mosFETs and Transient Voltage Protection (schematic Uploaded)

[Falcon69]

I don't think I got my answer clearly in previous post i did on a different subject. (Although, I changed the design and went with a OptoMOS as suggested instead).

But in another circuit I am working on, I will be using n-channel mosFETs.  Now, in some chips I see dual zener (or maybe TVS) diodes connected between Source and Gate (see the second chip in the picture).  The ESD protection is 2kV for HBM and 200V for MM according to the data sheet on a particular chip I am looking at.  Now is that enough, or do I need to add a TVS Diode?  Basically, the gate would be on a connector that gets plugged in and out all the time with a 10k resistor in series (or would 47k be better?. The mosFET would be sinking approximately 80mA to ground.

On mosFETs that DON'T have them built it, I understand adding a TVS diode is a must (see the first chip in the picture).

FYI, the voltage source is 5.2volts.

Also, in the same circuit, I have several logic gates.  Many only have a connection to ground, or to supply, but never both at same time on a single input of a 2-input logic gate. Now, if that input is ONLY connected to ground or supply, and nothing else, and the supply voltage is less than 5.5v, is there any reason to put a resistor in series between the ground/supply and the input? Would it be good practice to put a 100k or so resistor there just in case to help lower any current into the chip?

Thanks in advance.
Jason

[T3sl4co1l]

What's the rest of the circuit?

And, are you expecting ESD to be applied to the relevant nodes?

Also, in the same circuit, I have several logic gates.  Many only have a connection to ground, or to supply, but never both at same time on a single input of a 2-input logic gate. Now, if that input is ONLY connected to ground or supply, and nothing else, and the supply voltage is less than 5.5v, is there any reason to put a resistor in series between the ground/supply and the input? Would it be good practice to put a 100k or so resistor there just in case to help lower any current into the chip?

Certain families may require input resistors or special voltages (ECL for example).  The common stuff (CMOS, TTL...) can't care less.

You might have external reason to do so; it's good practice to put a small e.g. 100R-10k resistor at each input or output, so you can perform functional testing by forcing test points high or low.  Or even just 0R resistors, so you have the option of reconfiguring them, with some pain (oh noes, the blue wirewrap wire!).  More of a prototype / production thing though.

Tim

[Falcon69]

i dont think any esd on the drain/source.  just on the gate, as it gets plugged/unplugged.  The drain and source are stay connected on the circuit board to sink the LEDs to ground.
Yes, I will be using CMOS.  I'll go ahead and take them out then, Seems like extra $ to keep them there if not needed.

Thank you Teslacoil.

[poorchava]

gate oxide as such is very thin, and in small mosfets the gate has very small capacitance, so any voltage spike during handling or operation is a killer. This is even more true for low threshold mosfets, which sometimes have Vgsmax as low as 7V.

In bigger mosfets, that have typical 20V of Vgsmax and like a nanofarad or more of gate capacitance - this is not much of a big deal unless you really expect the gate to be exposed to external ESD events.

[Falcon69]

Okay, I understand, but original question still lingers.  This mosFET, http://www.mouser.com/ds/2/308/NTJD5121N-D-119599.pdf, for example has the protection diodes, but only protects up to 2kV HBM and 200V MM for ESD.  Would I need an external TVS diode as well with this chip?  Only the gate has a possibility of transient voltages, as the gate would be plugged/unplugged multiple times.

What ESD for HBM and MM is acceptable protection for normal operating conditions?

[T3sl4co1l]

Okay, I understand, but original question still lingers.
...
What ESD for HBM and MM is acceptable protection for normal operating conditions?

The question still lingers: What's the rest of the circuit? And, are you expecting ESD to be applied to the relevant nodes?

Or rather I should now say, *how much* ESD are you expecting, and where?

What a product needs for ratings varies by market and intent.  Automotive is different from consumer, which is different from industrial.  If you need reliability in general purpose or commercial / industrial settings, it would be a good idea to pursue something like IEC 61000-4-2, 8kV contact / 15kV air discharge.  The air discharge figure is very roughly equivalent to an HBM rating, IIRC, so a device rated for 2kV HBM might be 1kV with the contact method and test setup of this standard.

Needless to say, the zeners in those FETs are woefully inadequate for serious ESD -- they're mainly a handling precaution, and usually fall somewhere between pointless and detrimental in actual use (anything facing the outside world needs real protection anyway; a case where they're detrimental might be a low-bias amplifier where gate leakage is critical).

Tim

[Falcon69]

okay, that's what I needed to know.

I'm confused by what you mean nodes.

I will add a TVS diode to the circuit.  My circuit is used for Hobby CNC machines, and the plugging in and unplugging of switches I'm worried about.

[T3sl4co1l]

Nodes or nets, are sets of component pins which are connected together.  So, wires, lines, connections, traces, etc..

Tim

[Falcon69]

oh, no.  Not unless they are touched directly.  Mainly, I have switches wired to USB connectors.  The connectors will most likely be heatshrink tubed or using plastic casing around the USB.  I was just concerned with the plugging and unplugging of those connections. 1 pin on those supplies the power to the gate of the mosFET to turn it on.  If no switch plugged in, mosFet does not turn on, thus disabling that part of the circuit on the board.

[T3sl4co1l]

The procedure of an ESD test might not catch that (discharge is applied to any exposed metal surface -- a mated, shielded connector has no risk of getting struck), but a lower level test would still be helpful.

For example, suppose you plug in a cable that's sitting around, open circuit, that somehow accumulated a static charge: now you have charged metal wires plugged directly into your circuit.

Possible causes could be careless handling (I suppose it's possible someone could shuffle their feet over the carpet, then touch an inner wire or pin, maliciously or accidentally), ambient ionization, piezoelectricity (voltage from strain) or triboelectricity (voltage from friction) due to outside handling and movement, or strain and movement of the wires within the cable itself.

You might also consider something that's naturally more robust.  A 2N4401 might help -- I don't think I've seen ESD ratings on BJTs before, but as a PN junction device, they at least exhibit controlled (if not necessarily non-destructive) breakdown at high voltages.  With the MOS gate, voltage goes too high and, pfft, it's done, single shot, no warning, end of story.  Yes the internal protection thing helps with that, but obviously, not much.  Whereas a BJT might actually survive a few strikes.

Just as a matter of practice, I put ESD protection and EMC filtering on all connections to the outside world.

I have no idea what kind of requirements you have, for this specific connection or for the overall system, so I can't really be any more helpful..

Tim

[Falcon69]

So here's 1/8 of the circuit I am working on, meaning, there's 8 total switches.

The circuits in the blue box are the switch.  The rest is the main circuit (some of the circuit is shared by the other 7 switches).

As you can see, the stuff in the blue box has 5 connections that get plugged/unplugged from the main circuit. Voltage In, Ground, Signal to Comparator, Signal from Test Switch (which turns on all the LED's to test), and Signal to the mosFET in the Red Box.  That signal wire let's the main circuit know if a switch is plugged in or not, thus turns on or off the led's on the main circuit board based on whether or not a switch is plugged in.

The other mosFET is actually an optoMOS. From my understanding, I do not need any ESD protection on that.

The circuit is 5v-5.2v (i will be using 12v going into a LDO that puts out 5.2v.  5.2v to compensate for voltage drops of wire length to the switches.

So, My question is, this circuit, should I have a TVS diode for each connection to the switch?  Or just on that mosFET at the bottom right, in the red box?

If it is just the mosFET that needs it, seeing how the source goes to ground, would this TVS Diode Array work for it (and the other 7 switches as well)? http://www.mouser.com/ds/2/308/ESD7008-D-241517.pdf

If I need ESD Protection for EACH connection of the switch, then would this TVS Diode work for it? http://www.mouser.com/ds/2/115/D5V0L4B5V-218334.pdf ,and would I place it on the switch or the main circuit board?

Thanks in Advance

[Falcon69]

nobody has input on those TVS diodes?  Or suggest a better method?

[T3sl4co1l]

As you can see, the stuff in the blue box has 5 connections that get plugged/unplugged from the main circuit. Voltage In, Ground, Signal to Comparator, Signal from Test Switch (which turns on all the LED's to test), and Signal to the mosFET in the Red Box.  That signal wire let's the main circuit know if a switch is plugged in or not, thus turns on or off the led's on the main circuit board based on whether or not a switch is plugged in.

Ok, I think I understand what you've got here.

And, are you really actually using (/ trying ) Falstead for drafting?!  That's at least moderately insane to me, but, I give you credit for trying, I guess...

