So, I think I need to build myself a DC blocker. There are plenty of circuits available, and even bare PCBs available.

The question is, what kind of current capacity should I allow for? With the soft start circuit, I can run my amp with a fuse of about 5 amps, so if I went for 10A, this should give me enough headroom? 

The fuse rating was calculated using the following:

So if I build a DC blocker with 10A constant current capacity, it should be more than adequate?

I’ll add a little disclaimer - if you don’t know what you are doing with mains electricity, then don’t mess around with it. It can, and if given a good enough chance, will kill you.

I’ll be interested to know how you get on with your DC blocker. I’ve got a similar buzzing / humming on my Audiolab 8000p. It’s not at an intrusive level and doesn’t affect listening, but I can hear it when the music’s off. Seems to be quieter at some points then louder at others.

I’ve seen that ifi Audio and Audiolab have DC blockers available for around the £100 mark. So have considered this. As mentioned the toroid buzz isn’t that loud that I can hear it while listening to music. It’s just that nagging feeling that I know that it’s there. 

This seems to be the Audiolab one - 

Not much in it for £100. A bridge rectifier, and a few caps. Oh, and the certification of a product, and the Audiolab marketing costs.

I think I can build two for less that, based on the ATL HiFi PCBs.

I can assure you that "DC on the mains" is not DC, although it is possible to measure it as such.

Toroidal grumbling is due to the transformer's large inrush current as it has to adjust to the received waveform's sudden changes in shape.

The cause of this change is often found close to home or in the home. It is a "recent phenomenon" due to energy-efficiency of other mains powered equipment.

It is the firing of thyristors in energy-efficient products and one example is the automatic washing machine. If only one half cycle is used, the energy consumed is lower than if the whole cycle is used. It has been noted elsewhere (an electrical engineers forum) that even a half-speed hair dryer is capable of causing a change in mains waveform.

Obviously it also depends on the mains wiring impedance, which must be such to allow the sudden "attenuation" to one side of the waveform. Even where an electrical installation has two separate supplies from the meter tails, this waveform change still happens. As such, it might be seen that the electric meter's "shunt resistance" might be all the impedance required. It should then follow that the source of "interference" must be inboard of the electricity meter.

How this energy-efficiency affects a toroidal transformer is shown by monitoring its magnetising current. During an event, one side of the magnetising current waveform "disappears" while the other side "grows", and after the event, it returns to be equal for each half cycle.

The reason this happens to a toroidal transformer and seldom to any other type, is because toroidal transformers have much lower magnetising current. It can range from units of milliamps, to a few tens of milliamps (about 30mA).

Unless the transformer has primary taps to which it is optimised to the local mains, the manufacturer must be careful to ensure it doesn't reach saturation at the highest envisaged supply voltage.

The universal 230V EU voltage can have a spread between 207V and 254V.

To prevent saturation at 254V the magnetising current at 207V will be only just sufficient, and might be in the order of 5mA or lower.

Therefore it is possible that use on a continental 220V supply might result in a worsening of the noise.

The rectifier "DC filter" might work to reduce or stop the noise, but doesn't always work. Whether the large can electrolytic capacitors have any effect is debatable. As they are supposed to let AC pass unhindered, and the problem is the AC waveform, then all the rectifier does is drop the voltage by a volt.

One large American professional amplifier manufacturer simply used a bridge rectifier without the capacitors. I tried it and it seemed to do much the same. I note that the later models of the amplifiers no longer use them.

You can find more on this subject at https://sound-au.com/articles/xfmr-dc.htm - https://sound-au.com/articles/xfmr-dc.htm

The transformers were wound to my local voltage.

I am sure it is my heat pump causing the issue, so I put a DC blocker in place yesterday evening (using a 600V 10A bridge rectifier and a couple of 4700uF capacitors) to test, and the hum is gone. I have no issue in the summer, so it ties in that the efficient heat pump is causing the issues by putting all sorts of noise on my mains.

Based on this, I will be building a suitable DC Blocker to fit inside my amplifier, as it will be a neater solution than another external box. I have space in the case for something suitable.

I have the PCB on the way from Sweden, and the diodes and capacitors on the way from the States.

Finding somewhere with the diodes and capacitors in stock was a challenge for the parts I wanted.

Total cost is about €90 for the parts, but you could do this cheaper. I have gone big on this, probably too big, but this is what the PCB was built for. It has 16 diodes, and 6 capacitors. Apparently, it works well on large toroidal transformers.

It should not take long to solder up, but will take a little longer to fit to the amplifier, as I may have to redo the wiring to fit it in the case.

