Need linear supply 3.3V/15A or 5V/25A

[Joerg]

Hello Folks,

For an experiment a client needs a 3.3V/15A or larger supply, must be
linear. Checked the usual suspects and at 10A there's pretty much the
end of the rope these days. 5V/25A could be useful if I can get in there
and hack it, should have enough SOA to deliver 3.3V/15A.

Does anyone know where to get one? Can be used/salvaged as long as it
works and is linear.

A switcher might work if it has next to nothing in ripple. Most are
50-100mV, won't work. I want to avoid having to build a post-regulator.
But if nothing's there, man's got to do what man's got to do.

Dang, I had two supplies 5V/30A until my wife said it's time to clean up
the garage :-(

 

[Joerg]

Any reason not to gang LM338's as per data sheet application note?

Too much dropout voltage unless I got a 12-15V supply, and then we'd end
up with a pretty good space heater ;-)

 

[Joerg]

Ok. I'll watch this one, it might get some suggestions I want to try too.

Yeah, one of our issues will soon be that there is only 15 amps coming
out of a standard 120V outlet ;-)

 

[Joerg]

Why not just filter a switcher output?

If I have to do that then I'll do it. But I found that load change
reactions of such switchers when loaded with unorthodox loads (mine
somehow always are ...) will usually be in the same level range as the
ripple. No matter what the marketing guys put on the datasheet.

Women just don't understand the value of stuff.

Yeah :-(

But: When we walked the dogs this afternoon a neighbor offered us his
grandfather clock. I just would have to repair it. My wife was tempted
but I remained firm, I just don't want to have yet another project on
the never-ending honey-do list.

 

[Joerg]

Switcher plus low-dropout mosfet regulator? That would be another
project, of course. One vref, one opamp, one big fet, a few passives.

Yes, plus you'd have to crank up the switcher a little. Meaning they
either must release the schematic (which they won't) or it'll be a hack.
The usual, find PWM chip, hope it's not filed off, try to decipher part
#, get datasheet, flip red button on the Weller.

 

[DJ Delorie]

If I have to do that then I'll do it. But I found that load change
reactions of such switchers when loaded with unorthodox loads (mine
somehow always are ...) will usually be in the same level range as
the ripple. No matter what the marketing guys put on the datasheet.

What happens to the ripple if you chain linear LDOs? I.e. switcher
down to 12v, big cap, 7809, big cap, 7805, big cap...

(not that I'd use a 7805 to provide 15 amps, of course, just wondering)

 

[DJ Delorie]

My guess is a slightly more complex space heater than I was suggesting. >

Er... what does that have to do with *ripple* ?

 

[Palindrome]

If I have to do that then I'll do it. But I found that load change
reactions of such switchers when loaded with unorthodox loads (mine
somehow always are ...) will usually be in the same level range as the
ripple. No matter what the marketing guys put on the datasheet.



Yeah :-(

But: When we walked the dogs this afternoon a neighbor offered us his
grandfather clock. I just would have to repair it. My wife was tempted
but I remained firm, I just don't want to have yet another project on
the never-ending honey-do list.

Well, that's her next birthday present sorted - a clock repair tool kit.

You mean that /she/ can't design you a 3.3v 15A linear supply? I won't
ask what she *is* good for

3.3v is /just/ within the minimum voltage spec of a "standard" 15A lab
variable power supply. Do you not have Government Surplus Stores? You
might pick up one there for a song (and a hernia..).

It also, (if my memory at twenty to two in the morning serves me right),
is the voltage used on early Cray and CDC Cybers to power the ECL chips.
So, if you could pick up a surplus power supply for one of those, you
would have amps and plenty.

Talking of which, I do happen to have a couple of 5v 50A modules from a
Cyber mainframe. 110v in. Not much use to me here in the UK but, hey, I
have a barn and what else would I put in it? Hey..barn..oh, never mind.
it's late.

However, if you do go to the hack-a-5v-supply approach, you may simply
be able to offset the *ground* reference of the regulator by a couple of
(negative) volts. With any luck, you will be able to use the same
transformer/transformer winding as the main positive output, which will
mean that you will *decrease* the offset voltage as the load increases -
thus potentially *reducing* the ripple. You could even end up with a
power supply with ripple that decreases as load increases .

 

[Phil Allison]

"Joerg"

Too much dropout voltage unless I got a 12-15V supply,

** Complete drivel.

An LM338K's drop out voltage for 5 amps DC is only 2.5 volts - so you need
a minimum (including ripple) DC input of 5.8 volts.

Chose the transformer carefully and you will have no more than 15watts
dissipation per device.



....... Phil

 

[Joerg]

This is what transformers are for. > That aside (and the less said about
the size they'd have to be, the better), this sounded like laser drive the
first time you mentioned low volts and high current with no ripple, now it
looks even more like it (gas this time), though I think if it had been either
you'd have been after current control and some different voltages.

You could become Sherlock Holmes. Yes, it's for a laser but current
control happens later. And yes, another voltage, too, but that one will
be easy.

 

[Joerg]

"Joerg"



** Complete drivel.

An LM338K's drop out voltage for 5 amps DC is only 2.5 volts - so you need
a minimum (including ripple) DC input of 5.8 volts.

Chose the transformer carefully and you will have no more than 15watts
dissipation per device.

Well, sure, if I build the whole chebang myself I know how to do that.
But that's what we try to avoid. There used to be off-the-shelf supplies
like this and I actually had two of those (until a garage clean-up day
mandated by SWMBO).

 

[Joerg]

I second that one. It might even be possible to use some linear regulator for
the reference and internal op-amp, if it lets you get at them the right way.


I once tweaked a computer SMPU to get about 7 volts out of the 5V line. I
traced a pair of resistors that were a divider, and shifted the value with an
added preset pot. It worked ok. I looked into trying it with another, and no
such easy trace existed, so it's pot luck. Yes, really said that. Anyway, if
you can do this, you might tweak it high enough to let ganged LM338's work
fairly efficiently fed off the was-supposed-to-be-5V line.

Or, being an RF guy, turn it down to 3.3V and mount some humongous
ferrites to kill the ripple. At least it'll get less hot. Looks more
nerdy as well

 

[Joerg]

What happens to the ripple if you chain linear LDOs? I.e. switcher
down to 12v, big cap, 7809, big cap, 7805, big cap...

(not that I'd use a 7805 to provide 15 amps, of course, just wondering)

Those aren't LDOs. There is an imaginary sign here in the hallway: "No
cheap beer and no LDOs beyond this point".

A well designed regulator will completely kill the ripple in just one
stage, no need to chain them. All you really need is a LM317 and a few
beefy transistors. Plus a huge heat sink, #10 screws, and so on. That's
what I want to avoid.

 

[Phil Allison]

"Joerg"

Well, sure,

** So what you claimed above was a complete BOLLOCKS.

Blown straight out your arse like everything else.

if I build the whole chebang myself I know how to do that. But that's what
we try to avoid.

** We all need to avoid arrogant fuckwits like you.



...... Phil

 

[Joerg]

Well, that's her next birthday present sorted - a clock repair tool kit.

You mean that /she/ can't design you a 3.3v 15A linear supply? I won't
ask what she *is* good for

3.3v is /just/ within the minimum voltage spec of a "standard" 15A lab
variable power supply. Do you not have Government Surplus Stores? You
might pick up one there for a song (and a hernia..).

Not around here :-(

It also, (if my memory at twenty to two in the morning serves me right),
is the voltage used on early Cray and CDC Cybers to power the ECL chips.
So, if you could pick up a surplus power supply for one of those, you
would have amps and plenty.

That's the kind of machine my two supplies came from.

Talking of which, I do happen to have a couple of 5v 50A modules from a
Cyber mainframe. 110v in. Not much use to me here in the UK but, hey, I
have a barn and what else would I put in it? Hey..barn..oh, never mind.
it's late.

Linear supplies? Why does everything useful have to be 6000 miles away ...

However, if you do go to the hack-a-5v-supply approach, you may simply
be able to offset the *ground* reference of the regulator by a couple of
(negative) volts. With any luck, you will be able to use the same
transformer/transformer winding as the main positive output, which will
mean that you will *decrease* the offset voltage as the load increases -
thus potentially *reducing* the ripple. You could even end up with a
power supply with ripple that decreases as load increases .

Usually when I do that I also try to pry off some of the secondary
windings of the transformer, whether on a linear or a switcher. Makes
everything run happier and less hot.

 

[Joerg]

Joerg wrote:

Addendum: If anyone is aware of a huge surplus place in the US
(preferably with a web site) that still has mainframe era stuff let me
know. That's how I used to get such supplies but it was in Europe.

 

[James Arthur]

What happens to the ripple if you chain linear LDOs? I.e. switcher
down to 12v, big cap, 7809, big cap, 7805, big cap...

(not that I'd use a 7805 to provide 15 amps, of course, just wondering)

The 78xx-series are not LDOs, but...

Lower ripple at the output is the result. Each regulator adds xx dB of
ripple rejection.

No need for all those big caps, just the input cap will do.

Cheers,
James Arthur

 

[James Arthur]

Those aren't LDOs. There is an imaginary sign here in the hallway: "No
cheap beer and no LDOs beyond this point".

A well designed regulator will completely kill the ripple in just one
stage, no need to chain them.

True, but he was asking about 78xx ;-)

(They're spec'd near 70dB each, which ain't bad.)

Cheers,
James Arthur

 

[DJ Delorie]

Those aren't LDOs.

Yeah, I know, but I figured everyone would know what those devices
*did* for the purposes of my example.

With adjustable LDOs you could use a switcher to get to 6v (or less),
then two LDO stages to get to 5v, minimizing heat loss.

 

[neon]

find a transformer for 5v 6v at 20 amps [efficiency ] and then use lm117 to regulate down to 3v or lower. you will need some heavy bypass transistor to handle that kind of current. 2n3055 are cheap pluss a 2n2219 to boot