Anyway, a couple things happen when you plug that in:
- As soon as +5V and GND make connection, C1 starts at 0V but there's 5V somewhere else in the circuit.  The inductance of the lines (cable and connections) begins to charge, and current rises.  As current rises, C1 voltage starts to rise.  By the time C1 reaches 5V, current is at the maximum (I ~= 5V / sqrt(L/C) ), and C1 voltage continues to rise.  If C1 remains constant (e.g., a 0.1uF 50V 0805 X7R ceramic, or C0G, or film), it reaches no more than 10V peak, then rings down.  However, a low voltage / small size ceramic will experience a reduction in capacitance at higher voltages, leading to a peak higher than 10V.

Neither case is probably very good for the Hall effect switch (if it's a switch, do you really need a comparator after it by the way?) or logic gates (whatever they are; they're unmarked).

So, a 5V TVS across the supply would be helpful.  You can also add series resistance anywhere along the line (at the 'base' connector or on the attached thingy) to dampen the LC resonance and make it charge more smoothly.  Probably 10 ohms would be enough not to notice the additional drop, while damping most of this junk.  Also, if you put it on the base end, it provides short circuit or fusible protection in case some idiot starts shoving paperclips into your connector.

- The logic input is only pulled low with 100k, so it will be very noise sensitive, and the inputs are CMOS logic inputs (I guess), so they won't be the most robust.  You could get away with this as-is, but I'd rather have some input clamp diodes (e.g., a BAT54S from GND to input to +5V).  The hazard is against the chips' own input protection diodes (which are small), so although you could use a TVS here, it won't do much good against a spike if the 5V hasn't come up yet (which is likely to happen; during insertion, pins connect and disconnect and reconnect multiple times over a 1-10ms time span, probably in random order -- so, expect anything).  In that situation, the fact that C1 is still around 0V helps you, since there's nowhere for the input voltage to go except through one diode or another into ~0V!

- The sense / gate output thingy doesn't need to do anything here, but the same logic will apply to that line as well; in that case, the spike will be transmitted down the cable, then back up for a round trip.  You could dampen it with a 1k resistor or something like that.

- So, the FET in the red box will see all the ugliness that's around the wires.  It needs filtering and protection, I'd say.

- Also, any time your comparator is near threshold, it's exposed to, and amplifying, whatever noise is on the cable.

If it is just the mosFET that needs it, seeing how the source goes to ground, would this TVS Diode Array work for it (and the other 7 switches as well)? http://www.mouser.com/ds/2/308/ESD7008-D-241517.pdf

Well.. you could... but do you really want to deal with a grain-of-sand DFN / LGA thing?

You certainly don't need low capacitance here, and you probably don't want it.  You could use bog standard 1N4733 / 1N5232 zeners (1 / 0.5W, 5.1V), which aren't really specified for ESD, but it's a start.  Or you can use a proper device, like, I don't know any really tiny parts offhand, but the big ones start in e.g. the SMAJxxA series (equivalent to a through-hole P6KExx).

If you add, say, 1k series resistance to each logic input, followed by an optional capacitor of maybe 1nF or more to ground (I'm guessing you don't need any of these LEDs blinking in the MHz), and either a TVS to ground, or clamping diodes to the nearest (locally bypassed) supply, you'll have a good handle on both ESD and EMC (you won't be picking up stray radio waves).  This goes for logic outputs as well, though you can't necessarily tolerate the series resistance (sometimes even 10 ohms is too much), in which case a ferrite bead or (if it's a high current signal or supply) inductor is more useful.

Tim

[Falcon69]

Here is the schematic I am working on. It's BIG. Very Big.  But hopefully you can see what I am trying to achieve.

I have it working, I am just not sure how to go about protecting it from ESD or EMF.  Tim pointed out some good tips, but to put a series resistor and cap on each input of each logic gate is going to be a HUGE PCB footprint. 

As you can see, I already have a decoupling cap on each Logic IC, and I've taken measures to protect the jumpers for the normally open/normally closed input (sinking the gates to ground to perform that operation by plugging in the jumper). 

There are 8 Circuit Boards for the Switches, 1 Circuit Board for the Rear Panel, 1 Circuit Board for the front Panel LED Indication, and 1 Circuit Board for the Mainboard.....so a total of 11 separate circuit boards for this project.

I need to protect the circuit when the switches are plugged in and unplugged. The switches are in the upper right corner.  As you can see, each has 5 connections to the rear panel board. I was hoping that all i needed was maybe that one chip I linked above that would protect all of the mosFETs.



EDIT: Added Ref Numbering to the components.

[Falcon69]

I found another circuit. I redrew it though in a way I think I understand.

But, If this is what is needed on each input of each logic gate and mosfet, it's going to be expensive. As you can see from the schematic of my circuit, i have over 50 IC's, and several of them have 6 inverters/quad logic gates.

Maybe I am over thinking this or just don't understand this ESD/EMF/Transient Voltage Suppression.  I just want the circuit protected from damage from these things if it happens.

I've spent all day trying to understand what Tim said.

The circuit below shows MOV's (Metal Oxide Varistor), TBU's (Transient Blocking Unit), Resistors, and TVS (Transient Voltage Suppression) Diodes.

The Cap is a 0.1uF Ceramic type 16volt.  I'm not sure on the values of the rest of the stuff.

[Falcon69]

What if I put a Uni-Directional TVS Diode on each of the V-In, V-Out, S-Out, and EN pins of the switches (upper right corner of schematic)?  Would that be enough to protect the circuit from plugging and unplugging of the switches? I could use this one maybe to do that?  http://www.mouser.com/ds/2/115/DRTR5V0U4S-271613.pdf

The other connections will probably never be plugged in and unplugged once they are set-up. (i.e. 28pin, 20pin, and 8 pin connectors in the center of schematic).

[Falcon69]

Any Thoughts on that?  The chip is already getting pretty expensive. I can't afford to put the recommendation that Tim said on EACH and Every Input of the logic gates. Would the idea in the aabove me sufficient enough?

[T3sl4co1l]

Yeah, that looks nice: the zener protects +5V/GND, and the clamp diodes take care of the rest.  Plus, the loop between any given input clamped to power/ground, or to a TVS, occurs within the device itself, so there should be little overshoot.

I would still recommend a series resistor on the supply rail, or a larger zener/TVS (1N47xx or SMAJ range).  And then you'll be pretty much set.

What's expensive?  BAT54S is like pennies in quantity.

Regarding the circuit you found/drew: the input MOV stuff is unnecessary.  Just have some sort of protection at the input pin itself (clamp diodes or TVS), followed by one series resistor, going to all the logic pins.

That way, any overshoot or voltage drop in the clamp/TVS is dropped across the resistor into the logic pin(s), limiting the peak current by orders of magnitude relative to a naked input, and still by many times if they were just connected straight (a low-ohm resistor, or none at all).

Tim

[Falcon69]

Okay, so If I use that DRTR..... chip on each of the connection pins of the switch circuit board...ON THE MAINBOARD (not on the switch board, or should I put one on both?)....and all my input connections to the logic gates (i.e. A and B inputs) put a resistor in series, that should do it (or does a TVS need to be on each input as well, in addition to the series resistor)? What size resistor? 10k? 4K7? I remember you mentioned 10ohm.

And put the resistors on the logic gates on the other circuit boards as well?  Or just a single TVS Diode across the input of supply and ground just after the fuse?

By doing this, will it mess with the decoupling cap I have on each IC?

FYI, Just got done reading this. http://www.onsemi.com/pub_link/Collateral/TND335-D.PDF Most of it made no sense to me, but some does.  They have used a USB 2.0 device as an example, and in my case, My circuit uses the same voltage.  So, According to that, If I understood it correctly, I should choose a TVS Diode that is no more then 7V Vrwm and 7V Vbr.  The Logic Chips have a max operating voltage of 7V.  Since my circuit is about 5.2v, choosing a TVS Diode with an operating voltage of 5.5V and a Reverse Breakdown Voltage of 6V should be exactly what I need.  I am not sure on the Amperage rating though, I am confused on that one. My circuit, when the test led is switched and all LEDs are on, will only see about 2A, just alittle less then that.

You said a larger TVS on the supply line.  Does that go right before the fuse (or after) on the mainboard? I'm guessing if it does, a Bi-Directional one?  Same 5.5V Vrwm and 6V Vbr? What Amperage rating you think would be good?

What's Expensive?  If I went with that circuit I drew with the TBUs and MOVs plus the TVS....to put that on EVERY input of each logic gate would be VERY expensive.  It all adds up.

[Falcon69]

Tim,

You mentioned earlier about the comparator,  I have it there because I am concerned with the signal getting to close to the threshold and the gate following it may or may not switch to a 0 or 1 when it should.  I thought that adding the comparator may solve that problem if the wiring from the mainboard, along the machines cable tracks, to the switch point is too long.  If It is not needed, I will take it out.  The max distance the wire could be is 50 feet.  This is dependent on the size of the machine.  The machine I am building currently is a 4x8 table, but in the future, I;d like to build a 5x12, and in that case, some wiring will be 50 feet. So I wanted to plan for a voltage drop.  Also, I was told that a comparator helps take out any noise in the wiring, which would help with switching.