I will post more when it is complete.

I am following attentively.

So I guess that's three filter sections because of the six capacitors, but I'm not sure what the extra four diodes are for.

Following the circuit diagram which I assume cannot really differ much from all the others, it is a DC "passer" which drops a total of 6 diode drops in both positive and negative directions.

The AC voltage drop would be exactly the same but for the bypass capacitors.

By the way, figure 2 from Rod's site shows exactly the behaviour I described earlier, and can illustrate photographically on my 'scope if I have the spare time.

It is interesting to see Rod's figure 4 - I shall have to mock that up in the reality of the real world and also photograph it.

As Rod says, the plots are almost identical, but I would hazard a guess that the capacitors are adding a teeny weeny bit of phase lag.

Detail of the DC blocker is available in the link above.

The designer references the same text that you linked to earlier.

I went for this one based on the feedback on the DIY Audio forum.

I am sure it won't be much different from the one I have currently fitted externally.

You did get his written permission? I wouldn't want him to come after the website's owner (me).

Estimated cost: £10 ex case

Be bloody careful!

Nope.So I removed it.

It is available on his website for anybody that is interested.

Why not put a hyperlink to the image instead? Something in the order of "click here to see the circuit" using the hyperlink tool (^ up there in the reply composer). That would put it in another browser tab and be on his website, and it would still be available for all to see.

I know you're well meaning and you put the link in image tags which isn't really uploading but he wouldn't be able to see the HTML.

I wouldn't want anybody to think I was pushing my latter circuit contribution - it's just another idea that does the same number of DC stages.

With 50A bridge rectifiers, I assume that some kind of thermal dissipation should be provisioned?

Obviously, it won’t be pulling anywhere near 50A, but something like that could happily feed my entire system?

(Replying to the post before last)

Not really. There are a few lying about on my R&D bench, and they're overrated for the job, as you say, but they're highly popular for some reason at the moment, so are very inexpensive.

I've been using them in my amplifier experiment's power supplies because they're pretty "bomb proof".

Apart from switch-on inrush, the primary current to your amplifier's transformers will be less than two amps, with amplifier at full output, so the bridge's won't get hot at all. Mine have a metal base for heatsinking, which is really well insulated from the diode junctions, but even so, I'd insulate them when used on the mains.

And the voltage doesn't really matter, as the reverse current flow sees the opposing polarity diodes limit the reverse voltage, and vice versa.

The capacitors can be any voltage above say 6V, but 16V is common, and there's nothing to stop the constructor using a higher voltage rating. I think the combined series-parallel rating of what your building at 102,000 uF is a bit OTT. I too have used a couple of 4700uF, but maybe 10,000uF (giving 5000uF because of the series connection) might rid all worry about phase lag.

The parts I ordered at the end of last week have already been delivered.

Just waiting for the PCB now, which I will hope will arrive by the weekend, so I can build and test it, and hopefully find the space and time to install it.

So the PCB arrived today, I built it and then tested it as per the instructions provided by the manufacturer. All good, so on to the next phase.

Where do I put this in my amp? Oh, easy, at the back where I left some space, except the space was too small. So I had to move the PSU boards a bit, and the soft start a bit, and then I had enough space. So I fitted the board, then checked the connection to earth on the chassis, and to the IEC socket was a nice low resistance reading.

Then I connected the DC blocker to the power input, checked polarity and continuity between the connections. All good. Connected the DC blocker to the soft start, and checked continuity of the connections and polarity. All good.

Connected up the power cord, and nervously powered on the amp.

The amp powered up, no buzzing from the transformers, and more importantly, no magic smoke escaped.

It’s a win for me.

Would I do it the same way again if I started again? No. The finished board was a pain to fit, and the project, although fun, was more expensive than it needed to be.

Anyway, I am happy that my amplifier is running silently. It is the first time for a while.

The product works, but so would the circuit that Graham posted. I think it is going to take longer to power down the am now, and I am certainly not going to be putting my hands in the case very quickly after I have removed the power feed with that additional capacitor bank.

As always, proceed at you own risk. Mains can, and will kill.

I wanted to test the "DC blocker" in the real world, so I built what I drew earlier in this topic, and tested it on the 'scope using my current probe. I used a regular toroidal transformer rated 160VA. I also measured the AC current.

It does work, but not as good as claimed. The upper current waveform could be seen to sag ever so slightly. My test interference comes from a Candy auto washer.

However, without the "blocker" it sagged considerably more, so it's good?

Well, it has to have losses, obviously, and with the "blocker" the mag current read about 15mA, and without it 18mA.

