a "high voltage" op-amp

[bitrex]

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

<https://www.dropbox.com/h?preview=linear_tracking.PNG>

Feed it with +50/-50 from a two-output flyback and make two the
resistors a dual isolated digipot to offer some adjustment.

This is a single with rail-rail output and rated for 36 volt supply and
+125 degree C temperature range:

<https://www.mouser.com/ProductDetail/STMicroelectronics/TSB611IYLT?qs=sGAEpiMZZMtCHixnSjNA6AbHcyCnhuoxncw8GzVzlt0%3D>

cheaper than a CA3140. it would be OK if the amp wasn't low-power as I
learned the hard way that unless there's sufficient loading on the rails
to pull the regulators into regulation (or one of the DIP-style amps
isn't properly seated in its socket) the op-amp get whapped with the
full "high voltage", bye-bye CA3140.

 

[bitrex]

On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Feed it with +50/-50 from a two-output flyback and make two the
resistors a dual isolated digipot to offer some adjustment.

This is a single with rail-rail output and rated for 36 volt supply and
+125 degree C temperature range:

https://www.mouser.com/ProductDetail/STMicroelectronics/TSB611IYLT?qs=sGAEpiMZZMtCHixnSjNA6AbHcyCnhuoxncw8GzVzlt0%3D


cheaper than a CA3140. it would be OK if the amp wasn't low-power as I
learned the hard way that unless there's sufficient loading on the rails
to pull the regulators into regulation (or one of the DIP-style amps
isn't properly seated in its socket)  the op-amp get whapped with the
full "high voltage", bye-bye CA3140.

The old LM358 seems a lot more tolerant of over-voltage oopsies than the
CA3140 but it can't get within hardly 2 volts of its own positive supply
even with a light load at 30 volts fugedaboutit.

 

[bitrex]

On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Sorry, bad link:
<https://www.dropbox.com/s/uf29ubm0v1ox7pv/linear_tracking.PNG?dl=0>

 

[George Herold]

On Thursday, April 16, 2020 at 12:35:24 PM UTC-4, bitrex wrote:

On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Sorry, bad link:
https://www.dropbox.com/s/uf29ubm0v1ox7pv/linear_tracking.PNG?dl=0

What's the purpose? For home/ lab use? I've had this HP6235a triple
PS for ages, a workhorse for me.

George H.

 

[sea moss]

This is a single with rail-rail output and rated for 36 volt supply and
+125 degree C temperature range:

https://www.mouser.com/ProductDetail/STMicroelectronics/TSB611IYLT?qs=sGAEpiMZZMtCHixnSjNA6AbHcyCnhuoxncw8GzVzlt0%3D

Check out the TLV172; similar specs.

https://www.digikey.com/product-detail/en/texas-instruments/TLV172IDBVR/296-45069-1-ND/6562071

 

[Ricky C]

On Thursday, April 16, 2020 at 12:35:24 PM UTC-4, bitrex wrote:

On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Sorry, bad link:
https://www.dropbox.com/s/uf29ubm0v1ox7pv/linear_tracking.PNG?dl=0

Aren't the low voltage rails going to be a bit higher than half the higher voltage rails. You are supplying U4 adj pin with half the "30V" output, so the -15V output will be half the -30V output with the 1.25V reference added (or subtracted, depending on how you look at it). The resistors for setting U3 output voltage puts it at the -15V output plus 14 volts.

Is that intentional? The whole thing seems a bit cyclical. Why not have one regulator set for a fixed voltage independent of the others, then the rest can leverage off that. Putting so many regulators in series like this with two in a loop seems like a bad idea.

--

Rick C.

- Get 1,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209

 

[bitrex]

On 4/16/2020 5:21 PM, bitrex wrote:

On 4/16/2020 3:54 PM, Ricky C wrote:
On Thursday, April 16, 2020 at 12:35:24 PM UTC-4, bitrex wrote:
On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Sorry, bad link:
https://www.dropbox.com/s/uf29ubm0v1ox7pv/linear_tracking.PNG?dl=0

Aren't the low voltage rails going to be a bit higher than half the
higher voltage rails.  You are supplying U4 adj pin with half the
"30V" output, so the -15V output will be half the -30V output with the
1.25V reference added (or subtracted, depending on how you look at
it).  The resistors for setting U3 output voltage puts it at the -15V
output plus 14 volts.

