LT3083 Power supply schematic tips?

[FotatoPotato]

Hey everyone,

So I just finished my first revision of my 4 channel bench schematic. Please note that this schematic would be copied 4 times for each channel, so that's why there is only one schematic. Also, note that this design is heavily based on Dave's uSupply design, but unlike his design, mine will be controlled by 10 turn Pots and not an arduino and I will be adding a power transistor to the LT3083/3080 to enable a higher current output + a hard output on/off.

Anyway, my real question is that because my design is going to be meant for a much higher power application, and the fact that it is based off of Dave's comparatively low power design, is there anything I'm missing? and/or is there anything that I should add? Obviously I will be building a breadboard prototype but while I wait for components to arrive I would like to finish/ correct my schematics.

Hopefully everything is clear and easy to understand but if not, then my apologies and I will provide as much assisting as possible.

EDIT PLEASE READ: My apologies to anyone who has been or is confused by this post. I should have stated my design goals a wile ago and my choices for my design. Basically I'm building a 4 channel bench supply. There will be 2 high current, low voltage outputs (Call this channel A) and 2 high voltage low(ish) current outputs (Call this channel B). The high current outputs will do 0-12v at 0-10A and the high voltage outputs will do 0-30v at 0-5A. Each channel has its own transformer and I want the ability to be able to combine the two outputs of each channel to get more power Ex: Combine both 12v outputs of channel A to get either 0-24v at 10A or 0-12v at 20A. The same goes for channel B.

Again, sorry for any confusion. I hope this helps clear things up.

Thanks! 

[TurboTom]

- U2A's non-inverting input more posiitive than its supply voltage - input clamping diode will conduct and probably U2 will be destroyed (latch-up?).
- Check Q2's SOA. Maybe use two or three of these transistors in parallel? Current sharing resistors in emitter connections will be required.
- Shunt is a little high for 10A (2V drop at rated current).
- Maybe add resistors in Q3 / Q4 drain connections.
- What is IC1 supposed to do? I don't see its temperature output signal to be routed to any (protective?) circuitry.
- All of your settings relate to the accuracy of your 12V rail (provided by an 7812). If this is accurate enough for your application is soemthing you've got to decide.
- Your current limiting circuitry will likely cause oscillation. You've got a very high gain (running U3B as a comparator, directly driving Q5) and C11 will cause a delayed response. I may be wrong but that setup appears to be quite an efficient oscillator at some frequency.

[Kleinstein]

Starting with a voltage regulator is a poor idea for a lab supply. It is very difficult to add current limiting as an afterthought to such an regulator - it is usually easier and better performance to build the regulator from scratch, using a reference, OPs and power transistors.

By it's own for a well defined load it is already tricky to add extra transistors for more power to an voltage regulator - it is even worse for a lab supply where the load can vary.

For a 1st such circuit one should keep the power lower, so less magic smoke can escape if things go wrong.

[Zero999]

The LM334 will not work at output voltages below 0.9V.

A current mirror could be used which will work at voltages very close to zero.


The LM358's common mode range is being exceeded, so won't work.

How about using the LT3081? More than one can be connected in parallel for higher output currents, than 1.5A.
http://www.analog.com/media/en/technical-documentation/data-sheets/3081fc.pdf

[FotatoPotato]

@Hero999

How so? In Dave's design his supply goes from 0-20v if I remember correctly. Also I'm not quite sure what you mean by "common mode Range".

I'll probably use the LT3081, Its just that on the output of this specific channel I will be pushing up to 10A. So adding 6 or 7 LT3081's for each channel would get really expensive. That's also why I have Q2 connected between the input and output

@Kleinstein

Are you talking about the LM7812? Because it isn't being used for anything power related. It's only purpose is to provide a 12v rail to power the opamp's. 
 
I want to stay away from making the power supply using just opamps and transistors as using a pre-made high quality adjustable LDO will give me better stability and lower output ripple. Unless you can find a design that would work better than this one. In that case then I would be happy to use it instead.
The added transistor on the main regulator is pulled directly from the datasheet, I think its pg. 17. That’s also where I got the ON/OFF circuitry from.

@TurboTom

IC1 is an LM334 and it is an adjustable current source. Its needed because the LT3080/83 needs a constant current draw in order to be stable. So I have the LM334 on the output to always draw 1mA.
 
