Adding a current limit led to an LM10 Power Supply

[tedbar]

The TI application note AN-211 (and others) have a design for a fully adjustable (V and I) power supply using LM10 op amps. See page 17 of the PDF at https://www.ti.com/lit/pdf/snoa638

I built the circuit and have been impressed with how it performs. However, it does not have an LED to indicate when it is in current limit mode, which is something I enjoy having.

Since the circuit operates in floating mode there are two issues that have me stumped on how to add a CC LED:
1) The voltage across the IC's is below 2V and thus insufficient to drive an LED
2) The output from the current limit IC is always at the +ve supply voltage

The config I am hoping to use this in will be 0 - 40V. This means that the inputs to the current limit IC will vary from 0 - 40.2V. Presumably I could use a high voltage op amp to handle the common mode voltage but that seems like an expensive way too turn on an LED. There will also be a slight mismatch between the LM10 and an additional op amp (probably not significant).

Another possibility is to  add a couple of diodes in the three stage emitter follower to increase the supply voltage and then another LM10 or other low voltage op amp. This is also an expensive way to drive an LED.

Is there an elegant way of doing this?

PS - I am strictly a hobbyist 

[David Hess]

I do not see any simple ways to add a current limit LED.  As you point out, the almost 2 volt floating supply is not enough, so Vin and COM will need to be used to power the LED.

Here is what I might do.  I would replace R10 with a constant current source so drive to Q5 is constant at any output voltage.  Then I would sample the current at the collector of Q1 where it connects to common.  When that current is zero, then the current limiting is active.  This can be can with a single NPN transistor configured as a folded cascode, so when Q1 is conducting, it supplies the current shutting off the NPN transistor, and when Q1 is not conducting, the NPN folded cascode drives an LED connected to Vin.

[Zero999]

The OP might get more replies if they posted the schematic, rather than just a PDF link


Can't you add an NPN transistor in place of D2, so it acts as diode, then it can turn on a PNP connected to V_IN?

[David Hess]

Can't you add an NPN transistor in place of D2, so it acts as diode, then it can turn on a PNP connected to V_IN?

D2 and Q2 do not turn on during current limiting.  The output of A2 directly pulls the base of Q5 down.  D2 resets C3 when the current limit is active.  Q2 limits the rise time when leaving current limit mode.  The text discusses this.

[iMo]

What is important to write is the LM10 is not an ordinary OPAMP, but an opamp with a built it Voltage Reference (rather complex).

[Zero999]

What is important to write is the LM10 is not an ordinary OPAMP, but an opamp with a built it Voltage Reference (rather complex).

Because it has other features such a power supply current of  270 μA and being able to run from 1.1V to 40V.
https://www.ti.com/lit/ds/symlink/lm10.pdf

[David Hess]

In this case the important part is that the LM10 has a rail-to-rail output, so when A2 is not current limiting, its output follows its positive supply and both are connected to the input or base of the output driver transistor.  In this configuration, A2 operates as a shunt regulator.

A PNP transistor could be inserted in series with A2's output with its collector driving an LED to ground, but this will not work at low output voltages because the difference between the supply voltage and common is only about 3 x Vbe of the output transistors, or 1.8 volts.

I guess replacing the LED with a current mirror and moving the LED to between the output of the current mirror and the positive supply would work.  Now the PNP only sees the Vbe of the current mirror, about 0.6 volts but a little higher with degeneration, and the LED will see the entire input supply voltage.  Provisions should be made to limit the current but that is easy enough.  Now we are up to 3 transistors, which is not so bad.

If availability of the LM10 is a problem, Linear Technology used to be a second source, and may still have an improved replacement.  An alternative design could use common operational amplifiers but I am not aware of any such floating designs, but it could be done starting from this one.

[iMo]

The LM10 looks at the first glance as a normal opamp, so somebody might be tempted to replace them with an opamp  .
I too have looked for the references in the above schematics, until I've opened the DS and surprised it has got the 1.1V built in reference at pin 1 and 8.
That would not be an easy part to get today (also $$), imho..

[magic]

Here is what I might do.  I would replace R10 with a constant current source so drive to Q5 is constant at any output voltage.  Then I would sample the current at the collector of Q1 where it connects to common.  When that current is zero, then the current limiting is active.  This can be can with a single NPN transistor configured as a folded cascode, so when Q1 is conducting, it supplies the current shutting off the NPN transistor, and when Q1 is not conducting, the NPN folded cascode drives an LED connected to Vin.

This is perhaps the simplest option, although I think instead of folded cascode I would try an NPN inverter with the collector pulled up to Vin, and LED goes between the collector and ground.

[BillyO]

That would not be an easy part to get today (also $$), imho..

$3.50 in stock at Mouser.