If You think it is unnecessary, I will take it out, that would save a lot of board space if it is redundant.

[T3sl4co1l]

Okay, so If I use that DRTR..... chip on each of the connection pins of the switch circuit board...ON THE MAINBOARD (not on the switch board, or should I put one on both?)..

No, on any input, regardless of side.  And probably outputs too.

Your circuit showing the switchy bits being right up next to the mainboardy bits belies the fact that there's one to a hundred meters (dunno, you didn't specify) of cable picking up nasty environmental signals and static between the two.  Two ends of a given wire are NOT the same connection!

..and all my input connections to the logic gates (i.e. A and B inputs) put a resistor in series, that should do it (or does a TVS need to be on each input as well, in addition to the series resistor)? What size resistor? 10k? 4K7? I remember you mentioned 10ohm.

Yeah, I mentioned 10 ohm... for damping the 5V output to the switchy bit.

100 ohms might be typical for a logic output pin.  1k-100k might be fine for a logic input.  Depends, obviously, but your application doesn't look too critical.  This is in line with what's on your comparator, for instance (10-100k resistors).

And put the resistors on the logic gates on the other circuit boards as well?  Or just a single TVS Diode across the input of supply and ground just after the fuse?

A resistor per gate is not necessary.  The goal is to filter / absorb all the crap that happens at the connector.  Once it's on the board and protected and filtered, it can be considered 'sanitized' for use.

Like filtering database inputs so this doesn't happen.  But electrically.

By doing this, will it mess with the decoupling cap I have on each IC?

No, in fact the closer they all are together (TVS array, bypass cap and IC), the better it will work.

FYI, Just got done reading this. http://www.onsemi.com/pub_link/Collateral/TND335-D.PDF Most of it made no sense to me, but some does.  They have used a USB 2.0 device as an example, and in my case, My circuit uses the same voltage.

No, your circuit uses the same supply voltage, not the same signaling.  USB2.0 is a low voltage differential, high bandwidth signalling method and cannot tolerate filtering for ESD or EMC purposes.  Your application has full 5V CMOS logic levels and, as far as I know, couldn't care less about bandwidth.

So, According to that, If I understood it correctly, I should choose a TVS Diode that is no more then 7V Vrwm and 7V Vbr.  The Logic Chips have a max operating voltage of 7V.  Since my circuit is about 5.2v, choosing a TVS Diode with an operating voltage of 5.5V and a Reverse Breakdown Voltage of 6V should be exactly what I need.

This logic is still correct, however.  This is in regards to the TVS diode on the supply rail.  If you are using diode clamps (or the kind like that array device, which contains clamp diodes), this is where ESD will be shunted to, so you need a zener / TVS / breakdown device to prevent the voltage from going crazy.  The array device contains a small one, but it wouldn't hurt to add e.g. an SMAJ5.0A somewhere as well.  The location isn't very critical, but should be near the location of greatest stress (connectors, power input, who knows?).

I am not sure on the Amperage rating though, I am confused on that one. My circuit, when the test led is switched and all LEDs are on, will only see about 2A, just alittle less then that.

Transient amperage is unrelated to power consumption, it's an external source.

ESD typically applies peaks up to 16A, but only for a ~50ns pulse.  The charge is small, so it's easily clamped by sufficiently large capacitors (on the order of 1uF or so), and whatever's left can be absorbed by a TVS.  The downside is, it has a very fast risetime, so you can easily get pulses over 200V (for just a few nanoseconds) sneaking past even a rather burly TVS directly at the input.  You generally want to bypass the input clamp diodes (+5V to GND) at point-of-use for best results.

You said a larger TVS on the supply line.  Does that go right before the fuse (or after) on the mainboard? I'm guessing if it does, a Bi-Directional one?  Same 5.5V Vrwm and 6V Vbr? What Amperage rating you think would be good?

Fuse, as in power input?  Or was there a fuse going out to the switchy bits too..?

You always want a TVS across the circuit, and on the load side of a fuse, so that 1. surges are able to blow the fuse, 2. when the fuse opens up, no transients get into your circuit, and 3. your circuit is clamped and all voltages well-defined, all by itself, whether there are cables into and out of your circuit or not, whether the fuse is blown or not.

What's Expensive?  If I went with that circuit I drew with the TBUs and MOVs plus the TVS....to put that on EVERY input of each logic gate would be VERY expensive.  It all adds up.

If you're working down to every last cent, you'll have to have your own introspective discussion regarding reliability versus cost.  I can tell you the best way to build something, whether you want to follow is up to you...

Tim

[Falcon69]

Okay,

The schematic I know is alittle confusing at the switch points.  Basically, the 5 conductor wire will be soldered to the little circuit boards for the switch, then on the other end (could be up to 50 feet of cable) soldered to a Mini-USB 5 pin plug.  That will get plugged into the Rear Panel Circuit Board.  Then, from there, via a 20 pin and a 8 pin connector, via ribbon cable, the wiring goes to the mainboard via the same type of connectors.

The input for the supply and ground (bottom left of schematic) will be done via a 2pin Screw Terminal.  The other Screw terminals (8 of them) are for the switch outputs, via the OptoMOS.

So the switches themselves only have one connector that is plugged/unplugged.
SO, Even if the output of one logic gate goes to the input of another logic gate, still need to put a TVS/Series resistor between them?  Each Input?  Or only need to go for inputs that aren't a direct connect to another gate/inverter?

So, for instance, the 1a-Y Xor Gate to the 1b-A XOR Gate, there's a 360 resistor going to the OptoMOS, and a 100k pull-down.  That circuit/trace would need a TVS AND a series resistor between the two Gates?  But, from 1i-Y to 1a-A, no need, because it's a direct connection with nothing between those two logic devices?

Still kinda confused on what needs.

If you say EVERY input needs 100k in series with a TVS, and EVERY output needs a 100ohm and TVS, regardless of connection, then I understand and will draw it up that way.  However, for the Gates that have the output going to a resistor and LED, would I need another series resistor AND the TVS as well??

[Falcon69]

And put the resistors on the logic gates on the other circuit boards as well?  Or just a single TVS Diode across the input of supply and ground just after the fuse?

A resistor per gate is not necessary.  The goal is to filter / absorb all the crap that happens at the connector.  Once it's on the board and protected and filtered, it can be considered 'sanitized' for use.

Think I missed that part. SO only the items that are actually connected to anything that is plugged/unplugged needs to be protected.  But what about the connectors? the 8 pin, 20 pin, and 28pin?  I don't think those will EVER be plugged/unplugged.  Just the initial plug in when setting them into the control box.  Once they are in there, they won't be unplugged.  The 2pin screw terminals may be plugged, unplugged, but the optoMOS protects those, correct?

[T3sl4co1l]

SO, Even if the output of one logic gate goes to the input of another logic gate, still need to put a TVS/Series resistor between them?

Only where "circuit stuff" connects to "cable stuff".

So, for instance, the 1a-Y Xor Gate to the 1b-A XOR Gate, there's a 360 resistor going to the OptoMOS, and a 100k pull-down.  That circuit/trace would need a TVS AND a series resistor between the two Gates?  But, from 1i-Y to 1a-A, no need, because it's a direct connection with nothing between those two logic devices?

Still kinda confused on what needs.

When "circuit stuff" connects to "circuit stuff" you don't need to worry (unless you're expecting dirty static-charged beasts to lick your board in normal operation).

If you say EVERY input needs 100k in series with a TVS, and EVERY output needs a 100ohm and TVS, regardless of connection, then I understand and will draw it up that way.  However, for the Gates that have the output going to a resistor and LED, would I need another series resistor AND the TVS as well??

If the LED is on a cable, perhaps.

If the LED is mounted near a metal enclosure, doubtful.

The operational case would be, if the LED is close enough to the surface that a spark can jump to one of its leads, it might be a hazard.  But most holders are deep enough that that's extremely unlikely.  (The IEC 61000-4-5 air discharge test waves up to 15kV static over the surface, and zaps whatever it can find.  If it's all good solid plastic, no jumps means you're probably not going to even notice it going on.)

As for the full schematic... that's... 20 logic chips?  And a, what the hell, like 50 inch sheet of paper?  My browser viewer doesn't even want to zoom out far enough to see the whole thing!    This is in serious need of some logic reduction.  I think I was seeing you could dump about half of them with some simple rearrangements..

