EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: alank2 on September 04, 2014, 11:57:40 am
-
Hi,
I need to protect 4 data lines. They are not high speed. They are normally pulled high with a 20k resistor and switched to ground directly with a switch. vcc=5v. It looks like the top one has a tvs for vcc as well, and the bottom one lacks it so with the bottom one I'd implement a main tvs in addition. Which of these two do you like better?
SRDA05-4R2G
http://www.digikey.com/product-detail/en/SRDA05-4R2G/SRDA05-4R2GOSCT-ND/2705227 (http://www.digikey.com/product-detail/en/SRDA05-4R2G/SRDA05-4R2GOSCT-ND/2705227)
SMS05T1G
http://www.digikey.com/product-detail/en/SMS05T1G/SMS05T1GOSCT-ND/1139990 (http://www.digikey.com/product-detail/en/SMS05T1G/SMS05T1GOSCT-ND/1139990)
-
Right now all I get is
Server Error
404 - File or directory not found.
The resource you are looking for might have been removed, had its name changed, or is temporarily unavailable.
Your URLs are broken.
-
Sorry about that - fixed them.
-
I like the 4 channel integrated, but typically use the SOT-363 style, 6 pin device.
Your first one is higher rated than most I've used though. 10A vs 6A for ones from Diodes Inc, etc. It really depends on how much protection you need and if a $0.40 part that you can actually buy is better than a $1.20 part that Digikey has 0 stock of.
-
That is a good point about availability. What are your thoughts about the two types. One steers the signal between two diodes to a zener diode vs the other being 4 zener diodes... I hadn't look at the "steering" type before...
-
That is a good point about availability. What are your thoughts about the two types. One steers the signal between two diodes to a zener diode vs the other being 4 zener diodes... I hadn't look at the "steering" type before...
Steering TVS diodes is all I ever use. Is this a one off product or is cost going to be a factor?
I typically put on something like this: http://www.digikey.com/product-detail/en/DT2636-04S-7/DT2636-04S-7DICT-ND/4563484 (http://www.digikey.com/product-detail/en/DT2636-04S-7/DT2636-04S-7DICT-ND/4563484)
The chips are designed so you can hang them on the parallel run data lines, as close to the port as possible. In this instance, data lines for the other side of the chip run between the pins of the port side. The goal is minimizing inductance from the data line to the large ground plane connection.
Your data signal is low speed, but the ESD is very high speed. So you want the path through the TVS diodes to be as inviting as possible, compared to the rest of the board.
-
Steering TVS diodes is all I ever use. Is this a one off product or is cost going to be a factor?
Not a one off and also the first time I'm baking SMD. I've got lots of pcb space because of mechanical positions that need to be done so I prefer a larger smd package to a smaller one, but don't want to pay extra for the larger package,
The chips are designed so you can hang them on the parallel run data lines, as close to the port as possible. In this instance, data lines for the other side of the chip run between the pins of the port side. The goal is minimizing inductance from the data line to the large ground plane connection. Your data signal is low speed, but the ESD is very high speed. So you want the path through the TVS diodes to be as inviting as possible, compared to the rest of the board.
My input circuit runs through a 470ohm resistor to a pullup and this tvs and then through another 470ohm resistor to a uc pin. When you say make the tvs more inviting, should I use larger tracks to it than my normal ones (12mil)? If yes, just between the first 470 and the TVS with all the other tracks being normal size?
Thanks,
Alan
-
My input circuit runs through a 470ohm resistor to a pullup and this tvs and then through another 470ohm resistor to a uc pin. When you say make the tvs more inviting, should I use larger tracks to it than my normal ones (12mil)? If yes, just between the first 470 and the TVS with all the other tracks being normal size?
Two factors that influence inductance of a trace per unit length are width and height above ground plane (this is determined by substrate thickness, i.e. the fiberglass thickness between copper). That just influences the inductance per unit length.
Easy way to make inductance lower is to make the length as short as possible. Put the TVS package as close to the connection to the outside world as possible.
If you have the space for larger tracks to the TVS and them normal past, that isn't a bad thing. But a quick calc has going from 12 to 20 mil being 8.3 -> 6.1 nH/in. If you have traces that are 0.5in instead of 1.0in, you have inductance of 4.15 nH instead of 8.3 nH, a far larger factor than the trace width difference.
-
First, reduce the bandwidth as much as possible (RC or LC filter). Then add the TVS (the pin-resistor-TVS-resistor-MCU scheme you've selected is a good plan). Bandwidth and pin characteristics dictate what kind of TVS to use: clamp diodes, schottky, zener, etc.
ESD can be handled by capacitors alone: the most strenuous tests use 330pF at whatever voltage; to get that down to a less logic-exploding voltage, say 8V from 8kV, you simply need 1000 times more capacitance, or 330nF. Rather a lot, but quite acceptable for a slowly changing input.
Surges can't be handled by capacitors (not reasonably, anyway), so you really need TVSs for that. You're less likely to get a surge (8/20us or 10/1000us, with big amps) on a signal line, but more likely to get it on a power line (which is more likely to need to be well bypassed, like the above case, too!).
The SRDA device is made for medium to high speed communication: USB full speed, perhaps USB high speed as well (also PCIe, LVDS, HDMI, etc.). This will more than fully include anything your MCU can put on its pins, say if you wanted to make an SPI expansion port without restricting the maximum clock rate.
The SMS device has higher capacitance, so it already will serve some filtering. It's nonlinear capacitance, so you should want to swamp that with a bit more in parallel (say 470pF), so the attenuation is more even. For 470 ohms in series, that'd be a bit under 720kHz cutoff, which will cut well into the maximum transition rate from CMOS level signals (you aren't going to be running 10MHz SPI through this), but would be perfectly reasonable for lower bandwidth stuff (bit-banging, most serial comms say <= 230kbaud, etc.). It would also be good for analog inputs, if the source impedance is low (driving from an op-amp?).
If you have a supply handy, BAT54S (schottky: higher C, leakage) or BAV99 (PN junction: low C, leakage; higher Vf, exhibits forward recovery) are very commonly used for protection, clamping signals between +V/GND. Don't forget a bypass cap at the point of use. Discrete solutions have more inductance than monolithic (especially feedthrough) types, so some additional RC filtering can help keep things clean.
Also, for EMC, you'll do a lot of good just with the resistors and capacitors and diodes -- for lower impedance signals (logic outputs, analog I/O?), you may want ferrite beads as well, and inductors for supply pins (not FBs, they saturate with DC).
Tim