[Doctorandus_P]

I had to look for a bit what actually determines the power supply voltage for the LM10, as there is no other (fixed) voltage regulator, but it's pretty simple. R10 provides some base current for the triple darlington Q5, Q4, Q3, and then a bit more is added for the voltage over the current shunt.
With an extra diode between R10 and Q5 you can give the LM10 a bit more headroom, but at the expense of reducing the maximum output voltage a bit. You could fix that by modifying the triple darlington into a Sziklai configuration, but that may modify the regulation behavior, and thus need quite extensive testing.

Edit, Oops I missed the wire directly above Q2, which forms a direct connection between A2 and the darlington... 
The current limit is built around A2, which opens Q2 to starve the darlington of base current. A saturated BJT has a pretty low Collector - Emitter voltage. It may just be possible to add a LED at the collector of Q2. Another quick option is to use an opto coupler. IR leds in optocouplers have a lower voltage drop. Around 900mV and that may give you the headroom you need. The other side of the optocoupler can then drive your LED. A red led in the place of D2 probably does not work because the B-E junction of Q2 also has some 600mV, but an IR led instead of D2 may work.

[magic]

Using an optocoupler is a clever way out of this.

But A2 doesn't drive the darlington base through Q2, it connects to it directly. You could probably connect A2 on the other side of D2, but not 100% sure if it wouldn't affect circuit operation somehow (stability or whatnot).

edit
Well, it quite likely would, because A2 would now need to discharge C3. Perhaps yet another PNP emitter follower could be added to work around that.

[David Hess]

The LM10 looks at the first glance as a normal opamp, so somebody might be tempted to replace them with an opamp  .
I too have looked for the references in the above schematics, until I've opened the DS and surprised it has got the 1.1V built in reference at pin 1 and 8.

The built in reference is 200 millivolts.

[tedbar]

On the bench yesterday evening I added an LM358, configured as a comparator, connected directly across the shunt resistor and driving an LED from its output. If the input offset voltages mismatch (which they will) it will either activate too early or not at all. In my case it worked really well. The on/off occurred very close to the voltage across the shunt being 0V, <1mV, or within 10mA. A bit of luck helps

Looking at more suitable op amps I found the TLE2141. This allows rail to rail operation with a supply voltage of up to 44V and a common mode voltage > 40.2V. It also has offset compensation. Lastly it only costs around $1.50.

Given that it has offset compensation would it be reasonable to use that to ensure that the operating point is correct?

This would allow me to solve the problem with a single op amp and a variable resistor?

[Zero999]

On the bench yesterday evening I added an LM358, configured as a comparator, connected directly across the shunt resistor and driving an LED from its output. If the input offset voltages mismatch (which they will) it will either activate too early or not at all. In my case it worked really well. The on/off occurred very close to the voltage across the shunt being 0V, <1mV, or within 10mA. A bit of luck helps

I assume you weren't running this circuit at the full 50V, as th LM358 has an absolute maximum rating of 32V.

Presumably the shunt resistor was in series with A2's output. What value did you use?

The LM358 has PNP inputs, so the bias currents always flow out of the inputs. A high value (220k to 1M) resistor in series with one of the inputs can be added introduce an offset, larger than the 7mV offset voltage, to ensure it always gives the correct operation when used as comparator with tens of mV between its inputs.

Looking at more suitable op amps I found the TLE2141. This allows rail to rail operation with a supply voltage of up to 44V and a common mode voltage > 40.2V. It also has offset compensation. Lastly it only costs around $1.50.

Given that it has offset compensation would it be reasonable to use that to ensure that the operating point is correct?

This would allow me to solve the problem with a single op amp and a variable resistor?

The TLE2141 is unsuitable because it only works down to 4V and this circuit takes advantage of the fact the LM10 works down to 1.1V.

[tedbar]

Sorry. I forgot to mention a few points.

The circuit was run on 30V to accommodate the LM358. The LM358 was powered using the 30V and not the floating power used for the LM10s.

The intention with the TLE2141 is to run it from Vin i.e. 42V and not the floating voltage. The output from the TLE2141 swings from V- +0.1V to V+ -1V so happily drives an LED on/off.

The shunt resistor is R7, it is 0.1 ohm. Effectively I connected the LM358 to the same inputs as A2.

[jwet]

This app note was written by the amazing Bob Widlar, who among other things was the designer of the LM10- a pretty amazing part.  It was common for design and apps people to write app notes to highlight special features of new parts.  I worked for Maxim and 25 years and wrote a lot of these.  The real purpose of that particular circuit is to show off the amp operating in a floating mode.  The Vcc of the op-amp is derived from the triple Vbe drop of the output driver.  Though the supply is 50, the LM10's floats across this 3 * .7V an merrily does it thing.  There are really no limits to how high Vcc could go except the ratings of the output transistors.  This is thinly disguised merchandising- EE customers enjoy looking at app circuits and design ideas and they don't cost anything like an advert would.

