diyAudio (
-   Power Supplies (
-   -   Cancellator: a magic active CM noise canceller to upgrade any brick or module SMPS (

Elvee 15th August 2019 09:16 PM

Cancellator: a magic active CM noise canceller to upgrade any brick or module SMPS
6 Attachment(s)
It's Christmas again!!!!

Well, not quite yet, but it looks like it: the Cancellator is a simple, active circuit that cures one of the most unpleasant flaw of class II SMPS's, the common mode noise.

Class II supplies are attractive because the absence of earthing makes them floating, thereby eliminating a major source of ground-loop problems.

Unfortunately, there are no free lunches in engineering, and this advantage comes at a cost: they spit a significant amount of HF trash on the output circuit, and unlike differential-mode disturbances, no amount of filtering can eliminate CM noise. CM chokes should in principle be able to do it, but their effect is rather imperfect, due to relative impedances, and anyway, all SMPS's are already fitted with one or more.

Recently, Daniel sent me a supply example to evaluate its suitability for his pet amplifiers: EZmos and OldFashioned.

Attachment 775493

The supply delivered the goods, and would be sufficient for two amplifier channels, but I also noticed its horrendous CM noise. Here is the CM voltage picked up at the output GND:

Attachment 775494Attachment 775495

The oscilloscope is earthed, thus the waveform is measured wrt. earth. There is a large 50Hz component, which is normal, but regions of the waveform also show a significant HF hash, between 20 and 30Vpp. Such a residue can cause various symptoms in audio equipments: hissing, buzzing, whistling, etc.

The 100Hz modulation is caused by the conduction of the input rectifying diodes: when they conduct, the primary circuit is connected to the mains, and the level is heavily reduced. In-between, the converter circuit is fed from the filter cap and is left floating, causing a maximum of disturbances. It is not always the case, and sometimes an opposite situation is observed.

This converter module is particularly poor, which means it is an ideal test-bed for a correction circuit, the Cancellator:

Attachment 775496

The principle of operation is extremely simple: the output GND is taken as a reference, and an amplifier actively forces the input 0V (or neutral, or whatever) into a virtual ground by injecting a suitable counteracting signal. Note that in fact the opposite might be true, but action=reaction, and the amplifier does what is needed to make the two grounds equipotential.

As always, the actual details are a bit more gory: in particular, it is essential to do the above whilst respecting the integrity of the safety isolation barrier. To do that, two Y-rated capacitors are used: one for the input, one for the output. The output one is a 4.7nF, the maximum reasonable value for such an application, to keep the correction path as low impedance as practical. It is also necessary to provide a supply for the amplifier, something the 36V output does perfectly.

How does it work? Quite well: here are the before/after oscillograms, you barely notice a thickening of the trace on the after pic:

Attachment 775497

Here is the breadboarded prototype:

Attachment 775498

SemperFi 15th August 2019 10:16 PM

Interesting. How high up in frequency does it work? What is the frequency of the noise in your example?

Elvee 15th August 2019 10:40 PM

1 Attachment(s)
The base switching frequency is ~50kHz, but most of the feedthrough is from the edges, because it is a squarewave, meaning significant harmonics at 500kHz and beyond.

This is why I used a good old LM318: it has a good bandwidth, and accepts the 36V supply.

Many modern alternatives have a much wider bandwidth, but they are often more limited in their supply range. One could think of reducing the supply voltage, and it could be sometimes possible, but with this example, a really large correction amplitude is required: significantly more than 16Vpp.

This pic shows the correction signal (measured with a differential probe) along with the corrected signal:
Attachment 775515

The bandwidth only depends on the opamp used: with something better than a 318, it would be possible to go in the tens of MHz range, but it does not look particularly necessary: the cancellation level reaches almost 40dB, because the converter module includes basic measures like a CMC and snubber

trobbins 16th August 2019 12:50 AM

Yet again - super scheme Elvee.

Early days, but I can see a strong interest related to commercial plug pack supplies such as laptop supplies, but also 12VDC supplies that I have used that don't have mains protective earth connecting through to 12VDC output. Hassle with commercial plugpacks is that there is no mains side electrical connection available, unless some form of interface socket/plug is prepared for easy access to the local protective earth.