At the power input, J9: you have a somewhat vulnerable semiconductor, Q1.  A bidirectional TVS from its drain to ground (i.e., just after the fuse), rated for at most 80% of Q1 Vdss, would help it survive.  TVSs are usually rated to fail shorted, so although you're still not guaranteed anything (maybe it'll physically blow out anyway?), it's at least more likely that it fails shorted (reducing the voltage on the circuit even further) and takes the fuse out, making for an easy repair.  U4 is rated 35V, so it would make sense to use Vdss >= 35V for Q1, and, say, a 24V TVS.  Like an SMAJ24CA (the -CA parts are bidirectional).

The OptoMOS has isolation so you don't need to worry about anything on the LED (driving) side.  You will have to be weary that anything connected to J2-J9 is limited to 60V (transient or otherwise) and 100mA.  Resistive loads up to 48V are fine, or even moderately inductive loads.  I wouldn't recommend relay or solenoid coils, at least at full voltage.  Maybe derating to 24V or less would be better.  You can address that by adding a 48V bidirectional TVS across each switch output.  However, protecting them against current overload would be much more difficult.

The isolation barrier is only 1500V or so, which isn't much against say 15kV of ESD (if they can be exposed to such, and need to withstand it).  One simple way to address that: use a 1nF "Y1" style capacitor to sink the ESD charge.  This is how Ethernet maintains an isolated communication channel, yet passes ESD with only a 1kV barrier -- part of the reason it's so popular, reliable and powerful, despite its otherwise difficulty to implement and use.

Also, what good is the 2.2k pullup to +5 if the switch contacts are otherwise floating?  Where's the ground reference?  And if one side is intended to be grounded (I don't know), why use SSRs at all, why not an "open collector" (or open drain) type transistor output?

Connectors:
- Board to board connectors are generally considered OK.  J20, J21, J23, J24, J26, J28: no need to screw with.  (I will note: it's probably better to put the series resistors for the LEDs near the source pins, rather than on the display board.  Transmission line stuff.)
- Series 10 ohm resistors: J32, J36, etc. V_IN pin.
- All the TVS stuff you have a choice on: probably best to put it on the J32, J36, etc. inputs (and outputs where applicable), as well as J29, J33, etc. (same idea).  Or instead of J23, J36, etc., you can push it up to the main board, on J19 / J20.1.

I've got to admit, as schematics go, I've seen some bad ones, but as far as human readability, I think this one takes the cake -- it's quite sufficiently horrible to try to follow any single line, there are no ground or supply symbols, and despite that, everything is neatly laid out!  Seems like it was autogenerated from a netlist or something.  Weird.

Tim

[Falcon69]

great! you've been a great help Tim Thank You.  I understand now.

And yes, the schematic is big. I need to work on it.  It was all hand done, no netlist or anything.  I did it this way with no ground or voltage reference so that I can "convert to PCB' and it all goes over without a problem.


I will have a look and see if I can reduce some of those logic gates.  I think you are right and I can. 

[Falcon69]

Well, Been playing around with it all day.

If I remove one logic gate, it effects an LED. Or it effects the LED Test switch (all LEDs wont turn on, etc.).  In fact, I think If I add an extra inverter on each switch, at V-Out to 1d-B and eliminate the 100k and 10k resistor there, I would get a true 5v/0v input to that logic gate 1d.  For example.  Otherwise, I was getting like 3.18V/390mV.  From what I read, it is better to be below 1volt and above 4.5 volts on logic gates to keep them cooler and better.

So, I added another inverter, but at least I got rid of 2 Resistors, but, It will cost slightly more for the inverter than it did for 2 resistors.  I think it will be better overall though.

[Falcon69]

As for the full schematic... that's... 20 logic chips?  And a, what the hell, like 50 inch sheet of paper?  My browser viewer doesn't even want to zoom out far enough to see the whole thing!    This is in serious need of some logic reduction.  I think I was seeing you could dump about half of them with some simple rearrangements..

Yup, I know it's big.  I will cut it down later into several sheets if need be.  I tried lowering the number of Logic Chips, bur ended up making the circuit not perform the way I want.  It that Test LED switch that just messes up everything.  I want all LED's to turn on when it is pressed, regardless whether or not the SS441A has been tripped.  I think that is important for visual indication. If something is wrong, a Red LED should light on the front panel, if that LED is out, you won't know if anything is wrong or not.

At the power input, J9: you have a somewhat vulnerable semiconductor, Q1.  A bidirectional TVS from its drain to ground (i.e., just after the fuse), rated for at most 80% of Q1 Vdss, would help it survive.  TVSs are usually rated to fail shorted, so although you're still not guaranteed anything (maybe it'll physically blow out anyway?), it's at least more likely that it fails shorted (reducing the voltage on the circuit even further) and takes the fuse out, making for an easy repair.  U4 is rated 35V, so it would make sense to use Vdss >= 35V for Q1, and, say, a 24V TVS.  Like an SMAJ24CA (the -CA parts are bidirectional).

Fixed, I think.

The OptoMOS has isolation so you don't need to worry about anything on the LED (driving) side.  You will have to be weary that anything connected to J2-J9 is limited to 60V (transient or otherwise) and 100mA.  Resistive loads up to 48V are fine, or even moderately inductive loads.  I wouldn't recommend relay or solenoid coils, at least at full voltage.  Maybe derating to 24V or less would be better.  You can address that by adding a 48V bidirectional TVS across each switch output.  However, protecting them against current overload would be much more difficult.

The isolation barrier is only 1500V or so, which isn't much against say 15kV of ESD (if they can be exposed to such, and need to withstand it).  One simple way to address that: use a 1nF "Y1" style capacitor to sink the ESD charge.  This is how Ethernet maintains an isolated communication channel, yet passes ESD with only a 1kV barrier -- part of the reason it's so popular, reliable and powerful, despite its otherwise difficulty to implement and use.

Do you mean one of these on the pins 3 or 4 of the OptoMOSs? http://www.mouser.com/ProductDetail/EPCOS/B81123C1102M/?qs=%2fha2pyFadugX8qgvXXfJKysWqbQK%2fDgDcnK0zaeiQTk%3d or http://psearch.murata.com/capacitor/product/DE2E3KY102MB3BM02.pdf

If so, where exactly does it go? On pin 4 then to ground?

Also, what good is the 2.2k pullup to +5 if the switch contacts are otherwise floating?  Where's the ground reference?  And if one side is intended to be grounded (I don't know), why use SSRs at all, why not an "open collector" (or open drain) type transistor output?

The 2.2k Pull-up to 5 volts is if the circuit of a specific break-out-board needs to be pulled up to 5volts. Some break-out-boards need to be pulled up to 5 volts, some need to sink to ground.  In the case that it needs to be pulled up to 5 volts, then pin 1 of the screw terminal would not be connected.To connect, a jumper would be plugged in for the 2.2k pull-up.

Connectors:
- Board to board connectors are generally considered OK.  J20, J21, J23, J24, J26, J28: no need to screw with.  (I will note: it's probably better to put the series resistors for the LEDs near the source pins, rather than on the display board.  Transmission line stuff.)

That front panel LED is a direct connect to the mainboard via it's connector. There is NO wire or cable separating them. The connectors are there to just stand the LED board up off the mainboard, to allow more space and smaller overall size of the mainboard.

- Series 10 ohm resistors: J32, J36, etc. V_IN pin.
- All the TVS stuff you have a choice on: probably best to put it on the J32, J36, etc. inputs (and outputs where applicable), as well as J29, J33, etc. (same idea).  Or instead of J23, J36, etc., you can push it up to the main board, on J19 / J20.1.

I can do that, though, there is extra space on the rear panel board, so I'll put them there.  That Mainboard is going to be tight. 100ohm on each V-In on jumpers J32....etc.

I've got to admit, as schematics go, I've seen some bad ones, but as far as human readability, I think this one takes the cake -- it's quite sufficiently horrible to try to follow any single line, there are no ground or supply symbols, and despite that, everything is neatly laid out!  Seems like it was auto generated from a net list or something.  Weird.

No, no net list generator. I hand drew everything, copy and paste helped.  I left off the ground and Vcc symbols.  I connected it all with ratlines instead. For me it's easier to understand, because when I mouse over the line, it highlights everything it is connected to.  I think it does it as well if I used the ground and Vcc symbols.  I may redo the schematic later and fix it.  Don't open it with the browser, use the free Adobe Reader and you can see it all, and zoom-in/out.

Almost there, getting excited. I've work for about 2 years trying to do this thing.  I've had like 5 different circuit boards made, and have had problems with them all.  The last one worked great!  Except, that when the mainboard loses power, it did not shut the machine off when it tripped the switch. That was bad news. But as long as the mainboard still had power, it worked perfect.  That board I was using 3x ULN2803's.  This board is more complicated, but I've added extra features, Like ESD protection, Normally Open or Normally Closed switch selection, and a couple other things.  I plan on building a few machines, including my own pick and place machine and 3d printing machine, and these switches will come in very handy when finished I think.

EDIT: Added updated Schematic with fixes Tim suggested....I think
All except for the Y1 capacitors, not sure about those or where they go yet.

[T3sl4co1l]

Yup, I know it's big.  I will cut it down later into several sheets if need be.  I tried lowering the number of Logic Chips, bur ended up making the circuit not perform the way I want.  It that Test LED switch that just messes up everything.  I want all LED's to turn on when it is pressed, regardless whether or not the SS441A has been tripped.  I think that is important for visual indication. If something is wrong, a Red LED should light on the front panel, if that LED is out, you won't know if anything is wrong or not.

I don't really know what all the XORs and stuff are going on about, but as it's an unusual logic function (important when needed, but not often), it's very likely you're doing something very strange, i.e., the hard way, and just not seeing the clear path.

One trick that helps immensely with optimizing glue logic is diode gates and, as they call it in AoE2, "Mickey Mouse Logic", i.e., a little transistor here or there as needed, to provide an invert or buffer or wired-OR or whatever sort of function.

All of your LEDs, for example, can be diode-OR'd with a diode pair per LED, one coming from the true logic signal, one coming from the TEST node, and the cathodes going to the LED.

Note that all your LED cathodes need a test function too (U24, etc.).  I didn't look if they are ultimately wired into the TEST circuit or not.

You can wire more transistors in parallel to do that, or you can pull down the transistor drain pins with yet another diode (per pin), pulled by a single transistor (which may need to be larger, I didn't check the spec of what you're using, or how many LEDs that comes to).

Using diodes, at least, eliminates...all?  Or almost all your OR gates?  And distributes them among the LEDs, so you don't have a bazillion wires running around.

Do you mean one of these on the pins 3 or 4 of the OptoMOSs? http://www.mouser.com/ProductDetail/EPCOS/B81123C1102M/?qs=%2fha2pyFadugX8qgvXXfJKysWqbQK%2fDgDcnK0zaeiQTk%3d or http://psearch.murata.com/capacitor/product/DE2E3KY102MB3BM02.pdf

If so, where exactly does it go? On pin 4 then to ground?

Yes, that will work.  More traditionally ceramic, I think, like: http://www.digikey.com/product-detail/en/VY1102M35Y5UQ63V0/BC2374-ND/1983413
If it says Y1, it'll carry the agency approval that says it's okay for line-to-ground transients without blowing up or whatever.

Of course, the Y1 rating is for 120VAC+ and 2500V isolation, hardly what's present here (60V switch, max. 1500V isolation).  I mainly picked that because it's a common example.  And they make millions of them, so they're not very pricey, despite the agency mark.  Since you don't even need that, you could pinch another penny or two and find a "general purpose" 1-3kV rated part.

(And yes, so much thought easily goes into the ratings and selection of little more than a dumb capacitor!  It gets easier as you get used to it all, though I'm sure it might be a bit surprising or overwhelming at first.)

Anyway, you want that capacitor from the "switch" side to ground, either pin, doesn't matter.  You still need (or... want?) a TVS across the "switch" terminals, like an SMAJ48CA, so that if ESD hits the other pin, it has somewhere to go, other than through the (open) switch.

The 2.2k Pull-up to 5 volts is if the circuit of a specific break-out-board needs to be pulled up to 5volts. Some break-out-boards need to be pulled up to 5 volts, some need to sink to ground.  In the case that it needs to be pulled up to 5 volts, then pin 1 of the screw terminal would not be connected.To connect, a jumper would be plugged in for the 2.2k pull-up.

But the point is, would there always be one pin connected to ground?  Are these not "dry contact" terminals like you'd have on a PLC card?

'Cuz if they're always going to be one-side-common, you can just use a dumb transistor there, no need for optos at all!

If they're intended as "uncommitted transistor outputs", but only ever used unidirectionally, you can still use a transistor for that, though it gets a little bit trickier to implement, and you may be better off with some ICs or buffers or something as a solution.  But it's absolutely possible, and will save the bother of all those optos.

Also, if these are ultimately internal connections, ESD probably doesn't matter.  If it's going on a connector to the outside world, and to cables, it's probably still better with the ESD stuff.

That front panel LED is a direct connect to the mainboard via it's connector. There is NO wire or cable separating them. The connectors are there to just stand the LED board up off the mainboard, to allow more space and smaller overall size of the mainboard.

So, a board-to-board connector?

I meant, the resistors to terminate the traces.  On-board stuff.

Which...

Oh shit, just noticing the misch-mash of logic you've selected: AHC, ACT, AC, NC7SZ, HC, NL27WZ, LVC even!  Almost all of which are high to "ultra high" speed families, with (at 5.0V supply, the very margin of operation) output impedances in the 20 ohm or less range -- these will beat the shit out of any transmission line, and with single nanosecond risetimes, it doesn't take much trace length to get there!

Absolutely, positively, reduce ALL of those to nothing faster than 74HC.  All of these gates are available in SOIC or TSSOP packages (74HC flavor), reduced-gates format (74HC1G, etc.), and probably there's a TinyLogic equivalent (NL7xxx) if you prefer.

On the upside, you've got bypass caps per chip, which if they're routed closely, probably won't have the chips themselves shitting the bed.  But the logic signals entering them, hoo boy -- unless you inspect the circuit with a 200MHz+ scope, you wouldn't even know anything is going on!

So, just another thing to be careful of.  If you want the "well why didn't they fucking tell me this before I bought them", check the appnotes: in particular, there's a signal quality one I'm thinking of, see if I can find it...

Yayayah, found it, this was the one I was thinking of:
http://www.nxp.com/documents/application_note/AN246.pdf
They show the V-I characteristics of a few logic families (in other words, how much current the outputs can dump) (note the axes: these things are dumping nearly half an amp at voltage extremes!), and some waveforms that you'll see when driving a modest length transmission line (which is everything: traces, cables, coax, waveguide, anything).  And note the scale of the waveforms, a few nanoseconds here and there is what makes all the difference, so it could be easy to miss, and yet you can end up with a spooky circuit (extra transitions causing weird behavior, especially around flip-flops) or poor reliability (note the overshoot, which bangs into the input protection diodes!) and be none the wiser if you don't have the bandwidth to see what's really going on.

One final word about logic families: 74HC is the quintessential, general purpose, 3.3-5V logic family.  It's a little on the fast side even for what you're doing (you can run afoul of these transmission line effects on pretty large boards), and you'll get more noise immunity (important for a potential industrial environment?) with high voltage logic.  CD4000 series is still the best in that matter: it's slow as molasses (even at 15V, rise time is like 100ns), weak (so even if it were fast, it won't cause overshoot), and just as cheap and easy to work with*.

*I lied, a lot of the pinouts really suck.  But on a PCB, that's not a big deal.

- Series 10 ohm resistors: J32, J36, etc. V_IN pin.
- All the TVS stuff you have a choice on: probably best to put it on the J32, J36, etc. inputs (and outputs where applicable), as well as J29, J33, etc. (same idea).  Or instead of J23, J36, etc., you can push it up to the main board, on J19 / J20.1.

I can do that, though, there is extra space on the rear panel board, so I'll put them there.  That Mainboard is going to be tight. 100ohm on each V-In on jumpers J32....etc.
[/quote]

100?  I said 10...

100 isn't necessarily bad, but you'll see a lot of drop when those LEDs light up.  Maybe not enough to upset the logic, but it doesn't sound good to me.

Also, why the 4.7k supplying the logic (R124, etc.)?  That'll probably drop more than you wanted, even without the LEDs.

I've had like 5 different circuit boards made, and have had problems with them all.

We've looked at a lot of basic design issues here, have you tried building any on solderless breadboard first?  PCBs are serious investment of time and money, just for something that doesn't work (or ends up with problems).

You can do quite good work with solderless methods, obviously not long term reliable, vibration proof stuff... but just for checking out an idea and running with it a little while to get a feel for how it's doing, absolutely.  Better than doing five revs of a board.

You'll also gain an appreciation for those logic shortcuts, like not having to place every god damned jumper back and forth to every pissing little logic gate....

(The worst I ever did: http://seventransistorlabs.com/Images/Z80_Timer2.jpg blue = address bus, red = data bus, black = control signals; it's a Z80 microprocessor, RAM, timer, and parallel port programming interface.  A very simple computer, pretty easy to build as 80s tech goes.  But yeah, all those buses add up, and you go through a lot of jumpers doing it.  I can't even imagine debugging that thing, going straight from schematic to PCB without trying it first!)

Tim

[Falcon69]

Thanks TIm,

Still reading over everything.

The 74HC's and TInylogic I just looked up. They only put out 5.2mA to drive the LEDs. Going with those would mean another coupel footprints. Resister+Transistor to drive the LED.  That is why I went with the others, They put out 24mA and 32mA.  More than enough to drive the LEDs.

[T3sl4co1l]

Nahh, 74HC will more than gladly put out 20mA!

5.2mA?  You saw the current the outputs are tested at, from a supply of 6.0V, it looks like.  That current load caused a drop of no more than 0.2V from the respective supply!

The outputs are roughly near 50 ohms series resistance, so you can expect a short-circuit current of well over 50mA, and even with an external series resistor, well over 20mA into an LED.

If you need up to 50mA, I would suggest a bus driver type, e.g., 74HC245 or the like.

If you need *really* bright LEDs (more than 50mA load), *then* I would suggest a driver array (maybe something better than a crusty old ULN2003, but it'll do), or discrete transistors.

Tim

[Falcon69]

Okay, so they will keep the green light on forever, if need be?  As I need to tone down the brightness of the green LEDs to match the brightness of the red, there will need to be a constant supply of about 10mA to the green LED's, While the red ones, for briefly when tripped, will require close to the full 20mA for the maximum brightness.

So, Even though the data sheet shows only 20mA output, It can safely do more?

Gah, Data SHeets are so confusing.

[Falcon69]

HMm, you were right. I didn't look at the data sheet.  Mouser list columns of specs when you do a search, it said 5.2mA high and low output.

The Data sheet says they can sink 25mA, and 50mA to sink to ground through the chip (i think that's what it meant by Icc.

I will make that change and change all those chips to that.  However, I have already bought 500 of the single inverters of that model.  Will it be okay to use those, or keep with the 74HC series?

[T3sl4co1l]

Unless you want to design and route a high speed capable PCB, I'd suggest getting HC logic.  Maybe you can return them still, or get a buck from eBay?

Tim

[Falcon69]

Hey Tim,

Is this what you mean for the protection for the OptoMOS?

[Falcon69]

Or this?

[Falcon69]

Tim,

You mentioned that a 4K7 series resistors on the V-In (for the switches) to the supply lines of the Logic gates was too much. I didn't catch that, It would lower the amperage way too much thus lowering the mA available to light the LEDs.  So would I need to change those to 100ohm, 1k, 2K2?

Also, I had a look at using Diodes as an OR gate. It works, in simulation, but, it was weird. Sometimes I would have the correct mA going to the LED's, others I would have less or more.  I think it was dependent on what the hall switch was doing.  I put a transistor on each LED, with the two diodes, and that worked.  However, I got to thinking, with the added resistors, transistor, and diodes, I'd be looking at MORE of a footprint for the PCB board.  So the cost of the transistor, resistor, and 2 diodes ware probably about the same as a cost for a Quad OR gate, but, It would cost me more for the PCB size.  SO, I think I'll just stick with the Logic gates. Less components to place by hand as well.

[T3sl4co1l]

Or this?

The second one 

[T3sl4co1l]

You mentioned that a 4K7 series resistors on the V-In (for the switches) to the supply lines of the Logic gates was too much. I didn't catch that, It would lower the amperage way too much thus lowering the mA available to light the LEDs.  So would I need to change those to 100ohm, 1k, 2K2?

Zero.  You've already got the 10 ohm going to the thing, don't need any more.

Also, I had a look at using Diodes as an OR gate. It works, in simulation, but, it was weird. Sometimes I would have the correct mA going to the LED's, others I would have less or more.  I think it was dependent on what the hall switch was doing.  I put a transistor on each LED, with the two diodes, and that worked.  However, I got to thinking, with the added resistors, transistor, and diodes, I'd be looking at MORE of a footprint for the PCB board.  So the cost of the transistor, resistor, and 2 diodes ware probably about the same as a cost for a Quad OR gate, but, It would cost me more for the PCB size.  SO, I think I'll just stick with the Logic gates. Less components to place by hand as well.

You're neglecting the amount of space taken up by that huge matrix of traces.  To resolve that, you'll need either a 4 layer board ($$$) or more area ($$).  Distributed logic saves space in the end.  Also, you can put the "TEST" line and diodes out on the display board, where you probably have some area to spare anyway.

Tim

[Falcon69]

oh!  Okay, I though you had mentioned that it needed to be on BOTH sides of the connector. Good deal, I will take them out.

Ya, them 4 layer boards are pretty spendy, but.........

I have yet to play around with the layer features of DipTrace.  I might just do it to see how it works.  MAYBE.  LOL


So, I noticed that Panasonic makes a surface mount Class Y1 capacitor now, however, only in three values. None of them 100nF.   Any other companies do, or am I stuck with through hole on those?

[T3sl4co1l]

100n, where?

I think Y1 caps end around 10n anyway, which is for safety reasons (ground leakage).

But anyway, like I said, 1-3kV "general purpose" is also acceptable for the isolator things.

Tim

[Falcon69]

Sorry, I meant to say 1nF (1000pF)

I was having trouble finding the 1-2kv thing you mention.  On the data sheet you linked, what exactly am I looking for when you say that?

The one you linked is 1.5kV?  That should be fine, correct?

[T3sl4co1l]

Yeah, so there's that.  Or, like,
http://www.digikey.com/product-detail/en/S102K29Y5PN63J5R/1251PH-ND/2356788
is general purpose (and, not really much cheaper), or
http://www.digikey.com/product-detail/en/102R18W102KV4E/709-1029-1-ND/1556434
is SMT (1206 doesn't leave much clearance though), or
http://www.digikey.com/product-detail/en/202R29W102KV4E/709-1046-1-ND/1556451
is the cheapest 1808 (more clearance, if you need it)

Like I said, probably only the voltage rating matters, not actual agency ratings.

The cap I linked earlier, says 760VAC, which is 1074V pk / DC.  But they're actually 100% tested to 4kVAC (5.6kV pk).  They won't necessarily be useful at those voltages, but they'll do it without zapping.

Tim

[Falcon69]

Awesome!  You found some surface mount ones!

I'll finish making the changes to the schematic and post it soon. If You find anything else that could be changed, please let me know.  I need to get going on designing the PCB.  Fun Stuff.!

[Falcon69]

Okay, here's the updated Schematic.  Yes, It is still big. I will fix it once I know that everything is as it should be, as far as circuit.  Once it is, I will break it apart into several sheets to get it smaller.

How's it look now?  Anything I need to change now?

[T3sl4co1l]

Yeah, looks pretty good.  Should be pretty robust now!

Tim

[Falcon69]

Awesome, thank you for all the help Tim.  I'll start drawing it up and order the PCBs.  The LED Panel so far I have drawn up, and it is the same size as old ones. Yay!.  I'm hoping the others will be as well.  I've already got the panels made (left over from old batch of PCBs, that didn't seem to work right)

And to answer your question about before..... I tried doing this on the bread board, even made a BIG bread board, but had problems with it.  I ordered some of those jumpers from China, and the breadboards from there as well.  I just couldn't get anything working right, and I think it was the quality of the breadboards and jumpers. 

I am going to just order a very small quantity of boards this batch, and pray they work right.

Jason

[eetech00]

Okay, here's the updated Schematic.  Yes, It is still big. I will fix it once I know that everything is as it should be, as far as circuit.  Once it is, I will break it apart into several sheets to get it smaller.

How's it look now?  Anything I need to change now?

Hi

Just curious.....what is this circuit supposed to do?

eT

[Falcon69]

When you push the button, it giggles like the Pillsbury Doughboy.

It's a switch system for a hobbyists machine.  3d printer, Laser, Router, etc.

[Falcon69]

Hey Tim,

I was just thinking of something...

Here I am drawing out the PCBs, and I thought of something.

You mentioned that ESD charge could be upwards of 16amps and a whole lot of voltage.  What does that do to the traces?  Like for instance, I'm using 0.2mm traces.  That Trace width is rated for up to 370mA with a 1oz copper.  Now that is way way way below that ESD charge, but, way above any that is needed for my circuit.  So, what does the ESD charge do to those traces?  If I design the board to account for that, those traces will have to be 35.8mm wide to accommodate that 16A ESD charge.

[T3sl4co1l]

Nothing, physically.  I don't think anyone is concerned about mechanical or electrical failure on visible-to-the-naked-eye parts.  The mechanism for failure in semiconductors is microscopic thermal or molecular damage, which is why you want rated protection devices.  Other than that...

Outside of extremely large and rapid currents (in the regime of exploding bridgewires and electromigration), trace width current rating is ONLY a thermal matter.  So, for very short pulses, if the voltage drop is low enough for your purpose, it's thick enough.

