Author Topic: Weird switching topologies you don't see anymore  (Read 4339 times)

0 Members and 1 Guest are viewing this topic.

Offline T3sl4co1lTopic starter

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Weird switching topologies you don't see anymore
« on: July 23, 2019, 09:27:51 pm »
Here's a buck converter with what looks like an unusual flyback connection:
https://patents.google.com/patent/US3959716A/en?oq=7782002
Probably they opted for this, to get a common [AC] ground for the switch -- easier to drive, no bootstrap, and less level shifting needed.  Switching supplies were new back then, transformers were cheaper than transistors (well, maybe not quite that cheap), and so they weren't afraid of using coupled inductors.  We tend to avoid coupled inductors nowadays, just because of commercial availability, really.

Analysis:

We can guess the phasing of the coupled inductor (who needs to read the text?).  If the diode conducts while the switch is on, that would be bad (large peak currents).  Also, if it were phased so that this happened, it would only conduct if Vin > 2Vo (assuming a 1:1 turns ratio), which is weird and useless.

We conclude they are phased the same direction (dot on the same side).  The diode is then doing catch (flyback) duty.  The inductor voltages during turn-on, off and idle can be written out assuming one or the other (switch or diode) is shorted, or both open.



If we swap the switch to the Vin side, and the diode to the GND side, we find it's essentially a centertapped inductor, except it's not a center tap at all but a double end tap -- in other words, a short circuit.  (Actually a common mode choke, but not a very useful one as one end is shorted.)  We can remove one winding, shorting nodes A and B together, and we find we have an ordinary buck topology.

I hope this analysis has been educational.  Any other oddball-but-not examples?

Ooh, flyback transformers are a good one, come to think of it.  Anyone want to do the "B+ Boost" from ye olde television? ;D

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline rsjsouza

  • Super Contributor
  • ***
  • Posts: 6106
  • Country: us
  • Eternally curious
    • Vbe - vídeo blog eletrônico
Re: Weird switching topologies you don't see anymore
« Reply #1 on: July 23, 2019, 09:40:31 pm »
Quite interesting; thanks for highlighting this.

Check the 1983 design review of a 250W power supply by Unitrode. It has a coupled inductor feedback network.

TI still makes this document available for download at: http://www.ti.com/ww/en/power-training/login.shtml (you have to login)
Vbe - vídeo blog eletrônico http://videos.vbeletronico.com

Oh, the "whys" of the datasheets... The information is there not to be an axiomatic truth, but instead each speck of data must be slowly inhaled while carefully performing a deep search inside oneself to find the true metaphysical sense...
 

Offline chris_leyson

  • Super Contributor
  • ***
  • Posts: 1549
  • Country: wales
Re: Weird switching topologies you don't see anymore
« Reply #2 on: July 23, 2019, 09:51:23 pm »
Kind of reminds me of the fly-buck converter which can be a useful topology if you need an additional low power auxiliary output voltage. The only drawback is that the auxiliary output is not short circuit proof so needs additional current limiting hardware, a short on the aux output effectively shorts out the main buck inductor. EDIT: To be fair I think the fly-buck is a relatively recent topology.
« Last Edit: July 23, 2019, 09:59:09 pm by chris_leyson »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Weird switching topologies you don't see anymore
« Reply #3 on: July 24, 2019, 02:25:05 am »
Tektronix used an odd design for a line voltage buck converter in their 22xx series oscilloscopes.

The second two examples look suspiciously like the boosted current buck converter configuration which uses a center tapped inductor but without the 2:1 step-down ratio.
 

Offline planet12

  • Regular Contributor
  • *
  • Posts: 83
  • Country: nz
Re: Weird switching topologies you don't see anymore
« Reply #4 on: July 24, 2019, 07:45:14 am »
To be fair I think the fly-buck is a relatively recent topology.

Depends what we're meaning by "recent" - I've seen this technique in old MC34063 app notes as a way of generating an extra rail.
 

Online magic

  • Super Contributor
  • ***
  • Posts: 7454
  • Country: pl
Re: Weird switching topologies you don't see anymore
« Reply #5 on: July 24, 2019, 08:17:17 am »
Probably they opted for this, to get a common [AC] ground for the switch -- easier to drive, no bootstrap, and less level shifting needed.
That's a clear advantage over usual buck. On the other hand, leakage inductance is a problem. Probably needs to be snubbed, or is there some clever solution which recovers the energy?
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: Weird switching topologies you don't see anymore
« Reply #6 on: July 24, 2019, 10:43:57 am »
Probably they opted for this, to get a common [AC] ground for the switch -- easier to drive, no bootstrap, and less level shifting needed.
That's a clear advantage over usual buck. On the other hand, leakage inductance is a problem. Probably needs to be snubbed, or is there some clever solution which recovers the energy?

Yeah, you get rid of one diode but introduce leakage inductance problems. See the attached screenshot and sim file from LTSpice already showing nasty spikes and ringing in the coupled inductor waveforms with 2% leakage (K = 0.99), and if you really want an eye-opener, run the sim and check out the voltage waveform across the switch...

 

Offline T3sl4co1lTopic starter

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Weird switching topologies you don't see anymore
« Reply #7 on: July 24, 2019, 04:03:58 pm »
Nah, they're driving it with a triple darlington, it's slow enough not to care.

In the modern day of MOSFETs throwing off harmonics beyond 200MHz, we'd have to observe that the transformer can be a 1:1 transmission line, and so can be made of twisted pairs or quads, that the impedance should be near the switching impedance and the leakage inductance is proportional simply to winding length.

I've done a similar thing with "star quad" magnet wire on a #52 core, for a Ćuk converter; it had about 5nH total, barely any need for coupling capacitors.



(Star quad is where you twist four wires together, in a diamond pattern; you tie opposite corners together, so going around the sequence, they alternate positive and negative.  This reduces the characteristic impedance considerably -- it's about half the value of two twisted pairs in parallel.  So for a twisted pair being ~100 ohms and two in parallel being 50, this gives ~25 ohms.  It's a great deal.  Higher order configurations don't really help unfortunately, so if you happen to need a really low impedance, just wire more quads in parallel.)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Weird switching topologies you don't see anymore
« Reply #8 on: July 24, 2019, 04:09:04 pm »
Higher order configurations don't really help unfortunately, so if you happen to need a really low impedance, just wire more quads in parallel.)

You had me momentarily considering a quad of quads.
 
The following users thanked this post: cpt.armadillo

Online CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 5571
  • Country: us
Re: Weird switching topologies you don't see anymore
« Reply #9 on: July 24, 2019, 04:26:48 pm »
You have to remember these early switcher designs were competing with linear supplies.  A few lost watts here and there didn't lose the huge efficiency advantage.  Same thing in terms of power density.  A whole different world than today.
 

Offline soldar

  • Super Contributor
  • ***
  • Posts: 3595
  • Country: es
All my posts are made with 100% recycled electrons and bare traces of grey matter.
 

Offline bd139

  • Super Contributor
  • ***
  • Posts: 23099
  • Country: gb
Re: Weird switching topologies you don't see anymore
« Reply #11 on: July 24, 2019, 07:52:14 pm »
Does a Royer converter count? Just built one today to drag +/-15V out of a 5V supply.



There's only 4 discrete components on the primary and I'm measuring ~ 85% efficiency! I'm rather pleased with it.
« Last Edit: July 24, 2019, 07:54:00 pm by bd139 »
 

Offline T3sl4co1lTopic starter

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Weird switching topologies you don't see anymore
« Reply #12 on: July 24, 2019, 08:03:11 pm »
Not really, as it's a standard transformer-coupled push-pull topology?

It's another one on the list though.

An interesting case is synthesizing more alternatives; the V-I dual of the constant-voltage half-bridge is the constant-current push-pull.

The half bridge has two series transistors, and two series capacitors, with the load strung between the midpoints.
The CCPP has two parallel-acting transistors, and two parallel-acting (coupled) inductors (indeed, one centertapped inductor), with the load strung between the endpoints).  In other words, the Baxandall oscillator.  Mind, this is often misattributed as Royer, but it's not saturation commutated, but is resonant and current-fed (i.e., a large series inductor, or ideally a current source as such).

And the dual of the CVPP is the current-fed half bridge, which uses a centertapped inductor to serve the same purpose.  It's a really oddball one (and also not something you'll ever see, for the same reason), but interesting theoretically.  Hmm, I don't happen to have a schematic representation handy, for shame.  I'll have to re-synthesize it some time.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline chris_leyson

  • Super Contributor
  • ***
  • Posts: 1549
  • Country: wales
Re: Weird switching topologies you don't see anymore
« Reply #13 on: July 24, 2019, 08:39:43 pm »
The Jensen converter is an improvement on the Royer, the saturable transformer is moved to the base drive and a conventional transformer is used for the main transformer. See picture below, a. Royer, b. Jensen, c. current driven Royer and d. interleaved boost derived. https://www.researchgate.net/publication/3280666_A_Novel_Self-Oscillating_Boost-Derived_DC-DC_Converter_With_Load_Regulation
EDIT: "Modern DC-to-DC Switchmode Power Converter Circuits" Severns and Bloom 1985 has several topologies that you never see.

« Last Edit: July 24, 2019, 09:26:28 pm by chris_leyson »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Weird switching topologies you don't see anymore
« Reply #14 on: July 24, 2019, 09:21:34 pm »
The Jensen converter is an improvement on the Royer, the saturable transformer is moved to the base drive and a conventional transformer is used for the main transformer.

I did not know there was a special name for that configuration.  Tektronix used the two transformer Royer configuration in their 22xx series oscilloscopes.  Later it was replaced by a current driven synchronous push-pull inverter.
 

Offline T3sl4co1lTopic starter

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Weird switching topologies you don't see anymore
« Reply #15 on: July 24, 2019, 10:32:24 pm »
I used an analogous drive method in this buck converter,



Here the drive transformer couples load current into the base, forcing a ~constant hFE(sat), and also the inductive energy provides rapid turn-off (shunting of B-E charge).  Once triggered, it remains on (positive feedback), until the transformer core saturates, yanking it off.  The values shown give a pulse width of about 5us.  The control is then a simple variable frequency multivibrator to regulate output current.

A similar circuit was used in ATX power supplies for decades; there, a symmetrical (half bridge) arrangement is used instead, with two base windings are provided, and the feedback winding is in series with the common output.  Drive was provided from a shorting-mode circuit, so that base voltage is held at zero when the transistors should be off, and one or the other side is driven on when desired.  Typically a TL494 and pair of transistors (2SC1815 or such) performed this.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline lordvader88

  • Frequent Contributor
  • **
  • !
  • Posts: 935
  • Country: ca
Re: Weird switching topologies you don't see anymore
« Reply #16 on: July 24, 2019, 10:37:15 pm »
Nah, they're driving it with a triple darlington, it's slow enough not to care.

In the modern day of MOSFETs throwing off harmonics beyond 200MHz, we'd have to observe that the transformer can be a 1:1 transmission line, and so can be made of twisted pairs or quads, that the impedance should be near the switching impedance and the leakage inductance is proportional simply to winding length.


What do u mean by that part " the impedance should be near the switching impedance " ? What's the switching impedance ? Is it just what the overall mosfet impedance is at operational frequencies, and then wire pairs should have about the same valve?
 
The following users thanked this post: chris_leyson

Offline chris_leyson

  • Super Contributor
  • ***
  • Posts: 1549
  • Country: wales
Re: Weird switching topologies you don't see anymore
« Reply #17 on: July 24, 2019, 10:49:10 pm »
@David Hess. The original Royer was designed as a linear voltage to frequency converter and was not really meant to be a power converter. "A Switching Transistor D-C to A-C Converter Having an Output Frequency Proportional to the D-C Input Voltage" July 1955.
For the Royer converter, the self oscillating frequency is proportional to the input voltage, whereas in the Jensen, the time taken to reach saturation is determined by the base drive transformer and the required base drive voltage which is more or less constant compared to a potentialy varying input voltage.
So, for the Jensen converter the operating frequency remains reasonably constant. Also, you don't get the large collector current spikes that you do with the Royer. There are also proportional base drive versions of it as well.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Weird switching topologies you don't see anymore
« Reply #18 on: July 24, 2019, 11:59:00 pm »
So, for the Jensen converter the operating frequency remains reasonably constant. Also, you don't get the large collector current spikes that you do with the Royer. There are also proportional base drive versions of it as well.

I suspect the low current spikes were why Tektronix used it.  They required no post regulation although they did include pi filters on the low voltage outputs.  Proportional base drive allowed primary side regulation although they were selecting the inverter output transistors for something.

 

Offline T3sl4co1lTopic starter

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Weird switching topologies you don't see anymore
« Reply #19 on: July 25, 2019, 12:49:24 am »
What do u mean by that part " the impedance should be near the switching impedance " ? What's the switching impedance ? Is it just what the overall mosfet impedance is at operational frequencies, and then wire pairs should have about the same valve?

Yeah, an informal impedance: peak switch voltage divided by peak switch current, or something along those lines.

The loop stray inductance and switch (transistor/diode) capacitance also form a ratio sqrt(L/C) with the same units (ohms), related for the same reason.

In a transformer coupled topology (flyback, forward, SEPIC, Ćuk..), the leakage inductance is part of the loop stray inductance.