Elvee 16th August 2019 02:34 PM

2 Attachment(s)
Yes, practical connection details are going to be an obstacle.

However, since the DC output voltage is not always going to be sufficient to properly power the cancellator, an alternative needs to be found anyway.

With this example (OK, rather extreme), the DC supply couldn't be smaller than 20V, otherwise the correction amplifier would begin to saturate and clip.

Thus, I devised the reverse alternative, where the amplifier is referenced and supplied from the mains. The mode of operation remains exactly the same, but the supply and correction amplifier could now be housed in a box inserted between the mains and the brick (or adapter). It brings some additional complication, but it is workable:

Attachment 775637

The ~8mA required by the opamp is supplied by a special capacitive supply: it has a symetrical configuration to ensure a continuous low impedance path to the mains for HF signals. The rest of the circuit is essentially the same, except that now the output ground is servoed to the input, but the end result remains exactly the same. Here is the prototype:

Attachment 775638

Note that an actual implementation needs to be built properly, not improvised, and housed in an insulating case, to ensure an adequate insulation from the mains. It is also important to respect the Y and X ratings of the capacitors where indicated

Elvee 16th August 2019 06:25 PM

I have also tested other opamps, beginning with a 741.

You probably guess the result, but I had to do it and begin somewhere. It managed to improve the situation compared to nothing, but it was clearly lacking: a significant uncorrected residue remained visible.

Next one was a LF411. This was a surprise: it did barely better than the 741. By contrast, a LF356 did almost as well as the LM318. It is probably due to a tighter "grip" of the OP stage, compared to the LF411. Finally, a NE5534 did as well, or even slightly better than the 318, confirming that the robustness of the OP is paramount for such an application.

In very difficult situations, an additional buffer might become necessary.

Elvee 16th August 2019 09:06 PM

Some details and context about the measurements:

I made them in my lab environment, without attempting to change anything: the scope has a 6 feet mains cord, and is connected to the master lab power switch via an extension, probably 5 feet or so.
The probe is also 5 feet long, and the DUT is connected via a "suicide cord" 5 feet long, and another extension 5 feet long.

This is in fact very similar to a typical audio configuration (except for the suicide part, hopefully), meaning inductances and parasitic effects are reasonably realistic and should reflect a real life situation quite closely.

Elvee 18th August 2019 10:16 AM

2 Attachment(s)
The circuits I provided for both versions are "ideal", in the sense that the input neutral/0V reference is extracted from the average of the two mains wires.

Thus, even if they are not equipotential (in HF), the reference will be optimum.

In practice however, the level of differential disturbance will be small compared to the CM, and anyway, 100% of SMPS have a X-cap at the mains input. This means that taking the "quiet" reference from just one side of the mains is going to be perfectly sufficient.

This allows some simplifications: here are the pruned circuits:

russc 18th August 2019 03:05 PM

It is a common question on the forum how to use 2 power bricks for symmetrical supplies.
So in anticipation, I will ask, can the circuit be adapted for use with 2 power bricks for this purpose and how would the circuit look?

Elvee 18th August 2019 05:43 PM

2 Attachment(s)
Here is how I see things, for one, two or more supplies.

You would build the cancellator into a case of this style (or a multiple outlet plug or extension):

A supplementary flying lead fitted with an alligator clip would connect to the output ground.

In the case of multiple supplies, they would of course have their output grounds connected together, and that's where the alligator clip (or other connector) would be attached.

A single cancellator can correct multiple supplies sharing the same input and the same ground, it will generate whatever composite signal is required to correct all the supplies at once provided its dynamic range is not exceeded.

The scheme is applicable to bricks, wall-warts, etc

BTW, you could earn thousands of $$$ selling these gadgets to audiophiles (its only flaw is that it actually works, unlike snake oil supposed to do the same)

All times are GMT. The time now is 09:01 AM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 17.65%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
Copyright ©1999-2019 diyAudio