ESD rise time is short, so you do want to put the TVSs close to the connector, very well grounded (stitched ground plane or 4 layer), and with the connector side entering the TVS, and the circuit side leaving opposite.  Not, like, the TVS hanging off to the side on a branched trace.

Check out the appnotes for those TVS devices, they show layout.

Tim

[eetech00]

When you push the button, it giggles like the Pillsbury Doughboy.

It's a switch system for a hobbyists machine.  3d printer, Laser, Router, etc.

Hi

Didn't mean to upset you...just wondering what its function was...
Sorry..

eT

[Falcon69]

Didn't upset me.

But I did mention it earlier in the thread.

[Falcon69]

so,

I got the boards back and assembled this circuit.  Everything works as intended, except for one thing....

The tactile (momentary) switch I am using.  If I place my finger across the contacts across the back of the pins as I press the tactile switch and release, it works as intended, on/off.  But, I take my finger off, and it is, well, variable.  Sometimes in turns off, sometimes it stays on.  In this particular circuit, the tactile switch turns on ALL the LED's in the circuit, making it easy to check to see if an LED has burned out or not.

As you can see in the circuit I uploaded, I placed a 1k resistor in series with the tactile switch, along with a 0.1uF capacitor to ground. This tactile switch and related components are located on the "front Panel LED Indication' part of the schematic.

How can I fix this problem?  Could it be as simple as changing the value of the capacitor?  If so, what value would be best?  Or should I change the value of the resistor?  Or, Maybe, place a 1M ohm resistor across the leads of the tactile switch?

I have also noticed that when power is first initially turned on, all the LED's flicker on for a fraction of a second. Can that big fixed? Maybe a slow start on without effecting the operation of the circuit?



One other thing, I noticed the voltage regulator gets very hot.  On the circuit board I had made, I only used the recommended footprint pattern for that chip.  I can touch the chip, but only for a few seconds before it starts feeling like it's burning my finger.  Is this normal for these voltage regulators to feel like that?  It's a 3A voltage regulator, with only about 1A flowing through it (what the circuit draws, and max of 2A when the tactile switch is pressed to turn on all LED's).

Thanks in advance for any and all help

Jason

[T3sl4co1l]

Well, a pulldown on the other side of the switch wouldn't hurt...

I didn't look extensively (for aforementioned reasons), but I only saw CMOS logic inputs attached to that net, nothing at all to define the logic level otherwise.

The regulator is... making 5.2V or so, and... it's about the shittiest "low dropout" LDO I've seen... never less than 0.8V drop, any current.  Gee, thanks.  So, you need over 6.5V for operation of this thing.  Plus the bother of two divider resistors.  Why not just a boring old 7805?  They make them in DPAK too.  Then you don't have to worry about pesky capacitor ESR choices either (stability plot says >0.1uF should be < 0.08 ohm ESR... have fun with that).

Anyway, it's rated to 125C, and, I guess, power limited, somehow, but they don't happen to tell us how that is achieved.  Thanks again, ST.

If you literally used only the minimal footprint, with no copper pour, you're asking for problems.  If you have copper pour (about 1 in^2 worth) around the heatsink pad (no thermal reliefs), and/or vias to a bottom side pour, you're okay for about two watts under ambient conditions.  If it's going in a box, have fun with that, too...

Was this thing going to run from a 7.0V regulated source, or something?  1A is an awful lot to expect from a linear regulator with no heatsink, let alone if the supply is the full 24V permissible.

Consumption should vary with number of LEDs lit, but I would expect with, what, 40 LEDs or so at 5mA each, you're not much over 0.2A total.  So... better go find the other thing(s) that's hot...

And if you can, scope every node to see if 1. it's sitting still when it's supposed to be (not bat shit crazy oscillating or something), and 2. make it transition (press the corresponding buttons, whatever) and see if there's bounce or overshoot or whatever.

Tim

[Falcon69]

thanks Tim,

I'm not sure I understand. I thought 7805 regulators only output 5 volts. I want to see about 5.25 volts.  The logic chips accept up to 5.5volts, and I'm trying to compensate for the voltage drop along the length of wire for the switches.

With that ST voltage regulator I am using, I can't connect that big pad to the ground plain, as the big pad is the OUTPUT to the circuit. I could enlarge the pad though, if that helps.  Could you suggest another voltage regulator that works better?

The LED's on the switches (8 of them, with 2 LED's each) put out the full 20mA each, or close to it.  They all go on when the tactile switch is pressed. I am trying to keep those LED's as bright as possible, as they will be placed on machines for visibility.

I found this circuit for the Debounce problem.  I think this should solve it?

[Falcon69]

Well, been trying to figure this out. I understand what you're talking about with the heat dissipation and only 2W with a 1² inch of copper pour.  I can;t get that with the current space I have.

I've seen these on eBay  http://www.ebay.com/itm/5V-USB-DC-7V-24V-to-5V-3A-Step-Down-Buck-KIS3R33S-Module-for-Arduino-LM2596-LU-/271666790063?pt=LH_DefaultDomain_0&hash=item3f409a3eaf

How do they do these? They advertise 7-24v input, 5 volt out at 3A.?

Here's a picture of the current board, and a picture of the newly laid out one underneath for the voltage regulator.  The bottom copper pour is about the same size as well.  With the amount of traces I have, it's just hard to find space without adding more board, which is not something I'm wanting to do.

So, based on what I did with the copper pour, will that work okay?

Oh, and to answer your question from before Tim....the circuit will most likely be powered from a 12 volt wall power supply.  But, I want to give them the option of 9v to 24 volt, as some power supplies for machines these will be used on, output 24 volts.

Also, that schematic I posted for the debounce circuit for the tactile switch, Will that work, at those values?

Tim, you had mentioned a pulldown resistor on the other side of the tactile switch.  Which side we talking about?  The circuit side, or the power side?

Thanks again!

Jason

[DutchGert]

Well, been trying to figure this out. I understand what you're talking about with the heat dissipation and only 2W with a 1² inch of copper pour.  I can;t get that with the current space I have.

I've seen these on eBay  http://www.ebay.com/itm/5V-USB-DC-7V-24V-to-5V-3A-Step-Down-Buck-KIS3R33S-Module-for-Arduino-LM2596-LU-/271666790063?pt=LH_DefaultDomain_0&hash=item3f409a3eaf

How do they do these? They advertise 7-24v input, 5 volt out at 3A.?

Here's a picture of the current board, and a picture of the newly laid out one underneath for the voltage regulator.  The bottom copper pour is about the same size as well.  With the amount of traces I have, it's just hard to find space without adding more board, which is not something I'm wanting to do.

So, based on what I did with the copper pour, will that work okay?

Oh, and to answer your question from before Tim....the circuit will most likely be powered from a 12 volt wall power supply.  But, I want to give them the option of 9v to 24 volt, as some power supplies for machines these will be used on, output 24 volts.

Also, that schematic I posted for the debounce circuit for the tactile switch, Will that work, at those values?

Tim, you had mentioned a pulldown resistor on the other side of the tactile switch.  Which side we talking about?  The circuit side, or the power side?

Thanks again!

Jason

U seem to use a lot of hatched unconnected copper on your board. In general unconnected copper is a bad idea so I would use it as extra ground plane with some via's and make it a solid plane while you are at it

[T3sl4co1l]

Those headers are too close together... good luck assembling them at all if they're just too wide, and so much for retention clips if you put this on a high vibration machine. e.g.
http://www.digikey.com/product-detail/en/3505-8110/3M15483-ND/761036

The screw terminals should be okay though; 5.08mm pitch (and some others) are available in long rows, or as stackable pieces (mind the assembly order for the dovetailed kind).

That debounce circuit is fine, but: remove the diode and drop the capacitor to 0.1uF.  Maybe raise the pullup to 1M as well.  The Schmitt trigger is optional if you're not driving clocked logic (in which case exchange VCC for GND to get the same logic states).

Switching converters with http://www.ti.com/product/LM2596/technicaldocuments or other are a good idea.  The circuit is typical, or you can get a dev kit.

Tim

[Falcon69]

The headers are actually a single one in 2x17 pins. I pull-out 6 of the pins.  I just drew it that way, because I didn't feel like redrawing an IGS file or a new pattern.

So a circuit like this is what you mean for the debounce?


As far as the regulator, I think I'm going to forget it.  Cost too much, and many more components.  I'm going to just supply a 5volt wall plug style power supply with the board.  I just don't have the space to add all those components on this board without making it MUCH larger.

