CMRR

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
I may well be incorrect but RS232 although shown as a 'twisted pair' is not a balanced connection in that one line is transmitting and the other is receiving.
RS422 is (IIRC) 'balanced' and perhaps not too dissimilar to AES3 digital audio format as often seen presented on a XLR3 connector. Of course CMMR in itself is not a 'guaranteed' way to have interference free 'communication' as it takes considerable effort to maintain the 'magical' rejection properties from DC to 'light' bandwidth so a combination of practical implementation (cost effective) and proper out of (required) band filtering as maintaining effective 'rejection' at for example cellphone frequencies is a considerable exercise.
Matt S
 
After re-visiting my post it seems I have been a bit economical with what's been said by the authors and missed an important point. They do mention that the differential scheme works on twisted pair lines like RS232, but they also include audio which is not correct, unless of course, and as you mentioned, one decides to transmit audio over 35,000ft using wires at 20kHz. At the same time I am also partially wrong with my comments.

However, what I am not entirely sure about is how terminating the line with an impedance equal to the source can assure equal induction of common mode voltage on a say twisted pair. I have dug in a bit deeper on this and consulted a couple of good books, and I can not see any reference to it. So, it would be great if you could comment.
You may need a different book. "Transmission line" phenomenon deals with standing wave behavior inside a wire/line. I also wrote about this in my old audio mythology column 40 years ago. IIRC the analogy I used back then to visualize the phenomenon was a short length of rope. If you vigorously wiggle one end up and down to create a standing (sine) wave in the rope. If the far end of the rope is solidly secured to a fixed point that is a node. The full energy of the standing wave reflects back and that node looks like an effective zero crossing. Alternately if that far end is left completely free it will swing up/down with full amplitude. That open end is effectively an anti-node. Again the full energy gets reflected back into the rope. Transmission line theory describes a third termination with the line's (ropes) characteristic impedance. If this termination is properly executed there is full energy transfer into that termination with no reflected energy coming back into the line. This is desirable to prevent interference between driving signals and reflected energy. To visualize what characteristic impedance is, imagine an infinitely long line (rope). Terminated with this perfect characteristic impedance, the signal just keeps going with no reflection to interfere with signal. This phenomenon is a function of signal transit speed and line length. In our world we mainly see this in RF, video, and higher than audio signal terminations. IIRC 75 ohm is a popular characteristic impedance for video cables and interfaces.

We generally don't have to worry about this for audio interfaces but historically it first came to engineer's attention working with very long audio lines/cables for undersea telephone transmission. The legacy 600 ohm I/O terminations in old audio gear is a remnant from old telephone system engineering. The characteristic impedance for typical short audio lines is more like 50 ohms, but modern audio interfaces are "bridging " so transmission line effects can be ignored.
My thinking is that, and as you also rightly commented, as long as the wire pair is twisted properly (evenly) then the common mode voltage should induce (theoretically) equally on both wires. So, I see no effect of the termination resistor on this in bridge termination. In fact for single ended (on each line individually) and AC termination the resistors would require good matching to minimise the common mode voltage developed across each of them.
Noise is not just induced magnetically, but from electrostatic coupling also. A "bridging" termination is by definition 10x or more the source impedance. Even with this high impedance, inaccurate matching can cause different noise pickup in both lines that won't cancel completely in the following differential.

====

This is a bit of a veer from the topic of CMRR but it is the nature of web discussions to not answer the OP's specific question, but to also share random technical information vaguely related to the subject, in this case interfaces and accurate signal transfer.

JR

PS: I am a self taught college dropout so my description may not be textbook perfect. For my old magazine column I included some drawings.
 
@ Matt Syson,

Apologies. Indeed RS232 is a typo. 422/485

@JR,

Thank you for your answer but unfortunately it has nothing to do wit my question.

My question is how terminating a balanced transmission line with impedance equal to the source impedance makes the common mode signal equal on both wires.

As for veering from the topic CMRR, I am not sure what you mean. Is this not about understanding CMR?
 
I am not aware of any benefit from silver paint, it may kill germs. ;)
Do you remember the silver guy at NAMM?

So I designed this narrow band 3 channel room mode filter for a hi-fi company, it could do 1/10th or 1/20th of an octave, to get rid of room modes and standing waves in a room that was wrong, when the guy spent too much on hi-fi crap to afford treatments. The whole system was picking up a bunch of RF (usually public radio, as they are except from many FCC regs), so the owner of the company sent me 3 things to see if they helped. One: was some nickel impregnated cloth that was pretty cool. It was non-conductive and was a great trouble shooter. Two: was a little black thing that had a laser engraved logo on it that you are supposed to set on the preamp and it wards off all evil demons (a serious product that 'hi-fi' experts could hear the difference), and Three: was a bottle of this special silver paint that you paint on the chips in the unit and it prevents stray electrons from escaping and lowers errors and jitter by reflecting them back in. When I asked that company owner how the reflected electrons know where to get back in line, he hung up on me.

Turns out, the hi-fi owner lived on a hill that was a mile away from every radio station tower in Richmond VA, and his problem was because all his billion dollar hi-fi stuff was sitting on a steel bakers rack, and all the little odd shaped Eddie current "receivers" were funneling the RF into everything. You could lightly rub you finder on the edge and feel the field... Took his wife's wooden book shelf and fixed the problem...
 
@ Matt Syson,

Apologies. Indeed RS232 is a typo. 422/485

@JR,

Thank you for your answer but unfortunately it has nothing to do wit my question.

My question is how terminating a balanced transmission line with impedance equal to the source impedance makes the common mode signal equal on both wires.
it doesn't
As for veering from the topic CMRR, I am not sure what you mean. Is this not about understanding CMR?
this is the internet where people tell you everything they know (or think they know), not answer your questions.

Transmission line phenomena also affect signal interfaces but as I noted, generally not important at audio frequencies over short lines (but once important for undersea cables but they are probably no longer audio over copper).

JR
 
After re-visiting my post it seems I have been a bit economical with what's been said by the authors and missed an important point. They do mention that the differential scheme works on twisted pair lines like RS232, but they also include audio which is not correct, unless of course, and as you mentioned, one decides to transmit audio over 35,000ft using wires at 20kHz. At the same time I am also partially wrong with my comments.

However, what I am not entirely sure about is how terminating the line with an impedance equal to the source can assure equal induction of common mode voltage on a say twisted pair. I have dug in a bit deeper on this and consulted a couple of good books, and I can not see any reference to it. So, it would be great if you could comment.

My thinking is that, and as you also rightly commented, as long as the wire pair is twisted properly (evenly) then the common mode voltage should induce (theoretically) equally on both wires. So, I see no effect of the termination resistor on this in bridge termination. In fact for single ended (on each line individually) and AC termination the resistors would require good matching to minimise the common mode voltage developed across each of them.
You are absolutely right, termination of a transmission line (to prevent aberrations in the differential signal due to reflections in the line) have nothing to do with impedance balance (of common-mode impedances). They are entirely separate issues. IMHO, signal/data interfaces that use balanced lines should still adhere to the practice at lower frequencies (like audio) of keeping common-mode impedances low at the driver end and as high as possible at the receiver. Terminations at the receiver should always be differential (between the lines) and not split with the center grounded - this puts an extreme burden on tolerances to avoid seriously degrading CMRR of the interface!
 
However, what I am not entirely sure about is how terminating the line with an impedance equal to the source can assure equal induction of common mode voltage on a say twisted pair. I have dug in a bit deeper on this and consulted a couple of good books, and I can not see any reference to it.
You must analyze the balanced connection as a Wheatstone bridge. The rejection is optimum when the transmission ratio of one branch is exactly the same as the other branch.
That means that you may have a source with a wild unbalance, like say 10 ohms on the cold leg and 100 ohms on the hot leg, you may have perfect CMRR if the receiver as an impedance of 1k on the cold side and 10k on the hot side. Of course at HF, when stray capacitances start to be felt, it's not that simple.
This configuration is not that crazy; it' s what happens with balanced bus, where the bus feed resistors are about 10k and the ground sense resistors are about 100r.
One thing is certain: making the connection a matching one results in signal loss, and is no guarantee of a better CMRR. On the contrary, in the real world (one where resistors are not perfect) the higher the ratio of the receiving Z to the sending Z, the better the CMRR.
 
Last edited:
Do you remember the silver guy at NAMM?
I barely remember NAMM, but vaguely recall seeing pretty booth girls (eye candy).
So I designed this narrow band 3 channel room mode filter for a hi-fi company, it could do 1/10th or 1/20th of an octave, to get rid of room modes and standing waves in a room that was wrong, when the guy spent too much on hi-fi crap to afford treatments.
Too narrow could be unstable over changing ambient temperature/humidity.

The last monitor console I designed (last century) had sweepable notch filters on each output channel. I don't recall how narrow I made them but not silly, perhaps 1/6th oct. I even included a trick variant on my FLS LED indicators where I had three LEDs indicate when max audio energy was above, below, or inside the notch band. As often occurred the mixer was too cool for the room.
The whole system was picking up a bunch of RF (usually public radio, as they are except from many FCC regs), so the owner of the company sent me 3 things to see if they helped. One: was some nickel impregnated cloth that was pretty cool. It was non-conductive and was a great trouble shooter. Two: was a little black thing that had a laser engraved logo on it that you are supposed to set on the preamp and it wards off all evil demons (a serious product that 'hi-fi' experts could hear the difference), and Three: was a bottle of this special silver paint that you paint on the chips in the unit and it prevents stray electrons from escaping and lowers errors and jitter by reflecting them back in. When I asked that company owner how the reflected electrons know where to get back in line, he hung up on me.

Turns out, the hi-fi owner lived on a hill that was a mile away from every radio station tower in Richmond VA, and his problem was because all his billion dollar hi-fi stuff was sitting on a steel bakers rack, and all the little odd shaped Eddie current "receivers" were funneling the RF into everything. You could lightly rub you finder on the edge and feel the field... Took his wife's wooden book shelf and fixed the problem...
Damn,,, deja vu.... Back in the early 80s Loft sold a console to a studio in Richmond. The studio was in the main beam of an AM radio tower. I fly down expecting a couple hours at most day trip chasing RF out of the console. I even recall the AM station frequency. I found my signal ground and power ground swinging with over 1V AC at 960 kHz between them. I nuked it with ceramic caps added on each strip to stiffen up the signal ground.

The studio was located in the basement of the lawyer's building who backed the studio, so punting was not an option. After I fixed the console I helped them with several other problems as much as I could (on day two). They had AM radio pickup in their plate reverb and sundry other studio units.

JR
 
I highly recommend Electromagnetic Compatibility Engineering” by Henry Ott.
he goes through all kinds of techniques for reducing and preventing noise.
I have books by Ott and Morrison on my bookshelf from the 70s when I was trying to figure this stuff out. The basic physics hasn't changed since then, but the noise sources have certainly grown more pervasive.

JR
 
@sahib: The reason to terminate a long line with a resistor equal to the line's characteristic impedance is to prevent reflections at the receiving end. It has absolutely nothing to do with the induction of common mode currents.

Cheers

Ian
 
Lets say we wanted to make up a balanced probe cable for measurement purposes ,
can we use starquad cable to good effect ?

Lets assume the input resistances in the ADC are close tollerance , but could we also arrange to balance stray capacity across the two pairs so we dont loose CMRR at high frequencies ?

I dont have any starquad cable to hand ,in any case I measured a 150 cm XLR/XLR patch cable I made up myself ,

Pin 2 to screen 246.7pf
pin 3 to screen 259.8pf
pin 2 to pin 3 144.7pf
Nominal capacitance of the Belden 8412 cable itself is 110pf/M cond.-cond. so the xlr's add 35pf .

Can I add an smd trimmer cap between pin 2 and ground inside the xlr housing and adjust for balance ?

https://www.markertek.com/product/s...ts-mono-male-to-1-4-ts-mono-male-black-6-foot
Can unbalanced signals also benefit from starquad ?
 
Last edited:
Lets say we wanted to make up a balanced probe cable for measurement purposes ,
can we use starquad cable to good effect ?

Lets assume the input resistances in the ADC are close tollerance , but could we also arrange to balance stray capacity across the two pairs so we dont loose CMRR at high frequencies ?

I dont have any starquad cable to hand ,in any case I measured a 150 cm XLR/XLR patch cable I made up myself ,

Pin 2 to screen 246.7pf
pin 3 to screen 259.8pf
pin 2 to pin 3 144.7pf
Nominal capacitance of the Belden 8412 cable itself is 110pf/M cond.-cond. so the xlr's add 35pf .
35pF for 2 XLR's is a lot. Typically the capacitance between pin 2 and 3 is just a few pF. I suspect either your capmeter is optimistic or the cable is in the upper range of tolerance (or slightly longer than advertized).
EDIT: I'm not sure; 1.5 meter of 110pF/m cable should result in 165pF, not 144...
It doesn't fit...
Can I add an smd trimmer cap between pin 2 and ground inside the xlr housing and adjust for balance ?
Of course you can; it's common practice in some broadcast equipment to have a HFCMRR trimmer.
Can unbalanced signals also benefit from starquad ?
Yes; marginally; if the shield is connected on one side only. It can reduce induced parasitics. However it won't help with ground-loop induced problems.
 
Ok I forgot to x1.5 the 110 pf/m

The overall lenght of the cable itself is 136cm ,not including the xlr's ,
Thats calculates out at 149.6 pf ,
If we want to drill down further we could add in the fact that 15mm at each end of the cable has the screen paired back and the conductors are spread apart, so lets deduct another 30 mm from the equation to account for that .
146.3 pf ,
Originally I measured at 1khz ,
I took another measurement at 100hz on the meter it came out at 148.2pf , fairly close to the calculated value and allowing a couple of pf for the connectors , no ghost in the lcr machine .

The Belden starquad cable comes outs at 129pf per meter across the conductors so not much more than twin core.

So if we want to use two or four core screened cable unbalanced which is the correct end to connect the screen , instrument or amp end of the cable ?
Electric guitars are getting difficult to record with all the RF mush floating around , for the the extra freedom from noise it might be worth a try , ok we take a hit on cable capacitance compared to typical low noise single core guitar cables .

Just found this ,
http://www.shootoutguitarcables.com/guitar-cables-explained/capacitance-chart.html
 
its hard to give a good answer without knowing source impedance and equipment input impedance.

Scope probes target very low capacitance and high input impedance. I am not aware of any standard for bench test equipment, but would speculate 10kHz to 100kHz input impedance.

Star quad is great for mic preamp cables and fine for modest line level cables. Not so great for guitar cables or extremely long runs. Back in the day mic cable capacitance was such an issue on films using dynamic mics spaced thousands of yards away. To deal with HF loss from cable capacitance they used 50 ohm mics, to deliver decent HF response and back in those days they considered 7kHz high frequency.
==

in the context of mic preamps I've seen both HF and LF CM trims used together. The HF trim was small variable series resistance in one leg with a fixed resistor in the other leg. This trim was to tweak for stray capacitance. A LF trim to tweak for mismatched phantom voltage blocking capacitors used higher impedance trim legs to ground.

JR
 
Though the resistors in ADC input are well matched, and flat across frequency (low stray C) I think you will find high frequency CMRR limited by loop gain in the differential amp stage, which is rolled off by the op amp's dominant pole. If I'm wrong because a commonly used ADC has low-feedback diff amp (which could improve high frequency CMRR at the expense of accuracy) please set me straight with details.
 
Back
Top