I do agree that the shunt is a bit oversized and I did find a 35W one that will work much better.

I'm using the LM7812 as a place holder for a more accurate LDO that will supply the opamps so that’s also taken care of. I’m thinking of using this http://www.ti.com/lit/ds/symlink/lm2940-n.pdf

Thanks so much for the responses guys, I really appreciate it! 

[Zero999]

@Hero999

How so? In Dave's design his supply goes from 0-20v if I remember correctly. Also I'm not quite sure what you mean by "common mode Range".

I'll probably use the LT3081, Its just that on the output of this specific channel I will be pushing up to 10A. So adding 6 or 7 LT3081's for each channel would get really expensive. That's also why I have Q2 connected between the input and output

I'm not familiar with Dave's design, please post a link.

The common mode range is the voltage span an op-amp's inputs will work over and you've exceeded the LM358's by powering it from 12V and biasing its inputs at 16V. The inputs should be a couple of volts or so below (the exact figures are on the data sheet, which can be found using a search engine) the supply voltage for it to work properly.

[FotatoPotato]

Dave's Supply is part of a 9 video epic on how to make one. I would highly recommend watching. Just search "EEVBblog Power supply tutorial" Here is a link to his design. I'm using the REV C schematic and only the "PSU section of it, nothing else.

https://www.eevblog.com/projects/usupply/

Enjoy

[Zero999]

Dave's Supply is part of a 9 video epic on how to make one. I would highly recommend watching. Just search "EEVBblog Power supply tutorial" Here is a link to his design. I'm using the REV C schematic and only the "PSU section of it, nothing else.

https://www.eevblog.com/projects/usupply/

Enjoy

http://www.eevblog.com/files/uSupplyBenchRevC.pdf
Unless there's something I've missed, I don't see how U12A can work, at lower output currents. According to the data sheet, its inputs should be at least 1.5V below +V to work, but in real life it'll probably be a bit better than that.

The problem is the input stage of the LM358 (a simplified schematic can be found on the data sheet) has two PNP transistors ,Q1 to Q4, on each input, which require their bases to be below +V to turn on. The typical turn on voltage for a BJT is about 0.6V, but it will be less at these low currents and the 6µA current source will probably work with minimal voltage drop: a couple of hundred mV. The two transistor base-emitter voltages add up to 1.2V and the minimum voltage requirement for the current source makes up the rest, hence 1.5V. In reality it may be fine down to 1V but it will be temperature and device dependant.

http://www.ti.com/lit/ds/symlink/lm158-n.pdf

An op-amp with a rail-to-rail input should be used.

[FotatoPotato]

Would something Like this be better?: http://www.ti.com/lit/ds/symlink/lmc660.pdf

[Kleinstein]

The problem with Daves supply is that is did not really work and still had/has quite some issues.  It was a long project on the going, but not really finished because of the problems.  So in some sense one can learn from the mistakes / poor decisions made.

[FotatoPotato]

well if that's the case, is there any design that you know of that would be better suited for my needs? The only thing I can think of doing is using discrete op amps and power transistors to make my own linear regulator with current limit. But I'm just worried about stability, noise and the ability to go down to 0v without any problems. I'll keep looking for some more designs but if you know of any then I would highly appreciate it! Oh and btw, I do have like 4 LM339's and like 6 LM358's and a BUNCH or power transistors lying arround, so if you find a new design that i can build I might be able to make it today and give some feedback on it.

Thanks! 

[Zero999]

Would something Like this be better?: http://www.ti.com/lit/ds/symlink/lmc660.pdf

Sorry, not suitable. It does have a rail-to-rail output (I forgot that was also an issue with the LM358: Q5 and Q6 limit the output voltage) but its input common mode range is more limited, than the LM358!

The original circuit stands more of a chance of working without the unity gain follower. The current taken by the 18k input resistors on the differential amplifier will not affect the result much.

Going back to your previous post:

I'll probably use the LT3081, Its just that on the output of this specific channel I will be pushing up to 10A. So adding 6 or 7 LT3081's for each channel would get really expensive.

It's possible to configure a pass transistor with an op-amp to pass a multiple of the current flowing through the LT3081. I'll draw up a schematic later, as I don't have time now.

[FotatoPotato]

So I just dug up the o'l electronics bible and found this schematic. What do you think? If I use LM358's instead of the opamps listed would it work?

[karoru]

It wouldn't because it uses this funky Harris opamp compensation input for current limiting.

[Wolfgang]

Naah, just have  look at the op amp supply voltages !

[David Hess]

Starting with a voltage regulator is a poor idea for a lab supply. It is very difficult to add current limiting as an afterthought to such an regulator - it is usually easier and better performance to build the regulator from scratch, using a reference, OPs and power transistors.

By it's own for a well defined load it is already tricky to add extra transistors for more power to an voltage regulator - it is even worse for a lab supply where the load can vary.

I agree that it significantly complicates frequency compensation.  I would not recommend enclosing an integrated linear regulator with or without current boosting inside of a feedback loop unless the designer is prepared to test transient response under all operating conditions.  The circuit as shown is a disaster.

But an integrated regulator has the advantage of simplifying driving the power pass elements and with cleverness can implement thermal and SOA protection.

For a 1st such circuit one should keep the power lower, so less magic smoke can escape if things go wrong.

I agree.

The common mode range is the voltage span an op-amp's inputs will work over and you've exceeded the LM358's by powering it from 12V and biasing its inputs at 16V. The inputs should be a couple of volts or so below (the exact figures are on the data sheet, which can be found using a search engine) the supply voltage for it to work properly.

The ancient LM301A and many JFET input operational amplifiers make great high side current sense amplifiers because their input common mode range includes their positive supply.

However using a separate current sense amplifier and current limit amplifier is very tricky because of decreased phase margin.  It is better to combine the two into one operational amplifier operating at the high side voltage.

well if that's the case, is there any design that you know of that would be better suited for my needs?

Study the example shown below from National Semiconductor linear brief 28.

[prasimix]

Perhaps you can take a look at Frex's design on another DIY forum also based on LTC's regulators. If nothing else he claims in post #93 a spectacular noise level that beat even LTC's design for an order of  magnitude. So far I didn't find anything even closer when SMPS preregulation (and even PFC) is involved.

[Kleinstein]

So I just dug up the o'l electronics bible and found this schematic. What do you think? If I use LM358's instead of the opamps listed would it work?
....

The LM358 would not like a 60 V supply. In addition this design used some compensation / adjust pin to couple the current limit to the voltage control. So it would be a little difficult to change the OP.  This type of regulator looks already much better than Daves try, but still has some week points: changing the loop gain with voltage setting (which is a common weak point of many old designs) and using MOSFETs in parallel without balancing resistors (this might be OK with those old types, but won't work with most modern ones).

The circuit from linear brief 28 has some good aspects, but the LM395 is hard to get now. LM317 might work as a replacement. Still it's not that practical for high power, as there is quite some drop out (LM317+OP+balancing resistors).

The more typical circuit the the floating regulator like used in many cheap supplies (e.g. Korad3005, HY1803) and in a similar form in old HP supplies. Circuit diagrams for some of these are available.

[StillTrying]

http://www.eevblog.com/files/uSupplyBenchRevC.pdf
Unless there's something I've missed, I don't see how U12A can work, at lower output currents. According to the data sheet, its inputs should be at least 1.5V below +V to work, but in real life it'll probably be a bit better than that.

It's been mentioned a few times in the distant past, it'll keep coming around every now and again.
https://www.eevblog.com/forum/beginners/noninverting-amplifier-gain-in-micro-supply/msg986727/#msg986727

[David Hess]

The circuit from linear brief 28 has some good aspects, but the LM395 is hard to get now. LM317 might work as a replacement. Still it's not that practical for high power, as there is quite some drop out (LM317+OP+balancing resistors).

I am just using it as an example of the following; I would not use it directly either.

1. Current limiting is high side and combines the current amplifier and current error amplifier so there is none of that nonsense about multiple amplifiers in series lowering phase margin.

2. The output capacitance is very low (0.22 microfarads!) considerably improving constant current performance.

3. Voltage and current error amplifier gain is fixed easing frequency compensation.

4. The current error amplifier is clamped.

5. The output has an active pull-down.

I would not use the LM395 either but it is essentially the output stage of an LM317.  If the LM395 was available at the same price as the LM317, then I would use it.

If I were building something from scratch today, I would:

1. Move the high side current sensing to the output of the pass element before the output capacitor.  This can present some difficulties at low output voltages for programmable current limiting if used but it is better than the alternative of a separate current sense amplifier and it provides a more accurate current limit.  This basically duplicates the old Tektronix PS501/PS503 design and it is suitable for bipolar tracking outputs.

2. Moving the current shunt to the output before the output capacitor improves phase margin if AC feedback is taken before the shunt.

3. I would use an integrated regulator either for the output stage or for driving bipolar output transistors *if* this was not a problem for maximum current at low output voltages due to SOA (safe operating area) considerations.  If I was a problem, then I might consider a cascode design despite the complexity.  The alternative is significant derating of the output stage or including discrete SOA and thermal protection anyway.

4. I would at least consider clamping of the current and voltage error amplifiers for faster response however this presents practical difficulties with modern parts which do not provide for external clamping or compensation.  In the LB-28 example, the current error amplifier is clamped through its external compensation pin.

[FotatoPotato]

I think I'm also going to investigate the idea of making a DC-DC Buck converter instead with a really good output filter to reduce output ripple. After all, the efficiency will be much better, It will be easier to make as there are AIO IC's like the LM2678 that can easily do 5A at 1.2-30v and it will cost a lot less. So all I'll need to do is add a current limit circuit on the output and everything is hunky dory.  Also, this power supply doesn't have to have the lowest ripple of all time, as long as its under 50mv peak to peak at a full load then its good enough for me. The only obvious downside will be that I wont be able to go all the way down to 0v. Unless there is a way to specially control the LM2678 that I don't know of to get to 0V. What do you guys think?

Here is a pic of what I had in mind for the current limit.

[FotatoPotato]

BINGO! 

I found the answer to everything! The LTC3649 is a high efficiency buck converter that has current limit, voltage limit and a whole bunch of other stuff built in. Plus at the bottom of the datasheet it shows this schematic for a lab supply.

This looks like the perfect option

What do yo guys think.

[Zero999]

http://www.eevblog.com/files/uSupplyBenchRevC.pdf
Unless there's something I've missed, I don't see how U12A can work, at lower output currents. According to the data sheet, its inputs should be at least 1.5V below +V to work, but in real life it'll probably be a bit better than that.

It's been mentioned a few times in the distant past, it'll keep coming around every now and again.
https://www.eevblog.com/forum/beginners/noninverting-amplifier-gain-in-micro-supply/msg986727/#msg986727

Where is revision D? There's no mention of it in the thread you linked to or on the page linked aboveand editing the link to RevD.pdf goes nowhere.

http://www.eevblog.com/files/uSupplyBenchRevD.pdf

Would something Like this be better?: http://www.ti.com/lit/ds/symlink/lmc660.pdf

Sorry, not suitable. It does have a rail-to-rail output (I forgot that was also an issue with the LM358: Q5 and Q6 limit the output voltage) but its input common mode range is more limited, than the LM358!

The original circuit stands more of a chance of working without the unity gain follower. The current taken by the 18k input resistors on the differential amplifier will not affect the result much.

Going back to your previous post:

I'll probably use the LT3081, Its just that on the output of this specific channel I will be pushing up to 10A. So adding 6 or 7 LT3081's for each channel would get really expensive.

It's possible to configure a pass transistor with an op-amp to pass a multiple of the current flowing through the LT3081. I'll draw up a schematic later, as I don't have time now.

Here's the schematic. It enables the LT3081's current limiting and monitoring pins to be used. The ratio of R1 and R2 determine the ratio of current taken by the pass transistor and the regulator IC.



I_Q1 = I_U1*R1/R2
I_U1 = I_RL/(1+R1/R2)

In the example above U1 takes 1/(1+R1/R2) = 1/(1+0.12/0.02) = 1/(1+6) = ¹/₇ of the current taken by RL.

Note: U2's common mode range needs to include its positive supply for this circuit to work.

EDIT:
Q1 will probably need to be three or so transistors in parallel, with emitter resistors to help with current sharing.

[FotatoPotato]

Thanks for the info, that will definitely come in handy for some other projects. But I probably won’t use it for my bench supply as I have found what seems to be a really good hybrid Switchmode/linear design in the data sheet for the LTC3649. It looks like it will be the perfect fit for my design as long as I parallel up 2 of the LT3086’s on the output.

Here is the data sheet (pg. 20 has the schematic): http://www.analog.com/media/en/technical-documentation/data-sheets/3649fb.pdf


Thanks