I just looked on ebay, they have lm2596 chips for about 30 cents each.  I'm weary about buying IC's off eBay. If I can get that cost down, and design the boards similarly to the one i posted above, maybe it would work? I would need a footprint size of about 25 x 50cm.  I think I can manage that.  It looks like they use the whole bottom layer as a heatsink. The downside is, they've placed all their vias into the pad.  I'm not sure I could get that reflowed.

http://www.ebay.com/itm/1-DC-DC-Buck-Converter-Step-Down-Module-LM2596-Power-Supply-Output-1-23V-30V-/181409861491?pt=LH_DefaultDomain_2&hash=item2a3cdecf73

[T3sl4co1l]

Yup!

A switching converter takes less space than a linear regulator, at least if you use a high efficiency part in a usefully sized package.

Probably the only reason they offer DPAK-5s and such are because they're the old, low efficiency bipolar designs, for maintaining old designs, and new designs for ham-fisted solderers.. if that's you, sorry.

I've seen converters up to 5A in 'thermally enhanced' packages (SO-8 with pad, DFN/SON, QFN), and with operating frequencies typically around 400kHz, the inductors don't need to be large either (and often can't be, if the controller is peak current mode -- something to keep in mind).

I don't see that voltage is a big deal in this project, even with long cables on the switch thingies.  Find the minimum voltage tolerance on the furthest connection, and evaluate current flow and voltage drops all the way back to the power source (regulator / adapter).

Tim

[Falcon69]

I drew this up after reading the TI data sheet for the LM2596

If I understood it correctly, this should work if I chose to go this route?

This should allow a Delayed Startup which I hope would fix the voltage surge when the power is connected? That of which, made all the LED's blink on for a fraction of a second.

[Seekonk]

I haven't read all these pages, but your feedback should be on the output side of the inductor.  What you have drawn could be very interesting.

[Falcon69]

you are correct, I missed that on the data sheet. Fixing it now, thank you.

[Falcon69]

tim, on the debounce circuit, I was reading that putting the diode in results in the cap recharging faster.  But you had mentioned, since I had no clocked circuit, just cmos logic gates attached to the circuit, the was no reason for the inverter. By that same logic, there's no reason for the cap to recharge faster then, correct?  Just want to make sure I understand.

Also, I was reading somewhere that the switch contacts may fail prematurely if that diode is NOT in place. Can you verify?  I will need to fix that somehow. Although, that switch won't be pressed 24/7, so probably not a big deal.

[T3sl4co1l]

Diodes are for inductive contacts, doesn't matter here.

Maybe debounce doesn't matter, it's a nice-to-have in principle.  The pulldown anyway is most important.

Tim

[Falcon69]

Hey Tim,

I spent about 6 hours today looking at the schematic for problems, etc.

I found quite a number of logic chips that would be 'floating' in most conditions, i.e., switch off, or switch on, etc.

A redid the PCB and placed a 100k resistor on all those to ground.  That should solve the problem.

My question is, with those logic chips confused on what state they were in, cause a large current draw somehow that would be one of the main causes as to why that Voltage Regulator was heating up, other then lack of pcb footprint and heatsink to dissipate the heat?

It would certainly attribute to the tactile switch operating funny, I am guessing as well.

[T3sl4co1l]

Intermittent operation is more important.  It's unlikely but quite possible that current consumption increases with floating inputs, supposedly on the order of a few mA per input.  Whether all at once behave this way, seems unlikely.

Tim

[Falcon69]

Ya, I understand.

It has to be that tactile switch thingy.  I'll print up some new boards with the added fixes, and see if that works.  It should.  Other then the switch problem and the voltage regulator, everything else was working just fine.

[Falcon69]

That voltage schematic up above of the lm2596, anything I should change on the values or size of resistors?

I'm thinking I will switch the voltage regulator I am using for that one.  I don't want to take the chance of botching another circuit board.  But, I want to make sure that schematic up above is going to work.

If someone could help out with this, it would be greatly appreciated.  If Ya every come to Oregon, I owe ya a beer.

Jason

[T3sl4co1l]

Hmm, doesn't seem like anything obviously wrong or that can't be fixed with parts choice later.  The feedback capacitor is kind of odd though, you usually want to keep output ripple away from the feedback pin.  This input is usually part of a differential amplifier, which can get unbalanced by RF and therefore the voltage shifts in a weird way (dependent on supply voltage and load?).

But whatever the case, if you're following the datasheet/appnote suggested circuit and values, it should go fine.

You should learn the art of dead bug construction.  You could get all these parts, construct the circuit, and test it, alone and with your circuit, without having to order PCBs at all...

Tim

[Falcon69]

Okay, thank you Tim

The resistor values are all okay, 1/10 watt?

That 680uF capacitor, it is really large in a 35v.  Can that be reduced?  The max input to the voltage regulator would be 24volt. (probably more like 12volt realistically, but I want the option for 24volt)

[Falcon69]

hmm, I was just looking at the ONSemiconductor datasheet for that LM2596 chip.

They are using a 100uF instead of that 680uF.  That is a much more manageable size.  Can that value be used instead?

[T3sl4co1l]

Yes, but it needs to be rated for the ripple current, which is why they use an excessively large value.

I also see 120uF and 220uF suggested in the TI datasheet.

You could probably plop down 10-47uF 25V in aluminum polymer, if you're feeling fancy.  They are a little more $.

Beware that, because it's a good capacitor, it will tend to resonate with the power line inductance -- which the TVS helps with (clamps peak voltage), but the regulator input itself is negative resistance at low frequencies (current draw goes up as voltage goes down), so you want to avoid that as well.  Typical 6' twisted pair or 'zip cord' style cable is going to be a few microhenry, which would resonate with 47uF at 16kHz, and an impedance of 0.2 ohms.  Might be close enough to the regulation bandwidth to matter.  Nice thing about electrolytics, they come with ESR that dampens (100uF and under 1 ohm ESR would be fine I would think).  So does tantalum, but you don't really want to use that in a power input application (too vulnerable to peak currents, too large a value needed).

But anyway, 220uF 25V isn't big, are you really that pressed on space now?

Same idea goes for the output filter cap, except you probably don't want a low-ESR (polymer or ceramic) part there because the ESR is in the feedback loop (again, check datasheet/appnote).

A subsequent filter probably wouldn't hurt; it doesn't need to be much, maybe 1uH and 1-10uF (ceramic), just to help keep stray RF from the converter away from those long lines you've got.  The bypasses spread around the logic chips will help suppress that, too.

Tim

[Falcon69]

Okay, spent all day on it, think I got it.

Here's the pics for the finished board layout.  I was able to fit all the extra parts for the voltage regulator on there.

Here's the schematic for the voltage regulator part of circuit. 

Does everything look fine now?  I know the components are packed in there pretty tight, is that going to be a problem?

[Falcon69]

oh, wanted to add...

I am going to take off the copper hatched pour for the top side (even though I hate the plain look), but it seems everyone is saying that will hurt the performance of this PCB.  I may just get these in black solder mask in that case.

[Falcon69]

Any thoughts on the layout?  Does it look fine? Do you think I can go ahead and have the boards made?

I think it will work, I followed the data sheet info for the OnSemiconductor LM2596 chip. Though, to save money, I will try and buy the LM2596 chips off eBay and hope they work fine.

I am guessing, that since TI bought out National Semiconductor, that there are alot of legitimate LM2596 chips out there with the branded National Semiconductor name on them, but the chip still seems Identical to the TI chips.  In fact, the data sheets I have found for the National Semiconductor chips look like an exact copy for the TI data sheets.  Am I wrong?

Thanks in advance.  If everything looks good, I'll order the boards today.

I'm anxious to get these working right as I'll be using them on my own machine. Get this headache out of the way, and I can start working on building the machine again.

Jason

[T3sl4co1l]

TI/National are one in the same now, but that doesn't mean suspicious lots aren't suspicious...

Layout looks good I think, at least at a glance.

Hatching doesn't save you anything, unless you're doing toner transfer (solid toner areas don't usually maintain density).  If it's dense, it will carry heat as well as solid, but low density obviously won't.

Tim

[Falcon69]

Ya

They had mentioned, (in this thread or another) that stray signals or noise could be transferred through the hatched or solid pours if they are not connected to ground. That was my concern.

I'd like to keep the hatched look, I like it, but if it hurts the performance of the board, I don't want to. (fyi, the top copper pour/hatched, is not connected to ground plane, it's just floating for looks)

Tim, the components all packed tight like that isn't going to be a problem?  I have about .45-1mm between the components for the POWER traces and like .24mm for the traces.

[T3sl4co1l]

Don't leave floating copper, ground it.  Vias are cheap (free?).

[Falcon69]

Okay, that is what I thought, I will leave it off then. 

I thought about the soldermask color, and because the switches have LED's that light within encased resin, I think a White soldermask would amplify the color of the LED's light.  I really like the black look, but I think I will order them in white because of the LED's.  I so hate the color green. It's so generic.

I'll get these ordered up, and let you know how it goes.

Thank you Tim!

Jason