Hybrid PSU - negative rail preregulator

[atferrari]

After seeing the reply by MrAl to throbscottle's Hybrid PSU thread, proposing this circuit I felt I could ask for the equivalent for the negative rail.

The (ADJ) switchers LM2575 or LM2576 are available locally, but I do not know where to start. No LTspice model seems to exist nor examples of a negative buck converter in the datasheets I got. I believed the LT1070 family was kind of a replacement but it isn't.

With no load, after filtering, I have 36V DC per rail available. I do not expect needing more than 2A. Say 3A if you like.

Any help will be much appreciated.

 

[ronsimpson]

Here is a LM5088. +37V to -25V. You can take a high voltage buck and make a negative voltage. The switch will see the positive voltage plus the negative voltage. Amy HV buck can be tricked into this. The LM2576HV might work.

 

[atferrari]

I want to go from minus 36 to say minus 5 (or any value in between).
Negative buck converter negative in negative out

 

[ronsimpson]

Is the -36 volts "isolated"? OR is the -36 connected to a +36 volts? As in +/-36 volts.

 

[atferrari]

Is the -36 volts "isolated"? OR is the -36 connected to a +36 volts? As in +/-36 volts.

Sorry for not explaining the basics properly, Ron. It is a center tapped transformer.

I consider the preregulator requirement for the positive rail, solved with MrAl's circuit. I was dreaming to obtain the equivalent for the negative rail of my dual PSU.

Gracias for replying.

 

[ronsimpson]

The LM2586 is a boost PWM but I will use it in buck mode.
As long as the feed back pin is low the switch will be on hard and pull -V OUT down.
When the opto isolator turns on it will pull up on FEEDBACK and the IC will reduce the duty cycle. (use the 5 volts version of LM2586-5)

The linear regulator has a (Zener diode and LED of the opto-isolator) across it. The PWM is set to regulate the IN to OUT voltage across the linear regulator. The idea is to keep 4 to 5 volts of head room on the linear regulator. (must have a resistor)

When the linear regulator has good voltage across it the PWM will back the duty-cycle down.

Have not used the LM2586 in this move. I have used other parts in this mode. I probably forgot something.

 

[atferrari]

Sorry Ron but, locally, I have chances to get the LM2575, 76 or 77.

Thanks for spending so much time in helping.

 

[ronsimpson]

IF( LM2575 OR 76 OR 77) is the only choices
...AND The -36 volts is not isolated from the +36 volts
...THEN I do not know how to do this.

 

[MrAl]

Hello there,

I see the requirement for this negative switcher is that it must use the LM2576.

The LM2576 has positive output so to get a negative output we'd normally have to connect the output to ground which means we cant connect the input to ground. That means the input has to be 'floating'.

That's the 'normal' way to do it, which would require a separate transformer winding for the negative supply. So instead of using a center tapped transformer for both supplies we would use a dual winding transformer with two completely different windings although both the same voltage perhaps.

Using a center trapped transformer means we'd have to trick the LM2576 into working differently.
One idea that comes to mind that might work is if we can use the M2576 in boost mode to invert the output.
Another idea that might work is to level shift both the output and the feedback signal. This might even be better.

I am not completely sure yet however if either of these will work, but you can think about that and i'll take a better look a little bit later and get back here with more info. I thought i would post those two possibilities right away to give other members something to think about while i have time to look better myself. So give that some thought and i'll get back here a little later hopefully with some better information. I do know what whatever we come up with, if anything, will have to be bench tested thoroughly.

LATER:
One way might be to forget about the input negative supply, instead start by making the rectifier create a positive only voltage using the center tapped transformer. This typically uses two diodes, or four for higher output voltage. This provides an output that is positive only. This would power both the positive switcher and the negative switcher.
The negative switcher would be set up to create a negative voltage by grounding the output positive instead of the output negative. The input negative then gets connected to the negative output instead of the input ground.
I cant be sure yet if this works with the LM2576 though, but supposedly it works with some other IC's. I'll check around a bit more.

MORE LATER:
Ok the buck boost method above does work. The input is positive from a purely positive supply and the output is negative. The catch is that the output current is more limited than when using it as a regular positive switcher. In fact, it may be more limited then we'd like because the data sheet says 700ma and that is because the output switch has to switch a higher current than normal.
So doing it this way will mean simplicity, but the output will be severely limited in current.

