Author Topic: Transformer tap switching brings other issues: op-amp supply voltage switching!  (Read 1597 times)

0 Members and 1 Guest are viewing this topic.

Offline nemail2Topic starter

  • Regular Contributor
  • *
  • Posts: 206
  • Country: at
Hi

in this thread: https://www.eevblog.com/forum/beginners/transformer-tap-switching-between-serial-and-parallel/
I'm asking about help regarding the switching of transformer taps to reduce power dissipation on the series pass transistor of my lab power supply design.

With this being dealt, new issues arise: I'm feeding the op-amps from the transformer which is now being switched between 9V and 18V. Until now, I had a 7818 VREG which made a stable 18V for the opamps.
Now, when the transformer is being switched to 9V, that VREG will go wild, I guess, but while being at 9V on the transformer side, I won't need 18V at the opamps either.
So I thought I could maybe high-side switch the voltage supply rail of the opamp and put in another VREG (let's say 7809) and switch between the 7818 and the 7809, based on whether the transformer currently is outputting 9V or 18V AC.

Would this work or is it a bad idea?

Obviously I could use a separate transformer for the opamp supply voltage but I'm already planning to use a separate lower voltage transformer for the micro controller and stuff and I'd prefer not to have three transformers in the design, if possible.
Also, I don't want any switching regulator action in this design, that's what doesn't make things easier.

Thanks!
Boron rhymes with moron
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16553
  • Country: us
  • DavidH
The usual solutions are to either have the control circuits on a separate floating supply, usually because they are referenced to the high side of the output anyway, or to switch between transformer taps and always run the control circuits at the higher voltage.

I have also done it where the control circuits always run at the lower voltage, and a level shifter is used so that they could control a higher voltage pass element.

If I did switch the regulated voltage for the control circuits, then I would build that function into the regulator instead of using two separate regulators.
 
The following users thanked this post: nemail2

Offline nemail2Topic starter

  • Regular Contributor
  • *
  • Posts: 206
  • Country: at
The usual solutions are to either have the control circuits on a separate floating supply, usually because they are referenced to the high side of the output anyway, or to switch between transformer taps and always run the control circuits at the higher voltage.

Unfortunately my transformer's datasheet says that I can't use the windings independently, I have to always use them either in parallel or in series, so the control circuits will have to either have their own transformer or change their voltage with the tap switching. I'm not sure what a separate floating supply would be, exactly that (a separate transformer?)?

Would it work to have the control circuits on unregulated voltage directly from the bridge rectifier or would that lead to issues like unstable operation of the opamps and such? I guess I want my opamp supply voltage to be rather solid and stable, don't I?

I have also done it where the control circuits always run at the lower voltage, and a level shifter is used so that they could control a higher voltage pass element.
I read about that idea earlier somewhere else - but wouldn't the voltage shifter ultimately still need the higher supply voltage where it is shifting to and wouldn't that voltage also need to be stabilized (same issue, just moved somewhere else)?

If I did switch the regulated voltage for the control circuits, then I would build that function into the regulator instead of using two separate regulators.
Do you know of any off the shelf linear regulator (no switching action desired) which would be able to do that

Also wouldn't my idea switching the voltage rails for the control circuit work at all? Like with some capacitance which would buffer and smooth out the voltage change...
Boron rhymes with moron
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16553
  • Country: us
  • DavidH
The usual solutions are to either have the control circuits on a separate floating supply, usually because they are referenced to the high side of the output anyway, or to switch between transformer taps and always run the control circuits at the higher voltage.

Unfortunately my transformer's datasheet says that I can't use the windings independently, I have to always use them either in parallel or in series, so the control circuits will have to either have their own transformer or change their voltage with the tap switching. I'm not sure what a separate floating supply would be, exactly that (a separate transformer?)?

That just means that the voltage between the separate windings is limited to a relatively low value.  They can still be floated as long as the difference in voltage is small, and certainly up to the output voltage of the windings themselves.  But it would be a waste of current capacity to do that here anyway.

Quote
Would it work to have the control circuits on unregulated voltage directly from the bridge rectifier or would that lead to issues like unstable operation of the opamps and such? I guess I want my opamp supply voltage to be rather solid and stable, don't I?

Power supply decoupling is what matters for stability, and that does not require regulation.  Operational amplifiers have good power supply rejection so operation from an unregulated supply is only a problem in precision circuits where even this is not enough.

Quote
I have also done it where the control circuits always run at the lower voltage, and a level shifter is used so that they could control a higher voltage pass element.

I read about that idea earlier somewhere else - but wouldn't the voltage shifter ultimately still need the higher supply voltage where it is shifting to and wouldn't that voltage also need to be stabilized (same issue, just moved somewhere else)?

Usually that higher supply is the unregulated supply, or a separate higher voltage.  But for instance a PNP high side pass transistor requires no higher voltage to control.  Regulation is not strictly needed because the feedback loop will remove variation.

Quote
If I did switch the regulated voltage for the control circuits, then I would build that function into the regulator instead of using two separate regulators.

Do you know of any off the shelf linear regulator (no switching action desired) which would be able to do that

Most of them can.  Adjustable regulators like the LM317 would be the natural choice but the output voltage of a fixed regulator can be increased by adding a voltage in series with the common pin.

Quote
Also wouldn't my idea switching the voltage rails for the control circuit work at all? Like with some capacitance which would buffer and smooth out the voltage change...

That can work but it requires some attention. The tricky part is not changing the voltage too quickly because series inductance in the wiring combined with low impedance decoupling can cause high voltage ringing.  It is almost always better to use a regulator  to control the voltage and change the setpoint of the regulator instead.
 
The following users thanked this post: nemail2

Offline nemail2Topic starter

  • Regular Contributor
  • *
  • Posts: 206
  • Country: at
 |O I feel so stupid now, that was most obvious, to e.g. use something like a LM317 and just change the output voltage on demand. Thanks for pointing that out!

Would you do that using an NPN transistor or N-channel FET and an additional resistor which is being added/removed as needed or would you use something like a digital potentiometer (e.g. CAT5171)?

I won't be switching the supply voltage too fast, as I would control that through the microcontroller anyway and will be limiting the switching between parallel and series windings to at max. 1-4 times a second or so... Everything in my plan would be microcontroller controlled (tap switching and voltage switching for the LM317/control circuit) so I guess I would be able to time that carefully.

Feasable or BS? :-)
Boron rhymes with moron
 

Offline Wolfram

  • Frequent Contributor
  • **
  • Posts: 378
  • Country: no
To expand on my idea from the previous thread, any problem can be solved with enough diodes. It's a bit of a mind-bender, but simulation proves that it works.

Rsw represents the switch. Vaux stays at the lowest voltage level independent of the position of the switch.
 
The following users thanked this post: nemail2

Offline nemail2Topic starter

  • Regular Contributor
  • *
  • Posts: 206
  • Country: at
To expand on my idea from the previous thread, any problem can be solved with enough diodes. It's a bit of a mind-bender, but simulation proves that it works.

Rsw represents the switch. Vaux stays at the lowest voltage level independent of the position of the switch.

may I humbly ask you to attach the file? if you have already discarded it, no worries, I'll redraw it then. maybe analyzing the simulation will help me understand that.

Regarding the LM317 i have built the circuit in the attached schematic on a breadboard and it seems to work - would this be feasable as well or would I probably run into issues?
The Voltage is selectable between 18.3V/10.6V with this. Would be quite OK for pretty much any opamp if I want them to go to 8 or 9V or 16V respectively (which my current design's Vout(max) is). Especially with the OPA2197, as that is a R2R opamp.

edit: whoops, forgot to attach
Boron rhymes with moron
 

Offline Wolfram

  • Frequent Contributor
  • **
  • Posts: 378
  • Country: no
I've attached the LTSpice file. The 1G resistors are added to help with convergence issues. If you use a fancy MOSFET rectifier, the cost of adding the second rectifier would in most cases be larger than the added cost of the relays.
 
The following users thanked this post: nemail2

Offline nemail2Topic starter

  • Regular Contributor
  • *
  • Posts: 206
  • Country: at
any objections regarding the LM317 circuit from my last post? On the breadboard it is working but I'm not yet confident enough to put it into "the real thing"...
Boron rhymes with moron
 

Online xavier60

  • Super Contributor
  • ***
  • Posts: 2783
  • Country: au
The circuit I chose for my PSU can be described as Level Shifting. The 8V control rail is regulated from the full unregulated supply by an LM317.
Because it's lowside CS, another floating control rail would be needed to totally prevent unwanted current through the CS resistor.
Some discussion here,
https://www.eevblog.com/forum/projects/how-this-linear-psu-circuit-works-and-why-it-doesnt/msg3160486/#msg3160486
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 
The following users thanked this post: nemail2

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16553
  • Country: us
  • DavidH
Would you do that using an NPN transistor or N-channel FET and an additional resistor which is being added/removed as needed or would you use something like a digital potentiometer (e.g. CAT5171)?

Either an NPN transistor or n-channel MOSFET will work fine and is how I usually did it in the past.  I have also seen it done with the open collector or open drain output of a logic gate.
 
The following users thanked this post: nemail2

Offline nemail2Topic starter

  • Regular Contributor
  • *
  • Posts: 206
  • Country: at
Would you do that using an NPN transistor or N-channel FET and an additional resistor which is being added/removed as needed or would you use something like a digital potentiometer (e.g. CAT5171)?

Either an NPN transistor or n-channel MOSFET will work fine and is how I usually did it in the past.  I have also seen it done with the open collector or open drain output of a logic gate.
cool, so this will work not only on the breadboard?
Boron rhymes with moron
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16553
  • Country: us
  • DavidH
Would you do that using an NPN transistor or N-channel FET and an additional resistor which is being added/removed as needed or would you use something like a digital potentiometer (e.g. CAT5171)?

Either an NPN transistor or n-channel MOSFET will work fine and is how I usually did it in the past.  I have also seen it done with the open collector or open drain output of a logic gate.

cool, so this will work not only on the breadboard?

I have used that configuration before without problems.  The usual warning about allowing the input voltage to be lower than the output voltage if you switch the input voltage applies, but you have that covered with the diode.

There are other ways to do it with an adjustable regulator but that is about the simplest.
 
The following users thanked this post: nemail2

Offline wizard69

  • Super Contributor
  • ***
  • Posts: 1175
  • Country: us
I’m not sure if your transformer choices are written in stone but have you considered a multi tap transformer for logic and analog?   
 
The following users thanked this post: nemail2

Offline nemail2Topic starter

  • Regular Contributor
  • *
  • Posts: 206
  • Country: at
I’m not sure if your transformer choices are written in stone but have you considered a multi tap transformer for logic and analog?
they are not written in stone but unfortunately neither at mouser nor at rs-components I was able to find suitable (preferably toroidal) transformers. The Triad Magnetics ones which I'm using currently, all do say that the windings are designed to be used either in parallel or in series wiring, not independently. I have way too little clue to guesstimate whether *some* independent load on the windings would hurt or not.
Currently I'm thinking of using the large toroidal transformer + two small PCB soldered ones for the control stuff + fan, relays and the 2.7" OLED display. Those small transformers are quite cheap (EUR 3-10,-) and it would save me from additional complexity in hardware and in software, when having to change the voltage using a LM317 like mentioned before.

edit: two small ones instead of one, because I need 18V (analog) and 12V (fan, relay, oled display) as well as 3.3V (Teensy Microcontroller) and I 1) don't want to mix digital switchy and coily, fany, noisy stuff with the voltage supply of the control circuit and 2) the fan, relays, oled display and the Teensy demand quite some juice and I wanted to keep Vin as near as possible to Vout to reduce heat dissipation on the linear voltage regulators. According to my calculations, they'll burn around 1W each nonetheless, that's something one can easily cool and I didn't want to go way beyond that figure.
« Last Edit: April 25, 2021, 11:19:03 pm by nemail2 »
Boron rhymes with moron
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf