Mid Side M-S Matrix Uses No Precision Resistors

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Mid Side M-S Matrix Uses No Precision Resistors

Post by mediatechnology »

This is an overall block diagram of the MS Matrix with added functions:

Insert Bypass
Mute for both Mid and Side
Variable Mid SIde Level wit Bypass
Pre/Post Mid/Side Solo

Image

Cleaner gif file: https://www.ka-electronics.com/images/g ... iagram.gif
eweitzman
Posts: 5
Joined: Sun Jan 01, 2012 8:18 am

LP rumble filter crosstalk

Post by eweitzman »

Thanks for this clever circuit! I used it to build an LP rumble filter for my phono preamp. It has separate HP filters in the mid and side channel to take advantage of the fact that bass is converted to mono below some frequency (70Hz, 100Hz, 140Hz, 200Hz, depends on production) before the lacquer is cut. Mono-izing the bass reduces vertical modulation and record skipping. And it allows a rumble circuit that removes out-of-phase bass from the side channel at higher frequencies than the filter for in-phase bass in the mid channel.

My implementation has switchable 20Hz and 100Hz HP filters in the side channel and a 20Hz filter in the mid channel. I used the 1246 for the decoders to get unity gain through the circuit, and used half of an LM4562 as the inverter with closely trimmed feedback resistors to ensure unity gain. DC offset is around 4mV at both output after swapping ICs around.

I had a board made and the circuit works like a charm, except that the crosstalk is very high, around -35db with the 20hz side channel filter and -27db with the 100Hz filter. It's also not the same L->R and R->L, 10db different IIRC. But since the crosstalk is around the same or better than the channel separation of most cartridges (20-30db) my concerns are mostly academic, and I can't hear the crosstalk anyways. Still, I can't figure out why it is so high. Any ideas?

Thanks!
- Eric
circuit overview.png
circuit overview.png (26.99 KiB) Viewed 23415 times
populated board (small).jpg
populated board (small).jpg (90.6 KiB) Viewed 23415 times
RMAA summary.png
RMAA summary.png (17.54 KiB) Viewed 23415 times
RMAA crosstalk.png
RMAA crosstalk.png (23.43 KiB) Viewed 23415 times
ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: Mid Side M-S Matrix Uses No Precision Resistors

Post by ricardo »

LR -> MS -> LR xtalk relies on precise matching. If what you do to M is not the same as what you do to S, xtalk will increase.

If your 2 x 20Hz filters are not matched to much better than 0.1dB and also in phase, xtalk will increase. The phase has an effect at least a decade above the filter frequency. How closely matched are the Rs & Cs in your 20Hz filter?

When you reduce (filter) S, you are getting rid of the difference between L & R. So the large xtalk with your 100Hz S filter is precisely what I'd expect.

But this stuff is what you use a MS circuit for! If you want the signals to remain exactly as they were (no xtalk), why not bypass the circuit entirely.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Mid Side M-S Matrix Uses No Precision Resistors

Post by mediatechnology »

Using it as a rumble filter is cool.

How did you do the decoders with THAT1246's?
I've done it before and it requires some tricks since the internal resistors do not have equal value.
It added stages so I chose an alternate method using 1240s and attenuating the 2L and 2R terms by 6 dB elsewhere.
I'm wondering what's going on in the decode stage.

EDIT: Another thing to be mindful of is the source resistance driving the inputs of the 1240 used in the matrix. They really should be buffered so that the source impedances do not become part of the gain equation and reduce separation. In the MS board everything is buffered. I never checked it, but exposing the 1240 inputs directly to L and R or, more importantly, M and S, may provide unequal termination impedances and alter the MS ratio unintentionally.

EDIT 2: The more I think about it, the unbuffered L and R inputs could seriously degrade crosstalk. A slight imbalance in source impedance of L and R has siginificant consequences since it appears as part of the resistive MS encode network. I would definitely buffer them before feeding the 1240. The clue is that the crosstalk is broadband. There's a resistive mis-match somewhere.

WIth the filters out and doing a full encode/decode you should see crosstalk in the -70 to -90 dB range provided your soundcard measures that good. (It doesn't and they usually don't.)

Another point to make is that the filters should be exactly unity gain in the passband. For that reason I would recommend a unity gain Sallen-Key topology.

I would also not twist L & R together or you will see a HF rise in crosstalk.
eweitzman
Posts: 5
Joined: Sun Jan 01, 2012 8:18 am

Re: Mid Side M-S Matrix Uses No Precision Resistors

Post by eweitzman »

Lots to think about. Thanks.

The 20Hz filters are *not* precisely matched.

