I didn't know if I should post this in the Power Supplies forum or the Chipamps forum, but:
Has anyone tried designing a voltage regulator using a power chipamp IC, such as the LM3886, LM4780, LM3875, et al?
While researching regulated power supply options for chipamp-base power amplifiers, it occurred to me that I could use the same chipamp to make a voltage regulator, enabling me to get much more current than I could from, say, an LM338 OR LT1083.
I've played around with it, using LT-Spice (excellent free simulator from http://www.linear.com ), substituting TI's OPA541E model for the chipamp (since I couldn't find any models for LM3875 or LM4780), and have gotten some very encouraging results.
Since the LM3886 and LM4780 (and LM3875) can handle voltage rails that differ by 84V (and can push 11.5A max; or 6A max for LM3875), they could even be used as regulators with a 60 VAC(RMS) (84.84 V peak) transformer, if a small series resistance is inserted between the rectifier diodes and the smoothing capacitors, before the chipamp regulator, to keep the chipamp's power supply pins' differential voltage at least 10% or so below 84V.
I was initially wanting to produce +/-30VDC rails for a chipamp-based power amp. So I came up with a chipamp-based voltage regulator circuit that regulates to 60vdc, from a 60VAC then-rectified-and-smoothed input. It looks fairly good in simulations (with realistic ESR added to caps, etc), giving 60vdc with something like 10.5 uV ripple for any resistive load drawing up to 11 Amps (I was using 4X 4700uF/80V/.053-Ohm-ESR caps, in the rectification/smoothing stage, in that case.).
Actually, I then also added a "rail splitter" stage, using a circuit that's similar to the rail splitter circuit that's shown the datasheet for National.com's LM675(http://www.national.com/ds.cgi/LM/LM675.pdf), giving me +/-30vdc rails, at up to 11A each, and a new "ground" (i.e. the -30V "common"). With that setup, and with a (third) chipamp power amp used as a load that was producing 20vpp square waves (with 2.5us edge times) into 8 Ohms, the power rails had about 15mvpp ripple. I don't know if that's "good" or not. Anybody?
The voltage regulator stage's circuit is simple: The chipamp's power pins are across the smoothing caps. I apply a "DC Reference" voltage to the chipamp's + input and apply the divided-down output voltage to the - input. The DC Reference voltage is derived from the output, too, by running it through a LP filter and across something like a TL431 TO-92 2.5v voltage-ref device.
I guess I'm mostly wondering if this "chipamp-based regulator" approach could be economical, or "reasonable", to pursue. It does appear to buy me the ability to produce clean +/-30v@up-to-11A-each power supply rails, using a single-secondary 60VAC RMS transformer. But it would take either two LM3875 (about $6 each) or one LM4780 ($9.49 each) to do it. And it looks like the transformer would need to be able to provide at least 20.34 Amps RMS (+/- 46.7A peaks), if 2X 11A max 30v outputs were desired. (Of course, it might be a little "easier" if we don't really need a full 11A continuous DC current equivalent from the two outputs simultaneously. For 2X 8.5 Amps DC out max, the 60VAC transformer would only need to be able to provide about 16.4A RMS, +/-39.4A peaks (710W?)) I'm also a little troubled by the small resistance that needs to be placed between the recitifier bridge and the caps (to keep chipamp's power supply pins 10% or more below 84V), since with 0.25 Ohm (about the minimum that will work), I seem to need a maximum dissipation of about 88 Watts, from the resistor. I can easily find such resistors. That's not the problem. It just seems like quite a waste of power. I suppose I could add some series diodes, instead of, or in addition to, a series resistance.
The same type of power supply could be done with a lower-voltage transformer and lower-voltage chipamps, such as the LM1875, for currents up to maybe 4A, or with the same 60VAC but 2X LM3875, for currents up to 6A. But it seems like those options might be tending to get less-economical, compared to more-standard solutions (e.g. 2X LM338 reg, or switchmode PS, etc), although it might be somewhat-attractive just for the sake of "commonality of components", for chipamp-based power amps.
So, does anyone have any experience with or thoughts about any of this?
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Has anyone tried designing a voltage regulator using a power chipamp IC, such as the LM3886, LM4780, LM3875, et al?
While researching regulated power supply options for chipamp-base power amplifiers, it occurred to me that I could use the same chipamp to make a voltage regulator, enabling me to get much more current than I could from, say, an LM338 OR LT1083.
I've played around with it, using LT-Spice (excellent free simulator from http://www.linear.com ), substituting TI's OPA541E model for the chipamp (since I couldn't find any models for LM3875 or LM4780), and have gotten some very encouraging results.
Since the LM3886 and LM4780 (and LM3875) can handle voltage rails that differ by 84V (and can push 11.5A max; or 6A max for LM3875), they could even be used as regulators with a 60 VAC(RMS) (84.84 V peak) transformer, if a small series resistance is inserted between the rectifier diodes and the smoothing capacitors, before the chipamp regulator, to keep the chipamp's power supply pins' differential voltage at least 10% or so below 84V.
I was initially wanting to produce +/-30VDC rails for a chipamp-based power amp. So I came up with a chipamp-based voltage regulator circuit that regulates to 60vdc, from a 60VAC then-rectified-and-smoothed input. It looks fairly good in simulations (with realistic ESR added to caps, etc), giving 60vdc with something like 10.5 uV ripple for any resistive load drawing up to 11 Amps (I was using 4X 4700uF/80V/.053-Ohm-ESR caps, in the rectification/smoothing stage, in that case.).
Actually, I then also added a "rail splitter" stage, using a circuit that's similar to the rail splitter circuit that's shown the datasheet for National.com's LM675(http://www.national.com/ds.cgi/LM/LM675.pdf), giving me +/-30vdc rails, at up to 11A each, and a new "ground" (i.e. the -30V "common"). With that setup, and with a (third) chipamp power amp used as a load that was producing 20vpp square waves (with 2.5us edge times) into 8 Ohms, the power rails had about 15mvpp ripple. I don't know if that's "good" or not. Anybody?
The voltage regulator stage's circuit is simple: The chipamp's power pins are across the smoothing caps. I apply a "DC Reference" voltage to the chipamp's + input and apply the divided-down output voltage to the - input. The DC Reference voltage is derived from the output, too, by running it through a LP filter and across something like a TL431 TO-92 2.5v voltage-ref device.
I guess I'm mostly wondering if this "chipamp-based regulator" approach could be economical, or "reasonable", to pursue. It does appear to buy me the ability to produce clean +/-30v@up-to-11A-each power supply rails, using a single-secondary 60VAC RMS transformer. But it would take either two LM3875 (about $6 each) or one LM4780 ($9.49 each) to do it. And it looks like the transformer would need to be able to provide at least 20.34 Amps RMS (+/- 46.7A peaks), if 2X 11A max 30v outputs were desired. (Of course, it might be a little "easier" if we don't really need a full 11A continuous DC current equivalent from the two outputs simultaneously. For 2X 8.5 Amps DC out max, the 60VAC transformer would only need to be able to provide about 16.4A RMS, +/-39.4A peaks (710W?)) I'm also a little troubled by the small resistance that needs to be placed between the recitifier bridge and the caps (to keep chipamp's power supply pins 10% or more below 84V), since with 0.25 Ohm (about the minimum that will work), I seem to need a maximum dissipation of about 88 Watts, from the resistor. I can easily find such resistors. That's not the problem. It just seems like quite a waste of power. I suppose I could add some series diodes, instead of, or in addition to, a series resistance.
The same type of power supply could be done with a lower-voltage transformer and lower-voltage chipamps, such as the LM1875, for currents up to maybe 4A, or with the same 60VAC but 2X LM3875, for currents up to 6A. But it seems like those options might be tending to get less-economical, compared to more-standard solutions (e.g. 2X LM338 reg, or switchmode PS, etc), although it might be somewhat-attractive just for the sake of "commonality of components", for chipamp-based power amps.
So, does anyone have any experience with or thoughts about any of this?
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Not to pile on, but I think that power dissipation is a problem, too. If I wrong, I'm sure someone is going to tell me so, but here's my analysis:
Assuming Tj-c + Tc-hs + Ths-air = 2C/W, your first chipamp has 24 volts across it (84-60), so it's dissipating 24W / Amp output. That limits you to about 4 Amps of output per chip. The LM675 is even worse, it's dissipating 30W/amp, and it's total thermal impedance is more like 3C/W, so it's limited to ~1.5A. So you need 3 of those to keep up with a LM3886 or LM4780.
So cost comes into play here...you're spending 3x$5.75 + $6.18 = $23.43 (Digi-Key pricing) to replace 2 LM338's ($2.50). You could pre-regulate with another pair of LM338's to reduce your ripple to vanishing levels and still come out ahead. The other option is 2 LT1083 for ~$26.50 and have 7.5A on tap. You could also do a boosted regulator with either of these and have as much as 15A on tap.
Just my $0.02, but it doesn't make sense to re-invent the wheel.
--Greg
Assuming Tj-c + Tc-hs + Ths-air = 2C/W, your first chipamp has 24 volts across it (84-60), so it's dissipating 24W / Amp output. That limits you to about 4 Amps of output per chip. The LM675 is even worse, it's dissipating 30W/amp, and it's total thermal impedance is more like 3C/W, so it's limited to ~1.5A. So you need 3 of those to keep up with a LM3886 or LM4780.
So cost comes into play here...you're spending 3x$5.75 + $6.18 = $23.43 (Digi-Key pricing) to replace 2 LM338's ($2.50). You could pre-regulate with another pair of LM338's to reduce your ripple to vanishing levels and still come out ahead. The other option is 2 LT1083 for ~$26.50 and have 7.5A on tap. You could also do a boosted regulator with either of these and have as much as 15A on tap.
Just my $0.02, but it doesn't make sense to re-invent the wheel.
--Greg
pinkmouse said:
Thanks, pinkmouse.
I have PDFs of those datasheets, but had neglected to look at the apps sections.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Heat dissipation would be an issue at higher dropout voltages but even worse is that it would be pity to use such nice chips to do the dirty work!
Might I suggest using beefy discrete NPN pass transistors with something like the LM338 as the base drive/voltage regulation element? I posted a design here: http://www.diyaudio.com/forums/showthread.php?postid=1005645#post1005645 that Anonymous1 eventually cobbled together and got good results with. Shortly after I posted that design I found that I had not come up with an original design. I forget where I saw it but someone else had posted schematics of the same configuration a couple years back.
Might I suggest using beefy discrete NPN pass transistors with something like the LM338 as the base drive/voltage regulation element? I posted a design here: http://www.diyaudio.com/forums/showthread.php?postid=1005645#post1005645 that Anonymous1 eventually cobbled together and got good results with. Shortly after I posted that design I found that I had not come up with an original design. I forget where I saw it but someone else had posted schematics of the same configuration a couple years back.
gmikol said:
Greg,
Thanks! "Piling-on" graciously accepted. That's exactly the type of response I was looking for.
It was "just an idea". And I like playing with LT-Spice, anyway. But I do hate "re-inventing the wheel" (except for the sheer joy of it; same goes for "beating a dead horse".).
I didn't check your math. But I just checked it in LT-Spice: (Holding down ALT while left-clicking on a component plots its power dissipation. And holding down CTRL while left-clicking on a plot label calculates that plot's average.) While putting 10 AMPS through both the +30V and -30V loads, the regulator's chipamp was dissipating almost 109 Watts, and dissipated almost 80 Watts for 2X 5 Amps.
Bummer.
It apparently won't be feasible, this way, then.
Thanks again!
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
this is true of any linear regulator -- if you want to supply X amps at Y volts, but you have a source of Z volts, linear regulation, at best, gives X(Z-Y) watts of loss. this gets very high very fast.
its the same reason linear amplifiers don't have stellar efficiency either.
alternate proposals have been low dropout regulation, and capacitive multiplication. you can also set regulation set-point based upon how much voltage comes out of the secondary. in this way load regulation is acheived, while line variations are not tightly regulated. this keeps the voltage across the regulator low, and provides an output that has minimal ripple. (of course, under high-line conditions with a linear amp, you'd still have near the same power loss, this would just keep the losses of the regulator low)
if you want low loss, use a switching based design.
its the same reason linear amplifiers don't have stellar efficiency either.
alternate proposals have been low dropout regulation, and capacitive multiplication. you can also set regulation set-point based upon how much voltage comes out of the secondary. in this way load regulation is acheived, while line variations are not tightly regulated. this keeps the voltage across the regulator low, and provides an output that has minimal ripple. (of course, under high-line conditions with a linear amp, you'd still have near the same power loss, this would just keep the losses of the regulator low)
if you want low loss, use a switching based design.
theChris said:
Thanks for putting that in such perfect perspective. Sometimes it seems easy for me to lose sight of that. But this one was just "for fun", I guess.
Switchmode is where I almost-immediately headed when I wanted to design the power supply for my curve tracer product, several years ago. Nice tight PCB layout helps a lot, too. I do use a similar rail-splitter on that one, with an LM1875 (+/-17.5V rails, >1.5A ea), after an LT1084 post-regulator. It has always worked very well, and has been extremely reliable.
Thanks.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
jackinnj said:
jackinnj,
Good points, all. But, unfortunately, apparently the questions are moot, at this point, since the power levels won't work out to advantage, or even to feasibility. By the way, even though I am, relatively, a newbie, here, I would like to say, to you, that I've read many of your posts, in unrelated threads, on diyaudio.com, and have appreciated a lot of them. Keep up the good work, as they say. Thanks.
- Tom
why to use chip amps when u can use a single high power transistor + regular op amp?
power transistor the same size as the 3886 can deliver much more current and cost much less.
the chip amp got 2 small power transistors in its package which designed to amplify AC signals . when using the chip as a DC regulator ,only one of the power transistors is working and the other is there for no purpose.
if u got 100 peaces of lm3886 and u got nothing to do with them , then yes , use 'em for regulators
power transistor the same size as the 3886 can deliver much more current and cost much less.
the chip amp got 2 small power transistors in its package which designed to amplify AC signals . when using the chip as a DC regulator ,only one of the power transistors is working and the other is there for no purpose.
if u got 100 peaces of lm3886 and u got nothing to do with them , then yes , use 'em for regulators
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.