Mains supply and HiFi

[martin clark]

^ Yep.
I wish I could be cynical enough to promulgate such myths, and make & market such things ..but:

• The disturbance is in deliberately-upsetting/intterupting a non-technical people's approach to enjoying their music more.
  
• The cynicism follows the classic model of Religious conversion: - present a false dilemma; force a crisis of understanding - then, offer a/the way out. Count it a sale.

Cult-ish; or swap an l for an n as you feel appropriate. As an 'industry' /sales approach to making money its nothing less than an utter disgrace.

 

[Helen Bach]

I have the TV on ATM, actually a TiVo system, hooked up via a Cambridge Audio DacMagic 100 to the hifi (two CA 640v2 feeding a stereo pair of Usher 604 floor standers). My partner is a ceramicist, and her kiln in on, at the moment. That switches 26 amps on and off, and the voltage drops from 245v to 236v when it's on.

I cannot hear any switching, from the kiln or fridge or two deep freezes. The house wiring is 70 years old. No special mains just for hifi.

BUT in front of me is a coil of mains wire (TITANEX HO7-RNF 3 CORE 1.5MM 15A NEXANS BLACK RUBBER MAINS POWER CABLE), some new mains plugs and three Furutech FI-11 IEC Connector lookalike plugs, coloured braid and heat shrink.

I will soon make three mains cables, much like the Atacama Imperium Powercord UK Mains Power Cable, at £950 a pop.

It may not sound exactly the same, but I will have a smug grin...

 

[John Phillips]

...A true objectivist would know that industry, science and medicine all use mains filtering devices to improve the quality of mains supplies in critical environments - just take a look at Schaffner's catalogue for thousands of examples - all free of any hifi 'woo'. ...

Exactly. But if the competent objectivist engineer knows his equipment is still sensitive to things like RF on the mains even after he takes care of good power supply hygiene, then he builds the necessary filtering into the kit before it goes out of the door.

It's the cynical marketeer who deliberately omits to do that and looks for after-market sales to extract more money from the customer to cure the problem he left unfixed in his kit.

 

[John Phillips]

I think it depends on your mains supply. Where I live the sine wave is very pure with no distorsion and no noise. I live in the country – there’s no mains pollution at all, or any pollution for that matter.
Mains conditioners have no effect.
For people who live in crowded cities, near factories, that’s entirely different.

In the UK at least, and I am sure in France too, there are standards that the mains power provider has to meet. A competent equipment designer will know enough about these and be aware that he/she needs to cope with, as just one example, 5% individual harmonic distortion and 8% total harmonic distortion up to a certain frequency. These standards still apply in crowded cities, near factories etc.

The dirty secret is that mains providers do not get it right all of the time. But I suspect they do almost all of the time. In a very few cases, special arrangements may possibly be beneficial if you know exactly what is wrong and what specific arrangements may help. General-purpose solutions, marketed in such a way as to get customers to think they are cure-alls, are unlikely to actually hit the real target except in relatively few cases. But people are observably well-disposed to hear improvements.

Forums like this and other means are undoubtedly used (or abused) in order to increase the level of universal uncertainty and paranoia and so boost the sales of cures for problems a customer may not really have.

 

[a.palfreyman]

Dimensions in #58 for my earth plane weren't quite right. 27mm wide (assuming 35µm thick = 0.9mm2 CSA) and 65mm long. Pushing 7.6V (i.e. 1A) rms through a dummy load, I measured between the 'dirty' and 'clean' 0V points at opposite ends. At 1.2kHz, I got 2mV p-p sinusoidal (0.7V rms) on the scope plus some traces of the rectifier peaks. Unfortunately (with hind-sight) I didn't try any other frequencies. Reinforced the earth plane with 2.5mm2 CSA (mains, single strand wire along each line of cap tails) and re-measured. Now 0.5mV p-p at 1.2 kHz, so a 12dB improvement. It was only at this stage that I thought to check at 12kHz, which measured 1.5mV p-p and was reminiscent of an R-C charging / discharging trace. Anyway, initial impressions were of improved clarity, but didn't get enough time to assess properly. Hopefully will get some time next W/E.

 

[S-Man]

From the other mains cable thread, but THIS ^^^ is the nub of why I started this thread.
Now, if I can get some sort of oscillator / notch filter working properly, I MIGHT be able to measure what is going on. I suspect my earthing strategy around the power amp PSU is dodgy. Copper clad (circa 0.035mm thick by circa 20mm wide, about 0.7mm2 CSA) with 3 pair of caps, amp front-end ref at the 'clean' end with speaker return at 'dirty' end, adjacent to centre tap with these points being about 90mm apart. I suspect that this 0V plane has enough resistance to produce voltage differences between these two '0V' points due to the loading and charging currents. As a stop-gap, I might just arrange it so I can drive 7.5V rms from one channel of the power amp into a pair of parallel 15R 25W resistors as a dummy load (i.e. 1A rms current) and stick my o'scope earth to the clean end and input probe to the dirty end. At least that will show me what the 0V plane impedance is adding to the speaker 0V return if nothing else

I cannot see the logic of returning the speaker 0V to the dirty end of the supply.

