ICD3 inductor noise. Design flaw or feature? Maybe I'm just having bad luck?

[Arp]

Received a replacement ICD3 some days ago. Unfortunately the problem remains and it is still emitting an audible high frequency tone

I opened it up and it's coming from the two DRQ73 inductors marked T1 and T2. Probing those components a 4900 hertz signal can be seen. I played that frequency using an online tone generator, and it pretty much corresponds to what can be heard from the units.

Late at night. This noise can be pretty tiresome. I tested putting some weak magnets on them, and it seemed to dampen the noise. I reckon stronger magnets will work better. I have to investigate if that is a recommended way to solve this problem.  I'll leave them off for time being 

Anyone else here who own this programmer? How is your experience regarding this problem? Maybe it's completely silent?

[tom66]

Can you post a picture of these inductors?
4.9kHz is a very low switching frequency, this might be a bad design which is causing audible noise.

[Arp]

Can you post a picture of these inductors?
4.9kHz is a very low switching frequency, this might be a bad design which is causing audible noise.

Sure thing. I could also see periodical burst that had twice the frequency

[tom66]

They look like DC-DC converters, likely for providing target power and the VPP for the older PICs.

My guess is it's a case of bad design. Probably under light load those converters go into pulse skip mode or burst pulse mode. You can check this with a scope, if it is a square wave it is continuously switching (maybe duty cycle in range of about 5% to 95%), but it if it's only occasionally pulsing then it's skipping pulses. This is done to save power which probably isn't too important in a USB powered debugger.

One "bodge" way to fix this would be to give the converters a minimum load, enough to push them into continuous pulse mode. That would reduce the audible noise to near zero.

You would have to check the datasheet to find out what point the converter controller (U31, U32?) switches between the two modes. In many it's around the 1mA to 20mA mark.  Any schematics for these available?

[T3sl4co1l]

Applying magnets is a bad idea...

Some inductors simply make noise.  The reason is physical movement of the coil or core.  Usually, the coil is varnished in place, but the core can move due to sheer magnetization, an effect called magnetostriction.  Different core materials have different expansion coefficients, so it depends on what materials were used.  Simply replacing the inductors with off-the-shelf parts of equal value may fix the noise.  Otherwise, use acoustic damping, foam or tape or something.

Ceramic capacitors can also vibrate due to similar effects (electrostriction and piezoelectricity).  The AC voltage on capacitors in that kind of circuit should be small, though.

You can check noise sources by using a toothpick as a probe, attached to a small piece of paper as a diaphragm.

As for the frequency, as mentioned, a pulse skipping mode seems likely.  It could also be a sign of poor compensation (incorrectly damped regulator circuit), or an inferior switchmode controller (instead of trying to reduce current first, it just skips whole pulses, each pulse remaining near full power?).

Tim

[TMM]

Cover them in hot glue?

[amyk]

How about some acoustic insulation, like wrapping the whole thing in some sort of blanket? I can't imagine it would need much cooling.

[peter.mitchell]

Cover them in hot glue?

hot glue will probably not hold well enough to stop the resonance, but may add enough weight to dampen it to be quiet. personally, i'd just cover it in thick CA glue, that would stop it from vibrating.

[Arp]

Got feedback from Microchip. They claim that it is supposed to be silent, and that I had bad luck with the replacement, and will investigate the one I sent back and let me know.

No response on the schematics query

Cover them in hot glue?

hot glue will probably not hold well enough to stop the resonance, but may add enough weight to dampen it to be quiet. personally, i'd just cover it in thick CA glue, that would stop it from vibrating.

Have tested rosin flux before with limited success. Will test this, with maybe some extra isolation, if Microchip can't work it out

[fcb]

Have you tried the unit on a different PC (potential USB power interaction with the ICD3?)

But ultimately if it's new and you feel it is faulty, then ask for a replacement. Or turn the radio up.

[Arp]

Have you tried the unit on a different PC (potential USB power interaction with the ICD3?)

But ultimately if it's new and you feel it is faulty, then ask for a replacement. Or turn the radio up.

Yeah I tested it with a laptop on battery, and still pretty much the same.

I did some rechecking today, trying to record the tone with a condenser microphone hooked up to the oscilloscope. I think I was wrong about T1 and T2 being the source, it seems like it's the L1 marked in yellow. Have had those sound before on chinese LCD modules, so I reckon it's pretty much the same thing going on :/. No response from Microchip yet.

[tom66]

Since it has exposed inductor wire, you might be able to use varnish or glue on it to stop the inductor resonating.

[dannyf]

I would go with glue as well.

If you still need the schematic, take a look at PICKIT3.

[mazurov]

I checked my ICD3. The board is green but otherwise looks very similar to yours. It makes no sounds, ever.

