1044 Charge Pump Capacitor Specs

[spacecommandant]

Hi, a couple of questions regarding capacitor values for this voltage doubler circuit (I'm using TC1044SCOA).
I've read that 'low' ESR caps are recommended, but what is considered low enough for this application, and which caps in the circuit should use low ESR? I'd guess C3 is important, but any others?
Am I correct in assuming C1, C2 and C3 are okay using 25V caps but the others should be 35V?

One side note (I'm going to read up on this since I know high-pitched whines have been discussed plenty): I'm seeing but (not surprisingly) not hearing a steady, narrow frequency spike at around 19 kHz (in a frequency analysis VST in Reaper) when I use these doublers with a couple of simple boost pedals I made (they run at approximately 11mA and 19mA).

[anotherjim]

I've recently used 10uF MLCC for the pump cap and the first output cap. Really because I needed a small footprint and I had them already. Seems to work ok but it's only a neg supply for one dual opamp.

[ElectricDruid]

The datasheet says that with the Boost pin (pin 1) tied to +V, the oscilator should be at 45KHz. So I'm not sure why you're getting 19KHz, seems a bit wrong.

https://ww1.microchip.com/downloads/en/DeviceDoc/21348a.pdf

[anotherjim]

That's a good point, Tom. Hey, is voltage doubling a 2-step process? So it's actually doing 38Khz given tolerances and the diode pump actions add a half-frequency component while the PC audio interface is probably deaf to the "45Khz". I see no tolerance spec for variation in the frequency of the on-chip oscillator.

[spacecommandant]

Yeah, I was using an EHX 9V power adapter and switched to a different adapter but there was no difference.
One thing to mention is that the 11mA boost (using one TL072) has a 19.3kHz spike and the 19mA boost (using two TL072) has an 18.9kHz spike.

[spacecommandant]

Here's a clue: when I touch C3 the spike gets louder.
I'm currently using this 25V, low impedance cap for C3:

https://www.mouser.ca/ProductDetail/80-A765KG106M1ELAE35

[ElectricDruid]

That's a good point, Tom. Hey, is voltage doubling a 2-step process? So it's actually doing 38Khz given tolerances and the diode pump actions add a half-frequency component while the PC audio interface is probably deaf to the "45Khz".

That's a good point too. The noise could even be aliased back from whatever the sample rate of the interface's ADC is, so perhaps I'm being naive thinking 19KHz is weird. I should be thinking about what the sample rate is likely to be and what the hell happens to a 45KHz racket when you try and sample it at an audio rate!

I see no tolerance spec for variation in the frequency of the on-chip oscillator.

Another good point. I doubt this thing is tuned to the C above Top C! They're probably all over the place, it's not like it would matter much for most applications.

[spacecommandant]

Wow. Okay, based on what you were discussing, I dropped the 48 kHz sample rate of Reaper to 44.1 kHz and I no longer see the spike.
It may be time to invest in an oscilloscope.
Thanks so much for the input!

[Rob Strand]

Good debugging by ElectricDruid.

I guess the point is shifting sample rate proves what the problem is and confirms ElectricDruid's theory but it still means the problem is there.

What's causing it exactly can be tricky.

The incoming ground power needs wire directly to C1, C1 needs to be mounted close to the IC the ground side of C4 needs to be close to the IC.  You really want the input ground, C1 ground, IC ground and C4 ground all going to the same point.  All those grounds are noisy.   *Then* tap off that ground point to feed your audio circuit.  Don't tap off any ground along the noisy ground.

Beyond that output filtering.

If you are using the input jack to switch the power then you can get a lot of noise from the along the ground which is switched by the input jack.

[spacecommandant]

Good debugging by ElectricDruid.

I guess the point is shifting sample rate proves what the problem is and confirms ElectricDruid's theory but it still means the problem is there.

What's causing it exactly can be tricky.

The incoming ground power needs wire directly to C1, C1 needs to be mounted close to the IC the ground side of C4 needs to be close to the IC.  You really want the input ground, C1 ground, IC ground and C4 ground all going to the same point.  All those grounds are noisy.   *Then* tap off that ground point to feed your audio circuit.  Don't tap off any ground along the noisy ground.

Beyond that output filtering.

If you are using the input jack to switch the power then you can get a lot of noise from the along the ground which is switched by the input jack.