Loaded voltages in the sim with the LT1014 that is RRIO:

+30: 29.44
+15: 14.68
-15: -14.61
-30: -28.02

The regulators are within the negative feedback loop of the op amps.


Is that intentional?  The whole thing seems a bit cyclical.  Why not
have one regulator set for a fixed voltage independent of the others,
then the rest can leverage off that.  Putting so many regulators in
series like this with two in a loop seems like a bad idea.


The rails track each other and come up and go down in unison check out
the wave forms.

If you remove all the nanofarard-ish bypass caps it will oscillate or at
least looks like it's trying to in the sim there's a bunch of ringing on
startup.

It still seems to work OK with an amp like the LM338 without a RR-output
stage but the output ends a couple volts short of where it should be.
And if I'm not wrong the PSRR of amps without RR-output stages gets
degraded when their output stages have to operate up against their
Vcc-whatever limits.

LM358, rather, one section of each is wasted though

 

[bitrex]

On 4/16/2020 3:54 PM, Ricky C wrote:

On Thursday, April 16, 2020 at 12:35:24 PM UTC-4, bitrex wrote:
On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Sorry, bad link:
https://www.dropbox.com/s/uf29ubm0v1ox7pv/linear_tracking.PNG?dl=0

Aren't the low voltage rails going to be a bit higher than half the higher voltage rails. You are supplying U4 adj pin with half the "30V" output, so the -15V output will be half the -30V output with the 1.25V reference added (or subtracted, depending on how you look at it). The resistors for setting U3 output voltage puts it at the -15V output plus 14 volts.

Loaded voltages in the sim with the LT1014 that is RRIO:

+30: 29.44
+15: 14.68
-15: -14.61
-30: -28.02

The regulators are within the negative feedback loop of the op amps.

Is that intentional? The whole thing seems a bit cyclical. Why not have one regulator set for a fixed voltage independent of the others, then the rest can leverage off that. Putting so many regulators in series like this with two in a loop seems like a bad idea.

The rails track each other and come up and go down in unison check out
the wave forms.

If you remove all the nanofarard-ish bypass caps it will oscillate or at
least looks like it's trying to in the sim there's a bunch of ringing on
startup.

It still seems to work OK with an amp like the LM338 without a RR-output
stage but the output ends a couple volts short of where it should be.
And if I'm not wrong the PSRR of amps without RR-output stages gets
degraded when their output stages have to operate up against their
Vcc-whatever limits.

 

[Ricky C]

On Thursday, April 16, 2020 at 5:21:28 PM UTC-4, bitrex wrote:

On 4/16/2020 3:54 PM, Ricky C wrote:
On Thursday, April 16, 2020 at 12:35:24 PM UTC-4, bitrex wrote:
On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Sorry, bad link:
https://www.dropbox.com/s/uf29ubm0v1ox7pv/linear_tracking.PNG?dl=0

Aren't the low voltage rails going to be a bit higher than half the higher voltage rails. You are supplying U4 adj pin with half the "30V" output, so the -15V output will be half the -30V output with the 1.25V reference added (or subtracted, depending on how you look at it). The resistors for setting U3 output voltage puts it at the -15V output plus 14 volts.

Loaded voltages in the sim with the LT1014 that is RRIO:

+30: 29.44
+15: 14.68
-15: -14.61
-30: -28.02

Why are the voltages off? They may not be exactly 15 and 30, but shouldn't the opposite sign supplies track? Can you explain these voltages? In the sim it doesn't impose tolerance variations, so all the voltages should be pretty close, no?

I'm curious. Could you measure the voltages on the four ADJ pins? Maybe I'm completely missing how this circuit works.


> The regulators are within the negative feedback loop of the op amps.

I don't follow that. U5 is just a voltage follower with nothing in the negative feedback loop but wire. U6 is an inverting voltage follower. There's a cap in the feedback loop acting as a low pass. How are any of the regulators in the inverting feedback loop?

The op amps do remove the small offset from the ADJ pin current on the low voltage regulators versus using just the resistor divider. With the added 0.25 volts for the ADJ pin current the two high regulators should output 14 volts higher than the low side regulators. The feedback point for the low neg reg is half the high neg output which is

Vh- = Vl- + 14

The feedback voltage then is

Vf- = (Vl- + 14)/2

so the final low output should be 1.25 volts higher or...

Vl- = 1.25V + (Vl- + 14)/2

Vl- = 8.25V + Vl-/2

Vl- = 16.5V

The high output should be 30.5V, which is close to the 30 volts intended.

Is that intentional? The whole thing seems a bit cyclical. Why not have one regulator set for a fixed voltage independent of the others, then the rest can leverage off that. Putting so many regulators in series like this with two in a loop seems like a bad idea.


The rails track each other and come up and go down in unison check out
the wave forms.

Not saying they don't. I'm saying that to get the initial voltage on the negative side, the master if you will, I'd have to write and solve a couple of simultaneous equations, one for the low rail dependent on the high rail and one for the high rail dependent on the low rail.

Once you get those voltages you can then find the low positive voltage and then the high positive voltage. That high positive voltage depends on the operation of all three of the other circuits.

If you remove all the nanofarard-ish bypass caps it will oscillate or at
least looks like it's trying to in the sim there's a bunch of ringing on
startup.

It still seems to work OK with an amp like the LM338 without a RR-output
stage but the output ends a couple volts short of where it should be.
And if I'm not wrong the PSRR of amps without RR-output stages gets
degraded when their output stages have to operate up against their
Vcc-whatever limits.

The op amp shouldn't need RR outputs once it is operating, but U6 will be running the inputs at ground. Why not power it from +/- voltage?

Why not have the low negative voltage regulator regulate its own voltage rather than sensing the high voltage? Everything else in the circuit will be the same but it takes the neg high voltage regulator out of the regulation path for the two positive regulators. R1 = 1.1k and R2 = 15k will get you to within 1%.

Am I completely misunderstanding this design? I just can't see a good reason for putting two regulators in a loop regulating each other.

--

Rick C.

+ Get 1,000 miles of free Supercharging
+ Tesla referral code - https://ts.la/richard11209

 

[bitrex]

On 4/16/2020 10:35 PM, Ricky C wrote:

On Thursday, April 16, 2020 at 5:21:28 PM UTC-4, bitrex wrote:
On 4/16/2020 3:54 PM, Ricky C wrote:
On Thursday, April 16, 2020 at 12:35:24 PM UTC-4, bitrex wrote:
On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Sorry, bad link:
https://www.dropbox.com/s/uf29ubm0v1ox7pv/linear_tracking.PNG?dl=0

Aren't the low voltage rails going to be a bit higher than half the higher voltage rails. You are supplying U4 adj pin with half the "30V" output, so the -15V output will be half the -30V output with the 1.25V reference added (or subtracted, depending on how you look at it). The resistors for setting U3 output voltage puts it at the -15V output plus 14 volts.

Loaded voltages in the sim with the LT1014 that is RRIO:

+30: 29.44
+15: 14.68
-15: -14.61
-30: -28.02

Why are the voltages off? They may not be exactly 15 and 30, but shouldn't the opposite sign supplies track? Can you explain these voltages? In the sim it doesn't impose tolerance variations, so all the voltages should be pretty close, no?

I'm curious. Could you measure the voltages on the four ADJ pins? Maybe I'm completely missing how this circuit works.


The regulators are within the negative feedback loop of the op amps.

I don't follow that. U5 is just a voltage follower with nothing in the negative feedback loop but wire. U6 is an inverting voltage follower. There's a cap in the feedback loop acting as a low pass. How are any of the regulators in the inverting feedback loop?

The op amps do remove the small offset from the ADJ pin current on the low voltage regulators versus using just the resistor divider. With the added 0.25 volts for the ADJ pin current the two high regulators should output 14 volts higher than the low side regulators. The feedback point for the low neg reg is half the high neg output which is

Vh- = Vl- + 14

The feedback voltage then is

Vf- = (Vl- + 14)/2

so the final low output should be 1.25 volts higher or...

Vl- = 1.25V + (Vl- + 14)/2

Vl- = 8.25V + Vl-/2

Vl- = 16.5V

The high output should be 30.5V, which is close to the 30 volts intended.


Is that intentional? The whole thing seems a bit cyclical. Why not have one regulator set for a fixed voltage independent of the others, then the rest can leverage off that. Putting so many regulators in series like this with two in a loop seems like a bad idea.


The rails track each other and come up and go down in unison check out
the wave forms.

Not saying they don't. I'm saying that to get the initial voltage on the negative side, the master if you will, I'd have to write and solve a couple of simultaneous equations, one for the low rail dependent on the high rail and one for the high rail dependent on the low rail.

