Tech Square

Posted by Jacques (A) on October 02, 2006 at 08:21:55

In Reply to: Re: Wrong. And advice for gusser posted by fatbottle on September 14, 2006 at 06:49:56:

The supply will have a higher impedence with the diodes , peaks drawn by the amps WILL be compressed .

They would be if the LM3875 supplies were to lower to a point where they wouldn't be able to give the requested output voltage for the peak output power.
Is it possible?
The first question to answer is to determine the value of the voltages and currents we are dealing with.

• 
  The peak power @±34volts PS on Z=8 ohms is calculated (output dropout voltage=3V, P=U_RMS²/Z, where U_RMS is the max steady output) : 60W .
  Note I didn't considered 4ohm speakers, since the chosen 10A power supply would be unsufficient to the job on 5 LM3875),
  
  
• 
  Since any lowering of the ±34volts would lower the peak power, we need to specify some real-world peak power, in order to allow some lowering of the supplies.
  Let us accept difference in SPL of 1.0 dB, unnoticeable.
  The required power is 48W.
  
• 
  It implies a minimal power voltage of U_RMS²=PZ=19.5volts, plus the dropout voltage = 23Volts.
  We get those 48W thanks to a power current = 2.44A.
  

So, the question now is:
What minimum value for the local cap to provide those 23V under a load of 2.44A, while the mains instantaneous voltage is too low to charge again the cap?

To answer this question,

• first, how long will this time last at worse?
  Considering the xformer has secondaries @25.5V_RMS, you get peak of ±36V, to get 34V after 2 diodes dropout), and you get 23V+2 @ an angle of 44 degrees. So, the caps won't load between angles of ±44degrees on each half-wave, about 1/4 wave period, which is 4.1 milliseconds.
  
• 
  A current of 2.44A during 4.4 ms depleted the cap by 2.44*4.4.10^-3 = 10.74mC
  
• 
  For this charge depletion to give a voltage drop lesser than 34-23 = 11V, the cap shall be >10.74.10^-3/11, that is > 1,000uF
  

Since the chosen value (5600uf) is superior to this value by a very confortable margin, we can say there won't be any compression

Obviously, to make things sure, one should know do some simulations to check the analysis, then some measurement on the real thing, but I'm rather confident...

You forgot to mention that schottkeys should also be fitted for rectification

No, I didn't on purpose.
I just don't like schottkys directly in contact with mains :

1. they are prone to fail on main transients, even very short.
2. they have a very short inverse recovery, which is not a good thing in this app, generating short a short RF transient at each main commutation.
   This transient is hard to get rid off, since most power supplies are not laid out to deal with RF correctly.
   
3. A diode with lengthy inverse recovery time is much better fitted to rectify 50/60Hz. I wouldn't use them on SMPS @200KHz obviously, since the loss at each commutation would soon burn them out! But on low-frequency PS, they are perfectly fit.

And what about the commutation spike?
Paralleling the advised diode with a RC (R=50-80ohms, C=10nF, voltage rating be thrice the rectified voltage, 100V here) will dampen out any excitation from commutation of the diode. It can be completed by a 47nF cap across each secondary to shorten the secondary capacitance of the transformer. On the later, use caps with voltage rating 5 times the regular peak voltage to be insensitive to any likely voltage transient (so 160V for the current case)

Post a Followup: