3-Phase PSU for Class A amp?

Hi all,

I've just found this site and I'm considering several class A circuits for possible construction/experimentation.

Power supplies seem to be quite a problem for this configuration so I thought I'd start by building a bench supply.

These may be a stupid questions, but would a 3-phase (or more) supply, with each phase feeding the supply caps via it's own rectifier reduce the ripple a bit or just make it far more audible than the original 50/60Hz? AC motors use hefty caps to simulate a second phase from a single phase supply, would this work with torroidal transfomers? Has anyone tried this without being showered in little bits of foil?

Thank's for reading my first posting here,

Graham
 

grataku

Member
2000-12-31 9:31 am
-
This is what I remember about triphase systems, please check on some ee texbook so what I am saying may very well contain major ******** (I took classes on the subject 18 years ago and never used it since). To get the reduced ripple you need a tri-phase line, transformer and rectifier. When you are done you should get one DC line output. I think the phases on the AC line are well...out of phase by some multiple of 360 deg (either 120 or 60) so the ripple frequency is increased but it's intensity is dramatically reduced by some factor like 3 or sqrt(3) with respect to a mono-phase rectifier. It nearly looks like DC even without filtering caps.
I think in order to work this way the system as to be a pure tri-phase system so trick with capacitors to multiply phases won't do what you want, I don't think.
 
Thank you for your reply Grataku,

I think this is worth further investigation as the higher ripple frequency should be easier to filter out anyway.

Besides rotary converters, there must be some elegant three phase conversion circuits out there, if anyone knows of any please let me know.

I'll try creating three phases from a single transformer secondary first as this would save the expense of extra transformers. Here in the UK 55-0-55v 1000va transformers are reletively cheap though, packed in yellow plastic tubs as isolation transformers for power tools.

I'll let you know how it goes.
 
Sorry Hoffmeyer,

You posted as I was writing my reply to Grataku,

I'd be surprised if simulation caps did work, it seems too easy, the values and power ratings needed could be quite hard to find (especially on a 50hz/240vrms supply) and the distorted waveforms generated could cause no-end of trouble.

I'm still intrigued though.
 
You need 3 transformers or a 3ph transformer

You need three transformers. If you really want to do this well, why don't you rectify the mains and convert it into 400Hz and get some surplus aviation transformers (which run at 400Hz)?

3 phase is good, but the only advantage you get is a higher frequency (since your ripple is low) output which results in less ripple.

Typically 3 phase lines are also noisy since industrial equipment runs on it.

Do something radical and make a very good switched mode power supply. That would surely benefit this group as everyone could use it.
 

grataku

Member
2000-12-31 9:31 am
-
Petter
how do you propose to do ac remodulation to 400 hz without introducing a bunch of extra noise?
Personally, unless I had other reasons to have a 3-phase line to my house I wouldn't mess with it. Just use a nice pi filter or a cascade of them in your AMP power supply and rewire your house with a big dedicated line from your braker directly to the stereo outlet to try and keep noises form your household appliances, and resistive voltage losses to a minimum.
 
Yes, the ripple frequency is increased, and this reduces the ripple due to the decrease in time between charging pulses to the capacitor.

The rectifier conducts only briefly at the tranformer's voltage peaks. The ripple is defined by

i = C dV/dt

where i is the peak DC load current, C is the total filter capacitance, dV is the ripple voltage, and dt is the time between the charging peaks. This is a simplification, but represents 90% or so of the ripple. The finite time to charge the capacitor through the dynamic impedance of the rectifier is a secondary effect.

In effect, the value of dt (delta time) is divided by three when using 3 phase, which changes delta V by the same factor.

All that said, transformers are more expensive than capacitors. It would be cheaper to triple the capacitance than to go to 3 transformers.
 
A solution I have seen in a commercial high end power supply uses 3 transformers with double secondaries (central tap). I think they should be rated at one third the current of a normal power supply since the three currents add up.
What you do now.
Each phase goes to each transformer´s primary.
The neutral goes to all the transformers other free primary cable.
Each now transformer has it´s own rectifier bridge.
Center tap to ground and each output to bridges ac input.
So we have so far 3 identicle power supplies like we know them with + and - rails.
We now parallel all the + outputs from the bridges and all the - outputs.
We use smaller caps for each voltage and thats it.
For the RF noise we use snubbing caps around 100 nF parallel to each diode of the bridges (better if directly on the bridge).
Also we use a X type cap wired between live and neutral as close as possible to the inlet. 100nF - 470 nF.
 
mmmmm (scratches head after reading all your posts)

Here's an idea that probably won't work but it's a starting point for this experiment. Please rip it apart, especially if any of you lucky people have circuit simulation software to plug a few values into.

[IMGDEAD]http://freespace.virgin.net/public.stuff/graphics/sixpot.jpg[/IMGDEAD]

Thanks for reading.
 
How to do it without introducing too much noise

Frequency conversion is relatively simple, but as is noted, there is a great chance of introducing more problems than you take away.

Typically, one would make DC using perhaps some form of switched mode technique. Then one would ideally use pulse modulation (or resonant techniques) to create the desired frequency.

One method is to perhaps use Tripath's demo-boards. Then all you need is an oscillator at it's input -- to provide any waveform you desire. You can also change output filter down to further remove noise.

I think this is a pretty interesting way of decoupling the mains. I know Krell or Mark Levinson uses 400Hz. I assume that a similar method is employed by the top power conditioner companies.

Another interesting thought is to reduce effect of wiring BEFORE amp by inserting power factor correction circuitry to increase conduction length and enforce sinusoidal draw. All that is required is a controller IC, a FET switch an an inductor. I am looking into doing this down the road.

Any self-respecing book on power electronics will have information on how to do all this. Also, application notes from National, Linfinity, Maxim, IRF and others may be useful.
 
Only having 2 P-channel MOSFETS lying around, I decided to try a simpler version of the six-pot to see what happens.

This is what I ended-up with.

[IMGDEAD]http://freespace.virgin.net/public.stuff/graphics/fourpot.jpg[/IMGDEAD]

But while building it, I had another idea...

A row of caps all charge up together. The first begins to discharge into the load and once the voltage has dropped relative to the second cap, a mosfet connects the second cap to the load. Once the voltage of the second cap is lower than the third cap, a mosfet connects the third cap to the load. Once the voltage of the third cap... and so on.

If you like flashing lights, connect LEDs to the drains of each mosfet, to form a bargraph representing how many caps had to fire off during each cycle.

But seriously, the ripple current should never be more than the turn-on threshold of the mosfets, and the ripple frequency will increase with the loading current.

Watch this space...
 
A 3 pahse bridge rectfier power supply produces an unfiltered DC voltage with 360hz ripple with an amplitude of about 10-15% of the DC voltage. You will either need special 3-phase transformers or 3 single phase units. The cost seems very high to me unless your amp is so power hungry that house current can't handle the load. I saw a huge class A monster that used a cloths drier plug. That would be 220v at 30 amps or 6600 watts! The power compnay will love you. My expeariance is with large AM and FM transmitters that put out 5KW or more. Less powerfull transmitters are usally single phase.