Peter P. wrote: > Bob Axtell neomailbox.com> writes: > > >> This is similar to the reduction in utility of the 74HCxxx parts that >> occurred in the late '90s. The semi manufacturers, >> starting with Fairchild, made the series with smaller dies, increasing >> their profit margin. But inadvertantly, the >> change dramatically increased the speeds of the parts, requiring me and >> many others of redesigning the products >> so that more bypassing is installed to maintain proper intended operation. >> > > Yes, but that was a relatively small thing and it was not unexpected. This is > large imho. > > >> Luckily, the costs of small MOVs and TVSs are falling, and some are >> multiples. I especially like the >> Cooper-Bussman devices. >> > > As I found out the hard way, TVSs and MOVs do not have a low enough Vf to clamp > for the 'new' micros. In fact, at usual circuit currents in a PIC (say 10 mA - > ish inductive kickback from a relay coil), many small and inexpensive Schottkys > don't have a low enough Vf to be used. So this would mean (also for Wouter): > > Gee, peter, I have recently used new MOVs to allow a PIC to survive in a literal BATH of motor noise generated by a AC Motor driver at 40A PWM at 16Khz. After significant testing all PICs survived when the PIC signals were laid _alongside_ the driving signals. The MOVs were connected across the PIC's VDD and GND, and across the signal and GND, right at the PIC. Each line also had a 3.3 ohm series resistor. The devices I chose begin clamping at 8V. The 2-sided PCB measures 0.5" x 0.4" and is 0.031" thick. Mounts in a motor housing, for identification purposes. Almost continuous testing for a month, no failures whatever. Can you provide some clues for instances where MOVs and TVSs are failing for you? Seriously, I'd like to know why these are inadequate. --Bob > - suppression on any wire or PCB trace that is not shielded and longer than > about 2 inches, and may be high Z (including during reset) (to pass > susceptibility tests). > - all relay kickback catchers must be expensive, large Schottky to have Vf < > +0.3V during inductive kickback of 100 mA or more, if the relay is driven > directly by the PIC (5V, 3V etc). > - anything inductive or capacitive connected to a PIC pin needs the same > suppression. This includes piezo disks and speakers and even microphones and > 'ac' inputs or inputs from analog conditioning circuits that can overshoot or > undershoot a little (including during reset/powerup/powerdown, when coupling > caps charge). Any ground bounce or glitch travelling on the supply lines and > exceeding 0.3V will cause a problem on any decoupled pins, such as timing > inaccuracy. 0.3V is 6% of 5V. Thus the supply must now be filtered to 2% too. > - any PIC pin connected to a large logic level MOSFET gate and tristated will > 'see' ripple from the mosfet drain and this will cause trouble if it's larger > than 0.6Vpp at the gate (easily so if something fast is happening at the drain, > like a relay or motor kickback, or a glitch propagating through the unfiltered > supply line). This includes the time during reset when pins are necessarily > tristate. My tests with 16F84 showed that the chip won't exit reset at all if > such a condition exists. > - any floating tristated pins MUST be pulled to some logic level including at > all times during development (to permit same) AND reset (i.e. use resistors on > all pins a la MCS51). > - anything else I haven't thought about, such as small filter caps on IO pins > discharging while the power goes up/down to/from standby level and much much > more. > - also as I said, the Schottky leakage is at least 1 order of magnitude worse > than CMOS leakage, and gets worse with heat. Multiply that by 20 IO pins and > your battery will have to grow. > > My recent experience with the 'new' features on 16F54 lad me to believe that > even small pulses violating the 0.3 V rule will cause the chip to lose > instructions/timing. Any real time clock or other time dependent process should > be affected. The watchdog could be triggered periodically. When such a > condition exists there is no way to pull the chip out of reset reliably. > Checking for oscillation with an external watchdog does not work (the > oscillator works with reduced amplitude on 16f54 during the fault condition). > > All ICSP pins that lead to a header must be suppressed as above, as they are > connected to a long cable that is floating and will rectify mains ac from the > 'air' when the programmer is not driving them. Maybe the diodes can be mounted > on the cable end to avoid changing the circuit. > > >>> Q3 - what happens to small minimal parts count projects where the parts count >>> can double or triple due to the new clamping requirements ? >>> >>> >> Unfortunately this will be required to meet ESD requirements of UL, CSA, >> and EU. >> > > I know. That's why I am asking. It looks like the people who make $ 0.5 MCUs > are actually losing money to the people who make $0.05 Schottky suppressors. > Maybe it's a new trend in outsourcing ? > > Peter P. > > > -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist