Reflowing a previously soldered board

Go To Last Post
90 posts / 0 new

Pages

Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

In a week or two I'm going to bring up a new board I've designed based on an ATMega32A.  It has a 5V regulator fed from an external 12V supply, 16 MOSFETS, an RS485 driver, and some connectors.  It's all surface mount.  This is a large board for me, and with my previous, smaller boards, I'd hand-solder a piece (like the regulator), test, then hand-solder some more.  Once I'd validated the board, I'd make more by reflowing them in a toaster oven.  With this one, it would be nice to not have to hand-solder all those MOSFETs, so I was wondering what would happen if I reflowed them after hand-soldering other parts?  Would the hand-soldered parts come unsoldered and, if so, would it matter?  They're just going to be sitting there on their pads and the solder would solidify again once the board cooled.  I could also reflow the MOSFETS first, but I'd prefer not to do that since I may encounter a problem somewhere else on the board before I need those parts.  Thanks for any thoughts.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Post a schematic & we can look for trouble now...if none spotted,, just solder em all at once.

The solder should hold things in place, even with a reflow.

 

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

On my boards, after hand soldering, I hand re-flow using hot-air, the result looks like a board house built unit!

 

Good luck.

 

Jim

 

Mission: Improving the readiness of hams world wide : flinthillsradioinc.com

Interests: Ham Radio, Solar power, futures & currency trading - whats yours?

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Jim:  Hadn't thought about using hot air.  I can do that, I guess, though I prefer using it for removal, since you don't care about the temperature in that case.  For reflowing, you don't know what you're subjecting the part to.  Even if it works when you finish, you'd only see a problem in reduced lifespan, so that wouldn't show up for an unpredictable period.  At least in my oven (which uses a reflow controller), I have some semblance of a proper reflow profile.

 

avrcandies:  Thanks for the offer.  Always appreciate feedback.  I've attached the two pages of the schematic, plus the PCB.  Be brutal.  ;-)

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The 1k pull down resistor are a bit strong.

That is 16 * 5 = 80mA if all the MOSfet's are on.

And the only use for them is to hold the gate down during reset of the AVR.

They can easily be 100k or 1M.

 

Add some big pads to the voltage regulator (especially when using the 1k resistors).

Replace the 7805 with a small SMPS circuit. 12V -> 5V is quite a voltage drop, waste of energy.

This also more than halves the supply current to your circuit, which helps a lot if you use long cables (As I do with power over CAT-5).

A buck convertor has an inductor between input and ouput. This helps a lot in keeping conducted EMI through the power supply wires far away from your ATMEGA.

I would also add a Polyfuse in the 12V power supply. They are cheap and wonderfull.

 

 

How much current do your MOS fets switch? To they also need a thermal pad?

The GND trace to all the MOSfets is quite thin.

Ah: note of 2.7A...

It looks like that the GND trace to your MOSfet's has a lot of via holes drilled in it, which also make it thinner locally.

 

I would like to see a big fat continuous GND plane at least around the uC and the power supply section.

For the high current traces I would not trust a single via.

 

You need some way of making sure that the RS485 levels are within spec.

RS485 only has a common mode voltage level between -7 and +12V.

Normally this is done by adding a third GND wire (possibly with a power resistor in series.

I have used standard CAT-5 +RJ-45 for RS485 an dused 2 wires for distribution of 24Vdc.

Small SMPS on each AVR board, and you can power a lot of boards from a single power supply.

Putting Poly fuses in series with A & B make your RS485 robust against hard DC voltages (such as wrong 24V connection).

I would also not use RS485 any more witout added Transzorbs or at least very rugged RS485 drivers.

 

Pin 2 & 3 are both permanenty connected to Vcc. You can not switch between sending and receiving this way.

 

KiCad user here, unfamiliar with Eagle.

Why do some component numbers have a "$" in the middle, but your resistors and capacitors don't ???

 

I do not see a buffer capacitor on the board. Use something between 10uF and 100uF.

Oops, now I see it (10 minutes later).

For drawing schematics: they are a lot easier to read (especially for others, but also for yourself) if you keep to some conventions, as much as possible:

- Voltages from top to bottom.

- Signal flow from left to right.

- Schematic symbol pins grouped in logical order, not by physical pin location (You can always do a text search to find pin AF46 of a big BGA on page 23 of a big schematic).

 

For example:

Power input connector, then buffer cap, voltage regulator, decoupling caps Vcc & GND labels all neatly in a row (with the GND trace on the bottom.

Look at your GND trace wrapped all around your ATMEGA. GND pins should be on the bottom of your ATMEGA symbol.

All those "$" in components and even in symol pin names makes it hard to read.

Which of pin 2 or pin 3 is the enable for the driver and which is for the receiver in your RS485 chip?

 

Put the polarity of your diode on the silk screen.

C5 C9 C10 U$24 should all be physically close together, connected through a fat GND plane.

Your pcb design looks like you have just let an autorouter rip through it without giving it further thought.

 

Some people here think I am feeling "superior" and want to "show of" or "get a kick out of" pointing out other peoples mistakes.

I see it as having some experience and trying to help others to build better stuff.

It's up to you whether you want to implement any or all of the suggestions I mentioned.

Paul van der Hoeven.
Bunch of old projects with AVR's:
http://www.hoevendesign.com

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

By soldering iron or hot air pencil can have a repeatable and controlled profile such that parts aren't damaged.

Why :

The Science of Component Heat Damage

How Overheating Damages I.C.s

http://web.archive.org/web/20121016080444/http://www.emsciences.com/HTML/Heat.html

soldering iron training :

http://www.emsciences.com/training.html

hot air :

BGA Rework Station, BGA Hot Air System, BGA Alignment, BGA, Protoyping, SMD, Hot Air Station

Zephyrtronics®

Two Critical Benchtop Processes

by David Jacks

http://www.zeph.com/pap1.html

(pre-heat, cool down)

 

"Dare to be naïve." - Buckminster Fuller

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Ah yes, I forgot:

Original question was about the soldering.

Hot air is not uncommon, and the closed loop feedback from your eyes & brain from when the solder melts is a quite good indication of temperature.

Pre-heating the PCB helps.

Also the direction in which you blow the heat.

If there are parts you do not want to heat too much you can cover them with aluminimum foil and a few pieces of kapton tape.

 

A lot (most?) reliability problems are not from the peak temperature, but from a too fast heating profile.

The idea is that a lot of plastic's absorb a bit of moisture.

I have received orders from mouser in sealed plastic bags with a dessicant and a moisture indicator inside.

If the indicator is discolored you have to bake the components for SEVERAL HOURS at temperatures of around 120 to 150 Celcius to let all the moisture evaporate out of the plastic.

(Do not trust my numbers, do some research yourself). Different baking procedures are recommended for different components.

The regural black epoxy seems to be fairly moisture proof.

It is often clear components such as leds that are affected more by this.

(Also sensitive stuff, such as IC's for measurment equipment, very low offset opamps / inamps etc).

Paul van der Hoeven.
Bunch of old projects with AVR's:
http://www.hoevendesign.com

Last Edited: Tue. May 15, 2018 - 11:56 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Don’t worry Paul, your observations are on the money. I would not consider that pcb viable. It looks like its been done by an autorouter.
If you consider the pcb as a means of interconnect, then everything is connected, but the pcb is actually a vital part of the circuit. Consider if you draw some considerable current through the gnd tracks. What voltage drops can you expect? Would they affect other parts of the circuit? Most likely.
Also consider the power disipation of the regulator - the 485 transceiver can draw over 100mA, so what will this do to your regulator?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Wow, Paul!  Thanks so much for your detailed reply.  It feels like a master class in circuit design and PCB layout, both of which I could certainly benefit from.  I don't feel you're showing off at all, but, rather, sharing your experience, which is precisely what I asked for, and what a lot of people seek from online forums.  Unfortunately, some here only offer *destructive* criticism, not *constructive*.
Comments and questions below.

 

First, to add some context, the application is a piece of kinetic sculpture consisting of 1024 12VDC relays arranged in a square array of 64 panels, each containing 16 relays in a 4 x 4 array (something like in the attachment, though there are only about 600 in that render).  Each panel will contain (on the back) a 12VDC power supply (converting 110 - 220 VAC) and the board we're discussing here.  The relays are supplied with 12VDC from the power supply, and low-side driven via the MOSFETs (to which they're connected via the two RJ45 connectors).  Each relay only draws 170 mA, and will be on for no more than 1 sec. at a time.  The piece will be controlled by a master controller communicating with the panels via an RS485 bus at 300 kbps.

 

> The 1k pull down resistor are a bit strong.

 

Good point.  Changing.

> Add some big pads to the voltage regulator.

 

I expanded the pad already with a filled area, plua another one underneath it on the other side of the board.  That's what all those thermal vias are for.  How bigh should it be (see next)?

 

> Replace the 7805 with a small SMPS circuit. 12V -> 5V is quite a voltage drop, waste of energy.

 

That was my original plan, but 95% of the time the RS485 trasceiver (MAX3061E) will be inactive, making the MCU the main power consumer, drawing, worst case, 20 mA.  The RS485 trasceiver will be active the rest of the time, bringing total power consumption up to, maybe 70 mA.  The only time the circuit will draw more than that is if the RS485 bus shorts out, at which point the transceiver will max out at 250 mA.  That may never happen, but, of course, I have to plan for it anyway.  Unfortunately, all the SMPS circuits I looked at (admittedly not an exhaustive search) are very inefficient at the low end of their output power range.  So, it seemed to me that an SMPS would add cost and complexity for little or no gain in efficiency compared to a linear regulator (500 mA rated in this case) the vast majority of the time.  Does that make sense?

 

> This also more than halves the supply current to your circuit, which helps a lot if you use long cables

 

The power supply will be within six inches of everything it's powering.

 

> I would also add a Polyfuse in the 12V power supply. They are cheap and wonderfull.

 

What rating would you suggest?

 

> How much current do your MOS fets switch? To they also need a thermal pad?

 

As I said above, 170 mA.  It's not scientific, but I can hold one on a breakout board in my hand for the one second max. on time and it doesn't get warm.

 

> The GND trace to all the MOSfets is quite thin.

 

I was concerned about this, too.  I checked with a couple of online trace width calculators (e.g. http://www.4pcb.com/trace-width-...) and got the same number:  170 mA through 1 ounce copper yields a trace wdith of 1 mil.  I was going to bounce this off a forum, but forgot and ended up going off of the calculated numbers.  So, to be "safe" (and to stay within the PCB house's design rules), I used 8 mils for drain and source, 27 mils (21 calculated) for each of the two ground busses running to all the MOSFETs, and 56 mils (47 calculated) for the ground bus feeding these two back to the 12V input and regulator.

 

> It looks like that the GND trace to your MOSfet's has a lot of via holes drilled in it, which also make it thinner locally.

 

True, but how else should traces connect to it? 

 

> I would like to see a big fat continuous GND plane at least around the uC and the power supply section.

 

I could extend the thermal pad that's already under the regulator, as that's connected to ground.  It would be hard to do for the MCU, though, given all the routing going on within its footprint.  At least I'm notusing any analog functions.

 

> For the high current traces I would not trust a single via.

 

Please explain. 

 

> You need some way of making sure that the RS485 levels are within spec.

 

The MAX3061E claims to be reliable as-is (https://www.maximintegrated.com/...).

 

> Normally this is done by adding a third GND wire (possibly with a power resistor in series.

 

This is my first RS485 application, but I haven't seen anything about this.  There has been mention of using shielded twisted pair cabling with the shield grounded at one (and only one) end.

 

> Small SMPS on each AVR board, and you can power a lot of boards from a single power supply.

 

I thought about using a single supply for the whole piece, as I'm not wild about running AC to all the panels (even if using proper cords and sockets), but that would require a 174A, 2.1KW supply (without derating for margin); a $700+ single point of failure.  Another advantage of my approach is that I can split the AC into two circuits, allowing each to be supplied by a separate 15A AC circuit (there's no way to power it off even a single 20A circuit).  Of course, I could use multiple, smaller 12V supplies, and put half on one AC circuit and the others on another one, though I'd still run the risk of losing a large chunk of the piece if one of the supplies failed.  ALso, the piece is going to be 8' wide by 9' - 10' tall, so the voltage drop in the cables becomes a consideration, given their length and current.