All that being said, this is not a very good power supply at all.  The LM10 is a very low power amp with limited bandwidth, mediocre reference- the 200 mV is only fair.  Its an incredible clever part but I can't say I've ever found an app where it really fit. 

Overall, I applaud what you've done but there are much better ways to get there especially if the 50v doesn't matter- most any amp runs on 30v.  There is a long thread on the board about making a PS using a LM324 quad op amp which is a much better circuit.

Have fun.

[Doctorandus_P]

The overall design of a regulating circuit riding along on the top of the output voltage is quite common. Until recently I thought it was a bit weird or at least unusual, but after diving a bit deeper you find quite a lot of this type of power supplies. Often they have an extra mains transformer for the opamps and the regulation circuit.

It probably started all with Harrison:

http://hparchive.com/Journals/HPJ-1962-07.pdf

[David Hess]

The overall design of a regulating circuit riding along on the top of the output voltage is quite common. Until recently I thought it was a bit weird or at least unusual, but after diving a bit deeper you find quite a lot of this type of power supplies. Often they have an extra mains transformer for the opamps and the regulation circuit.

It probably started all with Harrison:

http://hparchive.com/Journals/HPJ-1962-07.pdf

Too bad they do not show more details about how a floating regulator is designed.  The Harrison/HP service manuals are worth studying.

I wonder how many people are left who know why those schematics all show PNP pass elements.

[Doctorandus_P]

The reason is of course simple and logical. With an NPN as an emitter follower, the output voltage of the opamp stays very much at the same level. But you may as well mirror the whole thing and use a PNP transistor.

Another reason is that though the '60-ies (and I think also the '70-ies) it was difficult (expensive, lower specifications) to make PNP BJT's. Probably similar to MOSfet's today. Due to the difference in mobility in the charge carriers (holes versus electrons) the "P" versions do not work as well. It's a fundamental principle of semiconductor physics, but I forgot the details.

I have also seen designs with a Push / pull stage, similar as used in audio amplifiers. The concept is also easily extendable to full 4-quadrant operation.

For variation, below is one with an N-channel MOSfet. It does not look particularly well designed though.

[David Hess]

Another reason is that though the '60-ies (and I think also the '70-ies) it was difficult (expensive, lower specifications) to make PNP BJT's. Probably similar to MOSfet's today. Due to the difference in mobility in the charge carriers (holes versus electrons) the "P" versions do not work as well. It's a fundamental principle of semiconductor physics, but I forgot the details.

It is because they were designing for germanium power transistors where PNPs were less expensive and higher performance.

[jwet]

On the floating business- true, there is nothing unusual about developing an additional supply with an extra secondary, etc and floating that above about the output.  The things that is very unique and shows off a  unique aspect of the LM10 is that the isolated supply isn't developed externally, its simply the voltage drop across the combined pass element- about 2v, this is pretty unique and is what Widlar was trying to show.  In the next circuit in that app note, he makes a 200V linear with this same trick that I think demonstrates the point better.  Saves a secondary, etc.

[mawyatt]

Another reason is that though the '60-ies (and I think also the '70-ies) it was difficult (expensive, lower specifications) to make PNP BJT's. Probably similar to MOSfet's today. Due to the difference in mobility in the charge carriers (holes versus electrons) the "P" versions do not work as well. It's a fundamental principle of semiconductor physics, but I forgot the details.

It is because they were designing for germanium power transistors where PNPs were less expensive and higher performance.

Hole mobility is highest in Ge, lower in Si and lower in GaAs, with electron to hole mobility of ~2, 3 and 21 to 1 respectively (why there is no GaAs CMOS, painful story after massive $ USG and Honeywell investments   )

Best

[David Hess]

On the floating business- true, there is nothing unusual about developing an additional supply with an extra secondary, etc and floating that above about the output.  The things that is very unique and shows off a  unique aspect of the LM10 is that the isolated supply isn't developed externally, its simply the voltage drop across the combined pass element- about 2v, this is pretty unique and is what Widlar was trying to show.  In the next circuit in that app note, he makes a 200V linear with this same trick that I think demonstrates the point better.  Saves a secondary, etc.

I really like those types of application notes when circuit operation is described in detail.

The LM10 could be replaced with a modern low voltage rail-to-rail output operational amplifier and 1.2 volt reference.  Total cost of the parts would be about the same.  Some of the suitable operational amplifiers are faster at the expense of higher quiescent current.

3-terminal regulators do the same thing, with the regulation circuitry operating across the minimum voltage difference between the input and output.

[jwet]

Don't forget the min Vcc for the LM10 is 1.1v!  This isn't too common in my experience!

Quote >>
3-terminal regulators do the same thing, with the regulation circuitry operating across the minimum voltage difference between the input and output.

Not quite- since they have three terminals, they have the full Vin range to work with.  There were some parts called "EZ-Droppers" that were two terminal regulators that were designed to make 3.3 from 5v (terrible parts BTW).