If that's not enough current then we are left with level shifting both the output switch and maybe the feedback. This would be something we'd have to design ourselves.
This would mean using external transistors to do all the switching, so maybe just boosting the output switch with one or two more transistors would work then using it in the buck boost configuration. The external transistors would provide more current. This wont be as simple as simply connecting the LM2576 however.

So how do you feel about being limited in current or else using external transistors?

 

[atferrari]

The LM2576 has positive output so to get a negative output we'd normally have to connect the output to ground which means we cant connect the input to ground. That means the input has to be 'floating'.

That's the 'normal' way to do it, which would require a separate transformer winding for the negative supply.

Now I see where the problem is. I was thinking of connecting the input to ground instead...

I think that I started to understand why the input should float somehow.

So how do you feel about being limited in current or else using external transistors?

I am afraid I am going to waste your time MrAl.

Albeit dated, I recalled the UA78S40. IIRC, it is adjustable (divider is accesible) prety much as the MC34063. Expensive but both available locally.

In few hours I am traveling some 720 Km to the place where I grew up, to meet my classmates after just 50 years. And precisely THAT girl will be there! She was beatiful..and still IS.

In the intervals, some datasheets and ANs have to be read. Gracias.

 

[ronsimpson]

The MC34063 is only good for 40 volts!
The error amp is no good in this case. See post #6 as how to monitor the voltage.
The circle is the MC43063.

to meet my classmates after just 50 years

Some of those people have gotten real old. Hair fell out, etc.

 

[atferrari]

The error amp is no good in this case.

Not sure why, so I have to read. Thanks for the warning.

Some of those people have gotten real old. Hair fell out, etc.

Yes, the sole comfort (?) is that we all are in that list.

 

[ronsimpson]

the error amp is trying to make the outpot 1.25 volts above the negative 36 volts. It is referanced to -36V not ground.

 

[MrAl]

Now I see where the problem is. I was thinking of connecting the input to ground instead...

I think that I started to understand why the input should float somehow.


I am afraid I am going to waste your time MrAl.

Albeit dated, I recalled the UA78S40. IIRC, it is adjustable (divider is accesible) prety much as the MC34063. Expensive but both available locally.

In few hours I am traveling some 720 Km to the place where I grew up, to meet my classmates after just 50 years. And precisely THAT girl will be there! She was beatiful..and still IS.

In the intervals, some datasheets and ANs have to be read. Gracias.

Hello,

No problem, and good luck with your reunion.

Ron said that the device is only good for 40 volts, so did i miss something where you might have said you need to go over 40 volts? I thought this was maybe 30 volts max. The MC device will be more versatile but might require a higher value inductor too.

 

[ronsimpson]

With no load, after filtering, I have 36V DC per rail available. I do not expect needing more than 2A. Say 3A if you like.

Ron said that the device is only good for 40 volts, so did i miss something where you might have said you need to go over 40 volts? I thought this was maybe 30 volts max. The MC device will be more versatile but might require a higher value inductor too.

The MC43063 is a 40V part and the raw power is 36V. If the power line is a little high some day......Do you trust the power line to never go +10%?
I would put a regulated 36V on a 40V part. This 36 is from the power line. For a hobby power supply---yes, but not for production.

 

[ronsimpson]

So you want a PWM that negative bucks 36V down to a linear regulator. This way you have the heat of a PWM and the noise level of linear.
I am using LT Spice so I used their parts.
U2 is a LDO linear regulator. R3 sets current limit. R4,5 sets the output voltage. You can use any linear regulator here.
Q1,R6,R2 is the feed back function.
>>The PWM is looking for 1.25 volts across R2 (1mA)
>>When (Vin-Vout-0.7V)=1.25V the transistor is on and 1mA is on R2 and R6.
>>The linear regulator will have about 2V across it. You can change R6 to get any voltage you want across IC2. Double R6 to get 2X voltage.
U1 is the PWM.
>>It thinks it is "boosting up" but I lied to it.
>>Use any boost PWM that can function at high duty-cycle.
I attached the Spice file.

Here is the input and output of the linear regulator. (at start up)