Clearing the bench then going to breadboard the LM358 and start playing around with the supply voltage.

 

@dl324 Gremlins have struck again Dennis. Was trying to get the diodes trimmed to get as close to 10V & -20V as I could and it quit working. So I put the diodes back as it was and still not working. Checked the op amp with an AC follower on each side and it is working. Coming out of the 1st op amp is the negative V and it is following the LM317. The second op amp is a comparator between the variable input and the negative rail but the value for pin 7 does not change. I am getting 8.92V for +Vs and -21.16V for -Vs. Could this be the problem? I'll check the PNP XSTR later and somewhat baffled as to it not having any resistors in it's section of the circuit.

 

Was trying to get the diodes trimmed to get as close to 10V & -20V as I could and it quit working.

The values for those supplies isn't critical. You just need to make sure they're enough to give you the common mode input range you need. Within 5-10 percent should be fine.

Once you understand how all of the circuit blocks work, you know what's important.

Checked the op amp with an AC follower on each side and it is working. Coming out of the 1st op amp is the negative V and it is following the LM317. The second op amp is a comparator between the variable input and the negative rail but the value for pin 7 does not change. I am getting 8.92V for +Vs and -21.16V for -Vs. Could this be the problem?

Could you mark voltages on the schematic that's closest to what you breadboarded? I think that'll be the one with the calibrate pot in the middle of the voltage divider, but a couple resistor values will be different.

I'll check the PNP XSTR later and somewhat baffled as to it not having any resistors in it's section of the circuit.

It doesn't need any because the BE junction is in the feedback loop of the voltage follower.

 

Shown as built without rectifier, 1000uf on +/-, Vs Zeners (3Vz, 13Vz), Vs measured @ pins

 

Pin 7 shouldn't be able to go below the opamp's negative supply.

Can you adjust R4 to get the output of IC1A to be -20.68V?

Just for grins, check the output of the LM317 to make sure there's no "ripple" (manufacturer speak for oscillations). When I was experimenting with my prototype of the simplification of this circuit, I found that the LM317 was oscillating when I reduced the output to minimum. I tried 2 other LM317 and found one that didn't oscillate. Putting a 10uF electrolytic on the output stopped the oscillations; which is about what the manufacturers suggest.

The problem manifested itself on a DVM as an output voltage below 1.25V.

 

Between R3 & R5 with R6 removed should be 10.35! Sry, typo I just caught. I'll get back on in after another cup of coffee and wake up a bit.

 

EDIT: The transistor failed. Tested and it was intermittent shorting, replaced and now working. The transistor tester did not show fault but built a test circuit and it was bad. Getting -0.04V to -18.7V. But still, see below, very quirky voltage...

The voltages have dropped a bit this morn so I let it warm up for a while and still lower. Pin 1 is now -20.41. LM317 output is flat until ~20V then a sawtooth. Already had 1uF on each side so added 10uF with no real change. Ended up with 2200uF and still not smooth.

Then it went away!! This is @ -21.96V with only the 1uF cap.

I had swapped out the LM317 at least 4 times turning the circuit off each time. The DK regulators were able to output the ~-22V and my other stock topped out slightly greater than -20V. After it quit, I put the DK part back in and still no ripple? This is screwy... Pin 1 @ -20.53 now and still not following.

 

Then it went away!! This is @ -21.96V with only the 1uF cap.

The datasheet recommends 1uF tantalum or 25uF electrolytic. Those are conservative numbers and I was able to stop oscillations with a 10uF electrolytic. All of the LM317 I've been using are STM; likely from the same lot because they were in the same tube.

The DK regulators were able to output the ~-22V and my other stock topped out slightly greater than -20V. After it quit, I put the DK part back in and still no ripple? This is screwy... Pin 1 @ -20.53 now and still not following.

With a 22V input voltage, I'd be conservative and assume a dropout voltage of at least 2V. Dropout voltage tends to decrease at higher temperatures at low currents.

To get things working, why not try for 15V? That should keep you away from problems caused by trying to be on the edge of things not working. Once it's working properly at lower voltages, you can push the limits until something stops working.

If you're not getting tracking voltages, adjust R4. If that doesn't help, normal troubleshooting technique would be to verify that the voltage from the LM317 was stable. Then work forward from something that's working until you find something that isn't; or start at the negative output and work backwards until you find something that is working.

 

@dl324 see above edit, now working!

 

@dl324 see above edit, now working!

Did you have a significant load on the transistor?

Once you understand how that circuit works, I can show you how to simplify.

Then I can show you how to add current limiting.

For my current limiting circuit, I opted to use LM393. They're inexpensive (like LM358) and I probably have over 100 of them. There are some components that I prefer to use and I buy them in quantity 100+ to get a price break.

When I first drew the current limit circuit, I was thinking about using some comparators with rail to rail inputs. They were hard to find (in stock at the likes of Newark, Digikey, Mouser - Jameco options were non-existent) and too expensive for my taste when I could find them. So I decided to work around the limitations of the LM393.

For the switch, I used some low threshold voltage MOSFETs. That meant I had to limit the gate voltage. I also opted to not use the full Vgs because there was little benefit in reduced on resistance and it gave me more flexibility in which P MOSFET I could use (I have 2 and wanted the circuit to work with either).

 

Did you have a significant load on the transistor?

No, I never connected a load other than the instruments but I did have it in backward originally and who knows what may have happened during the build.

Once you understand how that circuit works, I can show you how to simplify.

I know what a comparator is but that is about it. And I am weak on the transistor also. Next on my study list is Malvino's Electronic Devices to flesh out the solid-state components and circuits more than just the introduction from Floyd and Grob. I want to add the current limiting and put this on a PCB but will hold off until I finish Malvino and understand it better. Many thanks for leading me this far!

For my current limiting circuit, I opted to use LM393.

I have a few and some LM311, and LM339s but haven't worked with them as yet.

I'm going to leave this breadboarded and if I get the itch to put it on protoboard I will. I noticed you are etching your own boards old-school style and I may want to get into that. I can't see sending it out for fab as a single at the price they want. I'll call this done for now and many thanks again for all the help Dennis!

 

I know what a comparator is but that is about it.

A comparator is much like an opamp. The front end is a differential pair like opamps. They don't have a compensated stage and the outputs used to all be open collector; but that's changed now. Opamps are sometimes used as slow comparators and comparators are sometimes used as slow opamps.

I have a few and some LM311, and LM339s but haven't worked with them as yet.

They're both comparators. LM339 is the quad version of LM393. I have both in my supplies, but rarely use the quad version. LM311 is too good of a comparator for ordinary circuits.

I'll call this done for now and many thanks again for all the help Dennis!

Don't stop until you understand how the circuit works. That was the whole point of the circuit. It's more complex than it needed to be, but it combined several basic building blocks to make a tracking regulator. If you take the time to understand how each of them is working, you can use them as building blocks in other circuits.

The simplification I did combined three blocks into one.

I noticed you are etching your own boards old-school style and I may want to get into that. I can't see sending it out for fab as a single at the price they want.

OSH Park will do 3 copies of a design for what people say are reasonable prices ($5/square inch for 2 sided boards). I get more satisfaction from making my own boards.

 

Don't stop until you understand how the circuit works.

I understand the basics. R4 balances the Voltage divider between R3 &R5. Which feeds into the inverting DC 3X amp using the A op amp and R6 as the Ri and R7 as the Rf. Which is fed into the comparator using the B op amp and comparing the output of the A op amp against - V ril and using that to control the -Vo rail voltage with the 2N3906 being fed from the -Vo from the filtered rectifier. What I can't do at this point is the math for the comparator and the XSTR which is why I still need to do some further studying.

$5/sq in is a compelling reason to go SMD... Which at my age...

 

I understand the basics.

If you want to have those building blocks in your tool box, try to understand everything you can about them. Look at the values I used and see if there was any logic behind them.

Is there a reason why that voltage division was chosen for the divider? Originally it was 2:1, why did I change it to 3:1? Wouldn't it have been easier to just feed the output voltage of the LM317 to the opamp?

Why were those resistor values used?

Why was LM358 used? Could a uA741 have been substituted?

