[andyr]

I have been reading various posts on various Fora about:
* use of 2- or 3-cord mains cables,
* relation of signal ground to mains earth,
* unbalanced IC construction,
* signal ground connection options,
* IGP (Interchassis Ground Potential) and consequent hum.

As the philosophy which I adopt seems to be different to what I read in these posts, I thought I'd set out my philosophy re. - in particular - unbalanced systems and invite your critique as to what might be wrong with it. 

Firstly, mains cords.  IMO, unearthed, 2-core mains cables only exist for cost-saving reasons.  If you wanna do it right, then:
* the mains cord is 3-core
* whether it's a captive mains cord or you have an IEC chassis-socket, the earth wire/earth tag is connected to the (metal) case.  This:
a) acts as "safety earth", and
b) creates a shield against RFI for the signal processing going on inside the case.

Secondly, input & output RCA sockets.  These are always isolated from the case.  IE. signal ground does not touch chassis - hence never touches mains earth.

Thirdly, unbalanced IC construction.  Simple RCA-based ICs have the shield acting as as the 'ground' / 'signal return'.  Given that the shield picks up RFI, what you automatically have is RFI noise being injected into signal ground at the destination RCA socket.

IMO, what is a better idea is to have an IC made up of 2 wires ('signal' and 'ground') enclosed by a shield which is connected to the chassis at one end only (source or destination - I don't think it matters).

Fourthly, signal ground ... as mentioned in 2, above, signal ground floats throughout the system; it never connects to any chassis.

Fifthly, IGP.  Yes, which way round the primary winding of the power transformer is connected to the mains makes a difference to the chassis ground potential ... and the sound of the component.  (And the lower voltage reading gives the better sound.)

However, if you adopt the previous 4 articles of faith, it makes no difference to the level of hum.

Interested in your input, guys.

Regards,

Andy

[Daryl]

Firstly, mains cords.  IMO, unearthed, 2-core mains cables only exist for cost-saving reasons.  If you wanna do it right, then:
* the mains cord is 3-core
* whether it's a captive mains cord or you have an IEC chassis-socket, the earth wire/earth tag is connected to the (metal) case.  This:
a) acts as "safety earth", and
b) creates a shield against RFI for the signal processing going on inside the case.

Unbalanced systems cannot reject ground signals and it is important that their gounds be influenced as little as possible.

You do not want to ground unbalanced componets to the mains ground and that's why you see componets with two wire mains cords.

By not having a ground connection the only thing capable of influencing the componet ground is the primary/secondary capacitance of the power transformer and airborne electrostatic and electromagnetic fields and hopefuly the power transformer is chosen for low interwinding capacitance.

You want your componets grounds to be points of very high impedance so that the shields of your interconnects will easily be able to pull them to the same potential with very little voltage developed from end to end of the interconnects shields.

Your componets do not reject voltage differences between componet grounds but instead add them to the signal so anything influencing componet grounds can cause current through interconnect shields and a resulting voltage since these shields have a finite impedance.

With the componet ground isolated from the case as your are talking of below you could ground the chassis.

I would worry that this might add more capacitance between the componet ground and the mains ground but in some situations it will give an improvement and others not.

Secondly, input & output RCA sockets.  These are always isolated from the case.  IE. signal ground does not touch chassis - hence never touches mains earth.

Isolated inputs and outputs will stop the case from acting as an antenna for interference.

Isolated multiple inputs where the switching circuitry disconnects signal grounds from unsused componets will help minimize ground influences by having as few componets connected to signal ground as possible.

The componets circuitry will need strategicly located groundplanes/cages protecting it from from outside noise.

Thirdly, unbalanced IC construction.  Simple RCA-based ICs have the shield acting as as the 'ground' / 'signal return'.  Given that the shield picks up RFI, what you automatically have is RFI noise being injected into signal ground at the destination RCA socket.

IMO, what is a better idea is to have an IC made up of 2 wires ('signal' and 'ground') enclosed by a shield which is connected to the chassis at one end only (source or destination - I don't think it matters).

That is how you should construct BALANCED interconnects.

Unbalanced gear cannot reject ground differentials between componets unless it happens to have floating/balanced/differential inputs which you sometimes see (mostly in car stereo amplifiers).

With unbalanced gear you want each componets ground to float at as high an impedance as possible then you want the grounds of your interconnects to have as low an impedance (end to end) as possible.

A tubular ground shield in a quality coaxial cable (like LC-1 from Blue Jeans) has vastly lower resistance (due to the amount of metal used) and lower inductance (due to the tubualar shape) than the '-' wire in two wire interconnects and is much more effective at shunting ground differentials.

Low inductance for the '-' conductor is important so that componet grounds will be shunted together throughout the spectrum and not just at DC.

Also you get nearly double the cable capacitance when you drive a shielded twisted pair with an unbalanced source vs. driving it with a balanced source as would be appropriate for that type of cable.

A shielded twisted pair does offer the highest possible noise isolation with balanced gear and ground lift selectors (balanced gear only) will let you choose the most effective grounding scheme for the shield which will vary depending on the chief cause of interference.

Fourthly, signal ground ... as mentioned in 2, above, signal ground floats throughout the system; it never connects to any chassis.

Yes the rejection of ground loops in unbalanced systems will depend upon componet grounds floating at as high an impedance as possible and the ground shields of interconnects having as low an impedance (end to end) as possible.

Fifthly, IGP.  Yes, which way round the primary winding of the power transformer is connected to the mains makes a difference to the chassis ground potential ... and the sound of the component.  (And the lower voltage reading gives the better sound.)

Yes

There will always be some capacitance between the primary and secondary of the power transformer thus you will want the mains 'neutral' wire connected to the side of the winding which is closest to the secondary.

[andyr]

Thanks for your reply, Daryl, but some of what you said is beyond me ... so I need to make some comment.    Hopefully, we're on the same page!

Unbalanced systems cannot reject ground signals and it is important that their gounds be influenced as little as possible.

You do not want to ground unbalanced components to the mains ground and that's why you see components with 2-wire mains cords.

I absolutely agree with your statement but I suggest a better way of achieving this is to have components with 3-wire mains cords (with the earth wire connected to the chassis as "safety earth") but keep signal earth floating - so you never connect signal earth to the chassis.

Thirdly, unbalanced IC construction.  Simple RCA-based ICs have the shield acting as as the 'ground' / 'signal return'.  Given that the shield picks up RFI, what you automatically have is RFI noise being injected into signal ground at the destination RCA socket.

IMO, what is a better idea is to have an IC made up of 2 wires ('signal' and 'ground') enclosed by a shield which is connected to the chassis at one end only (source or destination - I don't think it matters).

That is how you should construct BALANCED interconnects.

Correct but the above is a "psuedo-balanced IC" ... for those folk who have unbalanced components.  My contention is that it delivers a better sound than a standard coax IC.

With unbalanced gear you want each component's ground to float at as high an impedance as possible then you want the grounds of your interconnects to have as low an impedance (end to end) as possible.

A tubular ground shield in a quality coaxial cable (like LC-1 from Blue Jeans) has vastly lower resistance (due to the amount of metal used) and lower inductance (due to the tubualar shape) than the '-' wire in 2-wire interconnects and is much more effective at shunting ground differentials.

You could use a thicker wire for the ground leg in a 2-wire-plus-shield IC.  This would give a low resistance and not pick up the RI that a coax shield will.

Also you get nearly double the cable capacitance when you drive a shielded twisted pair with an unbalanced source vs. driving it with a balanced source as would be appropriate for that type of cable.

"Double the capacitance" in ICs is not really a problem if Zin of the following component is >50 Zout of the preceeding component.

Regards,

Andy