[rcheshire (Rowland)]

Well this has progressed and you may have read about my agonizing efforts to find a quiet switching solution to eliminate EMI and get rid of those nasty lines in images. Digital cameras are very sensitive to EMI.

I have tried 4 or 5 different logic level n-channel very low rds on mosfets driving direct from the PWM pin and an MCP1415/16 gate driver with mixed results. Having come up with all kinds of snubber networks, poor, fair and bad, rigorous testing revealed flaws in all .... back to square one.

In a nutshell the easy method, producing the cleanest images is a 22uf Tantalum capacitor Gate to Source. This does not send the mosfet linear from what I can see. The system is slow and temperature control sloppy, although useable, which is to be expected. The NXP PSMN1R1-30PL runs cool - unlike others. It's smd cousin the 1R3-30LY runs in liquid solder - don't be distracted when testing these things - it stopped working when I tilted the board and it slid off. So much for home brew smd, for now.

The Android SPICE based program I been using seems OK and tends to verify practical observations. Images below. Time base is 20us. 12v side is 10mv resolution and offset ~5mv ripple. Low side is 1v resolution. The other Android simulator had a bug producing incorrect results. The developers appear to have fixed it.

I understand that a 22uf cap gate to source may not be optimal as far as switching losses are concerned, however, I am at a loss to find another workable solution. I thought I had EMI solved on several occasions, but there was always anomolies. I will be less hasty to cry Eureka in the future - DRAIN turn-off produces some nasty voltage spikes which are easily eliminated with a high speed diode source to drain, or so I have found.

If I have some spare time today, I will play with one or two notions that come to mind, based around the 22uf Tantulum capacitor, to improve the system response.