Power Supply Mains Filter

The thread is quite straighforward: I would like to have a schematic of a good general purpose mains filter for audio use, to build it myself. I have always seen types of L-C combinations, some use caps to ground, others don't.

I am far from being familiar with the subject, but I would like to know basic things about the philoshophy behind their design. For example:

- Why not use a filter that cuts everything above 100-200 Hz? We are aiming at a DC power supply after rectification.
Answer I have read a few times: This limits the power supply's reactions to energy demands of the amp. If the transformer (with the filter, of course) cannot supply current for frequencies up to enough kHz (50-100?) the amplifier will, arguably, experience a severe loss of dynamics. Thus, damages to both the slew rate and frequency response of the amp occur.

Of course, this implies that the capability of the DC power supply does not entirely depend on the reservoir cap. Should it?

The above sets for me the question of what a proper filter's frequency response should be.

Last but not least: many people say toroidal transformers our troublesome regarding polluted mains - little leakage inductance, little protection. So, why not put an inductor having a value that is commonly measured as EI transformer leakage? I now think that we go again to the first question of dynamics.

So, I will be happy to hear your opinions and share your experience on this! And good designs that have worked for you, should you have any. :)
 
Call me crazy, but I always like the un-regulated power supply for my class A/B amplifiers.I know that since I use fewer parts, I have to use the best parts as I can. if you see fig#3 in the link down...that is what I always use!. I never like "complicated" power supply. The simpler the better for me!
Solid State Power Amplifier Supply Part 1


and for the transformer: Last time I build two same amplifiers, one using EI transformer and the other a toroidal transformer, the one using toroidal transformer was quiet as a mouse but the other amp using the EI transformer was noisy mechanically and it had some nasty "white noise" that I could not get rid off for the magnetic field that i produced. I had to run the wires very far away from main inputs and amp boards and even doing that I could not get a "clean" result as with the toroidal transformer. With the toroidal transformer was a piece of cake and zero noise.

Low Weight
Because they are more efficient, toroidals can be up to 50% lighter, (depending of power rating), than traditional E-1-transformers. Low weight simplifies end product design by reducing mounting hardware and supporting enclosure requirements.

Small Size
Most toroidals are smaller than their E-1-transformer counterparts. Electrical and mechanical designers, when "painted into a corner" by a minuscule space allotment for power supplies, appreciate a toroidal's compact dimensions.

Flexible Dimensions
compounding the benefits of low weight and small size is the flexibility to vary dimensions. As long as the core cross section is held constant, the height and diameter for the toroidal may be economically varied to accommodate equipment design requirements - a great help when designing low profile, slim line equipment.

Easy to Mount
A single center bolt easily and quickly mounts the torodals, avoiding costly mechanical design and practical problems associated with traditional E-1-laminated transformers..., and three bolts are eliminated at assembly.

Low Stray Magnetic Field
Toroidals have no air gaps: primaries and secondaries are wound uniformly around the entire core. As a result, toroidals emit very low radiated magnetic fields. This makes the toroidal ideal for application in CRT displays, high quality amplifiers, and medical equipment.

Low Mechanical Hum
the core of a toroidal s formed from a single strip of grain-oriented electrical grade silicon steel tightly wound in the form of a clock spring with the ends spot-welded i place. the copper wire is wound over polyester film, forming s silent, stable unit without give or varnish coating.

Reduced No-Load Losses
Compared to traditional E-1-transformers, toroidals exhibit extremely low no-load losses. In applications where a circuit is in a "stand-by" mode for long periods, the potential cost reductions for power can be significant.
 
Last edited:
Ott tells us that the mains input filter must be bonded to the enclosing chassis for best performance.
He further points out that typical SMPS filters that are integrated onto the SMPS PCB can never optimise the attenuation. Some escapes from the filter when it is not screened from the rest of the circuit.

How would you enclose your filter inside a screening box bonded to the screening chassis?
 
Ott tells us that the mains input filter must be bonded to the enclosing chassis for best performance.
He further points out that typical SMPS filters that are integrated onto the SMPS PCB can never optimise the attenuation. Some escapes from the filter when it is not screened from the rest of the circuit.

How would you enclose your filter inside a screening box bonded to the screening chassis?
Using another chassis? A small aluminum box (a Hammond one etc.) using the fact that the bottom is mounted using bolts into the main chassis.
 

ticknpop

Member
Paid Member
2005-05-28 9:43 pm
toronto
Torroids are small, but not great power transformers. They are wide band devices and pass whatever line noise there is up several hundred kilohertz.
Getting a torroid built with an electrostatic shield primary to secondary reduces this bandwidth and noise (this isn't the external shield version and has a separate ground lead) - also AC filters as shown above on the primary side using a 2uf to 10uf AC LINE rated cap (same for low value caps to ground they MUST be line voltage approved) with appropriate IEC/UL approvals helps, smaller caps on the secondary side of the transformer and snub the soft recovery diodes, also use small chokes/resistors between DC filter caps sections.

EI core power transformers pass a lower bandwidth, but are larger, heavier etc, not common but electrostatic shields are available here too (the very earliest Krell KSA 100 used a shielded EI Core). R core are better but not found big enough for power amps. Richard Marsh I believe measured the power bandwidth of a Hafler DH 500 EI core power transformer and it was 60 kHz.
 
Last edited:
The schematic above has a resonance at around 3 kHz, and starts to cut well before 20 kHz. That is my main question - does it matter if the mains filter cuts things in the audio band?

The supporters would say "yes, because it also cuts audible noise and does no harm to dynamics, since these are solely defined by the following power supply".

The enemies: "no, because you need a response of up to 50-100 kHz to support amp slew rate and dynamics".

Which of them is true? Is there anyone who has done a side to side comparison to see whether there are losses in the sonic quality if the filter cuts before 20 kHz? :rolleyes:
 
I agree on that AndrewT!

This is what I think: you just don't want to put large inductors in series, because then you would limit the current pulse capability of the whole system, which could degrade dynamics. Other than that, you don't need any wide transformer frequency response.

It would be telling to see a study on the current pulse demands for a power amplifier of a certain wattage throughout the whole audio spectrum. This could safely create a margin of maximum smoothing inductance in series, that would allow the current pulses to always reach the maximum value the amplifier needs.
 
You agree and then tell us that inductors are bad ??????
What will the inductor do to the 50 or 60Hz charging pulses?

Go and read Gootee's very long Thread on determining the minimum capacitance required for the power amp.

Well, I did not phrase that so well!

I mean that you just can't go and add a huge inductor that could limit the total energy that is required by the supply below the minimum required.

I like the idea of inductors! :)

edit: Of course, I am talking about putting inductors before the primary. Or in the secondary, but before the reservoir.
 
Last edited:
Hello AndrewT,

I looked up that thread you said, but is is very big for me to understand where to look. I have started reading it by the way.
Does it iclude information on inductors for filtering? Could you point me at where such information exists in that thread?

Also, about the inductors. They limit the maximum current that can be supplied from the transformer, I assume. How do we know how much energy has to be transported to the power supply during the most demanding musical sequence through the amp? Shouldn't the inductors have such a value that would allow all of this energy to pass?

I say energy, because the current pulses have both a peak value and a duration. Should I use an inductor, I limit the peak value, but what about the total energy? How big can they be before they degrade the energy supply to the reservoir?
 
You will find all the information you need to convert the mains frequency to a DC supply for your amplifier.
You will probably not find much if anything about mains interference attenuation on the Thread, because the amplifier expects that to be done separately.
The mains interference attenuation should not affect amplifier operation.

Back to inductors.
No, an inductor does not limit the current.
It limits the rate of change of the current. That is quite different.
That is a good thing. The inductor can selectively change the ratio of low frequencies to medium, or high, frequencies in the mains waveform.

