Power supply design option, dual rail+/-30V) or two floating +30V. Pros vs Cons?

[David_]

Hello.

I've got a decision to make regarding the power supply topology for my lab power supply, ether I stick with a center-tapped transformer that generates a 0 to +30 rail and a 0 to -30V rail or I'll simplify the design somewhat since I have some time ago opened up the transformer and split the center-tap I could go for two identical floating 0 to +30V rails that could be connected to result in +30V/-30V. It would only need one extra digital isolator for the digital control signal to separate the common planes, but

Is there any pros in keeping it +/-30V?
And is there any cons in doing it the double floating channel way?

The circuits will need a -3V supply each, I have just begun to research the options for that but that might be a problem.

I appreciate all kinds of input.

 

[erikl]

You probably mostly will need the +/-30V option, but perhaps once you'll need 2 separate (isolated) power supplies. So I think it would be better to build the latter option - even if it is more complex - because it also includes the first option.

 

[David_]

I guess your right, it will be a easier PCB layout but I have to solve the -3 maybe -5V rails.
I have never worked with charge pumps and such, now that i think of it I just so happen to come across yesterday a article about generating a negative voltage out of a microcontrollers PWM signal. I should look that up strait away.

 

[crutschow]

What's the purpose of the negative rail? There are many power supply designs that need only one rail. For example the LT3083 is a low-dropout 3A Linear Reg that adjusts down to 0V output with a single supply and which can be readily paralleled for more current.

 

[David_]

Are you sure about that 0V? I have just read the data sheet and although I did not find anything pointing to it I do for some reason mistrust that the 0V spec, or it might have been the LT3081 that only gave 0V under certain load situations that made it in many cases not true.

The problem is that LT3083 only goes up to 23V, I had remembered otherwise but there you go. My recent change to the design with two floating positive rails opens up options that was previously not possible, as it is now I am using(only theoretically)the TPS7A4701 which gives out
+1.4V to +34V: http://pdf.datasheetcatalog.com/datasheet/texasinstruments2/tps7a4701.pdf

It has a very tricky package to solder but one I had searched for regulators till my eyes bleed before I found a suitable adjustable regulator that have an Enable pin and a negative counter part, I now only need a positive but its seams that this is my only option anyways. Although I have just assumed that as with other regulators such as LM317 this could be brought down to 0V with a negative voltage, if some regulator with high enough output voltage could do that without a negative voltage that would be so darn pleasing but as it looks now a PWM based "diode charge pump inverter" circuit presents me with so much to gain(0-30V instead of 0-23V) that that's what I'll go for. But I am not really happy with this hole thing, the project involves so many things and learning programming takes a lot of my thoughts so I have forgotten why the Enable function was so important...

I think I was/am going to have the rails adjustable down to 0V(no question about that) and have a momentary-push button right next to the output connector to enable/disable the rail, I have a function generator that has such a feature and it is very nice to have BUT now that I think of it my way would not be the same. Or it might, the thing is that it acts as if it has a relay just behind the connector so the output function is not terminated but ready to jump in, or out so to speak.
I get these unfounded ideas that i tend to stick to like for some reason I felt that since I'm going for as a precise instrument as related to my capabilities could possibly be produced it could not hurt to minimize all sources of possible errors no matter how tiny, so I preferred not to put the output through a relay. But as with many other things I realize that that's stupid and the output voltage will be calibrated through software anyway, but I'm not clear on if a active relay coil can have any at all effects on a close proximity voltage rail. Presumably it could only have any effect in the moment of on/off action?
But the enable function still serves a purpose seeing as all other parts(close to) have enable functions, a nearly complete sleep-mode could be implemented.

But I have a huge, enormous, catastrophically disabling problem with power dissipation, all the LT308X chips seems so very well suited to be paralleled. Originally I was going to use multiple LT3081/LT3090 to spread out the dissipation on a heat sink dedicated PCB with fans but the price of the chips killed that together with the functionality but I am weighing the benefits of the enable function against the complexity(or simplicity) of a power handling circuit. It isn't harder than to "bypass" the regulator with a couple of power transistors, no?

I don't know how I would proceed to share a PDF with the schematic here in the thread, it would be a big as nasty thing to put in a picture. If you have any idea for a suitable way of sharing a big schematic it would be fun to do so. //never mind, I know and it will be posted soon.

 

[David_]

This is by no means a finished schematic.

 

[schmitt trigger]

David;
indeed the TPS7A4701 is an extremely versatile regulator, but a QFN package is beyond my capabilities for hand soldering.

As I become older and my eyesight becomes weaker, how come electronic components continue to become smaller and smaller?

Anyways, to your original question: I would do two separate supplies. It would be more flexible that way.

