And if your amp has a decent PSRR than your power supply is not part of the signal path. (Add a couple of volt music signal directly to your rails and see how much shows up on your output. Try not to induce signals into the early amp stages from the huge currents this will take. Or try it in a simulation. ) from your line of thinking, the generator at the hydroelectric dam is in the signal path, maybe even the water that turns the turbine.
While that last approach would be possible with separate supplies for voltage and output stages of an amplifier module but that would depend on the voltage stage connections being available on a chip amplifier.
With regard to your first point, if you had suggested injecting a ripple voltage onto a power supply rail and shorting the amplifier input to ground and measuring the r.m.s. voltage present on the output that would emulate the setup for determining the amplifier rejection ratio - as described in Motorola Application Note AN483A from 1975 covering "Basic Design of Medium Power Amplifiers".
Interestingly, this says "The power transformer is a major component along with the filter capacitor affecting the voltage regulation of a power supply" and "The filter capacitor is also a major component affecting the percentage of ripple voltage on the power supply output."
Last edited:
I agree with this appraisal however seeing there has been no acknowledgment of any of your points and since the subsequent discussion has pursued a different course I am uncertain as to whether such points have been understood or taken as read.
The music signal that you hear from the speakers is literally, directly, current from/to the reservoir and decoupling capacitors. That's why I say that the PSU caps "ARE" the signal path.
The top plot is the current from the power supply capacitors and the bottom plot is the voltage across the speaker. It doesn't get much more obvious than that. (That's a single drum strike from the intro to AC/DC's "Highway to Hell".)
Voltage-centric thinking about power supplies will not help you to understand,
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Noise on the output only appears when the output voltage reaches the rails.
On the other hand noise into the LTP/VAS stages will modulate the output signal.
I always decouple the LTP/VAS stages well.
On the other hand noise into the LTP/VAS stages will modulate the output signal.
I always decouple the LTP/VAS stages well.
The energy stored in a capacitor is proportional to the square of the voltage applied and low powered amplifiers use relatively low voltage rails. You will still want them to be capable of driving 4 ohm loads without clipping.
For example an amplifier with a reservoir capacitance of 2200 microfarads charged to 70 volts will hold roughly as much energy as a 10000 microfarad one charged to only 30 volts.
The relevant formula can be found at Capacitors - Energy Stored
In theory the large capacitance allows the low powered amplifier to behave as well as high powered ones up to it's power limit.
The design of power supplies involves compromise with transformer design and capacitor values.
Since increasing capacitor values will draw more current from the transformer in increasingly briefer bursts and reduce efficiency, the one chosen in a commercial design will be the optimal value.
On the DIY scene one starts with a blank page which is where the all fun starts.
I have been horrified at some commercial amps when I open the box.
Some have very small transformers and capacitors.
Some have very small transformers and capacitors.
Hi guys, if we use 50V caps on amps that used 18 Vac for +/- 24V rails, will the the 50V caps get charged to it's full 50V capacity or will it only be filled up to 24V?
Also, will a cap perform better if its charged to nearly full capacity, compared, to say charged to only 1/2 of it's capacity?
Also, will a cap perform better if its charged to nearly full capacity, compared, to say charged to only 1/2 of it's capacity?
Last edited:
The charge Q stored in a capacitor is Q = C x V, where V is the voltage actually present across the capacitor, and C is the rated capacitance.
If only half the rated voltage is used, then only half the possible maximum charge is stored in it. It's best not to have a voltage too close to the rated
maximum voltage across a capacitor (especially for electrolytics), for reliability reasons. For example, If the rated voltage is 100V, you may want to
have no more than 80V on the capacitor, especially if the temperature is above ambient.
Last edited:
The Rotel RHB10 power supply has much the same aims but in a different implementation.
It has separate bridge rectified supplies for the output stages and low level stages. The supply rails on all stages are +/- 70 volts.
The output stage power supply capacitors are 2200 microfarad while those in the low level supplies are 6800 microfarad.
The output power is 200/330 watts into 8ohms/4ohms. Although I don't need this sort of power I find the power supply concept interesting.
only up to 24 volts....
not really, a rule of thumb is 80% of cap spec...so if rated for 50 volts, then about 40 volts....but, hey, if you decide to use those caps for another projectrequiriing higher rails, then you are already equipped...😉
The latter - the energy storage is a function of the voltage on the cap, not its rated voltage.
Depends what is meant by 'performance'. Its better utilized if its nearer its rated voltage (allowing some safety margin, which depends on the manufacturer.)
Interestingly a capacitor's volume is roughly proportional to the CV product (i.e. the charge stored) whereas the energy storage (what matters in amps) is proportional to the square of the voltage (half C * V-squared). So a 220uF/100V cap is roughly the same volume as a 2,200uF/10V but its energy storage is 10X greater. Bottom line - to get the most energy storage per litre of cap volume (which also roughly correlates with cap price), go for high voltage supplies. This then necessitates an OPT to make use of the high voltage rails if you don't have any high impedance speakers to drive.
Generally speaking capacitors with higher voltage ratings have lower equivalent resistance and will have a longer life if used at 50% of the rated working voltage. It gets better if you can source ones rated for 105 degrees C or for low equivalent series resistance - these will have better ripple ratings.
Also one can substitute two or more smaller capacitors in parallel making up the required value and get better ripple ratings and equivalent series resistance than one big unit.
If you are experimentally inclined you could make up a board to house three capacitors per rail and start with two a side - leaving an option to increase to three later. There is an optimal value for capacitance in relation to the rating of the transformer so that approach will provide some scope in finding a happy medium. If the transformer VA is 100 or more I would use 2200 microfarads per each.
As I wrote this Post 72 was in train. If you have not yet purchased the transformer it leaves more options open to you as discussed by abraxalito
Last edited:
I think a lot of it has to do with the design of the amp as well. If both the driver and output stages are running off of same supply rails it makes more of a difference than if there are separate supplies. Typically, the harder a transistor is driven, the more the collector/drain current is dependent on the collector to emitter voltage, and power supply ripple can show up more in the output signal.
I don't think more can ever hurt. So even though their bragging about how much capacitance they're dumping in may be little more than marketing, it's not going to do anything but maybe make the price go up a bit.
I don't think more can ever hurt. So even though their bragging about how much capacitance they're dumping in may be little more than marketing, it's not going to do anything but maybe make the price go up a bit.
Incorrect. The signal out of your amp is not directly connected to the PSU. There is an amplifier in between. That is what an amp is for. That is why there is a PSRR. This is basic stuff.
What would be interesting and informative would be to look at all those waveforms with the amplifier driven to clipping. And then see how those waveforms behave with small vs. large caps. PSRR is pretty good in most cases when the feedback is controlling the amplifier, not so good when it isn't.
Its good to have plenty of headroom for transients.
Transients can be many times the normal listening level.
One of my first designs was just a few watts but suffered clipping of transients badly at reasonable listening levels.
On subsequent designs I raised the power supply voltages and had much better listening experiences.
A picture can inform better than 1000 words and Gootee has provided that for interested parties to digest/interpret for themselves.
Part of power supply design process calls for determining the PSRR of an amplifier module. Motorola AN484A quotes the example of a 30 watt amplifier with an 80 dB hum and noise requirement allowing a maximum ripple on the output of 1.55 m.v. R.MS. The amplifier circuit dates from 1975 and the ripple rejection of the module is 40 dB allowing a maximum on the power supply of 155 m.v. R.M.S.
To achieve that result the remaining procedure is to consider capacitor size, transformer voltage and current capability and regulation under load.
While modern amplifier design allows better specifications amplifiers tend to be more powerful and need to be capable of delivering substantial currents - since recharging of a capacitor bank occurs in brief bursts this has more serious ripple implications.
I understand why Gootee thinks capacitors are an important and possibly neglected area of design consideration and I don't have any argument about his expressing views in a figurative sense.
Just saying that the psu caps are not in the signal path. The amplifier seperate the 2. Don't get me wrong, I know the psu is an important consideration, but it needs to be looked at properly in its place. If the caps can supply enough current to drive a 4 ohm load, for 1/120 second without sagging below the allowable ripple current, than they are sufficient.
This is an awesome discussion. I think the following might reconcile cbdb's point. The signal is in fact the current coming from the caps but that signal is being generated and conditioned by the amp circuit so whatever the caps introduce is subject to what the amp does with it. Now, things fall apart when the caps can't supply the energy demanded by the signal. This can happen for a variety of reasons and as you would guess, its nature depends on the amp circuit. In a class A design, the signal, ripple and charge/discharge is constant so the first priority is in maintaining sufficient energy levels in the caps, then the concern becomes reducing the noise. A sufficiently large, unstressed transformer seems more important here than huge capacitance at higher voltages but they begin to equal in importance as the designed voltage rail decreases and the current demand rises. The supply output impedance, which is so vitally important in other amp classifications, is not much of a concern. Naturally, things like pi filters and cap multipliers complicate these guidelines but the general rule put forward here seems right to me.
Now, when we go to high bandwidth, fast, class B and D designs the capacitance delivery speed trumps its size and the transformer slumping slightly under load is somewhat desirable. In this environment, the frequency and amount of the energy delivered must be tuned to the amplification process. The correct determination of these values are paramount and they are not intuitive because they are parabolic functions. In these topologies, big capacitance can be a big waste at best and a big liability at worst.
One further issue that applies to all caps in general is their rating. Very few caps meet their rating spec and some don't even bother to provide the specs you need to design your circuit properly so you can imagine the head scratching that can go on when your amp is clipping 3V below its rails and everything measures right. Add to this the huge manufacturing tolerances with caps and you have suspect number one of most amp mystery problems. This is my input and I hope it helped reconcile the different opinions put forward here. Keep this thread going, it's a good one.
Now, when we go to high bandwidth, fast, class B and D designs the capacitance delivery speed trumps its size and the transformer slumping slightly under load is somewhat desirable. In this environment, the frequency and amount of the energy delivered must be tuned to the amplification process. The correct determination of these values are paramount and they are not intuitive because they are parabolic functions. In these topologies, big capacitance can be a big waste at best and a big liability at worst.
One further issue that applies to all caps in general is their rating. Very few caps meet their rating spec and some don't even bother to provide the specs you need to design your circuit properly so you can imagine the head scratching that can go on when your amp is clipping 3V below its rails and everything measures right. Add to this the huge manufacturing tolerances with caps and you have suspect number one of most amp mystery problems. This is my input and I hope it helped reconcile the different opinions put forward here. Keep this thread going, it's a good one.
Last edited:
- Status
- Not open for further replies.