Beautiful convection oscillations

[maxwell3e10]

One can often see low-frequency noise or semi-periodic oscillations due to air currents and convection. But here I saw very stable oscillations which are most likely due to a convection cell in the enclosure. I was testing a DAC in a sealed metal box about 1 inch in height, with no fan. The board is near the bottom of the box. When the box is upside up (the board is on the bottom and most parts are mounted on top of it), there are very strong oscillations in the output. Rotating the box to any other orientation kills the oscillations.

I believe most of the sensitivity of the components are due to temperature gradients. When I take the board out of the box and put a fan over it, there is no significant change in output voltage, just more noise. Putting my finger over various chips also does not cause a permanent shift, just an initial transient and opposite transient when removed. So, I need to reduce temperature gradients in the box, but without adding a fan. Perhaps gluing small heat sinks to some of the hotter chips would help?

[blackdog]

Hi,

Try to find out which component is generating, it could also be a temperature effect of several components together.

Just adding a heatsink is usually not a good solution, what if the ambient temperature goes to 35C, like last week here in Amsterdam...
Does your heatsink method still work well then?

Maybe you have a power supply or opamp that has a too small Phase or Gain margin?
This is temperature dependent and can go wrong if you are on the edge.

Perhaps you are suffering from a mechanical failure component, that only occurs in a certain position of your device?

Kind regards,
Bram

[mendip_discovery]

More testing needed.

A thermal imaging camera would help in working out what is causing some of the issues but as previously said dont exclude mechanical issues.

[Stray Electron]

    I'm wondering if both internal and external air currents are having an effect when you reorientation the box or if it's just one and not both?

    I would try insulating the box on the outside to block external air currents and see what effect that has. Then put some partitions inside the box to block internal air currents and see what effect blocking the internal air currents has. (Keeping in mind that if any circuit components are getting very warm then they may need the air currents for cooling.)

   I'm very curious about what is causing that standing wave pattern after 500 seconds and why that wasn't seen in the two earlier tests with the box right side up.

   If you can, it might be worth putting the circuit in the box upside down and then retesting. That should tell you if the change in oscillations are due to the nature of the box (different thicknesses of the top and bottom?) or the orientation of the circuit.

[maxwell3e10]

Thanks for the suggestions. I put a small heatsink on the DAC chip, it reduced the amplitude of oscillations but did not eliminate them or significantly change their frequency. Taping pieces of paper over various components to block air flow I can mostly eliminate the oscillations, but not reliably. It seems multiple components are responsible for temperature sensitivity. Putting my finger over different components or just over empty PCB area can cause the output to shift up or down. This is a commercial board and is nicely build in a heavy-duty sealed aluminum box.
I think for now it will have to be operated upside down to drain excess electrons

[iMo]

I saw similar flow in my 34401A while measuring its internal temperature (a thermometer glued inside the box). The temperature noise was large, due to flows.
What eliminated fully that noise was an internal insulation put inside the box - a sleeve made of a thin foamy textile put just beneath the metal case, insulating the case from the internal space (and thus slowing down the thermal interactions with the case).
PS: see below the experiments:
https://www.eevblog.com/forum/projects/thermostating-the-hp-34401a-meter/msg2619108/#msg2619108