Can we talk about Charge Pumps? Design.

[R.G.]

I redrew the last schematic to help my brain cope with the whole situation.
I'm thinking that C1 and C3 can be merged. C4 and C5 too.
Q: Plain ceramic caps can be used with good results, right?

Here's some additional commentary.

Ground is not simply ground. You have to know where the current flows. In the image, I've used lines to represent actual wires on the path that would normally just be called "ground". They are wires, and have resistance. The input current flows into C1 on the incoming power +/- wires. C1's negative lead and one lead of C6 go to pin 3. When the IC switches current into C2, the current flows in a loop from + of C1/C6 to pin 8, out pin 2, into C2; the return current flows out of C2, into pin4, out of pin 3 and back to the (-) of C1. The pulse currents are high, and edges are sharp. The individual wires have voltage drops reflecting currents that flow.

When the oscillator flips the other way, C2 + is connected through pin 2 to pin 3; The other end of C2 is connected through pin 4 to pin 5, where it equalizes voltage by forcing charge into C7. This also has pulse currents flowing, and in this case they flow through the wire on the + end of C4 back to pin 3. Notice again that there are pulsed voltages on all these wires from the currents flowing at switching times.

The The output voltage comes from the (-) side of C4 and the (-) side of C1 through a separate wire. This wire has resistance, and some voltage drop by V = I*R. But notice that none of the switching currents flow through it, so none of the voltage drops has signal from the switching operation of the IC.  There is still some small component of the switching frequency caused by the resistance of the input wires and the charge/discharge of C1, but this is smoothed to a small triangle wave by the capacitance of C1, and is much smaller than the voltages caused by current spikes in the actual switching. And here's another reason to make C1 big. When C2 is connected to C1 through the IC, they equalize voltage. In capacitors,
charge Q = C*V, or V= Q/C. The voltages equalize, so charge flows equal to the ratios of the capacitances. If C1 is 10X C2, then making any voltage change to C2 required C1 to supply charge to C2; the amount of charge C2 can hold at any given voltage is 1/10 that of C1, so C1 changes voltage only 1/10 as much as it changes C2's voltage. The bigger C1 is compared to C2, within reason, the smaller the charge/discharge voltage is across it.

[davent]

Q: Is there a significant advantage to soldering the 100n directly on the IC pins as opposed to the above layout?

For that i solder the a little grain-a-rice ceramic cap to the machined socket pins so it sits under the socket but above the board. Makes it easy to add to found layouts that haven't put such a cap on the layout and saves a tad of space.

[italianguy63]

That's a slick idea for adding the 100nF caps!

For the record, playing with another circuit today (another off the shelf charge pump with FuzzFace).

Values:  C1 = 47uF, C2=10uF, C4=47uF, C6 & C7 added = 100nF, C3 and C5 omitted.  Quiet with the 1044.  Whines badly with the LC962.  Grounding the LV helps, but still whines.

MC

[ggedamed]

R.G., you're using the schematic as layout.  All the parts are just lines! This makes my head hurt  . OK, new version:




Now, seriously, my mind doesn't work very well with symbols, it's difficult for me to use the schematic for layout purposes. Would you be so kind to comment on the layout?

Mark, if you don't use R.G.'s PNP transistor trick, how do you manage the start-up of the circuit in a pedal? I added a pin 6 solder jumper, just as you asked.

About the reverse diode on the output: the voltage reversal on pin 5 happened to me not much time ago. I keep a LT1054 on one of my breaboards, because I like very much the noise improvement  brought by the raised supply voltage. And I'm nuts about distortions. At that time I had a mixture of op amps, JFETs and a CD4069 on the breadboard. LT1054 was in the inverter mode. I found the circuit would not start and LT1054 would start to heat up. If I cut the op amp's supply and/or reduce the supply current of the CD4069, it would start working. I found later in the datasheet (the datasheet comes always last, isn't it?) at the end of the page 5 that op amps can pull pin 5 positive. They even give a solution, but my problem was solved by a 1N5817 like in the schematic above.

[italianguy63]

Man that looks pretty great!

I will probably get flamed, but I just use the stereo jack trick.  I'm not getting noise-- so I didn't use transistor switching isolator thingy.  But, it is a good thingy, and if I had it on a board like this, I would certainly incorporate it.

If you push C1 and C2 apart slightly, you would have a perfect spot for the mounting standoff hole just below them.

On the reverse voltage thing... I was (unwisely) running my power LED on the negative side of the pump (because I had a convienient place to put it).  Anyway, the LED was lighting when the circuit was off!  WTF?!  That meant power was backflowing into the charge pump from downstream.  It is a mystery as to how or why.  I think it might have been A/C?  The diode fixed that.  Plus I saw the info. on the datasheet you mentioned, so it quickly became a "must do."

The twin ground pads make hookup a snap.

I even like that the board is square!  BRAVO!

Are you using Eagle to do that?  I am starting to play around with it (you might note the schematics I have been posting are Eagle).  I was using Visio prior-- because it was what I have.

MC

[italianguy63]

A couple more small thoughts.. (looked at it more).

If R.G. is really really intent on keeping the grounds isolated-- you would break the trace between pin 3 and the solder pad.  But, both grounds would have to be properly connected as they would not be common anymore.

Push IC1 to the right, to even it up with the edge of C4.  That makes the pad on pin 5 work better, and you can uncramp the solder pad on pin 6.  Then you have room to move C6 and C2 to the right slightly.  Gives room for the standoff hole.  Turning the labels would get rid of a lot of phenolic too.  Not nitpicking BTW.  I don't even know if this is the final desired design.  But, it looks good to my eye!

MC

[ggedamed]

Yes, I'm working with Eagle. When we say it's done I'll put the sch and brd files online.
But right now it's only a layout, not a fab order. The labels are there to let me know what to move.

How big is a standoff hole? I can fit there a 5mm (0.19685") hole. I usually use M3 (3mm hole) screws.
I'm reluctant to move the IC because the whole philosophy behind this version is to get everything as close to pin 3 as physically possible.
Of course, the ground pads have to be connected. I tried to make the ground trace pass through the capacitors before exiting the board.
Are there any problems with the solder jumper from pin 6?

This is how it would look with stand-up resistors and a 3mm standoff hole (the image is at 600dpi, the board is 1.35x0.9", around 34x23mm):

[italianguy63]

The standoffs I have been getting use a 5/32" (.15625") hole -  right at 4mm.  If that isn't a choice-- then 5mm.  Speaking of holes, I like the hookup holes and pads slightly bigger, as I prefer to use 22awg wire as opposed to 24awg, expecially on power wires.  

It seems counter-intuitive not to keep the grounds common.  The solder pad for LV is good.  It's there if you need it, but my recent experience with the LC962 and the LV connection means I probably won't go down that road anymore.  But, still nice to have I think (So what?  Leave it.).  I really like the solder pad options-- they are there if you need them, and normally "open" so not problematic.  You won't use them unless you know you need them.  And they are hidden under the IC.  I like it a lot.

Any reason you kept the signal traces on the left so thin?

I haven't figured out how to go from the basic board layout to making Eagle draw the actual traces (and rerouting them) yet.  Any suggestions on where to look?

[ggedamed]

OK, 4mm then. I also enlarged the pads and vias.
The traces on the left have actually normal thickness (24mils) because they are normal as in not-power traces. The solder pads traces are the same thicknes for the same reason. The power traces are thick, some 50mils, mostly 66mils. If you wonder what a mil is, take a look at this Wikipedia page.

For info on Eagle some good resources are:
http://gaussmarkov.net/wordpress/tools/software/eagle/eagle/
https://www.sparkfun.com/tutorials/109

It seems it is finished, so here are the files:

sch file
brd file
pdf layout (print using Scale: None)
png at 300dpi:

[italianguy63]

*YAWN*  Still waiting on the 1054's to get here from China for testing.  Why do some parts get here in a week or so, and others take up to 4 weeks?  You see them pass through the shipping facility, but then it is a crapshoot.

I've been working on this and tweeking it.  Going to have some boards made up to test/use.  I like it a lot.  I plan to use it on the builds I have been doing if it tests out (it should).  It makes a good "heart" for other circuits I think.

MC

[italianguy63]

Maybe helpful to others.  Green is good.  Yellow is marginal. Red is bad:

[tranceracer]

Sorry guys dumb question, but what is the SJ1 and SJ2 components and what is it for?

[italianguy63]

Solder Jumpers.  SJ1 turns the "Boost" functionality on.  SJ2 turns the "LV" (Low Voltage) functionality on.  That is the purpose of the chart I just posted.  Different chips have different functionality built in while others don't.  You have to set the jumpers to which chip you are using.  MC

[tranceracer]

Ahhh, Thanks Mark!

[gena_p1]

R.G. is right, as usually. Electronic choke based on emitter follower, as in Small Clone or Boss MD/MT, is very useful electronic block.

Another improvement is to use microchip's modern version of 7660 - TC7660H. High speed charge pump (120kHz instead of 10 kHz max1044 or usual 7660). It produce more "musical" voltage, even without a choke.

[italianguy63]

Thanks.. I will revise my spreadsheet and add the 7660H.  It too is at its limit at 10V though.  If anyone cares, I can add the current cost of each component from Mouser.  In looking this morning, based on cost, the TC1044S (EPA) is the clear winner at $0.93 a copy when buying 10.

MC

[gena_p1]

10 volts is limit of input voltage.
Standard stompbox supply is 9V, and protecting diode or resistor takes 0.2-0.7V.  So, all fine 8.5V < 10V.

For large current, it is good idea to take musopatamus A/DA flanger voltage doubler, based on LT1054.

[Govmnt_Lacky]

Your mA numbers are a bit low huh?

The LT1054 is rated to AT LEAST 50mA output 

[italianguy63]

Sorry, I didn't label it.. that was the max supply current.  The 1054 is 5.0mA at 15V.  I just took it off the spec. sheet.  MC

[Morocotopo]

I believe the 7660S can give 10 mA, not 2. Depends also on the configuration (voltage inverter or doubler).