Yeah, I made up a PCB and have ground pours which haven't isolated the power section ground as you've specified, so I'll try to adjust things. Thank you.

[anotherjim]

OK, if it was aliasing the samplerate, the numbers don't add up and why does it do it at 48Khz but not at 44.1Khz? I suppose the interface input filter must lower its cut-off with a reduced rate and is cutting the clock noise more effectively.

[ElectricDruid]

I agree it shouldn't disappear, just move. Let's have a think about it:

At 48KHz sampling rate, anything over 24KHz will alias. 19KHz is 5KHz back down from 24KHz, which implies that whatever went in was at 29KHz - not a frequency that we're expecting from anywhere. A 45KHz input would alias all the way back down to 3KHz!

At 44.1KHz, things will bounce back from 22KHz, so our supposed 29KHz noise should be at 22-(29-22) = 44-29 = 15KHz.

So I'm not sure it makes sense.

Anyway, the gist of it is that there's high frequency noise on the 18V supply output, and that noise is very likely ultrasonics. When you run those into something that samples at an audio rate, you get problems. One solution would be to make sure that any pedals that are using this boosted power have heavy filtering at the top of audio. For guitar, 15-18KHz would probably be plenty. The idea being that you make sure that only audio leaves the pedal and any ultrasonics from the power get filtered before they get to an audio interface.

It's not only audio interfaces though - you'd have the same problem if you used a digital reverb pedal last in your chain. Anything that samples and doesn't have enough anti-aliasing filtering will do it.

More filtering! Filtering on the power (can you cope with the volt drop on a small R/C filter on the boosted power?) and filtering on the audio (to remove any noise that does get through).

[anotherjim]

If we have the schematic for the audio output stage of the project, we can suggest a simple temporary low pass filter to tack on to prove the principle before working on a proper solution.
45kHz is probably not really high enough for simple ferrite beads over some wires/links, although there will be harmonics of 45Khz too which can be reduced that way. A commercial product would probably have to have some RF filtering to pass emissions regulations due to those harmonics.

[FSFX]

45kHz is probably not really high enough for simple ferrite beads over some wires/links,

Ferrite beads only give inductances of a few microhenrys so they are unlikely to have much effect below RF frequencies.



All of these inductances were measured at 200kHz on a Peak LCR45 tester.

[spacecommandant]

If we have the schematic for the audio output stage of the project, we can suggest a simple temporary low pass filter to tack on to prove the principle before working on a proper solution.
45kHz is probably not really high enough for simple ferrite beads over some wires/links, although there will be harmonics of 45Khz too which can be reduced that way. A commercial product would probably have to have some RF filtering to pass emissions regulations due to those harmonics.

This is the same boost I had some help with a few weeks back on this forum.
I made another one which is the same except for a buffer stage at the input, my thinking being it might "protect" the sweep of the volume pot in some scenarios, since I read that a passive volume pedal with a 250k or 500k pot in an amp's effects loop can alter the sweep, making it act more like a switch. Anyways, that was my thinking, right or wrong.
*Capacitor/resistor names have changed in this schematic compared to the initial power section schematic I first posted*

[anotherjim]

Make a filter with a cap from the output tip to 0v sleeve. Try 100nF to make a filter starting at 16Khz if R3 is 100r.

[spacecommandant]

Good debugging by ElectricDruid.

I guess the point is shifting sample rate proves what the problem is and confirms ElectricDruid's theory but it still means the problem is there.

What's causing it exactly can be tricky.

The incoming ground power needs wire directly to C1, C1 needs to be mounted close to the IC the ground side of C4 needs to be close to the IC.  You really want the input ground, C1 ground, IC ground and C4 ground all going to the same point.  All those grounds are noisy.   *Then* tap off that ground point to feed your audio circuit.  Don't tap off any ground along the noisy ground.

Beyond that output filtering.

If you are using the input jack to switch the power then you can get a lot of noise from the along the ground which is switched by the input jack.

Thanks Rob,

I soldered up a dead bug version of the power section by following the cap positions and ground routing you specified and the signal now looks totally clean; no high frequency spike.
Grateful for that bit of advice.

[Rob Strand]

I soldered up a dead bug version of the power section by following the cap positions and ground routing you specified and the signal now looks totally clean; no high frequency spike.
Grateful for that bit of advice.

Thanks for the feedback.   
It's surprising how easily the evil can get out of those things.