P3A Supply
P3A Amp Page: [1] [2] [3]


Half of a P3A supply

  Load R   Rails   Current   Ripple   Reduction
R1=1  
R2=2
R3=2

 150  46.285V  308mA  1mV  -38dB
60  41.640V  694mA  3mV  -35dB
30        
R1=0.5  
R2=1
R3=1

 150  47.344V  316mA  5mV  -24dB
60  43.700V  728mA  11mV  -24dB
30  39.290V  1310mA  18mV  -24dB
R1=0.25  
R2=0.5
R3=0.5

 150  47.878V  319mA  14mV  -15dB
60  44.791V  746mA  32mV  -14dB
30  41.008V  1370mA  55mV  -14dB
C only
C=25mF 

 150  48.415V  323mA  76mV  -0dB
60  45.931V  766mA  167mV  -0dB
30  42.87V  1430mA  288mV  -0dB
C only
C=.1F 

 150  48.660V  324mA  25mV  -10dB
60  46.410V  774mA  52mV  -10dB
30  433.620V  1450mA  87mV  -10dB
The table to the right was generated by successive runs of PSU Designer. Although the test supply is a single rail, the real split-rail supply is still modeled accurately, because we are primarily inspecting the filter, not the transformer. This P3A amp is biased at 300mA, so low ripple at and above that draw is necessary. At 100mA bias, less filtering would be required for low hum.

What do these numbers say? It's clear that the RC solution has much less ripple, and it should also be clear that it has much greater supply modulation for a 1 amp peak-peak draw. The real story is slightly more subtle; for one thing, although modulation can differ by a few volts, on a log scale this is less than 3dB of difference compared to the C-only filter.

More telling, we know that music tends to have a crest factor of no less than 6.67. (see Stereophile) This means that an amplifier which is conventionally rated at 60W RMS will only be outputting about 9W RMS power. (good news for our voice coils!) At 70% efficiency, this is a current draw of about 300mA...does that number sound familiar? As long as the final capacitor can store enough energy to handle the transients, the peak currents through the supply resistors will be close the RMS current, and little modulation will occur.

In a sense, this means we are (almost) getting something for nothing; a split-rail RC supply of 50,000uF outperforms a C-filter supply of 200,000uF (reduced ESR of parallel capacitors was taken into account). Not only does the RC filter provide a 20dB+ reduction in ripple, it also softens the hard triangular wave of a typical capacitor filter, reducing higher-order harmonics of the mains frequency. The major disadvantage of this approach is that continuous power is limited from the rail-implied 100W to about 60W. In reality this is not much of a problem; the higher rail voltage grants greater headroom on transients, and unless driven into clipping most music will not load the supply into collapse.

The downside to this approach is it does require some calculation as the numbers change with supply voltage, bias current, and target power handling.

Summary

An RC filter seems like a great idea even for a class AB amp like the P3A. It is somewhat more attractive at higher bias currents, as one can then minimize modulation. The numbers here will differ to some degree from real life, but the advantage of simulation is that we are already close to the final solution.


Other considerations

Doodads
DC protection should be a must-have for anyone who cares about their loudspeakers. ESP's Project 33 does nicely in this regard, and is what is used in my own amp. Current limiting (short circuit protection) should be considered, but may have to be rejected because of its limitations and effect on sound quality. (not used here) A clipping indicator could also be a nice thing to have.

MOVs
MOVs at the mains input are a nice idea. They provide some surge protection, and will absorb mains transients caused by equipment switching, etc. MOVs should be added between each pair of connections, so three are needed for Hot-Neutral, Neutral-Ground, Hot-Ground. Get the large 20mm kind; their cost is not that much greater, and their power and surge current ratings are superior. MOVs must be added after the mains fuses, as they fail as a short circuit.

Mains filter
A filter can also be a good thing, but shouldn't be applied indiscriminately. These filters are not as effective if your mains power is unbalanced (i.e. if it doesn't pass through an isolation transformer) and can actually make things worse in some situations. An advantage that holds true under any circumstance is that a filter will prevent power noise from the amp from getting into the mains.

Ground isolation
There are at least a couple of options available to prevent ground loops from forming. The one used in this amp is the ground loop breaker described here.

Inrush limiting
Some form of limiting may be required with toroidal transformers, which have higher inrush current than El type xforms, or when using extremely large capacitor banks. (>100,000uF) Failure to include limiting can cause fuses or even bridges to blow. ESP describes a relay-switching solution that puts a resistance in series with the input for a few seconds at turn-on. Another common solution (used by Nelson Pass) is a power thermistor in series with the input. Note that the RC supply described above does provide some limiting, but when expanded to five channels still displays very high inrush.
 
©2002 Tyler Thornblade  Contact Info  Disclaimer