Home |
Search |
Today's Posts |
#41
|
|||
|
|||
John Byrns wrote: In article , Patrick Turner wrote: As a follow up to the thread I started above, I have some test measurements which may be of interest to people such as John Byrns, who said that the resonant behaviour of the LC filter used with a class B amplifier would result in being forced to limit the signal output to 3dB less than theoretical maximum, or words to that effect. I never said words to that effect! My guess at what you might have been saying was wrong again, eh. What I actually said, complete with typo, was "reducing the available output power by as much a 3 dB on a dynamic basis." And on what dynamic basis? One has to say what one means, and mean what one says.... I didn't say that an LC power supply filter always reduces the output by 3 dB, Nobody said that's what you said. I said "by as much as 3 dB", it can be less depending on the design of the power supply and filter. Well OK, indeed..... I can't actually take credit for the 3 dB number, I took it from the literature on the subject. This is not a new effect, you need read something besides the RDH4, you might want to stop by the library to review the existing literature on the subject. I'd actually prefer to base my general judgements on the measurements I make on my own gear. I just ran a test on the LC input filter described here :- L is a swinging choke with L = 0.35H at 228 mA, and drops to 0.11H at 2 amps, and C = 10,000 uF, The power tranny winding for the full wave SS rectifier has 1 ohm dcR, and there is 65vrms from the winding, so there would be +57 volts at the C with Idc = 228 mA. Mains F is 50 Hz. The resistance to get 228mA was 250 ohms, and I then tried repeatedly adding 32 ohms, to make the supply voltage fall to 53v, and DC current increase to 1.87 amps. The speed at which I added/removed the 32 ohms was varied to generate the highest possible signal voltage across the C, and all was monitored with a CRO, and the voltage variation with an F around 4 Hz was 4.6 vrms, or about +/-7 peak volts, and I seriously doubt that any music signal condition could generate a greater disturbance to the supply. But where the F of adding subtracting the 32 ohms to cause the transient was reduced to less or more than around 4 Hz, the voltage swing of the supply was about 1/2 the 7 volts peak. So when the supply has a sudden 1.6 amp increase in load current, the rail dips to just under the final rail volts before stabilising at 53v, and when the current is reduced 1.6 amps, the rail voltage swings up to above the 57v rail before settling at 57v. There was no more than about 1/2 a "ring cycle". 1/2 a "ring cycle" is more than enough to cause trouble. Your results sound like your design has about a 1 dB loss of output due to power supply bounce, although your test results as presented are incomplete, so 1 dB is only an estimate on my part. 1 dB fits within the "as much a 3 dB" description I posted, obviously your design is not nearly as bad as it might be, although I don't think 1 dB is good enough to meet the FCC specs. for AM broadcast transmitters in the USA when the modulator and RF final amplifier share the same power supply as is typical. The substantially class B SS amp I am building will never be used anywhere near full power. So the odd dB here or there is of little concern. I will know more about the performance when I get the amp actually working... I figured I could eleiminate the bounce of the LC filter when tested with these crazy test signals by using a shunt regulator instead of a bleeder resistance. This reg need only be a power transistor, emitter taken to 0V, collector taken to the rail, with a series 10 ohm R to limit the current in the transistor to a max of 5.7 amps. From the rail to the base a 55 volt zener diode is used, with a 220 ohm base current limiting resistor, and all is set up so about 250 mA flows at idle, and so the dissipation in the power bjt is 14.3 watts, not much in a large amp capable of 650 watts from both channels Should the voltage at the C rise, the bjt conducts, and the current in the 10 ohms is enough to damp the filter. If it falls due to a sudden increase in current demand, the current in the bjt is cut off, thus tending to keep the voltage constant at the C. All that is needed is to keep the changes of rail voltages to what would occur due to rails sagging or rising that one would get with a passive supply. The situation for a class B amp is better if serious regulation is used for the rails in addition to reasonably sized supply capacitors, but the price to be paid is heatsinking and rugged pass elements, all of which I saw no reason to use as I explained in the text of another post. But for the normal intended operation, the LC filter poses absolutely no threat to the audio quality, and if anything improves the whole performance because a tiny fraction of the hum exists at the collector supply compared to a pure C input supply. The diode peak charge currents through the transformer are much lower than with no choke, hence there will be less switching signal junk in the earth paths. When I tested the supply with the chokes in the pots full of pitch, they were as silent as the grave, and the only hum seemed to come from the damned Phase Linear power tranny. Many high powered professional amps are unacceptably noisy, and worse for using normal cap input filters. I have a classic case here at present, where a dude bough the amp in good faith, but its noise can be heard 50 feet away. the fan blades whip past the holes in the panel, and cause a loud whine, like a siren, and the air splashes over the heatsink, itself a bean counter's special. With the fan off, the transformer hum and panel buzz is nearly as bad. With rubber mounting of the PT, the bobbin hum is still audible at 30 feet, and the tranny must be re-wound with more turns per volt to stop it vibrating so badly. The rails are +/- 85v, and there is 400 mA of idle through the 8 mosfets per channel, making 68 watts of diss at idle, and allowing about 0.6 watts of class A into 8 ohms. The heatsinks get stinking hot at idle without the fan. But this amp has been soldiering on for about 17 years like this; it has had much use, judging by the buildup of what is probably toxic dust and grime thoughout. It seems to have only been repaired once, but when that was done, 3 mosfets in one channel were Exicon, not Hitachi like the originals, and the idle current varies between 27mA and 120 mA. None of the bias problems would give poor sound, because the amp has the usual 75+ dB of global and local NFB. The guy who bought this on e-bay has to spend more fixing this amp than it cost him, and he had no idea that such an amp would be as crudely manufactured as it is, sellers never disclose the real story. It happens all the time.... Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |