Absolutly. One can design a simple 78xx based supply using two filter caps without understanding how it works but if you are building a muti-section pi filter youi do have to do a small amount of calculation, look at time constants and so on. When you do this the answer is going to be "bigger resistors". RC is well the product of R and C. People don't like it because of the voltage drop but you can work this into the design. With more effort you can replace the resistor with a choke. But high current chokes are big and expensice and only make sense on preamps or high voltage power amps (that run on low current)
For a preamp I think simply going to a one section "pi" reduces the 60Hz ripple by 40db. That and a (modern) regulator will get the ripple below the self-noise of the amp.
The other thing that you can do is use better quality caps. You don't need to spend a lot. You can parallel smaller low ESR caps with the big electrolytics. The best way is to make a figure out the size of the cap then parallel one the is 10% that size and one that is 1% that size. The smallist one will end up being a metal film cap
Another way is to have local bypass capacitors augmented by a resistor to form local R-Cs where you need them, leaving the main supply free of any R. That way you can add more modules as you go along, each with their own local R-C, as you see fit, without affecting greatly the common supply because its R is as small as possible (caps internal R).
A way I found works in many cases is for every sub-module/stage, you add a series R before the local bypass capacitor, so that the R is no more than 10% of the load of that module, which means the voltage drop across it will not be terrible and your original-calcs will not be thrown out completely.
Then, by experience, I have found you need to work out the local bypass electrolytic so that in conjunction with the series R it forms a 1Hz filter, no more than 1Hz. Going below 1Hz produces no measurable results on the oscilloscope or on the speakers that I could detect (perhaps I need to buy better oscilloscope and a pair of new ears!)
In most cases R is quite large and so C is quite small, in other words cheap. Problems may arise at buffer stages with very high loads (comparatively) and there a 10% R will be very small and the required C to hit the target of 1Hz will be quite large. That runs the risk of having local Cs in excess of the total capacity of the main filters at the PSU! Plus sparks everytime you connect a module plus an increased risk of instantly burning the discrete voltage regulator if its transistor is not man enough to take the jolt.
A practical exampe: a pre-amp stage uses 2 transistors with a total load of 15K, so you choose a 1.5K resistor in series with Vcc and you bypass it with a 100uF cap to achieve the 1Hz filter. The buffer stage of that pre-amp module however uses a 470R and so you choose a 47R resistor as the series R, which now requires a 3,300uF local capacitor to hit the 1Hz. This is not practical and also 3,330uF may be more than the total capacity of the filters at the PSU.
Therefore for buffer stages with high loads I leave them alone depending on the initial voltage regulator. After all ripple hum is worse at voltage amplifier stages, and much much less at buffer stages.
A point I am also trying to make is that in many cases you will have local bypass capacitors anyway, might as well add a small R before them to achieve local filters.
A way I found works in many cases is for every sub-module/stage, you add a series R before the local bypass capacitor, so that the R is no more than 10% of the load of that module, which means the voltage drop across it will not be terrible and your original-calcs will not be thrown out completely.
Then, by experience, I have found you need to work out the local bypass electrolytic so that in conjunction with the series R it forms a 1Hz filter, no more than 1Hz. Going below 1Hz produces no measurable results on the oscilloscope or on the speakers that I could detect (perhaps I need to buy better oscilloscope and a pair of new ears!)
In most cases R is quite large and so C is quite small, in other words cheap. Problems may arise at buffer stages with very high loads (comparatively) and there a 10% R will be very small and the required C to hit the target of 1Hz will be quite large. That runs the risk of having local Cs in excess of the total capacity of the main filters at the PSU! Plus sparks everytime you connect a module plus an increased risk of instantly burning the discrete voltage regulator if its transistor is not man enough to take the jolt.
A practical exampe: a pre-amp stage uses 2 transistors with a total load of 15K, so you choose a 1.5K resistor in series with Vcc and you bypass it with a 100uF cap to achieve the 1Hz filter. The buffer stage of that pre-amp module however uses a 470R and so you choose a 47R resistor as the series R, which now requires a 3,300uF local capacitor to hit the 1Hz. This is not practical and also 3,330uF may be more than the total capacity of the filters at the PSU.
Therefore for buffer stages with high loads I leave them alone depending on the initial voltage regulator. After all ripple hum is worse at voltage amplifier stages, and much much less at buffer stages.
A point I am also trying to make is that in many cases you will have local bypass capacitors anyway, might as well add a small R before them to achieve local filters.
This is exactly equivalent to what is called a "progressive filter". Look at almost any amp built in the 1940's or 50's and you will find a set of RC filters. The output straight off the first cap following the rectifier would go to power the main power output stage AND an RC filter, the output of that RC filter powers the "driver" stage AND to power another RC filter that in turns power a preamp stage.
With a stereo amp you have twice as many branches.
Here is an example of a progressive filter it works for an multi stage amp because the power requirements go down as you get closer to the input while at the same time the need for low ripple goes up. One other thing is that it provide isolation between the stages to prevent oscillation.
An externally hosted image should be here but it was not working when we last tested it.
The above schematic shows the filter stages all in one place but in a real amp many times the stages are physically close to the load so as to not have long runs of wire (that might pick up noise along the way.) The schematic above was deep linked from a great web site with good info on how to design tube tubes. T
The modern approach is to use separate power supplies for the logic, driver circuits, and power stages. That is because of the vastly different power requirements of each. (Typical voltages are 5v or less for logic, 12v or 15v for drivers, and the power stage voltage depends on design.) In particular, amplifiers with dynamic power stage voltage must have a completely separate power supply for the power stages since the power stage voltage is under logic control. It is even possible to completely turn off the power stage supply on a standby command or loss of input signal in order to eliminate idle switching losses.
Last edited:
1mF & 0.9mA meets your 10% and 1Hz criteria.
3.3mF & 3mA also meets your criteria.
Virtually any good audio opamp and all discrete opamp circuits and buffers exceed 1mA from a 15vdc PSU.
A 20mA opamp circuit would require 2.2mF & 75r. This current load is equivalent to a dual <10mA opamp.
29mA buffer would require 3.3mF and 51r.
Ignore the current load (of any amplifier) at your peril.
Last edited:
I find the easiest solution is to use a simple power supply and an amplifier with good supply noise rejection. Usualy achived with high difirential gain and low common mode gain and lots of feedback. When high currents are involved the power disipated by a linier regulator is cost prohibative in heatsinking and would be better spent on the amp circuit. Also high value highj current chokes are expensive! often more than a regulator.
Not a fair comparison. The switch mode supply is regulated while the simple linear supply is not. If you were to build a linear supply to the same specs as a switching supply the linear supply would be much larger and dump way more heat then the switcher. A regulated 600VA linear supply would absolutely require a large fan or liquid cooling. Try building a liner supply that can work on 100-240 volts 50-60 Hz without a selector switch. Yes, the typical switch mode supply has THAT good of line regulation.
Cost is another issue, the swiching mode supply uses less expensive parts, small transformers, small filter caps.
One way to think about this is that a linear supply is a bit like a class A amp while the switching mode supply is more like a class D.
- Status
- Not open for further replies.