How were the opamp power supply voltages selected? Why didn't I just use an opamp that could operate from the unregulated +/- voltages?

Which is fed into the comparator using the B op amp and comparing the output of the A op amp against - V ril and using that to control the -Vo rail voltage with the 2N3906 being fed from the -Vo from the filtered rectifier.

There are no comparators. One is an inverting opamp and the other is a voltage follower.

A comparator is usually used to compare voltages and give a binary output.

What I can't do at this point is the math for the comparator and the XSTR which is why I still need to do some further studying.

The transistor is in a voltage follower. How does that work? What limits the maximum negative voltage possible?

 

That certainly gives me something to think over...

 

That certainly gives me something to think over...

Following someone's design is quite different than designing something from scratch.

For a 2:1 voltage divider, you could use a pair of any same value resistors (1M, 100k, 30k, 1k, 1 ohm). Why would you chose one value over any other? Sometimes it's just what you happen to have on hand. But there are usually other considerations.

 

With my process control background, I see it as a control problem. Controlling -V based on the +V input signal. What I don't understand, and need to do some study on, is the final control element, the 2N3906. Without knowing how to control it (and PNP is still problematic for me) the rest of the control elements (primarily the 2nd op amp section) are meaningless. So give me some time to do a bit of research on PNP and digest this.

 

What I don't understand, and need to do some study on, is the final control element, the 2N3906.

You don't need to know how the last block works before you can understand how the ones before it work.

Would you be able to understand how the voltage follower works if it was an NPN transistor instead?

If you ask questions, I can try to provide details.

 

Set up a negative voltage and 2N3906 PNP. Put -8V on the emitter and measured the collector. With 0V on the base there was a bit of leakage but nothing significant until the ~-0.7V on the base then ramping up. Ramping up the -V on the base and getting -V on the collector and it is linear after overcoming base-emitter junction voltage. I expected this to be current driven but I built it without any resistors and the bench PSUs are indicating Voltage but 0 current? At -8V on the emitter and -4.6V on the base I am getting -4.14V on the collector without any current? So it is dropping -0.49V across the junction. I can put a resistor from the collector to GND and the voltage with drop a few -mV but still 0 Current? This must be something with the PSUs as I am using floating voltage and not earth GND referenced? And the meter GND is attached to the floating reference ground.

So this is why I thought the B op amp was a comparator. The B op amp is getting an inverter
-voltage from the A op amp and the -voltage from the -rail. So being a follower, instead of a GND, it is using the -rail voltage as a reference GND which changes? That confuses me... Or at least is new to me...

Enough for tonight Dennis. Back on it in the morn.

 

This must be something with the PSUs as I am using floating voltage and not earth GND referenced? And the meter GND is attached to the floating reference ground.

BJTs are current controlled devices and you need to limit the currents to appropriate values or the transistor could be destroyed. If you reverse biase a base-emitter junction by more than 5V, the transistor is toast as an amplifier.

I suspect the voltages you measured on the collector were phantom caused by the resistance of your meter.

There are 4 modes of BJT operation. Active, saturation, cutoff, and inverted. In the active mode, which the PNP in the tracking supply will be operating in for most voltages, the emitter-base junction will be forward biased and the collector-base will be will be reverse biased. If you try to force the collector-emitter voltage to close to 0V, the device will be operating in saturation mode and the collector-base junction will be forward biased.

Cutoff is when the transistor is off. Inverted is when the collector and emitter terminals are swapped. There's not much use in this mode because beta will be very low.

With the collector open, all you had was a diode.

EDIT:

So being a follower, instead of a GND, it is using the -rail voltage as a reference GND which changes? That confuses me...

In a voltage follower, the opamp tries to make the output voltage whatever is needed to make the inverting terminal the same as the non-inverting terminal. In this case, the output needs to be about -18.7V for that to happen. The opamp will sink as much current as necessary to make the emitter voltage -18V. Since the collector voltage is lower than the base, the collector-base junction is reverse biased, the emitter-base junction is forward biased, and the transistor is operating in the active region.


I made a correction in post #176. When I told you the input couldn't be ~10V, I was wrong.