Geofex voltage double with transistor switching... how does it work?

[Sage]

It is a standard detail. Usually some parts of the amp are more sensitive than others to power trash.

What specifically do you mean by power trash in this instance?  Are you talking about ripple, or something else?

I'd be tempted to move TL072 + power pin over to + of C8, so this R-C also cleans the TL072 power.

I sort of assumed the voltage regulator I'm putting in between the LT1054 charge pump's output and the TL072's input would clean things up enough.  Is that not the case?

[R.G.]

FTFY, Tex.

Fixed.

ACK! That's big. Sorry. It appears right size in the web page.

[PRR]

> ACK! That's big. Sorry.

Your web-page HTML probably uses height and width tags to set the display size.

You could image-edit a smaller copy for forum posts, but often not worth trouble.

There are similar size tags in the forum software. Let's see if I can do it....



Use "quote" to open this message and see the code I used.

Code Select Expand

[img width=511]http://geofex.com/circuits/voltpmp3.png[/img]

As you see, if you only specify one way, it also reduces the other way to preserve aspect ratio.

This does depend on browser re-sizing. You have not reduced the burden on my slim internet connection, and the result will be fuzzy. (Sometimes better if you can pick an integer reduction like 1534 to 511.)

[Sage]

Okay, so here's what I've got so far:

Some questions:

• The 0.1uF capacitor is standard for the input of the LM317, but here it's in parallel with that 100uF cap.  Seems redundant and/or unnecessary, like all it'll do is add to the total capacitance at that point.  The LM317 datasheet says it's only necessary if the input is more than 6 inches away from the power supply filter, and I'm way closer than that, but I'm not sure that 100uF cap counts as a "filter."  Should I just remove 0.1uF cap?  Everything seems stable on the breadboard without it.
• Do I really need the two protection diodes on the LM317 part of the circuit?  They're suggested in the datasheet for output voltages over 25V in case of short circuits, but that seems unlikely as I'm laying this out on a PCB.  What's the necessity here?
• Should I move all the LM317 circuitry (everything left of the 100uF cap) to the bottom ground wire, or even create a third ground wire to handle this part of the circuit?  A separate ground wire seems unnecessary since there's no audio signal in this part of the circuit, but I'm wondering if it would reduce noise to the opamp.
• Can I just remove that first R/C filter on the opamp as Keppy suggested?  There's already a 10k/47uF filter, and in theory the LM317 should be filtering noise on the voltage line --- even moreseo with the 10uF adjustment capacitor in place.
• Could I go even futher and remove the two 10k resistors and the second 47uF cap entirely?  Or do those components serve a secondary function?

[/list]

[Keppy]

• The 0.1uF capacitor is standard for the input of the LM317, but here it's in parallel with that 100uF cap.  Seems redundant and/or unnecessary, like all it'll do is add to the total capacitance at that point.  The LM317 datasheet says it's only necessary if the input is more than 6 inches away from the power supply filter, and I'm way closer than that, but I'm not sure that 100uF cap counts as a "filter."  Should I just remove 0.1uF cap?  Everything seems stable on the breadboard without it.

It's common practice to parallel large caps with small ones in power filtering, as the large caps sometimes have trouble with higher frequencies due to some non-ideal properties I can't remember (ESR? Inductance? Someone help me out here, it's late. ).

• Should I move all the LM317 circuitry (everything left of the 100uF cap) to the bottom ground wire, or even create a third ground wire to handle this part of the circuit?  A separate ground wire seems unnecessary since there's no audio signal in this part of the circuit, but I'm wondering if it would reduce noise to the opamp.

I would leave it as is. Best practice is to run a third trace/wire to the star ground point, but as Paul described, your power filtering is ample for the opamp.

Some perspective: The LT1054 runs above audible frequencies, so any whine that it creates should be inaudible, and you have no oscillator or clock circuits to introduce heterodyning issues that would creep into the audible spectrum. Even if you get a chip with the oscillator running near the 15kHz minimum spec (rather than the 25k "typical") you have multiple layers of filtering, plus a guitar speaker that can't reproduce that frequency even if it's audible in theory. The precautions in your diagram are ample in my opinion, and you could probably even use the earlier version of your design without issue. Of course, using good practice changes that "probably" to "almost definitely."

