48v power supply

Can someone tell me if this circuit will work.



Should there be a cap/caps before the regulator?

Also if C2 is increased will that filter out more ripple?

Ripple calculator here

> Should there be a cap/caps before the regulator?

You have the calculator. What is the ripple on the voltage going into the chip when C=0?

In fact that calculator uses a simplified formula that gives impossible answers for very small C. But try R=1K and C=2uFd: you get a low average DC voltage and a very high ripple. The regulator will trim the top of the ripple, but can't magically make power when the inpit voltage ripples lower than the desired output voltage.

YES, you need an input cap.

How big? Try for around 1V ripple, see if that is practical. For 1K load (several Phantom loads) and 48V*1.414= 68V peak input, you need about 150uFd. Even in a 100V rating, this is not an expensive part.

As a short-cut, use 1,000uFd for every 1A of DC load. A single Phantom load is supposed to be 10mA max, so 4 loads is 40mA. The regulator needs to be fed another 15-20mA. Say 60mA design load. 60mA is 0.060A, so use 0.060*1,000uFd or 60uFd. The calculator says 2.7V ripple. That may be OK for some purposes, will actually work here, but 150, 220, even 470 uFd is about the same price and less ripple.

> if C2 is increased will that filter out more ripple?

No. Or not enough to notice. The main ripple reduction is the amplification
inside the regulator. A heavy cap on the output does passively suck ripple, but it also loads-down the amplifer and reduces its ripple rejection. Typically you use 1-50uFd here. With a reasonable input cap, ripple rejection in most DIY is layout/wiring, mixing rectifier spikes into the regulated lines. And note that a Phantom supply really should not need absurdly low ripple.

There are several problems with this circuit. First, as PRR noted, you really need a cap at the input to cut ripple to something manageable.

Second, as noted in a thread over on The Lab, the 783 needs to have at least 15mA running through it to regulate properly. You're already drawing about 7mA with the 6.81k load resistor, so the resistor from Vout to Adj needs to be lower; 156 ohms will get you 8mA, for a total of 15mA. The voltage-setting resistor needs to be scaled accordingly (5.90k will get you ~48V; use at least a 1W resistor and preferably 2W.

Finally, somewhat surprisingly, TI recommends that you *not* use a bypass cap on the voltage-setting resistor. Anyone know why, and whether their recommendation is correct? Seems a bit counter-intuitive to me.

Peace,
Paul

Bo Hansen has a 48v psu (for phantom) using a TL783
http://web.telia.com/~u31617586/#48%20volts%20phantom%20power%20supply
I built it but only get 44-46 volts out of it..
There is another post around here which suggests adding a 500ohm trim pot in series with the 3k resistor to adjust
Problem for me is the 3k resistor is 1w - does the 500 ohm trim pot need to be this - and if so - the only ones I can find are 0.25w

From reading TI's documents it seems that they designed this device to track the load transients and account for them since they claim that adding this capacitor degrades load transient response. sounds like snake oil but may very well be just a well thought out design. i guess someone could build a test jig with both and test...

TI also stresses the need for an output capacitor and gives a handy formula in the datasheet for finding the optimium value, and also states that values larger than this value provide proportionately better transient response. Strangely enough this is almost in contrary to PRR's statement but I can only assume that this part may be an exception for a few reasons such as being an adjustable reg, being a different design, etc.

any more ideas?

Isn't a 48V ac trafo a bit much for a power supply like this?
I thought I understood 48V means 48V rms so it will peak to 48 * 1.414 = almost 68V.

The reference input will need to see the "output error" in order to regulate properly. If you filter the reference input, you will in effect prevent the regulator from being able to "see" the output error..

Jakob E.

[quote author="uk03878"]Problem for me is the 3k resistor is 1w - does the 500 ohm trim pot need to be this[/quote]

Maybe, maybe not... calculate the power dissipation of the trim pot and see...

Peace,
Al.

I think I'm gonna try this one.Parts are cheap.

> TI also stresses the need for an output capacitor and gives a handy formula in the datasheet for finding the optimium value, and also states that values larger than this value provide proportionately better transient response. Strangely enough this is almost in contrary to PRR's statement

No it is not. The original question said ripple. You jumped-track to load-transients.

Read the dang datasheet. There is a chart in there that shows ripple rejection unchanged with or without a cap for ripple inputs up to 100KHz. However the load transient response is different. It is mostly very good within the audio band as long as you feed plenty of raw voltage for the chip to work with. But above the audio band (or when you don't feed it well), load transient response is improved with an output cap.

> Isn't a 48V ac trafo a bit much for a power supply like this? I thought I understood 48V means 48V rms so it will peak to 48 * 1.414 = almost 68V.

How big is "almost"?

Anyway: to deliver 48V through silicon, we'll need at least 2 volts extra or 50V input. Actually most regulators need more: as a minimum, or to meet full regulation and rejection specs. Often 5V, and sometimes 10V to be fully happy. So we need 53 to 58V input.

On the other side, 48VRMS does look like 67.8V Peak. But that "48V" is really a ratio from the "120V" (or whatever) wall-power. In the US, nominal voltages run 115V to 122V, with +/-10% variations as total utility load sags or soars. Europe is converging 220 to 240V nominal voltages. And the transformers are not always perfect. Wise design allows at least +/-10%, even 15% or 20% variation in line and secondary voltage. So the "48V" might be 43 or even 39V, or up to 53-58VRMS. We have diode losses too, a few volts. We might end up at 53VDC to 80VDC. The 53VDC is just barely enough to keep a solid 48V output. 60VDC might be more typical low-line voltage, and isn't over-kill if you really want good regulation.

Think of it as "headroom". Nobody ever objects to headroom in the audio path. Why be stingy with the headroom in the power path that feeds the audio path?

I was searching a 48V PSU for my PM1000 chanel strips. I try a simple schematic found in the TL783 datasheet but i had problem when i charged this PSU to 500mA. I was using a 1000uf cap for filtering. The regalor quickly became very hot and the voltage fell down.

So tonight I try the Bo Hansen design and all seem to be good. I turn it on during 20mn and no problems ! The radiator is a little hot but i will use a bigger one.