The decoder is done as shown in this post, except that a LM4562 is used for the inverter instead of another THAT1240.
viewtopic.php?f=6&t=112&start=120#p2127

The driving source is a phono circuit which has an opamp output stage and a DC servo followed by a 100 ohm buildout resistor. I can remove the resistor and then the rumble filter will be driven directly. I've thought about moving the servo input after the rumble filter. but... none of this is in the test circuit!!! Instead, an EMU 0404 USB is driving the THAT1240s directly. I can't find any specs on the EMU's output impedance but it's certainly biasing the measurements. I'll have to add some buffers for proper testing.

I considered measuring the crosstalk with the filters bypassed but that meant cutting two traces that are hard to get at.

I suppose I should untwist the channel wiring, but they weren't part of the measurements.

That leaves the unbuffered filters driving the MS->LR stage as a potential problem.

- Eric
Last edited by eweitzman on Mon Jan 16, 2012 2:48 pm, edited 3 times in total.
eweitzman
Posts: 5
Joined: Sun Jan 01, 2012 8:18 am

Re: Mid Side M-S Matrix Uses No Precision Resistors

Post by eweitzman »

ricardo wrote:When you reduce (filter) S, you are getting rid of the difference between L & R. So the large xtalk with your 100Hz S filter is precisely what I'd expect.
I don't understand why removing the difference signal would show up as crosstalk. Would you mind explaining that? I guess it has to show up somehow...

Thanks,
- Eric
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Mid Side M-S Matrix Uses No Precision Resistors

Post by mediatechnology »

The driving source is a phono circuit which has an opamp output stage and a DC servo followed by a 100 ohm buildout resistor. I can remove the resistor and then the rumble filter will be driven directly. I've thought about moving the servo input after the rumble filter. but... none of this is in the test circuit!!! Instead, an EMU 0404 USB is driving the THAT1240s directly. I can't find any specs on the EMU's output impedance but it's certainly biasing the measurements. I'll have to add some buffers for proper testing.
I don't know about the EMU0404 USB but the 0404 Internal has a fairly high output impedance. However, the output resistors are likely 1%.
I think with the EMU0404 you got lucky.
In my very, very first work on this I got some fantastic crosstalk numbers driving the inputs directly from a Soundcard.
I quickly realized the need for a 0R source impedance once I went real-world.

I would definitely remove those 100Rs (if the preamp is not driving cable) or use buffers. The 100R build-outs will significantly degrade performance of the matrix.
On a 1240 the internal resistors are 9K 0.005% matched. Your 100R possibly 5% resistors degrade that accuracy.
I haven't looked at it but I suspect the loads on L&R are different.
(On an unpowered MS board one load is 14K the other 18K.)
In my earlier work on this I could insert a 1 Ohm resistor and degrade the cross-talk by a good 6-10 dB.

The circuit you used with a 1246 and 1246 precise-gain inverter does work but I later changed it.
The phase delay of the inverter reduced the symmetry and degraded HF cross-talk.
This change required 1240s and post-attenuation to maintain unity gain overall.

Using an op amp as an inverter without 50 ppm resistors, e.g. 1% will also blow cross-talk.
I would use the later MS matrix board schematic from A to C to encode before the filters.
Then use E to the 5532 output to decode. (note that the 5532's gain resistors are 1% but they are not in the MS domain. They affect channel balance.)
U7 and U8 MUST have precise gain accuracy as well.
If you don't use 1246's or 1240's there than by all means use a unity gain follower.

Image
I don't understand why removing the difference signal would show up as crosstalk.
Look at it this way. Complete cross-talk IS mono.
Removing difference makes something more mono.
Therefore, removing some difference makes something less stereo or more mono.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Mid Side M-S Matrix Uses No Precision Resistors

Post by mediatechnology »

I wanted to check back in and see how this is working out.

I think the concept of rumble/warp HPF in the MS domain has some merit.

Having a single HPF in mid produces a similar result to HP filtering L and R individually.
But, being able to HPF Side at 100 Hz or so makes lot's of sense to me.

Since the original disc recording is going to have a mono crossover there's no original LF Side information to loose.
Warp and rumble present during playback are going to produce significant Side artifacts.
As long as the Side filter is set at or below the mono crossover it seems like a win-win proposition to do it this way.

A question for Gold: What are the range of mono crossover frequencies do you typically see and use in cutting?
matsaddress
Posts: 2
Joined: Wed Nov 09, 2011 9:02 am

Re: Mid Side M-S Matrix Uses No Precision Resistors

Post by matsaddress »

I'm not Paul (obviously) but I happen to know a mastering house in London that uses the following frequencies as switchable settings for their EEs:

32.5, 75, 150, 212, 300

HTH,

Best,

Matt
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Mid Side M-S Matrix Uses No Precision Resistors

Post by mediatechnology »

Matt - Do you know what filter order is typically used? I'm thinking maybe -12dB octave.
Post Reply