No that's not it. The idea is that without a regulated supply and an average music load, the power supply is higher than under heavy constant max power output. So if a high level burst in a regular program comes along it can benefit from the max supply voltage. Of course, if the burst is long, the supply eventually sags and the advantage is lost.
There exists also a measurement protocol to measure this headroom and express it in dB; numbers you see are generally in the 0.5dB to 1.5dB range.
Jan
Last edited:
I agree. Every amplifiers have different PSRR.
But I prefer capacitance multiplier.
In DIY, there are no disadvantages as long we can afford it. Small improvement is matter.
Except you design for mass production, you should consider price per performance and target market.
But some religion said, all amplifier sound same. Then you should make as cheap as possible.
I agree again
It is simple than design good voltage regulator. If I can make good voltage regulator, maybe I will try it. But my current CM is better than DX regulator in listening test.
That for me is backwards. If you have a good reg, why spoil it with a cap multiplier? If you want the characteristics of a cap multiplier, why prefix it with a regulator?
Jan
From the PSU, which does not have time to droop.dotneck335 said:
There is another problem with regulators: their feedback loops may need to be designed as carefully as the main amp, and they may provide a supply rail with a complex impedance (typically, somewhat inductive - which may interect with decoupling capacitors).
If you only needed a few volts of regulation then you don't need a regulator. I don't think you have thought this through.
Yes, OK, I see your point. In my case, that would be about 1 db headroom difference. So, like all things electronic, it's a tradeoff. Additional costs of two regulators, four pass transistors, and some diodes and resistors---maybe $25. Certainly too much for a commercial design, but for a one-off DIY design power amp, it's doable without too much pain. Loss of 1 db of headroom. Maybe 80 db less ripple and noise. Which is more audible? I don't know for sure, but it seems to me that the quietness would be preferable to the headroom. YMMV.
A (high NFB) voltage regulator has excellent performance at low frequency. Very low Zout, very good ripple rejection / PSRR.
A (zero NFB) capacitance multiplier has excellent performance at high frequencies. Low and FLAT Zout, good PSRR.
Cascading the two might be viewed as an attempt to get the best of both worlds. Eliminate crud from one end of the frequency axis, then eliminate crud from the other.
I missed where OP stated actual Design Goals:
*For these goals*, super-low rail impedance is not needed. I remember when SOA really limited our designs (I thot those days were gone?). Recent decades and >999 products show that good honest PSRR makes PS ripple a non-issue for low-level work.
Leaving us at ripple modulation in *clipping*. Well, you shouldn't be clipping a High Fidelity amplifier. (It is a major flavor in guitar amplifiers.) But also: what have you gained? Say a choice between a raw supply rippling 44V-41V, or a smooth supply at 40V. Up to 39V of signal, there's really no difference. At "42V" output, the smooth supply is clipping flat and the raw supply is throwing 120Hz trash on 1% of signal peaks. Well, turn-down from "42V" to 39V peaks. The 0.6dB drop is really unimportant. And when you are stoned out of your mind and "enjoy intermodulation", the raw supply gives you a dB more plus "flavor".
This all assumes that, like most modern transistor amps, the output stage (with overall NFB) has lots of PSRR. It is typically the same sort of emitter follower you would use for a regulator, only with signal in it.
To reduce ripple one can also use a three phase mains supply. Without a smoothing capacitor, the output never goes below 86% of peak voltage, irrespective of load current.
Yet another way, is to use a mains supply that is not a single sine wave, but square wave with the first 10 harmonics present, and the rest removed. Iron cored transformers don't like the presence of high power high freqency harmonics.
Last but not least, use a DC mains power supply. This can be achieved mechanically using a motor and dynamo mounted on the same armature. I have seen this arrangement to provide a 3 phase mains supply from single phase. The owner operatures a milling machine and a lathe using this kind of power conversion.
These are all theoretical possibilities, albeit based in Physics and technology, but for mundane situations, costs play a very limiting role. Costs are often the first thing to consider in anything.
Yet another way, is to use a mains supply that is not a single sine wave, but square wave with the first 10 harmonics present, and the rest removed. Iron cored transformers don't like the presence of high power high freqency harmonics.
Last but not least, use a DC mains power supply. This can be achieved mechanically using a motor and dynamo mounted on the same armature. I have seen this arrangement to provide a 3 phase mains supply from single phase. The owner operatures a milling machine and a lathe using this kind of power conversion.
These are all theoretical possibilities, albeit based in Physics and technology, but for mundane situations, costs play a very limiting role. Costs are often the first thing to consider in anything.
Last edited:
How does one determine a power amplifier's PSRR? Although I do see this specification quoted on opamps and chip power amps, I don't see it on commercial amps. Can it be determined by a calculation from the schematic? I have seen it graphed for an LM3886 power amp chip as ~ -95db @120Hz from the -V supply. How does one design a power amp for excellent PSRR? I suppose that this is perhaps a separate discussion for a new thread.
A properly designed amplifier and suitable unregulated rectifier-bulk capacitance power supply would not need a regulator.
You would be chasing diminishing returns for higher cost...
The widespread use of current mirrors and constant current sources means noise on supplies is either cancelled or not passed on.
You would be chasing diminishing returns for higher cost...
The widespread use of current mirrors and constant current sources means noise on supplies is either cancelled or not passed on.
Last edited:
I would like to understand this further. How exactly do current mirrors and constant current sources avoid power supply noise and ripple from getting on the signal?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.