ProSoundWeb Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 3 [All]   Go Down

Author Topic: Chasing loudspeaker-level Crosstalk  (Read 6198 times)

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Chasing loudspeaker-level Crosstalk
« on: February 09, 2015, 04:42:22 pm »

I seem to be observing some crosstalk between Left and Right, on the order of -30dB. Confirmed it is downstream of amplification. Low impedance loudspeakers on multichannel class D amps - L and R live in discrete amp chassis and are driven on discrete AES3 drive lines. Iso Power and rack rails are shared.

 All runs are #10AWG THHN, physically parallel and tight bundled for ~30ft before hitting discrete EMT pipes per side.     

No change when driving only one side via analog straight into amp; acoustic output is still present at the undriven side.

Link to Transfer function of this behavior, measured acoustically at the undriven side in near field, referenced to the driven side : https://drive.google.com/file/d/0B0pQLbAUdKDhVUVuUzFac1luQ0E/view?usp=sharing 

Have looked for swapped neg conductors, found none. Have checked for drive voltage on (seperately solidly grounded) loudspeaker conduit grounds, none present. Powering off amplification for the undriven side eliminates undesired output. Perhaps signal is leaking into the noise floor of the undriven amps via G/N/Chassis-to-rails??

Would certainly welcome some input where to look next!

 
Logged

Luke Geis

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1857
    • Owner of Endever Music Production's
Re: Chasing loudspeaker-level Crosstalk
« Reply #1 on: February 09, 2015, 06:26:12 pm »

Are you certain the crosstalk is not present from the mixer?

Have you independently sent signal into each amp from a more direct source ( such as a small mixer or ipod located at the amp rack )?

It sounds as if there is one amp per side, or am I misunderstanding?

I find it hard to believe that there is any coupling in the speaker cables. There is either a leak of signal that is getting into amps or the mixer itself.
Logged
I don't understand how you can't hear yourself

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Chasing loudspeaker-level Crosstalkt
« Reply #2 on: February 09, 2015, 10:48:35 pm »

Are you certain the crosstalk is not present from the mixer?

Have you independently sent signal into each amp from a more direct source ( such as a small mixer or ipod located at the amp rack )?

It sounds as if there is one amp per side, or am I misunderstanding?

I find it hard to believe that there is any coupling in the speaker cables. There is either a leak of signal that is getting into amps or the mixer itself.

Yes, driving analog from a PC interface direct into one side only yields same result.

There are 2 x 4 channel amps per side.

I can unplug the load and plug a monitor wedge straight into the rear of the amps. There is nothing audible above the amp noise floor with no input present . There is low level audible signal on the R side loudspeaker lines with L side driven-  even when not loaded onto an amplifier.










Logged

Josh Millward

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 713
  • Meridian, MS
Re: Chasing loudspeaker-level Crosstalk
« Reply #3 on: February 10, 2015, 04:51:00 pm »

I seem to be observing some crosstalk between Left and Right, on the order of -30dB. Confirmed it is downstream of amplification. Low impedance loudspeakers on multichannel class D amps - L and R live in discrete amp chassis and are driven on discrete AES3 drive lines. Iso Power and rack rails are shared.

Interesting...

All runs are #10AWG THHN, physically parallel and tight bundled for ~30ft before hitting discrete EMT pipes per side.

Saying this makes me think that all the individual conductors are literally laid parallel to one another in a large bundle for 30 feet. Then they split into separate bundles to go into the pipes to the various loudspeaker locations.

This could absolutely cause what you are hearing.

What you need to do is cut that bundle apart and find the pairs of wire. Then you need to twist together the individual wires to form a two conductor cable. Twist them tight using a drill and they will not come apart. Honestly, you should probably pull all the wire out of the conduits and twist each pair over their full lengths. After all, you don't really want your LF signal leaking into the HF driver, even if it is way down in level.

Have a look at the attached photo for an example of where I had twisted pairs together using a drill. Granted, the wires you are looking at in this photo are only 16 AWG, not 10 AWG, but you can get the point. I have used a small cordless drill to do the same thing to wires up to 14 AWG, if I were planning to do this with long runs of 10 AWG, I would definitely use a corded 1/2 drill. Keep in mind, you are going to lose some length by twisting the wires, how much you will lose depends on how tightly you twist the wires. Put the far end securely in a bench vise, chuck up the other end in the drill, and go to town. Then you can pull the cables back into the pipes and bundle them back up. You will notice that now the cables will take up a lot more room than they used to since they are wrapped around one another.
Logged
Josh Millward
Danley Sound Labs

Luke Geis

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1857
    • Owner of Endever Music Production's
Re: Chasing loudspeaker-level Crosstalk
« Reply #4 on: February 10, 2015, 07:31:41 pm »

30' isn't very long and capacitive coupling should not be that stark, at least not very audibly. The only other way to test this out would be to run new independent speaker runs in as direct a path as possible to the amps and speakers, but you say that even bringing a monitor near the amp and connecting directly to it has the same results.

The above mentioned method will nearly eliminate any coupling due to the nature of the twist in the wire. It will cause it to naturally cancel out common modes. It will also require a longer length of cable as the twists will cause the cable to reduce length.
Logged
I don't understand how you can't hear yourself

Lee Buckalew

  • Classic LAB
  • Hero Member
  • *
  • Offline Offline
  • Posts: 1317
  • St. Louis, MO area
    • Pro Sound Advice, Inc.
Re: Chasing loudspeaker-level Crosstalk
« Reply #5 on: February 10, 2015, 10:30:50 pm »

30' isn't very long and capacitive coupling ...

This is an example of inductive coupling, not capacitive.

Lee
Logged
Lee Buckalew
Pro Sound Advice, Inc.

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Chasing loudspeaker-level Crosstalk
« Reply #6 on: February 11, 2015, 09:48:14 am »

This is an example of inductive coupling, not capacitive.

Lee

Wouldn't there be more low pass slope to the crosstalk spectrum, or could it just be happening above my measurement nyquist?
Logged

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 16379
  • Hickory, Mississippi, USA
    • Resotune
Re: Chasing loudspeaker-level Crosstalk
« Reply #7 on: February 11, 2015, 10:27:03 am »

is there a common contaminated ground between the two sides?

Is a speaker quiet when connected at the amp?

For test purposes if you run a separate ground lead from a clean ground at the amp out to the speakers and see if the speaker with crosstalk gets quieter when connected to a clean ground.

Speakers are differential devices so they can't tell the difference between signal at their + or - input.  If the grounds are common or ground currents get co-mingled it will degrade separation due to IxR voltage drop in the ground run.

JR

Logged
Don't tune your drums half-ass. Listen to what a properly "cleared" drum sounds like.   http://circularscience.com/

Josh Millward

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 713
  • Meridian, MS
Re: Chasing loudspeaker-level Crosstalk
« Reply #8 on: February 11, 2015, 10:39:12 am »

...also, I would be interested to know exactly what power amplifiers you are using in this application. You said they are four channel amps. Let's name some brands and models.
Logged
Josh Millward
Danley Sound Labs

Jason Lavoie

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 896
Re: Chasing loudspeaker-level Crosstalk
« Reply #9 on: February 11, 2015, 03:19:16 pm »

Powering off amplification for the undriven side eliminates undesired output.

Since we would expect that having the amp on but with no signal to be a load on the cable and theoretically reduce the crosstalk on the wire I would be suspecting either a fault in the amp or that the crosstalk is upstream of the amp.

does the problem go away if you turn down the input gain to the quiet channel?
you say it is being fed by AES but at some point in the amp that becomes low level analog and may be susceptible to a bad ground or bad design.
I'm also wondering if maybe the analog input is still hot while using AES (amp model would help)

Jason
Logged

Lee Buckalew

  • Classic LAB
  • Hero Member
  • *
  • Offline Offline
  • Posts: 1317
  • St. Louis, MO area
    • Pro Sound Advice, Inc.
Re: Chasing loudspeaker-level Crosstalkt
« Reply #10 on: February 11, 2015, 10:19:38 pm »

Yes, driving analog from a PC interface direct into one side only yields same result.

There are 2 x 4 channel amps per side.

I can unplug the load and plug a monitor wedge straight into the rear of the amps. There is nothing audible above the amp noise floor with no input present . There is low level audible signal on the R side loudspeaker lines with L side driven-  even when not loaded onto an amplifier.

Hayden, you say that you tested the output with a monitor connected locally, directly to the amp.  You then said "There is low level audible signal on the R side loudspeaker lines with L side driven-  even when not loaded onto an amplifier."
What do you mean by this?
We need to understand the full signal path when you are testing.

If you test this with 2 locally powered speakers does the one that is undriven exhibit the same crosstalk as the installed speaker?
If so then it's not the cabling.
If it does not then you need to find the fault in the wiring.  It could be inductive, it could be a ground problem as JR alluded to.  You need to isolate the problem one step at a time.

It may also help, as requested by another poster, to know what the amp make and model are.

Lee
Logged
Lee Buckalew
Pro Sound Advice, Inc.

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Chasing loudspeaker-level Crosstalkt
« Reply #11 on: February 11, 2015, 11:04:23 pm »

Hayden, you say that you tested the output with a monitor connected locally, directly to the amp.  You then said "There is low level audible signal on the R side loudspeaker lines with L side driven-  even when not loaded onto an amplifier."
What do you mean by this?
We need to understand the full signal path when you are testing.

If you test this with 2 locally powered speakers does the one that is undriven exhibit the same crosstalk as the installed speaker?
If so then it's not the cabling.
If it does not then you need to find the fault in the wiring.  It could be inductive, it could be a ground problem as JR alluded to.  You need to isolate the problem one step at a time.

It may also help, as requested by another poster, to know what the amp make and model are.

Lee

I had the local monitor paralleled to the installed loudspeaker. Unplugged the NL4 on the rear of the amp, put a NL4MM barrel on it, and connected the local wedge.

I've now observed the same behavior while loaded on both Powersoft M50 and d&b D80, even loaded one side on one brand and one on the other, to no effect.

Some sort of LC coupling seems more and more likely. I may try and get conduit all the way to the rack. Anyone know if flex might provide adequate shielding, or is rigid absolutely necessary?
Logged

Lee Buckalew

  • Classic LAB
  • Hero Member
  • *
  • Offline Offline
  • Posts: 1317
  • St. Louis, MO area
    • Pro Sound Advice, Inc.
Re: Chasing loudspeaker-level Crosstalkt
« Reply #12 on: February 11, 2015, 11:37:40 pm »

I had the local monitor paralleled to the installed loudspeaker. Unplugged the NL4 on the rear of the amp, put a NL4MM barrel on it, and connected the local wedge.

I've now observed the same behavior while loaded on both Powersoft M50 and d&b D80, even loaded one side on one brand and one on the other, to no effect.

Some sort of LC coupling seems more and more likely. I may try and get conduit all the way to the rack. Anyone know if flex might provide adequate shielding, or is rigid absolutely necessary?

This is not a shielding issue unless something is getting into the speaker cable and being rectified in the amp. 
It is most likely a lack of the speaker wires being twisted pairs or a common ground between the two channels of the amp (in the speaker wire run).  This could be a short (maybe high impedance) between wires from getting skinned slightly during a wire pull.  Could be a connection someplace that's frayed in the speaker cable.

Lee
Logged
Lee Buckalew
Pro Sound Advice, Inc.

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Chasing loudspeaker-level Crosstalkt
« Reply #13 on: February 12, 2015, 09:16:43 am »

This is not a shielding issue unless something is getting into the speaker cable and being rectified in the amp. 
It is most likely a lack of the speaker wires being twisted pairs or a common ground between the two channels of the amp (in the speaker wire run).  This could be a short (maybe high impedance) between wires from getting skinned slightly during a wire pull.  Could be a connection someplace that's frayed in the speaker cable.

Lee

I understand the utility of twisting the pairs, but short of that (i.e. pulling new copper)- if this is happening across those first 30 ft of parallel bundle without a direct conduction path, it seems counter-intuitive that discrete conduits all the way to the rack wouldn't provide some sort of magnetic brake. I was really hoping to grasp at a few straws before having new copper pulled!   

Have found no common ground. If there's a high-z short, it's tricking my fluke 87 DMM, which displays into the Mohm range before displaying open. 
Logged

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 16379
  • Hickory, Mississippi, USA
    • Resotune
Re: Chasing loudspeaker-level Crosstalkt
« Reply #14 on: February 12, 2015, 09:19:18 am »

This is not a shielding issue unless something is getting into the speaker cable and being rectified in the amp. 
It is most likely a lack of the speaker wires being twisted pairs or a common ground between the two channels of the amp (in the speaker wire run).  This could be a short (maybe high impedance) between wires from getting skinned slightly during a wire pull.  Could be a connection someplace that's frayed in the speaker cable.

Lee

+1... check the speaker ground return wiring... BUT I already suggested that.

If the two speaker grounds are shorted together somewhere that could cause it.

JR
Logged
Don't tune your drums half-ass. Listen to what a properly "cleared" drum sounds like.   http://circularscience.com/

Jason Lavoie

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 896
Re: Chasing loudspeaker-level Crosstalkt
« Reply #15 on: February 12, 2015, 10:18:28 am »

Some sort of LC coupling seems more and more likely. I may try and get conduit all the way to the rack. Anyone know if flex might provide adequate shielding, or is rigid absolutely necessary?

If conduit for the last 30' is going to solve the problem then simply separating the bundle and putting a few inches between the pairs should show a noticeable (measurable) difference as a test.

Quote
I've now observed the same behavior while loaded on both Powersoft M50 and d&b D80, even loaded one side on one brand and one on the other, to no effect.
Does the crosstalk go both ways? or is it only from speaker line A to speaker line B?

Jason
Logged

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Chasing loudspeaker-level Crosstalkt
« Reply #16 on: February 12, 2015, 11:53:48 am »

+1... check the speaker ground return wiring... BUT I already suggested that.

If the two speaker grounds are shorted together somewhere that could cause it.

JR

With amplification disconnected from load, checked for cross-continuity across all pairs. all open according to my fluke 87 which ranges to 50Mohm.

I have found a pair of negatives between sides reading ~120kohm with amplification connected and powered down. This jumps down to 160ohms with amps powered up! 
Logged

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 16379
  • Hickory, Mississippi, USA
    • Resotune
Re: Chasing loudspeaker-level Crosstalkt
« Reply #17 on: February 12, 2015, 12:06:41 pm »

With amplification disconnected from load, checked for cross-continuity across all pairs. all open according to my fluke 87 which ranges to 50Mohm.

I have found a pair of negatives between sides reading ~120kohm with amplification connected and powered down. This jumps down to 160ohms with amps powered up!

VOM Ohms readings with voltage present will not be accurate. The VOM imputes ohms by sending a reference current and reading the voltage rise, so if other currents and/or voltages are present you will read random results. Amps should not be dumping DC into output but that is a different issue.

What do you measure between amplifier grounds? While common speaker ground at the amps should cause minimal crosstalk, if at all.

OK all pairs are open with both ends disconnected (?), do any of the speakers have continuity to external metal parts and perhaps ground to ground path via speaker mounts? 

JR
Logged
Don't tune your drums half-ass. Listen to what a properly "cleared" drum sounds like.   http://circularscience.com/

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Chasing loudspeaker-level Crosstalkt
« Reply #18 on: February 12, 2015, 02:37:25 pm »

VOM Ohms readings with voltage present will not be accurate. The VOM imputes ohms by sending a reference current and reading the voltage rise, so if other currents and/or voltages are present you will read random results. Amps should not be dumping DC into output but that is a different issue.

What do you measure between amplifier grounds? While common speaker ground at the amps should cause minimal crosstalk, if at all.

OK all pairs are open with both ends disconnected (?), do any of the speakers have continuity to external metal parts and perhaps ground to ground path via speaker mounts? 

JR

 Measuring ~120k ohm between output grounds, loaded. Measuring ~60kohm from chassis to output ground, unloaded. Bonding conductors, chassis, and input pin 1's are all @ dead short to each other. 

Conductors are all open relative to flyware/conduit/tray.
Logged

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 16379
  • Hickory, Mississippi, USA
    • Resotune
Re: Chasing loudspeaker-level Crosstalkt
« Reply #19 on: February 12, 2015, 06:55:56 pm »

Measuring ~120k ohm between output grounds, loaded. Measuring ~60kohm from chassis to output ground, unloaded. Bonding conductors, chassis, and input pin 1's are all @ dead short to each other. 

Conductors are all open relative to flyware/conduit/tray.

tens of kOhm is effectively open circuit for speaker impedance loads.

If the signals are clean at the amp, and crosstalk occurs at the speaker end, it must be something. I ASSUme speaker grounds are similarly tens of K impedance separated?

JR
Logged
Don't tune your drums half-ass. Listen to what a properly "cleared" drum sounds like.   http://circularscience.com/

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Chasing loudspeaker-level Crosstalkt
« Reply #20 on: February 13, 2015, 09:54:22 am »

tens of kOhm is effectively open circuit for speaker impedance loads.

If the signals are clean at the amp, and crosstalk occurs at the speaker end, it must be something. I ASSUme speaker grounds are similarly tens of K impedance separated?

JR

Speaker grounds are 100+kohm to each other
Logged

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 16379
  • Hickory, Mississippi, USA
    • Resotune
Re: Chasing loudspeaker-level Crosstalkt
« Reply #21 on: February 13, 2015, 10:07:18 am »

Speaker grounds are 100+kohm to each other

If I understand your measurements that means the grounds are not common so crosstalk should not be from that path.

Troubleshooting involves divide and conquer. so how about removing another variable.

Connect one side amp to speaker. Connect the other pair of wires to other speaker. At the amp end of that other pair shirt the two wires to each other but do not connect them to the amp. 

Does the crosstalk go away... If not the wires are the guilty playa...

JR 
Logged
Don't tune your drums half-ass. Listen to what a properly "cleared" drum sounds like.   http://circularscience.com/

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Chasing loudspeaker-level Crosstalkt
« Reply #22 on: February 14, 2015, 10:01:05 am »

If I understand your measurements that means the grounds are not common so crosstalk should not be from that path.

Troubleshooting involves divide and conquer. so how about removing another variable.

Connect one side amp to speaker. Connect the other pair of wires to other speaker. At the amp end of that other pair shirt the two wires to each other but do not connect them to the amp. 

Does the crosstalk go away... If not the wires are the guilty playa...

JR


I'm 99.9% convinced the cabling is at fault, but will short one side (not the amp!) Monday and have a look.
Logged

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 16379
  • Hickory, Mississippi, USA
    • Resotune
Re: Chasing loudspeaker-level Crosstalkt
« Reply #23 on: February 14, 2015, 11:29:48 am »



I'm 99.9% convinced the cabling is at fault, but will short one side (not the amp!) Monday and have a look.

While it seems unlikely (to me), if speaker level current is being induced into the wires, shorting them at the amp end will replicate the low output impedance of the amp but completely without any possible amp signal, This will prove that it isn't crosstalk inside the amp but being induced into the wiring.

Good luck

JR
Logged
Don't tune your drums half-ass. Listen to what a properly "cleared" drum sounds like.   http://circularscience.com/

ProSoundWeb Community

Re: Chasing loudspeaker-level Crosstalkt
« Reply #23 on: February 14, 2015, 11:29:48 am »


Pages: 1 2 3 [All]   Go Up
 



Page created in 0.093 seconds with 22 queries.