Exact match. Not showing close matches.
'[EE] Diode Turn-on Time (was MOSFET)'
|At 07:12 PM 12/29/2007, you wrote:
Jack- interesting tests... you're getting significant current through the
diodes too.. 60mA or so assuming a 50R source impedance.
Here's the original 1970s data on the fr times:
Note the two curves. IIRC the higher voltage diodes are actually PIN
diodes, although some companies have apparently sold higher voltage
diodes as lower voltage types.
In practice, the 1N400x diodes are known to be okay-- millions of them
are in use. It would be interesting to see the dv/dt at the relay coil--
given the distributed capacitance in the coil, I'd be surprised to see
more than a few volts per microsecond.
Spehro Pefhany --"it's the network..." "The Journey is the reward"
interlog.com Info for manufacturers: speffhttp://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Thanks for the reference curve--the durations are generally consistent
with my measured data, extrapolating to the lower current in my test.
What not stated in the curve is the magnitude of the "bump" in Vf during
the turn-on period. I see it as around 1.2 V for a 1N4007.
It's been widely known amongst ham radio types that the 1N4007 is quite
usable as a cheap PIN diode for RF power switching, and many
transceivers have used them quite successfully at the 100 watt output
power level, including current production equipment such Elecraft's
I'll measure a real PIN diode (low power receiving type) later today and
add that to my diode measurements page.
Spehro Pefhany wrote:
> Jack- interesting tests... you're getting significant current through the
> diodes too.. 60mA or so assuming a 50R source impedance.
Measured the forward current today with a current probe at 68 mA.
Spehro Pefhany <speff <at> interlog.com> writes:
> In practice, the 1N400x diodes are known to be okay-- millions of them
> are in use. It would be interesting to see the dv/dt at the relay coil--
> given the distributed capacitance in the coil, I'd be surprised to see
> more than a few volts per microsecond.
I think that 'okay' should be used in the past tense here. I used to use 1N400x
to catch relay coil kickback etc. In some recent circuits I had issues with EMI
caused by the kickback that was not caught properly by the 1N400x. In all cases
putting in a Shottky diode or a fast recovery rectifier fixed the problem. The
problem is caused by very fast switching (fet switch, fast (die shrunk?) bipolar
etc), stray relay coil inductance (not the main inductance/capacitance) and more
sensitive/faster microprocessors. The first problems were seen on an atmel
mcs51 seris + 2n7000 drivers + relays on 12V more than 2 years ago. Inconclusive
measurements (I do not have a 2GHz scope) showed peaks in excess of 5Vpp and
much less than 50 nsec wide induced into the power and ground rails from the
otherwise very short 'book move' pcb mounted relays and their suppressor diodes.
Such things will go 'through' any kind of discrete filtering components imho.
The best suppression method was to use a Schottky diode directly on the relay
and add a 0.01u ceramic (preferrably smd) cap directly at the coil. Of course
this becomes expensive with several channels. If the pulses got out onto the
power or ground rails they were measurable *everywhere* and caused sporadic
malfunction. The pulses can be seen as a hair thin very tall (!!) needle pulse
riding on the leading edge of the kickback catching pulse (which is indeed
1-1.5V tall for a 1N400x Vf under those conditions). I do not know if something
also changed in the way relays coils are wound. Similar problems were seen in
the use of low voltage rectifier bridges as suppressor diodes for unipolar
drives (dc motors, electromagnets etc). These used to work but no longer do so.
The symptoms are the same as with 1N400x. The bridges were low voltage 0.5-2A
types, e.g. SM02, B40C1000 etc. Simulation with spice does lead to some similar
results using a small stray inductance and a 1N400x diode.
|At 07:52 AM 12/31/2007, you wrote:
IMHO, in such cases the edge should be controlled as a first
the stray inductance between the driver and the load (even if the load is
purely resistive) can cause problems. As well, of course, as conservative
EMC measures taken on the circuit that could be susceptible to interference
(particularly overall layout and filtering on all i/o).
Driving a MOSFET gate too hard can yield all kinds of problems including
transients leading to latch-up of the driver because of inductance in the
source return. A relatively large gate resistor, for example, can make a world
But, is it real? I had an engineer swear that there was huge noise at the
output of his SMPS design and was about to redesign it.. turns out that 95%
of the spikes were artifacts caused by the measurement.
I think part of the problem is that many microcontrollers are not
to EMI (and becoming less so as the drivers become a larger part of
the chip cost
and dies shrink), and there are market pressures to do things
cheaply, which is not
usually consistent with the most effective EMC practices.
Spehro Pefhany wrote:
I agree. I had to redesign many products when HC components on the
products were changed
from a larger DIE size to a smaller one, ESPECIALLY devices from
Fairchild Semi. The switching
speed (and sensitivity to noise) dramatically changed. This was in the
Some of my "belt and suspenders" clamping is designed to stop
inductive spikes from getting into
the controller sections; I have to use a 1N4005 plus an SMT varistor
to stop everything cleanly.
In a recent design, I had to place many SMT varistors at inputs to gate
changes of HC components
that were being fed from a optoisolator thru a IDC26 cable; spikes from
the cable (26" long) were
destroying 74HC86s; to be even safer, I replaced them with a cable
receiver (TI DS14C89A) and
no more issues arose during active testing.
The PIC16C series was, frankly, built like a rock; the newer 'F' parts
are not nearly so durable.
On one version, I had to enclose the PIC in a metal shield to stop noise
from interfering; but now
that the cables are suppressed properly, the shield was no longer
needed. Never had to do that
before using PIC16C devices.
More... (looser matching)
- Last day of these posts
- In 2007
, 2008 only
- New search...