What you have is likely sub-harmonic oscillation, it happens at light loads when inductor runs in discontinuous mode (each cycle current through it drops to zero). Try to switch target power to the debugger and play with the load - maybe you will be able to fix the issue by just drawing extra 30-50ua.

[Arp]

I checked my ICD3. The board is green but otherwise looks very similar to yours. It makes no sounds, ever.

It's still silent even when just connected to the USB?

I did some more testing with a free Android spectrum analyzator. It's picking up a >10kHz signal from the unit. My initial guess of 4.9khz was probably wrong.

*edit*

After unplugging it and reconnecting it, the intensity of the signal was much lower. Seems like after being connected for awhile it start to oscillate more and more. Gonna try to figure this out

[apelly]

You can check noise sources by using a toothpick as a probe, attached to a small piece of paper as a diaphragm.

Nice tip! Thanks!

[Arp]

I think I figured out when this happens by looking at the photo I posted and being boggled when the noise suddenly stopped

When active led is lit and status led is dark  the noise go up. Basically programming the PIC, then disconnecting the programmer USB and plugging it in again will make it oscillate 

[Arp]

Nothing new. Have tried the ICD3 on several locations, at home, at work, on laptop with battery. Still oscillates. Too bad it doesn't use the same DC-DC as on the Pickit 3.

Microchip doesn't seem to be able to reproduce the problem. Sent them recordings and whutsnot. But nope. Think I'll have to resort to glue and isolation material

Some videos I sent them. First has slightly distorted sound but contains some screwdriver wizardry where I do some pitchbending :

[Arp]

Took some measurement screenshots of the offending inductor.  I did some googling but couldn't find much on the subject. Is it normal that the frequency of the audible oscillation is twice of what can be measured on one of the inductor points? Still learning 

[T3sl4co1l]

Eww, hysteretic burst mode alright.

It is normal that the frequency is double, because magnetostriction is a square-law effect (the same direction and magnitude of expansion whether B field is + or -), but, that wouldn't be very important with this type of waveform, because it's very on-or-off.

Tim

[Arp]

Eww, hysteretic burst mode alright.

It is normal that the frequency is double, because magnetostriction is a square-law effect (the same direction and magnitude of expansion whether B field is + or -), but, that wouldn't be very important with this type of waveform, because it's very on-or-off.

Tim

Ah does that mean it could "audibly" oscillate in any frequency with this type of waveform? Trying to understand the last part

[T3sl4co1l]

Just that, well, consider the expansion, and what's causing it.  So, you run current through the inductor, which magnetizes the core, and it expands.  Turn it off, and it un-expands (despands..?!).

If the current were a +/- square wave, it would expand, contract then quickly expand again (as it crosses through zero), and so on; the contract-expand action only occurs for a split second, while it remains expanded for most of the time; the amount of acoustic power is small (though rich in harmonics, because it's a very spiky waveform).

Now, if current remains one sided (positive or zero, never negative), there's no reason for it to re-expand, so it should indeed be audible at the burst frequency, and not some multiple thereof (harmonics aside).

Tim

[Arp]

Ah I see. Thanks

[tom66]

It's in burst mode because power consumption is low. If you find the output capacitors and solder a 1k resistor across them, you would probably give it enough load to put it into continuous switching and then it wouldn't make (audible) noise.

[dannyf]

That is only for Vpp generation.

Load it with a resistor and see what happens. Try 10K first and then go down from there. If you have a pull up on mclr, adjust accordingly (make the pull-up 1/10 of the pull-down).

[Arp]

I think they're going to send me a second replacement. Maybe the replacement I got was an old returned unit with the same problem. If third ICD3 fails I'll see if I can manage to put a load on it. Lowering noise floor is more important than the warranty imho . Thanks for the advice

Btw, seems like I'm not the only one with this problem. After googling I found this from 2010 that sounds related:
http://www.microchip.com/forums/m476054.aspx

[Arp]

I received a second replacement today. It's an older revision with a green PCB instead of a red. Now when Power & Active is lit it's completely silent. Only when status it lit there is a slight squirm but nothing too depriving. Reckon Microchip got a bad second batch of components?

[AndyC_772]

Resurrecting this old thread because I too now have an ICD3 that whistles. The noise comes and goes, it's not really consistent or predictable.

Anyone else got this problem? Is there a simple fix that doesn't involve sending it back to Microchip?

[Sal Ammoniac]

I have two ICD3s (both have red PCBs) and neither of them makes any noise either connected to a target or just connected to a PC's USB port.

[heize0]

Quite a long time post ago. I came into the same problem. It's quite noise. Replacing C18(below U26) with a solid-state electrical capacitor solved the problem at all and give you a quiet world.