Once you get those voltages you can then find the low positive voltage and then the high positive voltage. That high positive voltage depends on the operation of all three of the other circuits.


If you remove all the nanofarard-ish bypass caps it will oscillate or at
least looks like it's trying to in the sim there's a bunch of ringing on
startup.

It still seems to work OK with an amp like the LM338 without a RR-output
stage but the output ends a couple volts short of where it should be.
And if I'm not wrong the PSRR of amps without RR-output stages gets
degraded when their output stages have to operate up against their
Vcc-whatever limits.

The op amp shouldn't need RR outputs once it is operating, but U6 will be running the inputs at ground. Why not power it from +/- voltage?

Why not have the low negative voltage regulator regulate its own voltage rather than sensing the high voltage? Everything else in the circuit will be the same but it takes the neg high voltage regulator out of the regulation path for the two positive regulators. R1 = 1.1k and R2 = 15k will get you to within 1%.

Am I completely misunderstanding this design? I just can't see a good reason for putting two regulators in a loop regulating each other.

You're making it too complicated I think you don't need to solve no
equations. U1 and U3 can just be replaced by +/-15 volt voltage sources.
at DC U4 and U5 function as a unit and acts like one big honkin'
power-op amp with two equivalent inputs, one at the output of U4 and at
U5's non-inverting input. There is an input to a three-terminal
regulator feedback loop there see e.g.:

<https://e2e.ti.com/cfs-file/__key/communityserver-discussions-components-files/196/4628.Untitled2.png>

Take an ideal op amp and connect its inverting input to its output, then
take the output and connect a 15 volt voltage-source in series with it.
Take a 0.5 resistor divider lower end to ground and mid-point back to
the op amp non-inverting input. What's the output of the op-amp gonna
be. What else can it be.

I think with ideal components you can redraw this circuit with two
voltage sources, two resistors and two unity-gain buffers, one inverting
and one non-inverting. that's all it is! (ideally)

There's something weird with the LM models connected this way perhaps
I've built the physical circuit on a breadboard seems to work fine but
I'll have to measure the voltages again, the "high voltage" rails are
closer than that IRL, IIRC

 

[Ricky C]

On Friday, April 17, 2020 at 5:50:29 AM UTC-4, bitrex wrote:

On 4/16/2020 10:35 PM, Ricky C wrote:
On Thursday, April 16, 2020 at 5:21:28 PM UTC-4, bitrex wrote:
On 4/16/2020 3:54 PM, Ricky C wrote:
On Thursday, April 16, 2020 at 12:35:24 PM UTC-4, bitrex wrote:
On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Sorry, bad link:
https://www.dropbox.com/s/uf29ubm0v1ox7pv/linear_tracking.PNG?dl=0

Aren't the low voltage rails going to be a bit higher than half the higher voltage rails. You are supplying U4 adj pin with half the "30V" output, so the -15V output will be half the -30V output with the 1.25V reference added (or subtracted, depending on how you look at it). The resistors for setting U3 output voltage puts it at the -15V output plus 14 volts.

Loaded voltages in the sim with the LT1014 that is RRIO:

+30: 29.44
+15: 14.68
-15: -14.61
-30: -28.02

Why are the voltages off? They may not be exactly 15 and 30, but shouldn't the opposite sign supplies track? Can you explain these voltages? In the sim it doesn't impose tolerance variations, so all the voltages should be pretty close, no?

I'm curious. Could you measure the voltages on the four ADJ pins? Maybe I'm completely missing how this circuit works.


The regulators are within the negative feedback loop of the op amps.

I don't follow that. U5 is just a voltage follower with nothing in the negative feedback loop but wire. U6 is an inverting voltage follower. There's a cap in the feedback loop acting as a low pass. How are any of the regulators in the inverting feedback loop?

The op amps do remove the small offset from the ADJ pin current on the low voltage regulators versus using just the resistor divider. With the added 0.25 volts for the ADJ pin current the two high regulators should output 14 volts higher than the low side regulators. The feedback point for the low neg reg is half the high neg output which is

Vh- = Vl- + 14

The feedback voltage then is

Vf- = (Vl- + 14)/2

so the final low output should be 1.25 volts higher or...

Vl- = 1.25V + (Vl- + 14)/2

Vl- = 8.25V + Vl-/2

Vl- = 16.5V

The high output should be 30.5V, which is close to the 30 volts intended.


Is that intentional? The whole thing seems a bit cyclical. Why not have one regulator set for a fixed voltage independent of the others, then the rest can leverage off that. Putting so many regulators in series like this with two in a loop seems like a bad idea.


The rails track each other and come up and go down in unison check out
the wave forms.

Not saying they don't. I'm saying that to get the initial voltage on the negative side, the master if you will, I'd have to write and solve a couple of simultaneous equations, one for the low rail dependent on the high rail and one for the high rail dependent on the low rail.

Once you get those voltages you can then find the low positive voltage and then the high positive voltage. That high positive voltage depends on the operation of all three of the other circuits.


If you remove all the nanofarard-ish bypass caps it will oscillate or at
least looks like it's trying to in the sim there's a bunch of ringing on
startup.

It still seems to work OK with an amp like the LM338 without a RR-output
stage but the output ends a couple volts short of where it should be.
And if I'm not wrong the PSRR of amps without RR-output stages gets
degraded when their output stages have to operate up against their
Vcc-whatever limits.

The op amp shouldn't need RR outputs once it is operating, but U6 will be running the inputs at ground. Why not power it from +/- voltage?

Why not have the low negative voltage regulator regulate its own voltage rather than sensing the high voltage? Everything else in the circuit will be the same but it takes the neg high voltage regulator out of the regulation path for the two positive regulators. R1 = 1.1k and R2 = 15k will get you to within 1%.

Am I completely misunderstanding this design? I just can't see a good reason for putting two regulators in a loop regulating each other.


You're making it too complicated I think you don't need to solve no
equations. U1 and U3 can just be replaced by +/-15 volt voltage sources.

Yes, but they are designed for 14 volt outputs, not 15. However the voltage at the feedback point between R1 and R2 is a function of both regulators.

Vout = Vref * (R7 + R8) / R7 = Vref * (1 + R8/R7)

Vout = 1.25V * (1 + 4.7k/470) = 1.25V * 11 = 13.75V

The ADJ pin bias is ~50 uA so the offset voltage is only 0.021 volts which can be ignored.

So U1 and U3 are 13.75 volt sources not 15 volts. This is straight from the data sheet and your resistor values.

at DC U4 and U5 function as a unit and acts like one big honkin'
power-op amp with two equivalent inputs, one at the output of U4 and at
U5's non-inverting input.

U5 is just a voltage follower. 1 volt in is 1 volt out, so it is actually transparent in this design other than sinking the ADJ pin current. The regulator is the same regulator it is without the op amp except the ADJ pin bias current is now absorbed by the op amp without affecting the voltage at the ADJ pin. Don't know why you need to try to change your thinking of the regulator to an op amp.

If the feedback were based on the output of U4 it would be completely obvious how it is working. By using the output of U3 as the feedback you now have to factor in the 13.75 volts along with the 1.25 volts of the internal reference and the voltage divider.

There is an input to a three-terminal
regulator feedback loop there see e.g.:

https://e2e.ti.com/cfs-file/__key/communityserver-discussions-components-files/196/4628.Untitled2.png

Take an ideal op amp and connect its inverting input to its output, then
take the output and connect a 15 volt voltage-source in series with it.
Take a 0.5 resistor divider lower end to ground and mid-point back to
the op amp non-inverting input. What's the output of the op-amp gonna
be. What else can it be.

You don't mention the 1.25 volt internal reference that must be added to the ADJ pin voltage. That's how the regulator operates which is not the same as an op amp... unless you treat it as 1.25 volts of input offset maybe.

I think with ideal components you can redraw this circuit with two
voltage sources, two resistors and two unity-gain buffers, one inverting
and one non-inverting. that's all it is! (ideally)

Sure, but you have to do it correctly factoring in the reference voltage internal to the regulators. That's how you get 14 volts out of U1 and U3. Take away that reference voltage and they will output zero volts.

There's something weird with the LM models connected this way perhaps
I've built the physical circuit on a breadboard seems to work fine but
I'll have to measure the voltages again, the "high voltage" rails are
closer than that IRL, IIRC

"Real Life" has many variation from tolerances. The simulations should not..

I don't see where you addressed my calculations. I'm pretty sure they are correct. It's just a consideration of how each part works without trying to apply undue simplifications which may or may not be valid. Why don't you try writing the equations yourself and see what you get?

I'm not trying to be a PITA. I just think this circuit isn't doing what you think it should. I also think making the feedback loop for U4 include the feedback loop for U3 is looking for trouble with stability. And I can't see any advantage to doing it this way.

--

Rick C.

-- Get 1,000 miles of free Supercharging
-- Tesla referral code - https://ts.la/richard11209

 

[bitrex]

On 4/17/2020 12:44 PM, Ricky C wrote:

On Friday, April 17, 2020 at 5:50:29 AM UTC-4, bitrex wrote:
On 4/16/2020 10:35 PM, Ricky C wrote:
On Thursday, April 16, 2020 at 5:21:28 PM UTC-4, bitrex wrote:
On 4/16/2020 3:54 PM, Ricky C wrote:
On Thursday, April 16, 2020 at 12:35:24 PM UTC-4, bitrex wrote:
On 4/16/2020 12:18 PM, bitrex wrote:

I'm thinking about making a lil board for my "doomsday" bomb-proof
linear tracking regulator as it's often handy to have four rails with
two that track halfway below the higher ones:

https://www.dropbox.com/h?preview=linear_tracking.PNG

Sorry, bad link:
https://www.dropbox.com/s/uf29ubm0v1ox7pv/linear_tracking.PNG?dl=0

Aren't the low voltage rails going to be a bit higher than half the higher voltage rails. You are supplying U4 adj pin with half the "30V" output, so the -15V output will be half the -30V output with the 1.25V reference added (or subtracted, depending on how you look at it). The resistors for setting U3 output voltage puts it at the -15V output plus 14 volts.

Loaded voltages in the sim with the LT1014 that is RRIO:

+30: 29.44
+15: 14.68
-15: -14.61
-30: -28.02

Why are the voltages off? They may not be exactly 15 and 30, but shouldn't the opposite sign supplies track? Can you explain these voltages? In the sim it doesn't impose tolerance variations, so all the voltages should be pretty close, no?

I'm curious. Could you measure the voltages on the four ADJ pins? Maybe I'm completely missing how this circuit works.


The regulators are within the negative feedback loop of the op amps.

I don't follow that. U5 is just a voltage follower with nothing in the negative feedback loop but wire. U6 is an inverting voltage follower. There's a cap in the feedback loop acting as a low pass. How are any of the regulators in the inverting feedback loop?

The op amps do remove the small offset from the ADJ pin current on the low voltage regulators versus using just the resistor divider. With the added 0.25 volts for the ADJ pin current the two high regulators should output 14 volts higher than the low side regulators. The feedback point for the low neg reg is half the high neg output which is

Vh- = Vl- + 14

The feedback voltage then is

Vf- = (Vl- + 14)/2

so the final low output should be 1.25 volts higher or...

Vl- = 1.25V + (Vl- + 14)/2

Vl- = 8.25V + Vl-/2

Vl- = 16.5V

The high output should be 30.5V, which is close to the 30 volts intended.


Is that intentional? The whole thing seems a bit cyclical. Why not have one regulator set for a fixed voltage independent of the others, then the rest can leverage off that. Putting so many regulators in series like this with two in a loop seems like a bad idea.


The rails track each other and come up and go down in unison check out
the wave forms.

Not saying they don't. I'm saying that to get the initial voltage on the negative side, the master if you will, I'd have to write and solve a couple of simultaneous equations, one for the low rail dependent on the high rail and one for the high rail dependent on the low rail.

Once you get those voltages you can then find the low positive voltage and then the high positive voltage. That high positive voltage depends on the operation of all three of the other circuits.


If you remove all the nanofarard-ish bypass caps it will oscillate or at
least looks like it's trying to in the sim there's a bunch of ringing on
startup.

It still seems to work OK with an amp like the LM338 without a RR-output
stage but the output ends a couple volts short of where it should be.
And if I'm not wrong the PSRR of amps without RR-output stages gets
degraded when their output stages have to operate up against their
Vcc-whatever limits.

The op amp shouldn't need RR outputs once it is operating, but U6 will be running the inputs at ground. Why not power it from +/- voltage?

Why not have the low negative voltage regulator regulate its own voltage rather than sensing the high voltage? Everything else in the circuit will be the same but it takes the neg high voltage regulator out of the regulation path for the two positive regulators. R1 = 1.1k and R2 = 15k will get you to within 1%.