The traditional rule of thumb "minimize inductance" is actually trying to do this: set 1/sqrt(Lstray*Csw) (the loop resonant frequency) higher than the switching speed, and set the impedance low enough, that you do not experience significant peaking or ringing.  (This is a rather outdated rule, as modern devices are faster than the strays that can be achieved with leaded components.)

But if you can't avoid that, then you can at least manage it, and that's where knowing the loop impedance comes into play.  What the impedance means is, for a step change in inverter state, and accompanying step change in voltage or current, what the respective peak current or voltage will be.  So the transistor carrying say 10A, turning off suddenly, will generate 10A * Zloop peak voltage.

Zloop also tells you what value resistor to use, in an RC damper.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Online Circlotron

  • Super Contributor
  • ***
  • Posts: 3362
  • Country: au
Re: Weird switching topologies you don't see anymore
« Reply #20 on: July 25, 2019, 01:15:21 am »
One topology I independently invented myself about 15 years ago (I'm not imagining that someone else hasn't also done elsewhere) Is a boost reg +12V nominal to +16V where there is a pair of boost inductors running antiphase driven by a SG3525 (yeah, I know  ::) ) and the tops of  the inductors, instead of being fed directly from +12V were instead attached to either end of a centre tapped single winding transformer. The idea was that if the left side inductor was charging, the right side inductor would be discharging, but at the same time the right side inductor would be being fed with somewhat more than +12V because of the transformer action of the tranny feeding it. At the same time, the left side inductor would be charging from somewhat less than +12V. The whole purpose was to reduce the slope of the inductor charge and discharge currents and consequently the input and output ripple currents. Never did follow through with it. I should do that.
 

Offline T3sl4co1lTopic starter

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Weird switching topologies you don't see anymore
« Reply #21 on: July 25, 2019, 07:45:29 am »
No shame in the SG3525, matter of fact it's kind of surprising come to think of it, that there aren't more applications/notes out there proposing such dual-phase controllers for phase-interleave purposes instead.  That said, major downside: <50% max duty cycle (per output) limitation.  (Well, that and the lack of tracking (e.g., for average current mode, independent current error amps) per channel, basically limiting this operation to voltage mode as well.  Which wouldn't be outside the scope of their original intent, but is decidedly passe for good reason.)

Your proposal works in CCM, but falls over badly in DCM I think?  Because, one inductor can't charge into the other when it's not carrying any current.  I should probably draw it out to get a better feel for behavior.  It may be the kind of thing that works well in a narrow range of conditions; maybe not exactly the conditions you were considering, but some low-numbered ratio between Vin and Vout, I think, and preferably where CCM is maintained.

The DCM-CCM transition should almost act like a "swinging choke" in a choke-input filter does, I think?  Because in DCM, the transformer (magnetizing inductance) dominates, but in CCM the transformer acts to unify the two chokes as acting in series.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Online Circlotron

  • Super Contributor
  • ***
  • Posts: 3362
  • Country: au
Re: Weird switching topologies you don't see anymore
« Reply #22 on: July 25, 2019, 01:34:05 pm »
Okay, this is a sim of it running in continuous mode 12 Volts in, 18 volts out, 220mH inductors, 20kHz, but with no feedback so the current is ratcheting up... but you get the idea. Red is the switch voltage, green is the left side inductor current 330mA min, 540mA max, and blue is the left diode current going right down to zero with the red switch voltage then follows the inductor current. Interesting part is where the left side inductor and diode current turns back and starts to rise again while the left side switch is still off because the right side switch is now on.

In discontinuous mode it might be a bit of a flop but I think it then acts more like a forward converter because for example if the left side switch turns on the top of the right side inductor is pulled up to double the supply voltage. Less so in continuous mode.
 
The following users thanked this post: T3sl4co1l

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: Weird switching topologies you don't see anymore
« Reply #23 on: July 25, 2019, 08:10:48 pm »
Okay, this is a sim of it running in continuous mode...

Looks like a "dual inductor boost" converter except with the DC feed via a coupled choke; if you are interested in comparing the two just delete the coupled choke and tie the two boost inductors directly together.

 

Online Circlotron

  • Super Contributor
  • ***
  • Posts: 3362
  • Country: au
Re: Weird switching topologies you don't see anymore
« Reply #24 on: July 26, 2019, 11:27:39 am »
Put together a real version of the above cct and it's a bit lame... You cannot wind the input voltage down as far as you can using two normally connected boost inductors. Maybe an interesting mod might be to wind a step-up secondary onto the transformer and feed it via a full wave rectifier straight to the output with no filter inductors of course because it would be a current source with a current related to the boost inductor current.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf