Recommended internal cables for amplifier kits

[hyperknot]

I'm trying to build / mod some Class D kits (ICEPower, Tripath modules). So about internal cabling, what are exactly the recommended cables for power, signal and speakers? I have some questions.

- power: I don't understand why is it better twisted than straight. In the walls, in the power cords, it's always just parallel wires, why should we twist it once it enters the amplifier chassis?

- signal: for unbalanced, I've read that coaxial is the right choice. But do I need a special solid core coax cable, or I can just cut and reuse any RCA interconnect?

- speakers: I was thinking about using the very same wire (11 AWG) I'm using for the external speaker cables, untwisted. What is the point in using thinner wires but twisting them inside the chassis?

 

[DonH56]

Hand-waving explanation...

A class D amplifier generates switching noise that generally extends beyond 1 MHz. Inside the chassis everything is exposed to it and thus the recommendation to twist and/or shield wires to provide better noise rejection. Outside the chassis, assuming it is grounded, radiation is blocked by the chassis itself so not an issue (hopefully).

A lot of us twist out of habit; it was often the case in the past that power and input/output jacks had small caps to block external RFI, and twisting helps maintain that rejection from anything inside the box. And it makes for a cleaner look plus easier to dress a twisted pair rather than two loose cables.

Any RCA interconnect should be fine. Stranded conductors are more flexible so easier to route. Just watch out for stray wires.

HTH - Don

 

[ajawamnet]

Not so sure about twisting unbalanced cables... that can lead to common mode signals cross-coupling, increased capacitance and some other nasty stuff like oscillations. Twisted pair is used for balanced signals to insure tight coupling for common mode rejection and differential characteristics.

https://www.maximintegrated.com/en/app-notes/index.mvp/id/4218

http://www.bluejeanscable.com/articles/balanced.htm
"Confusion of these concepts is fairly common, and understandable. One will often hear in audio discussions that "twisted pairs provide superior noise rejection," because it's often assumed that it is the cable construction itself, rather than the equipment circuitry, that accounts for common mode noise rejection; as we've discussed above, it's really the combination of the two which account for the phenomenon. This misconception sometimes leads to people using shielded twisted-pair balanced audio cable as an unbalanced interconnect; they will ground one of the two signal wires at both ends of the cable, and then ground the shield--sometimes at both ends, but sometimes only at one end, causing a loss of shield effectiveness. The problem with this sort of construction is that it dramatically increases the capacitance of the cable by adding the shield to the one side. Instead of just the capacitance between the two conductors, one now has the total of (1) the capacitance between the two conductors, and (2) the capacitance between the signal wire and the shield. As capacitance in audio cable is very definitely an enemy, this is a serious sacrifice to make, especially when there is no noise rejection benefit. Coaxial cable, not twisted pair audio cable, is the right choice when connecting unbalanced components. "


BTW - a refreshing cable company https://www.bluejeanscable.com/aboutus.htm :
"We entered this business after a bit of frustration in shopping for audio and video cables for home theater -- everywhere we looked, we found nonsense claims about the characteristics of what ought to have been relatively straightforward products. Our background is not in the airy heights of audiophilia, and not in the anything-to-make-it-cheaper world of importing containers full of Chinese goods, but in the reality-based, engineering-based world of broadcast-quality communications cabling. Our original mission was bringing standard broadcast cabling products into home theater installations; as we've grown over the years we've taken on an increasing amount of custom product made to our own specifications, but our insistence upon solid engineering-based quality, as opposed to fluffy sales talk, remains the same. We won't try to sell you anything you don't need, and when the cable you already own is probably doing what you need, we won't try to upsell you. And with that orientation toward the broadcast world comes another principle: we price our cables in a manner that reflects pricing of similar assemblies in the professional broadcast market, rather than trying to scale the pricey peaks of the boutique cable landscape. "

 

[DonH56]

Twisting will increase capacitance, true, but for interconnects inside the box that is usually in the mud and should certainly not cause oscillation. A twisted pair typically has roughly the same capacitance as coax IIRC, remembering coax puts the shield all around the center conductor though spacing may be larger from center to shield.

Common-mode signals will couple into the pair regardless; twisting will improve CM rejection depending on the turns ratio. Whether that matters depends upon the application and aggressor signals, natch.

I agree that coax is the best choice for shielding, and differential connections with an outer shield are even better.

 

[Krunok]

Outside the chassis, assuming it is grounded, radiation is blocked by the chassis itself so not an issue (hopefully).

I just tried an interesting experiment: I set my son's cellphone on autoanswer, put it in the microwave owen, clsoed the door and called it from my phone. It ranged and picked up the line. So, it seems that shielding of the microwave owen which is supposed to block 2.2MHz signal from the magnetron isn't able to block 900MHz signal from the GSM network. Just to mention that my microwave owen is not some cheap crap but a relatively expensive Sanyo product. So much about Farraday cage stuff..

 

[ajawamnet]

Twisting will increase capacitance, true, but for interconnects inside the box that is usually in the mud and should certainly not cause oscillation. A twisted pair typically has roughly the same capacitance as coax IIRC, remembering coax puts the shield all around the center conductor though spacing may be larger from center to shield.

Common-mode signals will couple into the pair regardless; twisting will improve CM rejection depending on the turns ratio. Whether that matters depends upon the application and aggressor signals, natch.

I agree that coax is the best choice for shielding, and differential connections with an outer shield are even better.

On some SMPS (switch mode power supplies) I've designed (mainly DC-DC) we've encountered issues with stability depending on the topology (SEPIC seemed a bit more sensitive). When we went to parallel conductors the oscillations/thumping went away. We had to rewire around 50 chassis (and the Molex microfits kinda suck to de-pin) I heard about it from the techs for a few weeks...

 

[DonH56]

I just tried an interesting experiment: I set my son's cellphone on autoanswer, put it in the microwave owen, clsoed the door and called it from my phone. It ranged and picked up the line. So, it seems that shielding of the microwave owen which is supposed to block 2.2MHz signal from the magnetron isn't able to block 900MHz signal from the GSM network. Just to mention that my microwave owen is not some cheap crap but a relatively expensive Sanyo product. So much about Farraday cage stuff..

Microwaves don't have that great of shielding... http://hyperphysics.phy-astr.gsu.edu/hbase/Waves/mwoven.html The seal around the door tends to get dirty and leak plus the glass can be spotty. I have not put a field meter on one in a long time but I remember thinking "shielding" barely applied. It (probably) won't cook you... And cell phones have fairly sensitive receivers.

I would advise you to refrain from turning the oven on whilst the cell phone is inside. The results may be... undesirable.

 

[DonH56]

On some SMPS (switch mode power supplies) I've designed (mainly DC-DC) we've encountered issues with stability depending on the topology (SEPIC seemed a bit more sensitive). When we went to parallel conductors the oscillations/thumping went away. We had to rewire around 50 chassis (and the Molex microfits kinda suck to de-pin) I heard about it from the techs for a few weeks...

Well, that sucks! I am NOT an SMPS designer but am responsible for designing the filters from them (also DC-DC) to the analog circuitry on our chips so do get looped in on the design reviews at times. I would not have imagined the output of an SMPS is that sensitive to twisting the wires. How does it handle decoupling caps? I do have to add snubber Rs in series with the larger caps to keep noise peaking down, but AFAIK we have not had one be unstable snubbers or no (just exceed our noise mask). That said the SMPS designs I've worked with recently are all on PCBs so no long wires (but sometimes lousy planes).

