On Sat, 14 Jan 2012 01:56:59 -0800, Bill Sloman wrote:> On Jan 14, 5:36 am, Tim Wescott <t...@seemywebsite.com> wrote: >> On Fri, 13 Jan 2012 17:44:04 -0800, Bill Sloman wrote: >> > The Baxandall parallel class-D oscillator "squegs" in real life if >> > you build it with a high value inductance in the current feed and >> > bipolar switches. An LTSpice circuit using the Gummel-Poon model for >> > the bipolar transistors doesn't show this - it settles down to a >> > stable sinusoidal oscillation within less than a hundred cycles. >> >> > The obvious effect of using a high value inductor is that the current >> > through the inductor over-shoots dramatically at start-up and >> > actually reverses polarity as it recovers, so the bipolar transistors >> > briefly run inverted (emitter and collector diffusions swap roles). >> > In real life this clearly makes the circuit behave very oddly, but >> > the Gummel- Poon model equally clearly doesn't capture this >> > particular oddness. >> >> > VBIC might do better, but it doesn't with the model parameters that I >> > bodged out of Gummel-Poon parameters. Somebody who had a better - >> > some - understanding of the VBIC parameters might well be able to do >> > better. >> >> (LTSpice _schematic_ (Phil) snipped) >> >> Since it's balanced, why do you need a monster inductor there? Why not >> use a resistor, or put a resistor in series with the L to dampen the >> squegging? > > Using a big inductor reduces the level of odd harmonics (3rd, 5th etc) > in the sine wave generated. Using an inductor at all is the kern of the > Class-D oscillator > > http://home.planet.nl/~sloma000/0344_001_Baxandal.pdf > > Adding more resistance in series with the inductor makes the circuit > less efficient. Jim William's versions get over 90% efficiency - Linear > Technology application notes AN45, AN49, AN51, AN55, AN61, AN65. > > If it were big enough, it might damp the overshoot on start-up enough to > prevent the initial reversal of current in the inductor, which might - > in turn - stop the squegging from starting.I know that putting a resistor in makes things less efficient -- but what do you want? An astonishingly efficient oscillator that wastes some power, but still works? Or an oscillator that is is astonishingly efficient and useless? -- Tim Wescott Control system and signal processing consulting www.wescottdesign.com
Baxandall class-D oscillator squegging and the VBIC model
Started by ●January 13, 2012
Reply by ●January 14, 20122012-01-14
Reply by ●January 14, 20122012-01-14
On Sat, 14 Jan 2012 01:56:34 -0800 (PST), Bill Sloman <bill.sloman@ieee.org> wrote:>On Jan 14, 5:36�am, John Larkin ><jjlar...@highNOTlandTHIStechnologyPART.com> wrote: >> On Fri, 13 Jan 2012 17:44:04 -0800 (PST), Bill Sloman >> >> >> >> >> >> >> >> >> >> <bill.slo...@ieee.org> wrote: >> >The Baxandall parallel class-D oscillator "squegs" in real life if you >> >build it with a high value inductance in the current feed and bipolar >> >switches. An LTSpice circuit using the Gummel-Poon model for the >> >bipolar transistors doesn't show this - it settles down to a stable >> >sinusoidal oscillation within less than a hundred cycles. >> >> >The obvious effect of using a high value inductor is that the current >> >through the inductor over-shoots dramatically at start-up and actually >> >reverses polarity as it recovers, so the bipolar transistors briefly >> >run inverted (emitter and collector diffusions swap roles). In real >> >life this clearly makes the circuit �behave very oddly, but the Gummel- >> >Poon model equally clearly doesn't capture this particular oddness. >> >> >VBIC might do better, but it doesn't with the model parameters that >> >I �bodged out of Gummel-Poon parameters. Somebody who had a better - >> >some - understanding of the VBIC parameters might well be able to do >> >better. >> >> It seems to wobble a bit on startup, but doesn't actually squeg even >> with a full henry as the feed inductor. > >That's what I'm complaining about. A real circuit would almost >certainly squeg.I doubt it. The whole thing is domnated by the Q of the tuned circuit. It's not a higher-order system. A cap in the base bias network could push it over the edge, like a super-regen receiver. But why don't you build one and see? John
Reply by ●January 14, 20122012-01-14
On Sat, 14 Jan 2012 03:00:55 -0800 (PST), Bill Sloman <bill.sloman@ieee.org> wrote:>On Jan 14, 11:09�am, John Devereux <j...@devereux.me.uk> wrote: >> BillSloman<bill.slo...@ieee.org> writes: >> > On Jan 14, 5:36�am, John Larkin >> > <jjlar...@highNOTlandTHIStechnologyPART.com> wrote: >> >> On Fri, 13 Jan 2012 17:44:04 -0800 (PST),BillSloman >> >> >> <bill.slo...@ieee.org> wrote: >> >> >The Baxandall parallel class-D oscillator "squegs" in real life if you >> >> >build it with a high value inductance in the current feed and bipolar >> >> >switches. An LTSpice circuit using the Gummel-Poon model for the >> >> >bipolar transistors doesn't show this - it settles down to a stable >> >> >sinusoidal oscillation within less than a hundred cycles. >> >> >> >The obvious effect of using a high value inductor is that the current >> >> >through the inductor over-shoots dramatically at start-up and actually >> >> >reverses polarity as it recovers, so the bipolar transistors briefly >> >> >run inverted (emitter and collector diffusions swap roles). In real >> >> >life this clearly makes the circuit �behave very oddly, but the Gummel- >> >> >Poon model equally clearly doesn't capture this particular oddness. >> >> >> >VBIC might do better, but it doesn't with the model parameters that >> >> >I �bodged out of Gummel-Poon parameters. Somebody who had a better - >> >> >some - understanding of the VBIC parameters might well be able to do >> >> >better. >> >> >> It seems to wobble a bit on startup, but doesn't actually squeg even >> >> with a full henry as the feed inductor. >> >> > That's what I'm complaining about. A real circuit would almost >> > certainly squeg. >> >> Well there's one way to find out isn't there? :) > >The last one I built with an over-sized oscillator (which was the >first one I ever built, more than forty years ago) certainly squegged. >Building another just to prove the point would be something of a waste >of time.If the behavior of this oscillator is a waste of time, give it up. If it were interesting to me, I'd have one running in about 30 minutes, on a handsome, dremeled, labeled slice of copperclad. John
Reply by ●January 14, 20122012-01-14
On Jan 14, 7:16=A0pm, John Larkin <jjlar...@highNOTlandTHIStechnologyPART.com> wrote:> On Sat, 14 Jan 2012 03:00:55 -0800 (PST),BillSloman > > > > > > > > > > <bill.slo...@ieee.org> wrote: > >On Jan 14, 11:09=A0am, John Devereux <j...@devereux.me.uk> wrote: > >> BillSloman<bill.slo...@ieee.org> writes: > >> > On Jan 14, 5:36=A0am, John Larkin > >> > <jjlar...@highNOTlandTHIStechnologyPART.com> wrote: > >> >> On Fri, 13 Jan 2012 17:44:04 -0800 (PST),BillSloman > > >> >> <bill.slo...@ieee.org> wrote: > >> >> >The Baxandall parallel class-D oscillator "squegs" in real life if=you> >> >> >build it with a high value inductance in the current feed and bipo=lar> >> >> >switches. An LTSpice circuit using the Gummel-Poon model for the > >> >> >bipolar transistors doesn't show this - it settles down to a stabl=e> >> >> >sinusoidal oscillation within less than a hundred cycles. > > >> >> >The obvious effect of using a high value inductor is that the curr=ent> >> >> >through the inductor over-shoots dramatically at start-up and actu=ally> >> >> >reverses polarity as it recovers, so the bipolar transistors brief=ly> >> >> >run inverted (emitter and collector diffusions swap roles). In rea=l> >> >> >life this clearly makes the circuit =A0behave very oddly, but the =Gummel-> >> >> >Poon model equally clearly doesn't capture this particular oddness=.> > >> >> >VBIC might do better, but it doesn't with the model parameters tha=t> >> >> >I =A0bodged out of Gummel-Poon parameters. Somebody who had a bett=er -> >> >> >some - understanding of the VBIC parameters might well be able to =do> >> >> >better. > > >> >> It seems to wobble a bit on startup, but doesn't actually squeg eve=n> >> >> with a full henry as the feed inductor. > > >> > That's what I'm complaining about. A real circuit would almost > >> > certainly squeg. > > >> Well there's one way to find out isn't there? :) > > >The last one I built with an over-sized oscillator (which was the > >first one I ever built, more than forty years ago) certainly squegged. > >Building another just to prove the point would be something of a waste > >of time. > > If the behavior of this oscillator is a waste of time, give it up. > > If it were interesting to me, I'd have one running in about 30 > minutes, on a handsome, dremeled, labeled slice of copperclad.A bit extravagant for a 16kHz oscillator. If I get enthusiastic enough to build one I'll put it on a perforated prototyping board. I rather like the one Farnell sells that comes with a "collander ground plane" on the component side of the board. The copper rings around the holes on the track side can be chopped up to support surface mount parts if you want to mix and match. It will take a bit longer than half an hour - winding transformers is tedious, even when you've got access to a coil winding machine. -- Bill Sloman
Reply by ●January 14, 20122012-01-14
On Jan 14, 5:47=A0pm, John Larkin <jjlar...@highNOTlandTHIStechnologyPART.com> wrote:> On Sat, 14 Jan 2012 01:56:34 -0800 (PST),BillSloman > > <bill.slo...@ieee.org> wrote: > >On Jan 14, 5:36=A0am, John Larkin > ><jjlar...@highNOTlandTHIStechnologyPART.com> wrote: > >> On Fri, 13 Jan 2012 17:44:04 -0800 (PST),BillSloman > > >> <bill.slo...@ieee.org> wrote: > >> >The Baxandall parallel class-D oscillator "squegs" in real life if yo=u> >> >build it with a high value inductance in the current feed and bipolar > >> >switches. An LTSpice circuit using the Gummel-Poon model for the > >> >bipolar transistors doesn't show this - it settles down to a stable > >> >sinusoidal oscillation within less than a hundred cycles. > > >> >The obvious effect of using a high value inductor is that the current > >> >through the inductor over-shoots dramatically at start-up and actuall=y> >> >reverses polarity as it recovers, so the bipolar transistors briefly > >> >run inverted (emitter and collector diffusions swap roles). In real > >> >life this clearly makes the circuit =A0behave very oddly, but the Gum=mel-> >> >Poon model equally clearly doesn't capture this particular oddness. > > >> >VBIC might do better, but it doesn't with the model parameters that > >> >I =A0bodged out of Gummel-Poon parameters. Somebody who had a better =-> >> >some - understanding of the VBIC parameters might well be able to do > >> >better. > > >> It seems to wobble a bit on startup, but doesn't actually squeg even > >> with a full henry as the feed inductor. > > >That's what I'm complaining about. A real circuit would almost > >certainly squeg. > > I doubt it. The whole thing is dominated by the Q of the tuned circuit. > It's not a higher-order system. A cap in the base bias network could > push it over the edge, like a super-regen receiver.Doubt all you like. It squegs anyway, and always has, if you make the feed inductor too big http://home.planet.nl/~sloma000/0344_001_Baxandal.pdf See Baxandall's footnote on page 752.> But why don't you build one and see?I've built several over the years. I've got no urgent need to build one now. -- Bill Sloman, Nijmegen
Reply by ●January 14, 20122012-01-14
On Jan 14, 5:36=A0pm, Tim Wescott <t...@seemywebsite.please> wrote:> On Sat, 14 Jan 2012 01:56:59 -0800,BillSlomanwrote: > > On Jan 14, 5:36=A0am, Tim Wescott <t...@seemywebsite.com> wrote: > >> On Fri, 13 Jan 2012 17:44:04 -0800,BillSlomanwrote: > >> > The Baxandall parallel class-D oscillator "squegs" in real life if > >> > you build it with a high value inductance in the current feed and > >> > bipolar switches. An LTSpice circuit using the Gummel-Poon model for > >> > the bipolar transistors doesn't show this - it settles down to a > >> > stable sinusoidal oscillation within less than a hundred cycles. > > >> > The obvious effect of using a high value inductor is that the curren=t> >> > through the inductor over-shoots dramatically at start-up and > >> > actually reverses polarity as it recovers, so the bipolar transistor=s> >> > briefly run inverted (emitter and collector diffusions swap roles). > >> > In real life this clearly makes the circuit =A0behave very oddly, bu=t> >> > the Gummel- Poon model equally clearly doesn't capture this > >> > particular oddness. > > >> > VBIC might do better, but it doesn't with the model parameters that =I> >> > bodged out of Gummel-Poon parameters. Somebody who had a better - > >> > some - understanding of the VBIC parameters might well be able to do > >> > better. > > >> (LTSpice _schematic_ (Phil) snipped) > > >> Since it's balanced, why do you need a monster inductor there? =A0Why =not> >> use a resistor, or put a resistor in series with the L to dampen the > >> squegging? > > > Using a big inductor reduces the level of odd harmonics (3rd, 5th etc) > > in the sine wave generated. Using an inductor at all is the kern of the > > Class-D oscillator > > >http://home.planet.nl/~sloma000/0344_001_Baxandal.pdf > > > Adding more resistance in series with the inductor makes the circuit > > less efficient. Jim William's versions get over 90% efficiency - Linear > > Technology application notes AN45, AN49, AN51, AN55, AN61, AN65. > > > If it were big enough, it might damp the overshoot on start-up enough t=o> > prevent the initial reversal of current in the inductor, which might - > > in turn - stop the squegging from starting. > > You might also consider putting diodes in series with the transistor > collectors, to force the transistors to only work one way.One in series with the feed inductor makes more sense than one in series with each of the switching transistors - see today's version of the circuit in my response to John Devereux. it doesn't help the efficiency. -- Bill Sloman,Nijmegen
Reply by ●January 14, 20122012-01-14
On Jan 14, 5:38=A0pm, Tim Wescott <t...@seemywebsite.please> wrote:> On Sat, 14 Jan 2012 01:56:59 -0800,BillSlomanwrote: > > On Jan 14, 5:36=A0am, Tim Wescott <t...@seemywebsite.com> wrote: > >> On Fri, 13 Jan 2012 17:44:04 -0800,BillSlomanwrote: > >> > The Baxandall parallel class-D oscillator "squegs" in real life if > >> > you build it with a high value inductance in the current feed and > >> > bipolar switches. An LTSpice circuit using the Gummel-Poon model for > >> > the bipolar transistors doesn't show this - it settles down to a > >> > stable sinusoidal oscillation within less than a hundred cycles. > > >> > The obvious effect of using a high value inductor is that the curren=t> >> > through the inductor over-shoots dramatically at start-up and > >> > actually reverses polarity as it recovers, so the bipolar transistor=s> >> > briefly run inverted (emitter and collector diffusions swap roles). > >> > In real life this clearly makes the circuit =A0behave very oddly, bu=t> >> > the Gummel- Poon model equally clearly doesn't capture this > >> > particular oddness. > > >> > VBIC might do better, but it doesn't with the model parameters that =I> >> > bodged out of Gummel-Poon parameters. Somebody who had a better - > >> > some - understanding of the VBIC parameters might well be able to do > >> > better. > > >> (LTSpice _schematic_ (Phil) snipped) > > >> Since it's balanced, why do you need a monster inductor there? =A0Why =not> >> use a resistor, or put a resistor in series with the L to dampen the > >> squegging? > > > Using a big inductor reduces the level of odd harmonics (3rd, 5th etc) > > in the sine wave generated. Using an inductor at all is the kern of the > > Class-D oscillator > > >http://home.planet.nl/~sloma000/0344_001_Baxandal.pdf > > > Adding more resistance in series with the inductor makes the circuit > > less efficient. Jim William's versions get over 90% efficiency - Linear > > Technology application notes AN45, AN49, AN51, AN55, AN61, AN65. > > > If it were big enough, it might damp the overshoot on start-up enough t=o> > prevent the initial reversal of current in the inductor, which might - > > in turn - stop the squegging from starting. > > I know that putting a resistor in makes things less efficient -- but what > do you want? =A0An astonishingly efficient oscillator that wastes some > power, but still works? =A0Or an oscillator that is is astonishingly > efficient and useless?Jim Williams' application notes AN45, AN49, AN51, AN55, AN61, AN65 make it perfectly clear that the circuit can be both efficient and useful. A bigger feed inductor would just reduce the (odd) harmonic content of the output. There's no perceptible even harmonic content. It has been claimed here that using MOSFET switches rather than bipolar transistors lets you use a big feed inductor and still avoid squegging. My primary interest is trying to work out what causes the squegging, rather than in coming up with a cure for the squegging when it does happen. If you can live with a slightly less efficient circuit, I think I've got a variant that keeps the harmonics some 80dB below the fundamental, which is better than you could do with any practical inductor (short of a high temperature super-conductor coil in liquid nitrogen). -- Bill Sloman, Nijmegen
Reply by ●January 14, 20122012-01-14
On 1/14/2012 3:56 AM, Bill Sloman wrote:> On Jan 14, 5:36 am, Tim Wescott<t...@seemywebsite.com> wrote: >> On Fri, 13 Jan 2012 17:44:04 -0800, Bill Sloman wrote: >>> The Baxandall parallel class-D oscillator "squegs" in real life if you >>> build it with a high value inductance in the current feed and bipolar >>> switches. An LTSpice circuit using the Gummel-Poon model for the bipolar >>> transistors doesn't show this - it settles down to a stable sinusoidal >>> oscillation within less than a hundred cycles. >> >>> The obvious effect of using a high value inductor is that the current >>> through the inductor over-shoots dramatically at start-up and actually >>> reverses polarity as it recovers, so the bipolar transistors briefly run >>> inverted (emitter and collector diffusions swap roles). In real life >>> this clearly makes the circuit behave very oddly, but the Gummel- Poon >>> model equally clearly doesn't capture this particular oddness. >> >>> VBIC might do better, but it doesn't with the model parameters that I >>> bodged out of Gummel-Poon parameters. Somebody who had a better - some - >>> understanding of the VBIC parameters might well be able to do better. >> >> (LTSpice _schematic_ (Phil) snipped) >> >> Since it's balanced, why do you need a monster inductor there? Why not >> use a resistor, or put a resistor in series with the L to dampen the >> squegging? > > Using a big inductor reduces the level of odd harmonics (3rd, 5th etc) > in the sine wave generated. Using an inductor at all is the kern of > the Class-D oscillator > > http://home.planet.nl/~sloma000/0344_001_Baxandal.pdf > > Adding more resistance in series with the inductor makes the circuit > less efficient.Can't the resistance be added in parallel with the inductor and achieve a lower Q?
Reply by ●January 14, 20122012-01-14
On Sat, 14 Jan 2012 10:59:47 -0800 (PST), Bill Sloman <bill.sloman@ieee.org> wrote:>On Jan 14, 7:16�pm, John Larkin ><jjlar...@highNOTlandTHIStechnologyPART.com> wrote: >> On Sat, 14 Jan 2012 03:00:55 -0800 (PST),BillSloman >> >> >> >> >> >> >> >> >> >> <bill.slo...@ieee.org> wrote: >> >On Jan 14, 11:09�am, John Devereux <j...@devereux.me.uk> wrote: >> >> BillSloman<bill.slo...@ieee.org> writes: >> >> > On Jan 14, 5:36�am, John Larkin >> >> > <jjlar...@highNOTlandTHIStechnologyPART.com> wrote: >> >> >> On Fri, 13 Jan 2012 17:44:04 -0800 (PST),BillSloman >> >> >> >> <bill.slo...@ieee.org> wrote: >> >> >> >The Baxandall parallel class-D oscillator "squegs" in real life if you >> >> >> >build it with a high value inductance in the current feed and bipolar >> >> >> >switches. An LTSpice circuit using the Gummel-Poon model for the >> >> >> >bipolar transistors doesn't show this - it settles down to a stable >> >> >> >sinusoidal oscillation within less than a hundred cycles. >> >> >> >> >The obvious effect of using a high value inductor is that the current >> >> >> >through the inductor over-shoots dramatically at start-up and actually >> >> >> >reverses polarity as it recovers, so the bipolar transistors briefly >> >> >> >run inverted (emitter and collector diffusions swap roles). In real >> >> >> >life this clearly makes the circuit �behave very oddly, but the Gummel- >> >> >> >Poon model equally clearly doesn't capture this particular oddness. >> >> >> >> >VBIC might do better, but it doesn't with the model parameters that >> >> >> >I �bodged out of Gummel-Poon parameters. Somebody who had a better - >> >> >> >some - understanding of the VBIC parameters might well be able to do >> >> >> >better. >> >> >> >> It seems to wobble a bit on startup, but doesn't actually squeg even >> >> >> with a full henry as the feed inductor. >> >> >> > That's what I'm complaining about. A real circuit would almost >> >> > certainly squeg. >> >> >> Well there's one way to find out isn't there? :) >> >> >The last one I built with an over-sized oscillator (which was the >> >first one I ever built, more than forty years ago) certainly squegged. >> >Building another just to prove the point would be something of a waste >> >of time. >> >> If the behavior of this oscillator is a waste of time, give it up. >> >> If it were interesting to me, I'd have one running in about 30 >> minutes, on a handsome, dremeled, labeled slice of copperclad. > >A bit extravagant for a 16kHz oscillator. If I get enthusiastic enough >to build one I'll put it on a perforated prototyping board. I rather >like the one Farnell sells that comes with a "collander ground plane" >on the component side of the board. The copper rings around the holes >on the track side can be chopped up to support surface mount parts if >you want to mix and match. > >It will take a bit longer than half an hour - winding transformers is >tedious, even when you've got access to a coil winding machine.You can wind a pot core in a few minutes, by hand. ftp://jjlarkin.lmi.net/Z206_pcb.JPG Or buy one of those dual-winding inductors for about a dollar. Or not. John
Reply by ●January 14, 20122012-01-14
On Jan 14, 9:39=A0pm, John S <Soph...@invalid.org> wrote:> On 1/14/2012 3:56 AM,BillSlomanwrote: > > > > > > > > > > > On Jan 14, 5:36 am, Tim Wescott<t...@seemywebsite.com> =A0wrote: > >> On Fri, 13 Jan 2012 17:44:04 -0800,BillSlomanwrote: > >>> The Baxandall parallel class-D oscillator "squegs" in real life if yo=u> >>> build it with a high value inductance in the current feed and bipolar > >>> switches. An LTSpice circuit using the Gummel-Poon model for the bipo=lar> >>> transistors doesn't show this - it settles down to a stable sinusoida=l> >>> oscillation within less than a hundred cycles. > > >>> The obvious effect of using a high value inductor is that the current > >>> through the inductor over-shoots dramatically at start-up and actuall=y> >>> reverses polarity as it recovers, so the bipolar transistors briefly =run> >>> inverted (emitter and collector diffusions swap roles). In real life > >>> this clearly makes the circuit =A0behave very oddly, but the Gummel- =Poon> >>> model equally clearly doesn't capture this particular oddness. > > >>> VBIC might do better, but it doesn't with the model parameters that I > >>> bodged out of Gummel-Poon parameters. Somebody who had a better - som=e -> >>> understanding of the VBIC parameters might well be able to do better. > > >> (LTSpice _schematic_ (Phil) snipped) > > >> Since it's balanced, why do you need a monster inductor there? =A0Why =not> >> use a resistor, or put a resistor in series with the L to dampen the > >> squegging? > > > Using a big inductor reduces the level of odd harmonics (3rd, 5th etc) > > in the sine wave generated. Using an inductor at all is the kern of > > the Class-D oscillator > > >http://home.planet.nl/~sloma000/0344_001_Baxandal.pdf > > > Adding more resistance in series with the inductor makes the circuit > > less efficient. > > Can't the resistance be added in parallel with the inductor and achieve > a lower Q?It's no less inefficient. Any resistor carrying current is dissipating heat. -- Bill Sloman, Nijmegen
