Linear AC/DC Power Supply Design - Basic questions

[Smillsey]

Hi all

I have a project where I need to provide +/- DC rails @ around 50V.

I will be going for a linear design, for simplicity with a toroidal transformer (unless you believe this is a stupid idea and I am not "up" on the latest switching technology".)

Desired performance criteria;

+/- 50V rails
10A per rail
Regulated, smoothed output

I thought this would be simpler than it is! My thoughts were to use a linear regaultor to regulate that unregulated 50V +/- 20% DC coming from the transformer/recitifer/filter stage but quickly I have realised there are no "off the shelf" IC's to do this (unless I am missing them) and the power loss even if only 5V is dropped across the linear regaultor will be 50W! thats hot!

I do not care about efficiency, I just need a nice 50V DC rail.

Any guidance on a topology to use for this would be greatly appreciated.

Thanks

millsey

 

[KlausST]

Hi,

I´d use two 48V 500W switch mode supplies. Outputs connected in series.
Cheap and simple. and low heat generation

Klaus

 

[zenerbjt]

Yes, put a diode across the output of each 48V supply. (as you know, Cathode to the higher voltage)

If you want to get exactly 50V, then add a boost converter to the 48v outputs.
There are plenty of DCDC modules on digikey and farnell that do this.

 

[Smillsey]

Thanks both

Ok, but are these supplies going to be quiet enough? Or am I in the dark ages thinking all SMPS are noisy?

 

[KlausST]

Hi,

There are noisy ones and quiet ones.
I´d read the datasheets.
I guess the cheaper ones may be more noisy.

The ripple frequency will be higher with SMPS .. but the ripple voltage surely is less than with a low frequency supply.

Reducing SMPS noise should be rather simple. But in detail it depends on your requirement.

Klaus

 

[Smillsey]

It will provide the rails for an apex analog MP111FD

https://www.mouser.co.uk/ProductDetail/Apex-Microtechnology/MP111FD?qs=TiOZkKH1s2SckWaBxt%252B7Wg==

I found some AC/DC 48V converters front Artesyn but that are £100 each!

 

[zenerbjt]

you mean audible noise?.....there will be fans whirring away yes.
You want 1kW...but is it a guitar amp power supply?...if so their averag epower is only about <1/8 max

 

[Smillsey]

Hey zenerbjt

No I’m not worried about audible noise, this is a lab amp with an output up to 1MHz so I’m more worried about switching noise messing with the output.

although the apex op amp has good noise rejection itself - maybe I am worrying unnecessarily?

 

[zenerbjt]

Yes, just filter the noise out. Be careful if you use an SMPS...there are elimits on external capacitance you can use......also, certain combo's of Ls and Cs in any filter that you add may cause instability in the power supply, so keep a check on vout with a scope probe..
But the SMPS that you buy should have an outptu filter anyway.

Keep your current loops as small in area as possible, and use a bit of filtering, and maybe a common mode choke.

 

[Easy peasy]

If you are really worried about noise you can put the switchers in a metal box, e.g. 19" rack mount case & add extra filters to the mains and to the DC - esp CM chokes & caps to earth - then earth the chassis well too - this will give you the cheapest low noise +/- 50V supply, don't forget the diodes and some extra electro's on the o/p ....

 

[Smillsey]

Nice thanks for that.

Caps directly from the 48V live to the casing? Or returning to the 0V of the dual rail I will create with the two switchers?

 

[Easy peasy]

Both, per the text above

 

[Smillsey]

Ok - I will go for that then.

I will post results in the next months.. if you are interested

 

[KlausST]

Hi,

Noise / ripple of SMPS.

Mind: Often the lowest nouse is not "without load" and not at "full load".
* with low or no load they may enter discontinous mode, also called hickup mode or burst mode. Here the ripple frequency is rather low and the ripple amplitude rather high. Not easy to filter. Maybe add a dummy load.
* at full load you may see the switching frequency. It is rather high thus should be rather easy to filter.

Buy a power supply from a reputable manufacturer, read specifications about ripple/noise and the according test conditions.
Select a power supply which fits your power requirements with some headroom. Not one that is rated 5 times the expected load.
Maybe the manufucturer provides additional documents on filtering.

Klaus

 

[Smillsey]

Got it

i can measure AC component accurately (excluding DC) up to 1MHz with the equipment I have here.

I will first try with minimal filtering and then with filtering in steps, I’m a fan of actually measuring a circuit to prove the addition of components is actually helping the situation in

I’ve worked with Artesyn supplies before and I am happy they are good supplies -let’s see what the measurements say.

I will look at an internal dummy load resistor on the bus.

--- Updated Dec 24, 2020 ---

This is the PSU I’m looking at, 480mV of ripple :-(

**broken link removed**

 

[Smillsey]

Ok, I am going for 2x48V supplies, I also now require the need to switch to other ranges such as +/-5V and maybe even +/-20V....

I have two options;

Have two more sets of Ac/DC converters inside the same unit, but I haven't thought through all of the implications of this yet like leakage currents at the input and safety....

Or I could us DC-DC converters (isolated) to step the 48V down...

What do you guys think? I will need to switch the rail voltages between 48V, 20V and 5V using relays.

It seems AC/DC converters are cheaper than DC-DC ...

 

[wwfeldman]

do you need to switch from +/- 48 to +/- 5, or +/- 20,
or do you need the different voltages at the same time?

 

[BradtheRad]

need to switch to other ranges such as +/-5V and maybe even +/-20V

This negative voltage doubler converts a positive supply to negative. Whatever amplitude is the initial supply, the output is opposite polarity at the same amplitude (minus a few volts). You need only change the positive supply, and this circuit automatically does the rest.

Clock voltages need to be same amplitude as the supply. An astable multivibrator can generate clock signals.

Two voltage doublers are included in the circuit. They are the same construction but clocked at opposite phase, so that continual current is taken from the power supply.

If you wish to simplify things you can omit the right-hand section, but in that case your supply must provide extreme bursts of current.