We have had a few layout issues (they are tricky as you know; must really watch the loop currents) and one layout had the output feedback trace run right under the output switching inductor (yeah, that worked -- not!) but generally I have not seen one go unstable like that. Makes me wonder exactly what caused the instability and seems like something to watch out for. Our designs drive a lot of stuff and myriad of capacitors and filter beads but are also low-voltage (albeit high-current, e.g. 1 V at 20+ A) and pretty decent power pours to limit IR drop so maybe just has not been an issue in our application. I have seen them do flaky things under light load (like a dead chip that doesn't draw much power). I do have to measure the noise and that is usually a surface-mount SMP connector to a long-ish coax into a DSO so maybe 100~200 pF net. No instabilities seen but that is also after the analog filters.

My experience with external SMPS designs is very limited, mostly to playing around with eval boards and rarely HV (50 - 250 V) designs that drive longer wires to the amp/load, so I may have gotten lucky and just not seen the problems you've experienced. And my background is more high-speed/RF oriented so load tolerance and stability is critical but again I am not designing high-power circuits (that's for the front-end guys).

 

[Krunok]

I would advise you to refrain from turning the oven on whilst the cell phone is inside. The results may be... undesirable.

LOOL How did you know I was tempted?

 

[DonH56]

LOOL How did you know I was tempted?

Actually, depending upon how much you like the cell phone and/or microwave oven, the results may be desirable...

Ages ago my grandmother put a fine china tea cup in her (new) microwave to warm the tea and burned off the gold-leaf rim. She was quite vexed.

 

[g29]

Actually, depending upon how much you like the cell phone and/or microwave oven, the results may be desirable...

Ages ago my grandmother put a fine china tea cup in her (new) microwave to warm the tea and burned off the gold-leaf rim. She was quite vexed.

My mom put a metal pot into her microwave and fried it. Luckily, only one part fried inside and it was an easy fix (replace the scorched thingy) once the correct part was sourced.

 

[ajawamnet]

Well, that sucks! I am NOT an SMPS designer but am responsible for designing the filters from them (also DC-DC) to the analog circuitry on our chips so do get looped in on the design reviews at times. I would not have imagined the output of an SMPS is that sensitive to twisting the wires. How does it handle decoupling caps? I do have to add snubber Rs in series with the larger caps to keep noise peaking down, but AFAIK we have not had one be unstable snubbers or no (just exceed our noise mask). That said the SMPS designs I've worked with recently are all on PCBs so no long wires (but sometimes lousy planes).

We have had a few layout issues (they are tricky as you know; must really watch the loop currents) and one layout had the output feedback trace run right under the output switching inductor (yeah, that worked -- not!) but generally I have not seen one go unstable like that. Makes me wonder exactly what caused the instability and seems like something to watch out for. Our designs drive a lot of stuff and myriad of capacitors and filter beads but are also low-voltage (albeit high-current, e.g. 1 V at 20+ A) and pretty decent power pours to limit IR drop so maybe just has not been an issue in our application. I have seen them do flaky things under light load (like a dead chip that doesn't draw much power). I do have to measure the noise and that is usually a surface-mount SMP connector to a long-ish coax into a DSO so maybe 100~200 pF net. No instabilities seen but that is also after the analog filters.

My experience with external SMPS designs is very limited, mostly to playing around with eval boards and rarely HV (50 - 250 V) designs that drive longer wires to the amp/load, so I may have gotten lucky and just not seen the problems you've experienced. And my background is more high-speed/RF oriented so load tolerance and stability is critical but again I am not designing high-power circuits (that's for the front-end guys).

Actually it was the input... it was an LT part.

 

[DonH56]

Actually it was the input... it was an LT part.

Ah, seen similar situations (not quite that sensitive), but a topic for another thread. Good input (no pun intended) though, and a reminder to look at the broader picture.

Thanks! - Don

 

[hyperknot]

So the short conclusion is to twist the power cables, coax the unbalanced signal cables and what should I do with the speaker cables? Try to twist them as well? And what about using some oversized wires instead, like the same 11 AWG ones I used for the room?