• Can I just remove that first R/C filter on the opamp as Keppy suggested?  There's already a 10k/47uF filter, and in theory the LM317 should be filtering noise on the voltage line --- even moreseo with the 10uF adjustment capacitor in place.

You can, but as Paul mentioned it will help clean up any noise that happens to escape the regulator (likely not much). Now that we've got your full power scheme those components make more sense to me. They're a power filter for the regulator output, which is not always necessary. Notice, though, the 317 datasheet has an optional 1uF cap on the output to improve transient response (read: filter the output in response to a changing load). The 100R/47uF power filter serves the same purpose, but much more so. I would ditch the 1uF regulator cap as it doesn't seem to be doing much compared to the RC filter.

• Could I go even futher and remove the two 10k resistors and the second 47uF cap entirely?  Or do those components serve a secondary function?

Those create the 1/2 supply bias voltage for the opamp. They are absolutely necessary.

[Sage]

It's common practice to parallel large caps with small ones in power filtering, as the large caps sometimes have trouble with higher frequencies due to some non-ideal properties I can't remember (ESR? Inductance? Someone help me out here, it's late. ).

I think it's ESR and/or Impedance.  I read something to that effect anyway, but this is my first experience manipulating the power stage so ESR is a totally new concept to me.  Thanks for the tip, very good to know.

Some perspective: The LT1054 runs above audible frequencies, so any whine that it creates should be inaudible, and you have no oscillator or clock circuits to introduce heterodyning issues that would creep into the audible spectrum. Even if you get a chip with the oscillator running near the 15kHz minimum spec (rather than the 25k "typical") you have multiple layers of filtering, plus a guitar speaker that can't reproduce that frequency even if it's audible in theory. The precautions in your diagram are ample in my opinion, and you could probably even use the earlier version of your design without issue. Of course, using good practice changes that "probably" to "almost definitely."

Right on.  I just figured I'd take the opportunity to learn some best practices here.

Notice, though, the 317 datasheet has an optional 1uF cap on the output to improve transient response (read: filter the output in response to a changing load). The 100R/47uF power filter serves the same purpose, but much more so. I would ditch the 1uF regulator cap as it doesn't seem to be doing much compared to the RC filter.

Is there any benefit to having both, or is the 1uF pretty much redundant at that point -- that is to say, would the RC filter anything the 1uF would have filtered anyway?

• Could I go even futher and remove the two 10k resistors and the second 47uF cap entirely?  Or do those components serve a secondary function?

Those create the 1/2 supply bias voltage for the opamp. They are absolutely necessary.

I suspected as much, thanks for confirming that for me.

Any thoughts on removing the protection diodes on the LM317?

[PRR]

> thoughts on removing the protection diodes on the LM317?

What is cheaper? A 10-cent diode now, or a blow-up before a career-critical performance and a messy repair the next day?

Generally, if an extra nail *might* keep the shed from collapsing in a blizzard, I'd favor the extra nail. (Drove about $13 of "extra screws" in my shed last fall, and this winter I have been glad I did.)

[Sage]

What is cheaper? A 10-cent diode now, or a blow-up before a career-critical performance and a messy repair the next day?

I'm more concerned with board space than cost, but you make a good point.

I've managed to lay everything out and still have separate ground lines for the charge pump, voltage regulator, and opamp sections:



We'll see how it works once I have the boards in hand.

[Sage]

So does anyone have any suggestions on what brand/type of capacitors I should be using for each stage of this circuit?

If I understood PRR correctly in my last thread, lower ESR is desired for the caps on the charge pump because it reduces ripple (it also reduces voltage loss but that's not an issue).  I was considering the Panasonic FR series for the 100uF and 10uF caps there, as they have low ESR and high lifetime (5000h).

I'm thinking Nichicon FG for the 1uF and 10uF caps in the opamp stage, as those are directly in the audio signal path.  Lifetime is only 1000h on those, though... is there a better alternative?

What about the caps on the voltage regulator, and the 47uF caps on the filter and supply bias for the opamp?  What characteristics are most desired on those, and why?

[PRR]

> What characteristics are most desired

Two legs.

Don't over-obsess. You aren't building a nuclear detonator. Not even an iPad. Your little buzzer is very low-tech and low-stress compared to many commercial products today. "Any" capacitors should do. If they don't, it's your layout. Maybe a too-low value.

Pleasant color is nice.

[PRR]

> Lifetime is only 1000h on those

At what voltage and temperature?

I bet if your stage is anywhere near as hot as the cap rating, you'll have heat castration. Maybe skin-burns.

Life goes up VERY fast with lower temperatures.

Staying below max voltage also helps.

The life of those caps has not been an issue in audio. (Things can be different in heavy machinery controllers.)

[Sage]

Don't over-obsess. You aren't building a nuclear detonator. Not even an iPad. Your little buzzer is very low-tech and low-stress compared to many commercial products today. "Any" capacitors should do. If they don't, it's your layout. Maybe a too-low value.

Good point, as always.  There are probably plenty of factors that have a greater impact than what brand/series caps I choose.

> Lifetime is only 1000h on those

At what voltage and temperature?

I bet if your stage is anywhere near as hot as the cap rating, you'll have heat castration. Maybe skin-burns.

Life goes up VERY fast with lower temperatures.

Staying below max voltage also helps.

The life of those caps has not been an issue in audio. (Things can be different in heavy machinery controllers.)

They're rated for 85 degrees and 50V.  I assumed the lifetime was an estimate of how long they last under load.  That number changes based on internal temperature and voltage?  What about ambient temperature?  I'm sticking this into a tube amp that can get a little hot.

[PRR]

85 deg C is not-quite boiling.

35 deg C is more likely, unless you jam the caps too near the power tubes.

Many chemical reactions run 2X faster every 10 deg C hotter. Or vice-versa, 35 deg C (50 deg C cooler) is 2^(50/10) or 32 times longer, 32,000 hours. At 2 hours a night every night this is 43 years. I suspect you will have many-many other wear-outs in this amplifier before this.

[Sage]

Well, I got the boards printed and all the parts wired up.  The good news is that increasing the voltage powering the effects loop to 30V does allow me to prevent the opamp from clipping even when the amp is cranked.

The bad news is that it introduces an audible clicking sound, even when removed from the signal path.  I powered my circuit from the top and bottom of C38 shown here:



Maybe that NE555 is interacting with it somehow...?  I don't even know what that component does, except that it's some kind of timer.

Perhaps I should just hang another voltage doubler off the top and bottom of C35 and forget the LT1054 altogether.

[R.G.]

First, don't use an NE555. They are notorious for producing large glitches in both power and ground. Use a CMOS version, unless you can't.

And you can't. Looks like you're using the 555 for driving your multiplier.

Given that, you're going to have to do an *immaculate* job of decoupling the 555 and multiplier from your other grounds, or you're going to get ground noise bleeding into the audio path even when it's removed from the signal path - which you say you do.

Grounding issues are always about where the *current* goes.

[PRR]

I *think* that '555 is the amplifier's logic's power-up reset mechanism.

It senses raw AC (pre filter cap), delays, and then *thunk* resets all options to defaults?

If so, it would be quiet after that thunk.

[R.G.]

Of course you're right!   

[Sage]

Sure enough, I just tested hanging an extra voltage multiplier off the tops of C29 and C35, and it appears to work perfectly.  I should've gone this route to begin with instead of fooling with the charge pump.  With a single extra 1n4007 and 1000uF capacitor I can get 30.5V.  With two of each, I can get 37V, enough to use an LM317 to get a regulated 30V ouptut.

Should I even bother with the LM317 at this point?  30.5V is exactly what I need to power the opamp... it seems like a waste to push the voltage higher only to regulate it back down, but I thought the LM317 would help with noise rejection.  Any thoughts on this?

[R.G.]

Cascaded multipliers sag badly under load. Unless your 30.5V is measured under your maximum (and almost constant!) load, it will wander around. That may or may not be OK, depending on the circuit.

If it's not OK, you'll need to regulate it back down to keep it mostly constant.

[Sage]

Hm. The TL072 dual buffer I'm powering doesn't use much current, but I don't know if it's consistent:



Would the power draw change based on the signal going through it, or can I expect it to present a consistent load?