These are rough measurements plus or minus a few fractions of a milliamp, but shows there is a loss.

My interest in this is for the '70s amp so as to prevent deterioration of sound quality during interference. The SQ is excellent in the dead of night, but not as good when the automatic washer or other daytime things are on.

Now, the '70s amp has a high mag current transformer for reasons of regulation, but it is still a toroid - a rather special one the transformer manufacturer has worked hard to develop. It has 80mA mag current, so should not be taken so "one sided" as a regular toroid, but as the SQ depends on this mag current, then a partial collapse, would, I think, worsen the SQ.

It will take the total mag current down slightly as demonstrated above, but perhaps only by a few mA, so should be OK. The important thing I think, is waveform symmetry.

Does it work? First impressions are it does, but I'm not so easily kidded, and will give it time.

I'll do the 'scope shots shortly.

'Scope shots as promised

Normal mag current waveform without blocker and no interference (except random noise that is).

Mag current waveform without blocker with interference source on mains supply.

Mag current waveform with blocker and interference source. I pressed the stop button to show the slight one-sidedness which comes and goes with the interference source. Still, quite a good result.

Apologies for the mucky 'scope screen.

Making a safer "DC blocker" (or transformer primary current centraliser)

Not much room in the amplifier case, so the blocker had to be made compact enough to fit in any spare space away from sensitive parts of the circuit.

Using 4,700uF capacitors instead of 10,000uF improves the blocker's response. I chose capacitors rated for 6,000 hours at 105C, with the hope they'll run cooler extending their lifespan to years.

I also chose the 25V rating as the ripple current is about 250% larger than the largest expected peak current, for this particular capacitor.

The proximity of the rectifiers to the capacitors was a worry in that the rectifiers can dissipate 1.5W each in this application. However, that is at full amplifier power which only ever happens on the test bench when testing for maximum output. Normally the rectifiers will dissipate less than one watt.

I decided to attach a thermocouple to monitor the temperature in actual use.

The insulation has to provide 100% coverage meaning encapsulating the capacitors in heat shrink tubing. I would not normally recommend this as the safety vent is bunged-up, and if the capacitors were ever to fail and liberate their contents, it would create a very loud bang. However, I am the "Guinee pig" and will test this over a long period.

If the temperature is kept well below 105C there should never be a liberation. As it would be contained inside the amplifier case I would venture to say it should be safe.

The best I could imagine to fit the minimum space

Connections soldered (using another short piece of stiff wire through the middle hole to solder the common capacitor ends together either side).

The capacitors are insulated with some 3-1 heat shrink tubing with as little heat applied as possible.

The thermocouple held temporarily in place with a rubber band

I made use of an IEC connector insulating boot to cover all the live connections, held in place with a cable tie.

This is the finished compact "blocker" ready to be located inside the '70s amp.

The compact fully insulated "DC blocker" subjected to maximum amplifier mains supply current for over one hour ranged from 55C to 72C with 72C being recorded with the worst interference, which causes the rectifiers to conduct most of the time.

Additional heating came from the amplifier output stage and the transformer, with the case getting unbearably hot, indicating an internal temperature of more than 50C.

At 1WPC, however, the "blocker" cooled to 37C. With no interference and no signal, it cooled to 31C.

From the above I think it safe to suggest the maximum temperature in normal use could be 55C.

The load life of the capacitor, a Rubycon XYG is 6000 hours at 105C at 4.22A ripple. The ripple is rated at 100kHz and also is stated in r.m.s. At 100Hz it is de-rated by 0.7, making it 2.95A. The maximum current in this application is 750mA, and the temperature from above, is 55C.

We need a lifetime graph to estimate the lifetime, but none is shown in the data sheet, so I borrowed that data from https://www.farnell.com/datasheets/2244110.pdf - https://www.farnell.com/datasheets/2244110.pdf which will give a good enough approximation.

The actual ripple divided by the rated ripple is 0.75/2.95 = 0.254, and we can see this corresponds to a 15x multiplier. 6000 x 15 = 90,000 hours, which is just over 10 years.

I am happy with the lack of humming traffos now. Does it make a difference to the sound quality?

I have no idea…

Or even the reduction in noise.

My amplifier is based on the LM3886, and apparently has the following specs:

With that amount of NFB - about 30dB available at 20kHz for a voltage gain of 20 - then distortion and noise simply vanishes (fig.47 LM3886 datasheet). The point though is what it does within the first watt, where most of the detailed listening takes place. No matter how good a circuit is, external factors can be more prominent. Not that a power supply should be considered an external factor, but is often considered that way.