Current limiting, more usually at start up, is usually achieved by inserting a temporary resistor into the current route.
 
You will find all the information you need to convert the mains frequency to a DC supply for your amplifier.
You will probably not find much if anything about mains interference attenuation on the Thread, because the amplifier expects that to be done separately.
The mains interference attenuation should not affect amplifier operation.

Back to inductors.
No, an inductor does not limit the current.
It limits the rate of change of the current. That is quite different.
That is a good thing. The inductor can selectively change the ratio of low frequencies to medium, or high, frequencies in the mains waveform.

Current limiting, more usually at start up, is usually achieved by inserting a temporary resistor into the current route.

One by one:

(1) That is what I suppose for filtering - I mean, 2-10mH inductors or chokes before the primary to cut the noise down should not bother. Still, no experience on that, that is why I ask.

(2) Of course and I know that it limits the rate of change instead of the peak value etc - but, I do know that in theory and for high power applications (not audio) that don't bother with dynamics. This is why I was reluctant to accept it for audio use. And reading people like you or anyone who answers that tells me that it is good for an audio power supply, gives me more experience on the subject. This comment is to help you realise that I am not unaware of the physics behind this, yet I have minimal experience on whether it does good or harm in audio applications.

(3) I am thinking of installing a soft start circuit that uses just that, resistors in start up.

AndrewT, a question: a large inductor before the reservoir (think it as a big leakage inductance, that I have installed) will smooth the current pulses. I have seen this in practice for inductive loads. I would like to know how it ensures that the proper energy is always transfered to charge the cup. I mean, the pulse will not have the same sharp edge as before, but I deduce from what you tell me that it will not limit the charging of the reservoir. By making the pulse wider?
I mean, I play with some simulations in Pspice, and adding large inductances or small resistances always limits the ripple current peak. My question is, up to what point are we safe to do that limiting?

EWorkshop1708: Thank you for the advice, I always wondered whether dynamics is all about the capacitors. Have you also done any comparisons in real life to realise this? Was there an amp of yours using large and good reservoirs plus high inductance for smoothing that had wide dynamic range, compared to an other one with zero leakage and inductance but worse caps that lost the fight?
 
Last edited:
................. By making the pulse wider?
..........
No,
by making the pulse "rounder", i.e. by selectively attenuating the higher harmonics that are the interference and having almost no effect on the mains 50/60Hz.

But you are still associating musical dynamics with the charging of the smoothing capacitors.
In my view this is wrong.
The dynamics are the CURRENT variations that pass from the decoupling and smoothing capacitors to the speaker, nothing to do with the much slower charging of the smoothing at a 100pulses/second rate.
I play with some simulations in Pspice
if you don't understand the mechanisms you are investigating, then how can you determine the question/s you need to ask the simulator?
 
Last edited:
if you don't understand the mechanisms you are investigating, then how can you determine the question/s you need to ask the simulator?

My point is that you can't have an enormous inductor, just because it is an inductor! Even at 50 Hz, it will have a reactance, and that would mean that the regulation of the transformer/inductor combination would become worse. Loss of power even at 50 Hz. I think you imply that for this to have a serious effect on the charging of the reservoir, the inductor needs to be huge.

I am not talking about dynamics here. But the reservoir demands a certain amount of joules to charge (still, no dynamics here) during each cycle, and if you lose that energy on transformer and inductor power losses per cycle, you will not be able to adequately charge it! This is what I suppose.

For this to become true, one will say "oh you really want to put a huge inductor in place". That's ok - I am just trying to assume how big it has to be to cause problems.

During my studies, I have learnt that big transformers come with 5% regulation, and in KVA ones, this comes mainly from leakage.
- Why don't we make it 1%? Because then, currents during a fault (short circuit etc) could be enormous, thus threatening.
- Why not make it 20%? Because then, we would lose too much energy across the leakage! This is what I am talking about.

I hope I have made myself clear. :)