Am I completely misunderstanding this design? I just can't see a good reason for putting two regulators in a loop regulating each other.


You're making it too complicated I think you don't need to solve no
equations. U1 and U3 can just be replaced by +/-15 volt voltage sources.

Yes, but they are designed for 14 volt outputs, not 15. However the voltage at the feedback point between R1 and R2 is a function of both regulators.

Vout = Vref * (R7 + R8) / R7 = Vref * (1 + R8/R7)

Vout = 1.25V * (1 + 4.7k/470) = 1.25V * 11 = 13.75V

The ADJ pin bias is ~50 uA so the offset voltage is only 0.021 volts which can be ignored.

So U1 and U3 are 13.75 volt sources not 15 volts. This is straight from the data sheet and your resistor values.


at DC U4 and U5 function as a unit and acts like one big honkin'
power-op amp with two equivalent inputs, one at the output of U4 and at
U5's non-inverting input.

U5 is just a voltage follower. 1 volt in is 1 volt out, so it is actually transparent in this design other than sinking the ADJ pin current. The regulator is the same regulator it is without the op amp except the ADJ pin bias current is now absorbed by the op amp without affecting the voltage at the ADJ pin. Don't know why you need to try to change your thinking of the regulator to an op amp.

If the feedback were based on the output of U4 it would be completely obvious how it is working. By using the output of U3 as the feedback you now have to factor in the 13.75 volts along with the 1.25 volts of the internal reference and the voltage divider.


There is an input to a three-terminal
regulator feedback loop there see e.g.:

https://e2e.ti.com/cfs-file/__key/communityserver-discussions-components-files/196/4628.Untitled2.png

Take an ideal op amp and connect its inverting input to its output, then
take the output and connect a 15 volt voltage-source in series with it.
Take a 0.5 resistor divider lower end to ground and mid-point back to
the op amp non-inverting input. What's the output of the op-amp gonna
be. What else can it be.

You don't mention the 1.25 volt internal reference that must be added to the ADJ pin voltage. That's how the regulator operates which is not the same as an op amp... unless you treat it as 1.25 volts of input offset maybe.


I think with ideal components you can redraw this circuit with two
voltage sources, two resistors and two unity-gain buffers, one inverting
and one non-inverting. that's all it is! (ideally)

Sure, but you have to do it correctly factoring in the reference voltage internal to the regulators. That's how you get 14 volts out of U1 and U3. Take away that reference voltage and they will output zero volts.


There's something weird with the LM models connected this way perhaps
I've built the physical circuit on a breadboard seems to work fine but
I'll have to measure the voltages again, the "high voltage" rails are
closer than that IRL, IIRC

"Real Life" has many variation from tolerances. The simulations should not.

I don't see where you addressed my calculations. I'm pretty sure they are correct. It's just a consideration of how each part works without trying to apply undue simplifications which may or may not be valid. Why don't you try writing the equations yourself and see what you get?

I'm not trying to be a PITA. I just think this circuit isn't doing what you think it should. I also think making the feedback loop for U4 include the feedback loop for U3 is looking for trouble with stability. And I can't see any advantage to doing it this way.

It is good to investigate it. Hey at least it's on-topic!

But I don't understand where your calculations come from. Two of the
regulators are configured as voltage sources; if the op-amp inside them
were ideal and they drew no input current then they would be essentially
ideal.

If I make R9 and R8 non-symmetric and change R8 to 5.2k to try to get
the voltage-source regulators more symmetric in isolation because the
LM317 and '337 aren't identical mirror-images of each other. Then I get:

+30: 30.62
+15: 15.86
-15: -15.79
-30: -30.54

The bottom -15 regulator has its reference inside a the feedback loop I
don't think it matters what it is exactly so long as it's approximately
similar between the '317 and '337. The op amp U5 tries to make the
voltage at its non-inverting terminal and the voltage between the -30
and -15 rails equal. Set up like this the voltage between those two
rails is -14.78. The voltage at the non-inverting input off the dividers
is -14.58. The voltage of the negative voltage-source regulator in
isolation with R8 tied to ground is -14.76.

If I then multiply the voltage at U5 non-inverting terminal by -1 and
add 1.25 to it then the +15 rail must be approximately equal to the -15.
if the two internal references are approximately equal. Then if the
upper voltage-source regulator is approximately the same as the bottom
one then its output must be approximately the same, too.