 

[David_]

I have never soldered a QFN package before but it should be no problem, I guess its a advantageous time to have young inexperienced eyes.

The two separate supplies idea shows more and more advantages, Negative rail current sensing was a headache that was solved by LT6105 but now that the negative rail is no more I can switch to INA226 and put it after the regulator right before the output connector and get precision readings of voltage, current and wattage.

I have made the choice to go with two separate positive rails but how about the ground connections?
there needs to be different choices of connecting the rails.
Scenario 1: two completely isolated 0-30V rails: no connections.
Scenario 2: two separate 0-30V referenced to one system but still floating: both rails common is connected.
Scenario 3: two separate 0-30V referenced to one system with PE-ground: both rails common connected with PE-ground.
Scenario 4: +/- 0-30V floating: Channel 2 output connected to Channel 1 common.
Scenario 5: +/- 0-30V PE referenced: Channel 2 output & Channel 1 common connected to PE-ground.
Scenario 6(not needed but if its no hassle why not enable it): separate 0 to -30V: both Channel outputs connected to one common PE or not.

How does I get 0 to +60V and 0 to -60V?
Is it scenario 4 and 5 using both extremes?

If I use SPST relays for each of the output terminals, two sets of banana connectors arranged physically ch2,com ch2,V+ --- ch1,com ch1,V+, every terminal can then be connected to a peace of copper/PCB to act as a junction, that junction does also have a SPST relay to PE ground.
Then ATmega2560 will activate the relays two at a time, or one at a time with the option of activating PE in all cases, I could have monetary push buttons right next to each terminal activating its relay with a programmed fail safe to ensure that no more than one relay on each channel could be activated at any single moment plus a button for PE ground, that would enable me to freely connect the channels in all possible configurations.

If I would use a relay for each channel to control its availability at its output connector, would it suffice to only break/make the V+ line and always leave the COM connected in order to disable the channels?

Thanks for your inputs.

- - - Updated - - -

I wonder what the difference would really be comparing soldering the QFN package without its big pad underneath soldered, and with the pad solder to the board. It could be a quite substantial difference in heat transfer rate could it not?

 

[schmitt trigger]

You have to solder the thermal pad in the bottom to the board. This is a must, it acts as a heatsink. Without it, it will overheat, or will not operate correctly as heavy ground currents may flow from the power pad.

 

[David_]

My plan was to solder it but I did not think of it as that critical, Had not even considered ground current capabilities of those other tiny ground pins.
I might have good soldering skills and but I sure have some way to go in proper electronic reasoning

About how to actually solder the pad, I assume that it is intended to be soldered in a real production environment like a solder oven. But first I thought that I'll drill a hole, as small as possible but big enough so succeed in the center of the pad. And with one of those precision type solder tips that have a very very fine tip and with a curve reach through and heat the pad and a appropriate amount of solder and then with fines fit the package and keep tip of the pen in place long enough to enable the solder to properly make contact with the pad, the tip would need to be quite hot so time is paramount in order to not damage the part but still spread the heat around the whole pad. And in the end pull the tip out while fixing the part in place, that algorithm is however not a easy thing to pull of but I will work it out. How hard can It be I keep thinking, well I am certain that I will get the part solder in place but regarding those solder temperature specs in data sheet, I wonder how critical they are in order to not damage the part at all.

I have begun contemplating on a exact design for a external pass-element something like a MJ3001 with the base connected to the output and regulator input voltage to collector, and emitter will be connected to a current sense resistor and finally the output terminal .
If so the soldering of the pad would not be as important and I would not need to go to such length to solder it fully. just to be sure there's a proper connection somewhere on the pad.

But my knowledge have been stretched to its limits and I need some guidance in choosing a topology for the pass-element, I was going to open another thread but I might as well ask here first.
The TPS7A4701 does require certain amount of output capacitance to be stable, so there would have to be output capacitors between the output and base of the pass element, the minimum required output current is so miniscule. Could the base current be enough perhaps?
The output would with this design be the raw voltage rail through the NPN darlington, could there be any need for output capacitor after the emitter?
or would it matter to increase the regulator input capacitance seeing as they would also be connected to the collector?

A resistor following the emitter, in what way would the value of that resistor impact the current capabilities of the transistor?

If anyone have looked at the schematic, is the resistor following the NPN/Zener regulator really needed?

 

[D.A.(Tony)Stewart]

a Split rail with an option to track or control independently is useful.

The efficient way is of course to use SMPS with a single rectified DC source perhaps with phase control and a dual SMPS converter one positive and one negative.

The transformers will therefore operate at the switch rate and not 50/60Hz.

 

[schmitt trigger]

Adding an external darlington NPN to the LDO regulator output, defeats the low dropout capabilities.