RE: [SI-LIST] : Diodes, EMC, ESD Protection networks

Charles Hill ([email protected])
Mon, 11 May 1998 21:54:11 -0600

Peter,

I have a few suggestions for you.

----------
From: Peter Baxter[SMTP:[email protected]]
Sent: Monday, May 11, 1998 2:31 PM
To: [email protected]
Subject: [SI-LIST] : Diodes, EMC, ESD Protection networks

Hi,

Sorry if this is not strictly SI issues.

1) With regard to diodes in Switch Mode Power Supplies. Do Ultra Fast
Recovery Diodes produce more EMI than Schottky Diodes? I've heard
conflicting stories.

Fast recovery diodes have storage time where schottky diodes do not. The
snap off of the fast recovery diode can produce very high frequency energy
similar to a step recovery diode. The schottky diode can also produce high
frequency energy due to its fast switching time. Pay attention to the
currents in the circuit and try to keep the loop areas small. I would then
just experimentally try each diode and see.

2) I have a design in which a buffer IC, 74VHC574, is being clocked at
100MHz. It creates a lot of 2nd harmonic radiation at 200MHz which
radiates badly on all input pins (and probably output pins). I have a
cable from an external device comming in via a connector to all 8 of
the 74VHC574's input pins. As you can imagine, this cable radiates a
lot of EMI. My solution is to put a ferrite clamp on the cable which
works and I pass emissions testing.

Would running that track (2"//50mm) from the connector to the IC's
input pin as a strip-line (internal) make any difference over
micro-strip (surface) or is this a problem that can only be solved by
filtering?

I think it is necessary to have filtering at the connector either
differential or common-mode. I like the ferrite bead.

3) A Multiple VCC Supplies//Partial Power Down question. Think of it
as a Bus Analyser sitting on a CPU bus.

I have a CPU circuit (50MHz clocks...) and a monitoring Bus Analyser
(not a product as such but a circuit I am developing). If power is
applied to both devices, all is fine. However, if power to the Bus
Analyser fails (or is turned on after the CPU circuit is turned on),
it
then loads the data lines on the CPU circuit corrupting operation. The
loading is primarily done by the Bus Analyser's Input IC's ESD
protection network, clamping the input to 0.7 volts above VCC, which
in
this case is 0 volts. The Bus Analyser's power supply has a very low
DC
resistance when the power is switched off due to other things.

The only solutions I have seen (so far) is from Texas Instruments
Application Information.

A) Use large value series resistors in series with the inputs. I can't
put 4K7 resistors (for example) in series with the Bus Analyser's
inputs as that in conjunction with the ICs input capacitance/PCB
strays.. will kill high frequency response (at 50MHz).

B) Put a diode in series with the Bus Analyser's input IC power supply
pin. This then moves the problem from the input IC to the following IC
(IC's) in sequence after the input IC. The input IC is free to "float"
but is clamped by the following IC's ESD clamp diodes once again.

C) Adding a diode in series with the input (cathode on the CPU bus,
anode to the Bus Analyser's input pin) and a pull up resistor from
anode to VCC. This would need a fast/low capacitance schottky diode
(Vf=0.4volts) which is fine. I've tried this and it does work but is
a bit messy.

My question is:

There must be thousands of circuits out there doing exactly what I'm
trying to do. What do others do?

Some logic families such as 74ALS and 74F do not have a diode to VCC on the
input. So I suggest using a 74F245 to interface to bidirectional lines,
and if you want to limit the effects of the analyzer plugged into a
circuit, put in some small (33 ohm) series resistors in the lines. These
dampen the additional resonances introduced into the circuit by the shunt
capacitance of the Bus Analyzer's IC inputs.

Thanks

Peter Baxter
[email protected]

Charles Hill, consultant
[email protected]