There are two more logical places to return the speaker 0V:
1) To the clean end of the main smoothing caps = the cleanest point with very low impedance
2) To the local 0V on the power amp PCB where the decoupling caps for the output stage meet = electrically closer to the the signal 0V - which is what you want the output referred to.

If you do it by method 1 or 2, then you don't need to worry too much about the impedance of the "0V plane".

 

[a.palfreyman]

S-Man,
I believe that this is how it was done with (Avondale) Cap 6 / Mini Cap boards, and is I believe how many do it, namely front end ref to 'clean' end, speaker return to 'dirty' (centre-taped) end.
NCC200 so no local decoupling caps on-board the amp PCB.
I am going to put a new return into the centre of my 0V plane (as this will be even lower impedance as it will sit in the centre of the cap-bank rather than at one / other end) and return ALL to the same point.
Cheers,
AP

 

[Pete MB&D]

The current comes out of the transformer and through the amplifier board/speaker so needs to go back as close to the transformer as posible, to avoid voltage drops across tracks in the circuit board/wires.

Pete

 

[John Phillips]

I cannot see the logic of returning the speaker 0V to the dirty end of the supply.

There are two more logical places to return the speaker 0V:
1) To the clean end of the main smoothing caps = the cleanest point with very low impedance
2) To the local 0V on the power amp PCB where the decoupling caps for the output stage meet = electrically closer to the the signal 0V - which is what you want the output referred to.

If you do it by method 1 or 2, then you don't need to worry too much about the impedance of the "0V plane".

Yes.

There's no such thing as an "equipotential ground". In other words, a voltage - including a signal voltage - is another name for a potential difference. So you always need to identify the right two points in a circuit's layout between which a signal really appears. They will be different at different stages. And one of those points is never a generic "ground".

As a new graduate let loose after hours amongst a research lab's well-stocked test kit to make my own audio equipment, I found that reality quickly punctured my over-confidence, and taught me that designing essential hygiene in power supplies, power grounds, signal paths and signal grounds (note the plural) was essential before I got to the clever circuit design stuff.

 

[Jim Audiomisc]

I learned a lot about the matter of 'ground' when using 2 kV Klystrons. 8-]

In one case because an idiot at Cambridge had failed to reconnect the ground return and I got a serious shock. Apparently I did a back-flip. Fortunate, though, as that was better than a 'death grip'! But I, erm, 'discussed' the issue with the relevent idiot afterwards..

The other was getting shocks from a telescope at 14,000ft. The dry air and mounting of the telescope, dome building construction, etc, made the concept of 'ground' a bit vague in practical terms. So the telescope used to drift up to a few kV when no-one had touched it recently. 8-]

 

[S-Man]

Yes.

There's no such thing as an "equipotential ground". In other words, a voltage - including a signal voltage - is another name for a potential difference. So you always need to identify the right two points in a circuit's layout between which a signal really appears. They will be different at different stages. And one of those points is never a generic "ground".

As a new graduate let loose after hours amongst a research lab's well-stocked test kit to make my own audio equipment, I found that reality quickly punctured my over-confidence, and taught me that designing essential hygiene in power supplies, power grounds, signal paths and signal grounds (note the plural) was essential before I got to the clever circuit design stuff.

Not really sure what your point is. Where did I mention "equipotential ground"? The whole point of this discussion is that grounds are not equipotential - as AP has measured.
Perhaps I have assumed that the reader knows that the "signal 0V" is the point where the input 0V and the feedback 0V meet. This is the voltage to which the input AND output ought to be referred to.

 

[John Phillips]

Not really sure what your point is. Where did I mention "equipotential ground"? The whole point of this discussion is that grounds are not equipotential - as AP has measured.
Perhaps I have assumed that the reader knows that the "signal 0V" is the point where the input 0V and the feedback 0V meet. This is the voltage to which the input AND output ought to be referred to.

I was just agreeing with you with a bit of generalization. No more.

 

[S-Man]

Apologies if I misunderstood, John.

 

[martin clark]

..becos a pic is worth,etc:

Here is how Doug Self describes it; and this matches the as @S-Man describes: for minimum distortion, you need input signal loop directly compared with feedback from the output-signal loop. An ampifier is a measuring device, reproducing the input signal at a defined gain. So, ask where is that gain most-exactly-defined..?

If the speaker 0v return is elsewhere, say midway/wrong(dirty) end of a close-chain of 'deliberately-decoupled from each other reservoir caps' - even via just the pcb trace impedance linking each cap in series, then there IS a noise voltage superimposed on this diagram on the output-return side. And - that relies on limited feedback to subtract; and the precision of gain definition is lost.

Such a layout is denying your amplifier the best performance it could have - regardless of architecture, or whether or not you agree with Doug Self: this stuff is inarguable algebra.

 

[S-Man]

Thanks Martin, the pic makes things much easier!

Fig 25.1 is pretty much my #1.
#2 would be with "F" moved close to the "Local HT decouple". That's how my NCC300 is configured. The logic is to keep the HF loop local decoupling/output devices as tight as possible, however the impedance of E-F then goes up a bit - which is like having slightly higher ESR/L main caps (if I understand correctly).

But the key point is your 2nd paragraph.

 

[S-Man]

Contentiously relating this back to post #1...

Once you have the internal amp layout well optimised, then you have to consider what happens when 2 channels and a few boxes are connected together. Each box is (usually) tied to mains earth and you have a ratsnest of 2nd and 3rd order couplings and impedances - electrostatic, magnetic, capacitive (e.g. output transistor collectors to heatsink) etc.
It seems possible that the impedance and physical layout of mains cables could affect this.

 

[martin clark]

Thanks Martin, the pic makes things much easier!

Fig 25.1 is pretty much my #1.
#2 would be with "F" moved close to the "Local HT decouple". That's how my NCC300 is configured. The logic is to keep the HF loop local decoupling/output devices as tight as possible, however the impedance of E-F then goes up a bit - which is like having slightly higher ESR/L main caps (if I understand correctly).

But the key point is your 2nd paragraph.

To my way of thinking - keeping F-G-H short and stout is the key; Kirchoff's 1st rule and all that - and quickly followed by ideally, that being closely-coincident with a fat T, at point 'E' - not on the transformer- charging side: but from geometric centreline of 'least charging current impressed on main reservoir caps'; to be clear: definitely the 'output' end on a multi-cap reservoir pcb.

 

[chiily]

I don't quite get that diagram for the signal gnd.

I once wired a 323MC board that the input signal gnd went directly to the star earth of the PSU, an ALWSR. It sounded terrible. When I moved the input signal ground directly to the 323MC's gnd pin and a wire from that point to the star earth of the PSU it sounded so much better. That diagram, if I read it right says the opposite for the input gnd.

TBH this thread is making me rethink the earthing in my Stasis amp, which has a bit of a buzz.

 

[a.palfreyman]

@chiily,
If I've understood you correctly, you had all your return currents in series with your return path of a very low output MC cart by doing it that way. In this situation I always take the signal and return to the circuit input and provide a local trace back to the star earth point, just for the input part of the circuit, plus any local feedback cap.
@martin clark,
Interesting diagram. I'd also pondered using local decoupling for HF directly at the board, but the thing that concerned me is that the load (speaker) should return to the centre of the local decoupling for best effectiveness (this was discussed in another thread, but can't place it now) to avoid the inductance of the loop 'H' and the speaker return wire. However, this gives you two different 0V points, one for HF signals at the local decoupling caps and one at the main psu for LF signals, separated by the impedance of 'H' that will have higher LF currents on it. In my case, 'H' would need to be circa 130mm long and circa 8mm CSA to keep Z really low.

 

[S-Man]

To my way of thinking - keeping F-G-H short and stout is the key; Kirchoff's 1st rule and all that - and quickly followed by ideally, that being closely-coincident with a fat T, at point 'E' - not on the transformer- charging side: but from geometric centreline of 'least charging current impressed on main reservoir caps'; to be clear: definitely the 'output' end on a multi-cap reservoir pcb.

I agree. However, in my NCC300 I could not keep this short due to buiiding my "double decker" arrangment and the size of the pcb. I elected to move F & G towards H, which means E to F is long. As mentioned before this is effectively the same as increasing the ESR of the caps by the resistance of the E-F wire. Not ideal, but the best compromise.... I hope.

I don't quite get that diagram for the signal gnd.

I once wired a 323MC board that the input signal gnd went directly to the star earth of the PSU, an ALWSR. It sounded terrible. When I moved the input signal ground directly to the 323MC's gnd pin and a wire from that point to the star earth of the PSU it sounded so much better. That diagram, if I read it right says the opposite for the input gnd.

TBH this thread is making me rethink the earthing in my Stasis amp, which has a bit of a buzz.

I think this breaks the rule that the input ground and feedback ground should be tee'ed off from the power star. Single rail designs are even more succeptible to common impedance coupling than dual rail designs in general. The ALSW is a very low impedance device, which is a double edged swrod and potentially means much larger currents flowing in supply rails. Naim "softened" everything with a 27R resistor for (some) good reason.