode/decode is three IC's in total, if one ignores the balancing IC's, which are present as the board is laid out to cover everything up to and including stand alone operation.  If you include the basic circuit within another design enclosure, without the in/out IC's, it becomes very short, straight, and elemental.   Simple, elemental, minimalist, inexpensive, elegant, and as foolproof as any option I've seen.  I think one would be a fool to dismiss it's capabilities without trying it, especially while chasing any more esoteric and expensive implementation.  

It only takes a very small mistake with passives to end up with much worse crosstalk; right down to the wiring on switches.

I like seeing your approach as well, Twenty Log.   But those $10 resistors start to hurt me quickly.   I think the devices I will insert into any M/S matrix will send all reasonable attempts at stereo precision to hell anyway, and will almost always have their own imprecise gain controls.   I expect to have to find unity, or any other point, by measuring tones. 

I am still using an encode/decode matrix made of Jensen transformers, which is also tough to beat, if one is inclined to feel spendy, and I don't do anything of any noted musical precision anyway.  

Bonus prize to identify both quotes.

but, like, it's got electrolytes!

Click to expand...

noise reduction?!  Why would you want to reduce the noise?!

Click to expand...

Totally true... as long as it sounds good, it is good! And to me it sounds perfect and i really love the Wayne's MS matrix.  8)
For the figure freaks, i just did some testing again and at the moment with MS i have 108,6db dynamic range and -87,3db xtalk, so now it even sounds better!  ;D

Indeed... $10 resistors are not necessary unless playing the specsmanship game...

The resistors I am planning on using are $0.845 each and are 10k 0.01% and 5ppm from Stackpole... available at Digikey...

The problem is that I need to buy 1000 units

How do you came to your -70dB figure?

Click to expand...

I breadboarded the matrix/rematrix and tried different chips (ssm, ina, that).
Separation measurement was simple: 0dbu (1v rms) from Boonton 1120 to left,
right out to analyzer's input. Right input terminated with 600 ohm at input.
IIRC THAT chips had better performance, but in general I did not liked 2 things:
I can't have matrix/rematrix with unity gain and I can't adjust the crosstalk
down to 90db.
I got some db down your results because actually it was crosstalk and noise.
At your measurement system, my matrix will show -90db at 10k at least.
I agree with people who says the crosstalk down -80 is fine,
in any case, if we use ms we insert here something to get some effect.
Less parts always better.
I did some experiments with OPA1632 and found them more usefull in
ms applications than balanced receivers.
Anyway, K-1 M/S matrix is different than Wayne's circuit (too much chips IMHO)
and not using OPA1632's

Hey dagoose... I am curious about the push buttons you are using; who makes them, with ability to have multiple colours...  Cheers, -chris

NKK type switches i think, not cheap but sweet:

http://www.nkkswitches.com/series.aspx?stid=8&id=5003

I make use of saia burgess switches, expensive but really nice!
http://nl.rs-online.com/web/search/searchBrowseAction.html?method=getProduct&R=339-358
I'll use a white cap and standard lamp and cut the neon lamp out and replace it with a 2 color 3mm led which changes color when you switch polarity.
This way you will have 2 colors without modding the switch itself, just the cheap lamp and the mechanism is really simple to mod.

dagoose said:

I make use of saia burgess switches, expensive but really nice!

Click to expand...

What happened to those nice switches from stuut and bruin? Are they out of stock?

I guess you bought them all right?  ;-)

radiance said:

dagoose said:

I make use of saia burgess switches, expensive but really nice!

Click to expand...

What happened to those nice switches from stuut and bruin? Are they out of stock?

I guess you bought them all right?  ;-)

Click to expand...

hahahaha! yes.. sold out.. i guess they still have some round ones and non latching though.
They where a bit smaller then the ones i use right now but the new ones feel a bit better.

I'd love to know more about how the inserts are done, and connected.  I'd like to build a simple 8-11 insert switcher with relays/EAO's, and just a lorlin for selecting 1 of 3 inputs.

Thanks!
Sig