 

> Putting Poly fuses in series with A & B make your RS485 robust against hard DC voltages (such as wrong 24V connection).

 

I'll have to look into that.

 

> I would also not use RS485 any more witout added Transzorbs or at least very rugged RS485 drivers.

 

Ditto. 

 

> Pin 2 & 3 are both permanenty connected to Vcc. You can not switch between sending and receiving this way.

 

I'm aware.  These are all only being used as receivers. I'd originally thought I might want to go both ways, changed my mind, but didn't change the part.

 

> Why do some component numbers have a "$" in the middle, but your resistors and capacitors don't ???

 

Eagle does that automatically.  It's been a while since I've read that section of the manual (and I'm using a very old version of Eagle), but I *think* it uses the $ sign for complex dvices that might have multiple gates (e.g. a 7400); then the $ helps designate different gates within the device.  It doesn't know an MCU doesn't work like that.  Some of my connectors have a $ in their names because I screwed up when I created the part.  I've gone back to fix it, but the fix doesn't propagate to the schematic and board when I force an update, for some reason. 

 

> Put the polarity of your diode on the silk screen.

 

At least I did this right.  It's the white line to the right of the left pad.

 

> C5 C9 C10 U$24 should all be physically close together, connected through a fat GND plane.

 

I thought the bulk capacitor (C5) could be anywhere on the board.  I guess not.

 

> Your pcb design looks like you have just let an autorouter rip through it without giving it further thought.

 

Sad to say, no.  This is my doing as the autorouter in Eagle (at least my version) does worse. 

 

Thanks again.

 

Karl

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

You have no reservoir caps on your AVR supply, pins, just some measly 0.1uf caps...what will supply trasients at the avr & IO circuits (such as fet gate pulses)?

 

Double all of your fet trace widths they are too small, especially when you have multiple heaters packed together.  75 mil for the 2.8amp fets  Otherwise you are just pouring your copper down the drain...why?  Trace calculators aim to prevent the traces from burning up, not for high performance.

 

That's an absolutely horrible fet (STN4NF20L), unless you need to switch 200V...do you?
There are many other multi channel fets that will do a much better low voltage job...Also, what are you doing about short circuits? 

All your fets should connect to a gnd plane section on the bottom...put your gate traces on the top.  Other arrangements are possible...but lots of current totals up into your gnd

 

It's a ridiculous layout to put all the gate resistors way over ...there....why are you creating a million extra traces?  Put each resistor at the fet where is connects to G & S!!  Note the big center tab is also D

The labeling of the MAX3061 is complete bogus...you can't tell what those pins do, same for the AVR...label the pins with their functions.  $P3$38 doesn't cut it!

Why is the 12V  connector 2 inches above the 5V regulator? Shouldn't it feed INTO the regulator?  If it is an input connector it belongs on the left side of the schematic.

Why would you control 1000 relays using fets?  Could you simply use the fets?  Of course in some cases the relay is needed...but you should check carefully.

I'd add a 1-2ma led at each gate, so you can tell what is going on...later you could simply stop populating them.  But you will probably find they are too valuable to get rid of.

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The trace calculators only tell you the trace width for a given temp rise. They don't factor in inductance which is critical in your application. Just recently I was debugging a pcb (8 layer) that had bad power routing and the mosfet would oscillate at around 60MHz when switched with a load. This was due to the inductance of the ground track to the mosfet. This caused what is known as ground bounce. You also need to consider what happens when the loads are drawing current and you get a voltage drop across your ground wiring and your microcontroller is in the middle. Bad juju is what happens. So, keep your load ground separate from the ground to your ics. Also consider the paths the ground follows from your connectors - like the RS485.

 

A regulator like the Recom R785 series is a 7805 drop-in that solves the temperature rise problems. It is a miniature switcher. Not super cheap, but it solves a common problem.

If you're using tantalum caps, don't put them on circuits that go to the outside world. They like to go up in flames.

Having input protection on your power is always a good idea. Use a polyfuse in series and transzorb in parallel. That way reverse and over voltage is taken care of.

 

High value ceramic caps have special considerations - they tend to be voltage and temperature sensitive as in the capacitance decreases with voltage and temperature. Add in ageing too for some fun. The datasheet needs to be read carefully. eg a 100uF 6.3V X5R only gives around 20uF at 4V. Even less if we factor in temperature and ageing.

 

There's mention of the mosfets controlling relays - the relays will need catch diodes on them. Other solutions are using TPIC6595 chips that have 8 drivers and a shift register in them. These have clamp diodes built in. There's also the ULN2803/4 drivers. These have catch diodes built in. These devices will simplify your pcb design.

 

Your board looks autorouted as there is strange routing - like around the crystal. These tracks should be short and direct without going through vias if possible. You also want to keep other tracks away from them and have a groundplane on the bottom layer. 90 degree angles on the tracks are a contentious issue. Personally I don't like the look of them, so I use 45degree angles. Most pcb packages do this for you. There were stories of acid traps and impedance problems, but these seem to have been dismissed.

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> just some measly 0.1uf caps ...

Standard bypass caps. The fets only draw a few uA, so I thought the .1s would be fine. What would you suggest?

> Otherwise you are just pouring your copper down the drain

I don’t understand this comment.

> Trace calculators aim to prevent the traces from burning up, not for high performance.

I’m more concerned about things burning up than performance. The fets won’t be fired faster than 15 Hz. I’d also like to keep the board as small as possible for cost reasons. Based on your, and other, comments, though, it looks like I’ll be taking another pass at it, and if it has to be larger, so be it.

> That's an absolutely horrible fet (STN4NF20L), unless you need to switch 200V

What’s so terrible about it? It was one of a handful that came up in a parametric search on Digikey. My requirements are logic level gate drive, 12V, 350 mA, 4W capacity, small size, smd, low cost. I don’t need high performance for switching a relay. What would you suggest?

> All your fets should connect to a gnd plane section on the bottom

Wouldn’t it be better to put the ground plane on top so I could use it for heat dissipation?

> It's a ridiculous layout to put all the gate resistors way over ...there

At the time, it seemed like there would be complications wherever I placed the resistors (I even considered using a packaged array). I’ll try as you suggest in the new layout.

> The labeling of the MAX3061 is complete bogus...you can't tell what those pins do, same for the AVR...label the pins with their functions. $P3$38 doesn't cut it!

Already fixed for the MAX. AVR next.

> Why would you control 1000 relays using fets?

It’s a kinetic sculpture. The appearance, and sound, of the relays are essential.

Thanks for the feedback!

Karl

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Thanks for the tips, Kartman, particularly the ULN2803. Completely forgot about those things, but they could solve a lot of problems (especially cost and size), if they can handle the power of all 8 relays simultaneously. Have to check the data sheet. Btw, don’t MOSFETs have check diodes built in? I’ve been using them for all kinds of inductive loads for years without external diodes for just that reason. Never had a problem.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> just some measly 0.1uf caps ...

Standard bypass caps. The fets only draw a few uA, so I thought the .1s would be fine. What would you suggest?

***Put a 4.7uf cap at the AVR to handle any bulk loads  0.1 good for noise

 

> Otherwise you are just pouring your copper down the drain

I don’t understand this comment.

****when you use thin traces ...where do you think all the etched copper ends up...down the drain (or hopefully recovered).  Thick traces use the copper you were given.  What do you gain by throwing it away?

 

> Trace calculators aim to prevent the traces from burning up, not for high performance.

I’m more concerned about things burning up than performance. The fets won’t be fired faster than 15 Hz. I’d also like to keep the board as small as possible for cost reasons. Based on your, and other, comments, though, it looks like I’ll be taking another pass at it, and if it has to be larger, so be it.

*** you have plenty of room...you are far far from being size limited.

 

 

> That's an absolutely horrible fet (STN4NF20L), unless you need to switch 200V

What’s so terrible about it? It was one of a handful that came up in a parametric search on Digikey. My requirements are logic level gate drive, 12V, 350 mA, 4W capacity, small size, smd, low cost. I don’t need high performance for switching a relay. What would you suggest?

****So why pick a 200V part...all you get is extremely high ohms (1.5ohms!!!)***...even 0.1 ohm is pretty poor these days.

****almost any other fet will be better...pick again  (search for around a 40V part)

also see ZXMS6004 dual & ZXMS6005 protected fets...higher resistance, but no burn out!

https://www.mouser.com/ProductDetail/Diodes-Incorporated/ZXMS6004DT8TA?qs=sGAEpiMZZMshyDBzk1%2fWi8Ex7o5z%2f%252bUs%2fymCgplIxYg%3d

 

> All your fets should connect to a gnd plane section on the bottom

Wouldn’t it be better to put the ground plane on top so I could use it for heat dissipation?

**** could be, but you'll get plenty cooling on the bottom.   Keep room on top for your traces.  Use a 0.1 ohm fet or less & heat will be 15x lower than 1.5 ohm fet

 

> It's a ridiculous layout to put all the gate resistors way over ...there

At the time, it seemed like there would be complications wherever I placed the resistors (I even considered using a packaged array). I’ll try as you suggest in the new layout.

**** goody

 

> The labeling of the MAX3061 is complete bogus...you can't tell what those pins do, same for the AVR...label the pins with their functions. $P3$38 doesn't cut it!

Already fixed for the MAX. AVR next.

> Why would you control 1000 relays using fets?

**** the question is , why use relays at all...MUCH less reliable...maybe they are already present & no choice.

**** You will likely not need diodes, the fet's built in zener should easily handle the flyback, If you were driving 10 amps at 10KHz might be a different story, even then perhaps not.

Have used larger power fets at 40amps for motor PWM, no diodes needed.

 

It’s a kinetic sculpture. The appearance, and sound, of the relays are essential.

****Maybe add some buzzers, which can be made to "click"

Perhaps these are open relays, where you can see them in action? (see 1:50  https://www.youtube.com/watch?v=n3wPBcmSb2U)

 

 

 

Thanks for the feedback!

Karl

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Writing such a post from what I see from a (not so clear) schematic / pcb is quite time consuming.

I am perfectly aware that any sort of GND connection is often "forgotten" on simplified RS485 examles.

Have you looked at documents such as "10 ways to bullet proof RS485" (I believe an old AN from national instruments, but quite good).

Twisting wires is always good for RS485. Shielding is ... not needed for such short runs.

Quite fancy / flashy photo, is each grey blob on the foto a RS485 node? You can not put so many nodes on a single cable.

 

If you put everything so close together, the ruggedness of the RS485 drivers becomes less of an issue.

170A over distributed power supplies is however a point for carefull consideration.

Solid GND & power connections becomes very important.

You do not want any signal lines to conduct a significant part of that current.

That will melt your wiring and give you a bad name.

But I do not have enough overview here for a complete overview of your product.

 

About SMPS:

A nice little chip such as for example LM2841 has an quieschent current of 1.5mA and a typical efficiency of 85%.

You do the math :)

I am a bit unpleasantly surprized by the mouser price though.

There are however hundreds (thousands?) of different small SMPS circuits nowadays, and you can buy build PCB's (working, with chip inductor and everything) from prices of 50 cent or so from Ebay / Ali. My guess is that Mouser / Digikey / Farnell / etc have proffit margins of 300% or more or have not seen a need to optimize their workflow costwize for 30 years or so. (I am just guessing here, such a remark can easily backfire). But the price difference with asian shops is ... astounishing.

But compared to your 12V 170Amps the power requirement of your microcontollers is ... almost nothing.

The "bulk" capacitor can indeed be placed almost anywhere on the PCB, I skipped a beat there.

If power consumption of the relays is an issue there are loads of ways around that, such as:

- Switch Relay @ 12 V and then PWM an output to about half that voltage (1/4 of the power to relay).

- Use bi-stable relays.

- Why use relays? What are they switching? Those things are noisy. Is that an issue?

(Too much micro management from my side, probably I should not even ask such questions).

 

(Poly) fuses should of course be calculated from the current they need to pass. How would I know what fuse you need?

These are wonderfull "self repairing" devices, but have some gotcha's. They open up when heated (Also by an exernal heat source) at quite low temperatures. They are also not very accurate and really only designed for preventing gross faults under severe overload conditions.

Experiment with them before you implement them.

 

Trace size calculations are often only based on temperature rise of the trace (often 10 Celcius?) You also have to consider voltage drop over the trace etc.

It's very simple: Wide traces are better for big currents.

Narrow traces for high current traces with also a via in the middle is bad practice.

You can very easily draw stubs on the GND trace and move all the vias to the MOSfets to those stubs, to at least preserve the with of your trace.

(Especially small) via's have a limited current handling capability. It is very common to use "via stitching" (Google that) with multiple vias to get a high current trace from one side of the board to another.

But really, the whole PCB design is ... very marginal at best. You clearly have to learn some tricks there.

Browse a bit around. Have a look at other PCB designs.

Optimal PCB design is also an art in itself and the art in decent engineering is far to ofthen not acknowledged.

Or even invisible, which is a compliment in itself (it just works).

Paul van der Hoeven.
Bunch of old projects with AVR's:
http://www.hoevendesign.com

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

lautman wrote:

The fets won’t be fired faster than 15 Hz.

 

It's not the repetition rate that's the problem but the rise-time of edges. Remember that a square wave has a frequency response that extends out to infinity. Think of an inductor (ie length of track) as a bell. Hit it with a hammer (ie edge of a pulse) and it'll ring for a long while.

"This forum helps those that help themselves."

"How have you proved that your chip is running at xxMHz?" - Me

"If you think you need floating point to solve the problem then you don't understand the problem. If you really do need floating point then you have a problem you do not understand." - Heater's ex-boss

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Now hit that bell 15x per second with a hammer ...

(But don't do that when I am in the neighbourhood).

 

Something often used here is to use series resistors in the gates.

They form a low pass filter with the internal capacitances of the FET to make them switch slower on purpose to make the switching flanks (and thus frequency content) more gentle.

It's like hitting the bell with a padded rubber hammer instead of a steel hammer.

 

 

Paul van der Hoeven.
Bunch of old projects with AVR's:
http://www.hoevendesign.com

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

 

Something often used here is to use series resistors in the gates.

Yep---I always put some resistance in the fet gate (usually placed at the fet rather than at the micro) , maybe 1K ohm...then if the fet shorts (such as when I drop my screwdriver), it doesn't just blow 20 amps into my AVR.

 

But compared to your 12V 170Amps the power

I believe OP is at 170mA, what's a factor of 1000x between freaks.

 

I recently saw on a post here (or somewhere) some newer polyswitches that have the tranzorb/zener combined in one...that solves some "issues" since a reacting (very hot) zener then helps to "pop" the polyswitch quickly (before the zener fries).  In the old days, you'd pick a massive zener that could take the heat, waiting around for the polyswitch to kick in.

 

 

 

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

avrcandies wrote:
... that have the tranzorb/zener combined in one.
Polymer-enhanced Zener diode or PolyZen :

Littelfuse

PolyZen Devices for Overvoltage-Overcurrent Protection

http://www.littelfuse.com/products/polyzen-devices.aspx

avrcandies wrote:
In the old days, you'd pick a massive zener that could take the heat, waiting around for the polyswitch to kick in.
Simple and effective especially for signals where some impedance is acceptable.

Rugged Circuits

Rugged Circuits

10 Ways to Destroy an Arduino

https://www.rugged-circuits.com/10-ways-to-destroy-an-arduino

...

Method #3: Apply Overvoltage to I/O Pins

...

 

"Dare to be naïve." - Buckminster Fuller

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> also see ZXMS6004 dual & ZXMS6005 protected fets...higher resistance, but no burn out!

 

Yes, those are much better.  It looks like each channel is built around a MOSFET in the datasheet, but they don't show the body diode.  I count on those for protection when using MOSFETs with inductive loads.  Do you know if there is one, or I'd need to provide one?  Also, how do you suppose it would compare to a ULN2803, as Kartman suggested?

 

> Perhaps these are open relays, where you can see them in action?

 

Correct.  Like this one.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Whilst the mosfets have a body diode, where does this the energy go? You rely on the impedance of the circuit, but there’s nothing specific to waste the energy. Also consider the loop that is formed by your relay circuit - this will most likely radiate. You’d be better off putting diodes or varistors on the relays themselves.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> Writing such a post from what I see from a (not so clear) schematic / pcb is quite time consuming.

 

I know.  And I greatly appreciate the time you've put into helping me.

 

> Have you looked at documents such as "10 ways to bullet proof RS485"

 

Not that particular one, but I found it and added it to my stack of reading.

 

> Twisting wires is always good for RS485. Shielding is ... not needed for such short runs.

 

I'll definitely be using twisted pairs, but the runs will only be about 18 inches.  While short, I'm alittle concerned about all the RF radiated by so many relay coils.  Still, swapping the cables for shielded ones, or just starting with them, wouldn't be difficult.

 

> Quite fancy / flashy photo, is each grey blob on the foto a RS485 node?

 

No, each blob is a relay (like this:  http://tinyurl.com/yabsrzst).  Each 4 x 4 arrangement of 16 relays will be a node, 64 nodes total.  Each driver presents a 1/8 unit load, so the bus can handle up to 256 of them, in theory.

 

> You do not want any signal lines to conduct a significant part of that current.

 

The intent (and, hopefully, design and implementation) is that the current consumed by each relay flows from the power supply to the coil to the FET to ground.  It should never come close to a signal line, except at the FET.

 

> About SMPS:

 

I'll revisit this.  

 

> But the price difference with asian shops is ... astounishing.

 

I agree, and for my own, one-off stuff, I use those little buck/boost boards all the time.  I'd never build one into a product like this, though, due to quality concerns.  Plus, I'm not wild about adding PCBs to my PCBs if I can build the same functionality directly onto the board.

 

> Why use relays? What are they switching? Those things are noisy. Is that an issue?

 

As a piece of kinetic sculpture, the appearance, movement, and sound of the relays are essential components of the piece.  They're actually not switching anything.  Ultimately, I hope to hire a percussionist to compose "music" to be played on it.

 

> (Too much micro management from my side, probably I should not even ask such questions).

 

Not at all.  There's no telling what may come up from "such questions."  ;-)

 

Last Edited: Thu. May 17, 2018 - 12:23 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

 

Whilst the mosfets have a body diode, where does this the energy go?

It all flows back to my house to help lower my electric billssurprise 

Actually, the body diode looks like a zener of the mosfet's rated voltage (say 40V)...so when it avalanches it looks like a 40V drop conducting the current that was previously flowing (say 1.23 amps)  so that V times A  is those watts disappearing. Of course, doing this at 2KHz can heat up the fet (diode), but it is a power fet!  Some smartfets have active clamping to actively quench the transient by slightly turning back on (sort of applying the brakes).  Since the excess energy is doing the turn on, as it dissipates, the fet goes back to being shut off.

Most normal fets seems to do pretty well in this regard, but there are specific "avalanche rated" fets that have toughened (and/or more fully characterized) body diodes....more imperative at high currents/fast repetitions.

 

https://www.mouser.in/pdfDocs/diodes-inc-automotive-mosfets.pdf

 

 

The ZXMS do not require diodes ---they have the active clamping and several other protections:

• Short circuit protection with auto restart

• Over voltage protection (active clamp)

• Thermal shutdown with auto restart

• Over-current protection

• Input Protection (ESD)

• High continuous current rating

 

also see fig12

https://www.onsemi.com/pub/Collateral/AND8202-D.PDF

 

 

 

 

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

As impressive as the ZXMS is, the ULN2803 suggested by Kartman is looking better and better. It takes less space than 8 FETs, has flyback diodes and pulldown resistors built in (further reducing space), requires only two traces per relay (instead of three for a discrete FET, simplifying routing), it’s cheap, and can handle all my current (2.5A rated vs. 1.4A in my application). It takes more current to drive than a FET, but still only 1 mA/IO, or 16 mA total for me, well within the Mega’s ability. Any thoughts about this?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

You have to be carefull with the maximum dissipation for the ULN2803.

It can not handle a continuous total current of 2.5A. Where does that 2.5A come from?

Different brands have differences in datasheets, make sure you get the right one.

A motorola datasheet I looked at has no info on this (didn't look to carefully)

A toshiba datasheet https://cdn-shop.adafruit.com/datasheets/ULN2803A.pdf limits power dissipation to 0.5W for the whole package with a Vce saturation voltage of around 1V, which means 1/2A total continous current for the whole package.

Paul van der Hoeven.
Bunch of old projects with AVR's:
http://www.hoevendesign.com

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I’ve been referring to the TI data sheet for their ULN2803A(http://www.ti.com/lit/ds/symlink/uln2803a.pdf). It implies a max. current sink of 2.5A, thought I’m confirming with TI.

Last Edited: Thu. May 17, 2018 - 06:17 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

As Paul suggests, you need to be aware of the thermal limitations. You can’t take one spec in isolation. Mosfet specs are one too look out for - the spec might say good for 100A, but do the loss and thermal calcs might suggest it will do 100A on an infinite heatsink for a microsecond!
Also, your relays look like big ones, so you need to consider if the internal catch diode is man enough for the job.
You could always go to 24VDC and halve the relay current.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The ULN2803 seem a bit marginal, if you need to keep the relays all on for any time...maybe ok if just "clicking them"...at 170ma, you'll drop about 1V...relay might take more current, if it's resistance is not precise (or your 12V  becomes 12.5V)

 

Might consider  active clamp https://www.diodes.com/assets/Datasheets/DMN61D8LQ.pdf  ..it has an awful Rds, but at 170ma, still only about 300mv drop.

It's cheap, small (sot-23) & lets you spread the heat out (plus @5V logic, the total heat is about 1/3 of the ULN part)

 

There are probably similar parts, with lower Rds

 

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

That DMN actually looks like an excellent part for this application, much as I’d miss the simplicity of the ULN. It certainly has a lot more margin than the ULN. (1.8 x .17) x .17 = 52 mW means temperature wouldn’t be an issue, plus the active clamp and integrated pulldown resistor. I ordered some to test. Thanks for the pointer.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

note you can also get a dual version:

https://www.diodes.com/assets/Da...

 

By the way, how will you hook up 1000+ relays with the rs485---it is limited to 32 nodes, are you using multiple uarts? 

If the relays are located close together, you could implement a board to board shift register/spi approach.

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> note you can also get a dual version:

 

 

Jeez, your part search Fu is way better than mine.  ;-)  That part may actually take up more space than individual FETs, though, given the fat traces that have to be routed for drain and source.  I'll play around with it to see.

 

> By the way, how will you hook up 1000+ relays with the rs485---it is limited to 32 nodes, are you using multiple uarts? 

 

 

The piece will be assembled out of modules, each containing 16 relays and a controller board based on an ATMega32A.  16 of the 28 IOs go to the relay FETs, 7 go to a dipswitch to set the RS485 address, and the UART goes to the RS485 transceiver.  The transceiver I'm using presents a 1/8 unit load, so I could have up to 256 modules on the bus; 64 shouldn't be a problem.  A host controller will send 3-byte packets to all the modules on the bus:  1 byte for address, and 2 bytes representing the state of the relays for the associated module.  Worst case, I should only need about 300 kbps (the transceivers handle up to 500 kbps), but I think I can get it much lower.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

A host controller will send 3-byte packets to all the modules on the bus:

 

Think carefully about fault tolerance on your project.

 

You might well wish to send your data in a small packet, with a Start of Packet character, your three data bytes, and then a CheckSum byte, (or a CRC Byte, instead of the CheckSum).

 

That would allow the receiver to validate its data reception prior to (re)setting the relays.

 

The next level of data checking would be to have the remote units send an acknowledgement back to the main controller...

 

JC 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

All good ideas.  Just haven't gotten that far into the project.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

You can also have an address that commands all units, then easy to clear all, set all, or set all to the same pattern.

 

You can also have an address/value that sets the specified unit into always rcv (so the selected unit takes action regardless of the address)...you could set a few such modules , then  those modules would respond to all patterns sent out.  some variation  would bring it back to normal address-checking operation.

 

 7 go to a dipswitch to set the RS485 address

Bleh....I use a button & store the address identity in the EEPROM...one command says "config unit 19"...any button press & that (or those) units are now marked as being unit 19.  I also have a slightly different command that autoincrements the call #, upon getting a buttonreponse from anyone.  So you give AUTOID13 command & then just start hitting buttons on modules...first one pressed will now be stored as #13, the next as #14,.... #15, ...etc.  Also, certain commands are responded to by all modules, regardless of their address (such as system standby command).

 

Since the address is being set in software, you could also possibly give a module more than one address (maybe allow a list of up to 8 addresses per module)..so by sending to a shared address, multiple boards couldinstantly  respond (might be an interesting effect).

 

You MUST  include an led on your module, so you can see that it is responding to the present command.  You will severely regret not having one!  Also add a pushbutton.

 

The next level of data checking would be to have the remote units send an acknowledgement back to the main controller...

 

I'd prob avoid that initially unless you are using full duplex.  Half duplex requires extra thinking/fiddling, which I did , since it was done by a half-duplex radio link.

 

When in the dark remember-the future looks brighter than ever.

Last Edited: Sun. May 20, 2018 - 10:26 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

DocJC wrote:
That would allow the receiver to validate its data reception prior to (re)setting the relays.
lautman wrote:
All good ideas. Just haven't gotten that far into the project.
Really something to think of before you you have your PCB's manufactured.

The last revision of the PCB I saw was receive only for the "remote" nodes.

 

Also:

If there is a "master" node who controlls all, I would probably add a bunch of RS485 chips to a single UsART there. ( 8 is a nice round number).

You can tie them all to the same UsART by toggling the driver & receiver enable lines in a smart way.

This will considerably reduce the loading on the RS485 bus and will also simplify trouble shooting.

Paul van der Hoeven.
Bunch of old projects with AVR's:
http://www.hoevendesign.com

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

These are great ideas, avrcandies, and I'd love to get away from the dipswitch.  My concern is that the piece is likely to be assembled and disassembled several times during its life, and I may not be present for each assembly, so I have to anticipate what may go wrong.  Of course, it's essential the modules be installed in the frame in a precise order, but, even though each will be labelled on the back, that may not happen.  If one module is out of order, then two will be out of order, and some (potentially a lot) of disassembly/reassembly will be required.  On the other hand, if each module's address could be changed in the field, via the dipswitch, disassembly/reassembly wouldn't be required.  That said, it might be safer overall to just disassemble/reassemble than have unskilled labor fiddling with dipswitches on PCBs.  Plus, any readdressed board will no longer match the order number printed on their panel.  As you can see, I'm still working through this, so I appreciate the nudge.

 

As for an LED, I could more easily just use a relay to signal feedback.  Maybe LEDs on the Rx/Tx lines, though I'm bringing those out to test points for debug during development.  Not sure how valuable LEDs would be once the thing is in the field.  If it stops working, and I have to go fix it, I'm going to need a scope and a bunch of other stuff anyway.

 

I certainly should a button, though.  Thanks for that.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> The last revision of the PCB I saw was receive only for the "remote" nodes.

 

True, as I didn't think I'd have the bandwidth for a resend anyway.  Now I think I will so I'll enable half-duplex.

 

> If there is a "master" node who controlls all, I would probably add a bunch of RS485 chips to a single UsART there. ( 8 is a nice round number).

 

Excellent idea.  Thanks.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

This project seems like a PLC(programmable logic controller). Why not just use one of them? The scale of the project suggests there’s a few dollars involved, so failure or unreliabilty is not going to make the customer happy. Just wiring the system together is a big task, some consideration here is going to save time and money as well as contribute to the overall reliability and maintenance of the system.
Personally, i don’t think RJ45s are going to cut it. I’d be using plug in screw terminals. More expensive, but easier to terminate and more robust.
I’d look at the following:
24V relays to drop the current requirements down.
Pcb mounting - maybe use DIN rail mountings/enclosures. A lot of industrial gear uses this method. Cuts down on screws etc and makes it easier to design, build and maintain.
Serial drivers like the tpic6595. Makes your pcb design easier and less i/o required.
Blinken lights. Add leds so you can see what is working. For rs485, a tx and rx led is a great diagnostic tool.another led for ‘heartbeat’ is also another great diagnostic tool.
Consider the implications of managing the relay back EMF. Having a diode across the fet protects the fet, but shunts the energy back into the supply. Multiply that by 100 relays and you’ve got a recipe for disaster. Best to put a diode or varistor on each relay on its terminals. The energy gets wasted as heat and is spread over the number of relays. Your board never gets to see it.
Voltage drops can ruin your day.don’t discount the amount of current that might be flowing that will cause voltage drops. The size of the conductors and how your system is wired is crucial to reliability. A star technique is useful here.
As the Doc mentions, your serial protocol is inadequate. You might want to look at an industrial protocol like Modbus.
If you’re in the western world, you’ll most likely have to consider EMC. If your system radiates and causes interference, then you’re in for trouble. I’ve had the regulatory body knock on my door.. Seems the cell providers don’t like interference with their frequencies. Seems a pcb track and mosfet can make a nice transmitter. It was only meant to switch a relay.

Last Edited: Sun. May 20, 2018 - 11:21 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I'm aware of PLCs but have never used them.  I'm not sure something like this would be a good learning exercise for them.  While this will certainly be the biggest thing I've ever made, more of it is within my range of experience than if I went with PLCs.

 

As for PCB mounting, I'm planning on putting them inside off-the-shelf enclosures, tbd pending some idea of how big my board will be.  I haven't decided yet if I'll have the four cable jacks penetrate the enclosure, or just bring all the cables inside through a single cutout and then plug into the jacks on the board.  The former would be easier to assemble/disassemble in the field, while the latter would provide better ingress protection from dust.  In either case, all penetrations of the enclosure will be pointing down.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

As for an LED, I could more easily just use a relay to signal feedback

Don't bet on it...you want to perhaps see a specific condition/thing occurring...an led will tell...listening for a relay to click (out of hundred), you'll wish for an led.  The led can blink, signaling it is ready & waiting for you to push the identity button.

 

Note, even without a dipsw, each module can be addressed (configured) as previously mentioned & then given to the the assemblers...the address is just read from the EEPROM rather than from the dipsw.

example:

load module with new software

send SETID23 command

led starts blinking

Push button....module has been set & saved as unit #23

All done...remove power & program next module

===============================

Optionally, later on (finished system)

Send SETID18 command

All modules start fast blinking

Push button(s) of module(s) desired to have address #18

Send SETID22 command

Push button(s) of module(s) desired to have address #22

Send ENDID  ...all modules leave identity setting mode (no fast blink)

 

===============

Remember, each relay requires 2 wires..will you bring 2 per from your board...you should if they are  in wide-ranging locations (making daisy chaining a poor choice).

Draw up your PHYSICAL cabling carefully BEFORE designing your connector  system!!!

 

Spring loaded connection blocks better than screw terminal types...springs won't work loose!  Push the button, insert wire, DONE

 

 

 

 

 

 

 

 

When in the dark remember-the future looks brighter than ever.

Last Edited: Mon. May 21, 2018 - 01:03 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> listening for a relay to click (out of hundred)

 

I figured each module would have its address flashed per your protocol during assembly (of the module, not the sculpture).  Since I'd be assembling the modules one at a time, and each module only has 16 relays on it, it would be pretty easy to hear the relay.  I was originally thinking this way because I was running out of IO.  If I omit the dipswitch, I'll have plenty (though I could put the LED on one of the relay IOs, I guess).

 

> Remember, each relay requires 2 wires..will you bring 2 per from your board...you should if they are  in wide-ranging locations (making daisy chaining a poor choice).

 

There will be 16 relays on a board roughly 12" x 12".  My plan was to daisy-chain the 12V through the relays and just bring the other side of the coils back to the PCB.  I was also planning on buying RJ45 cables, cutting them in half, and using two halves to serve as the return lines to the PCB, using the RJ45 connectors to connect to the board.  Much denser than spring-loaded or screw connectors, and the cabling is cheap and already made up (not sure if/how I'd terminate the cut end that attaches to the relays).  Of course, these cables would be heavily strain-reliefed so somebody can't yank them out of either the relays or the PCBs.  Once assembled during manufacturing, though, they wouldn't have to be disassembled (hopefully), unlike the RS485 cabling, which would have to be removed each time the piece gets broken down for transport.  Still not sure about the right connector for that (RJ12?).

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

AVRCandies,

 

I just purchased a handful of 4-terminal spring  connectors with through holes mounting pins from Spark Fun.

 

I looked for them at Mouser and I couldn't come up with the right tern to locate them there.

 

Do you have a vendor / manufacturer / model number for the spring terminal block you posted the photo of?

 

Thanks,

 

JC

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

 

Do you have a vendor / manufacturer / model number for the spring terminal block you posted the photo of?

 

Here you go, Doc....$0.54 @100

https://www.digikey.com/products/en?keywords=277-7675-ND

 

Also, click on terminal blocks, wire to board  at the top to see others

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Given a board with, say,32 wires, would you like to change it over? Whilst it is easy in your workshop, removing 32 wires, replacing the board then ensuring you put the correct wires in the correct terminals whilst on site is probably not something you really want to experience. Plugs are more expensive, but there is no comparison when you need to change a board on site. For me, on site usually meant 30+C, having to squat down and having sweat from your brow drip into your eyes and/or whatever you're working on. After sufferering whilst replacing a board with 50+ connections, I made a vow to use plug in connections. I never regretted it.

 

 

Is CAT5 cable rated for the current the power supply can source? It is most likely a legal requirement that you follow the local wiring rules. These usually require that the wire and insulation used is suitable for the worst case current flow and temperature of operation. Therefore if your power supply is 10A, any wiring connected to the power supply needs to be suitable for that current. If not, you need to provide a fuse or circuit breaker, again, rated to the capacity of the wire. The relay may only be 170mA, but if your power supply can deliver , say 10A, then wire accordingly. Remember - "a fuse is to protect the wire, not the load".

 

Consider where the unit is going to be mounted. Who is going to install it? Who is going to maintain it? If there is fixed wiring, then it may require an electrician to install it. If it has mains wiring in it, the enclosure may need to comply with certain regulations. If you look at switchboards in the area you are going to install it in, see what enclosures they use. The electrical regs are in place to ensure you don't burn the building down or electrocute or injure someone. What you do at home won't cut it in a public area. There are 'standard' ways of doing things and products are built around this. Do something different and anyone schooled in the art will be immediately suspicious. You may want to obtain the services of an electrician or suchlike to advise you.

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

but if your power supply can deliver , say 10A, then wire accordingly.

That might be a reason to use something like the ZXMS6004...since it would limit current per output, in case of a short.  Be careful to not use telephone wire that is more like "tinsel"  ...extremely flexible, but wouldn't be much good for current. 

It is also avail as a dual device:

(much cheaper at Mouser $0.44, than Digikey $0.74  @100):

https://www.mouser.com/ProductDetail/Diodes-Incorporated/ZXMS6004DN8-13?qs=%2fha2pyFaduinE%252bJ69nksx2pvWRG5ceARoqJwK14HJv8lL%252b7W5miUKA%3d%3d

 

https://www.digikey.com/product-detail/en/diodes-incorporated/ZXMS6004DN8-13/ZXMS6004DN8-13DICT-ND/5724360

 

 

 

 

 

When in the dark remember-the future looks brighter than ever.

Last Edited: Mon. May 21, 2018 - 03:42 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> Is CAT5 cable rated for the current the power supply can source?

 

Power-over-Ethernet permits 600 mA/Cat5e pair.  The smallest wires used in Cat5e are 24AWG which is rated at 577 mA.  I'm using 170 mA/wire so it seems OK.  Certainly I'll fuse the main supply line to each panel which will have a nominal consumption of around 3A.  As an artwork that will be doing some travelling, I'm just going to have to take my chances with local regulations, but it's a good idea to run the high-voltage wiring past an electrician, even though it's being designed to just plug into regular mains.

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> That might be a reason to use something like the ZXMS6004.

 

I like that part but I can't figure out from the data sheet if the gate has a pull-down resistor on it like the DMN61D8LQ, which also comes in a dual package.  I left a message for Diodes about this last week and they never called back.  Just tried again and they didn't even answer.  Sent an e-mail to Digikey to see if they can find out.  The dual version also lays out better than the dual DMN61D8LQ, but nobody has stock (in the US) and there'a 16 week lead time.  I also asked Digikey to verify the part is really Active, as they say it is.  The dual DMN61D8LQ has a lot of protection already, so it may be adequate.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Power-over-Ethernet permits 600 mA/Cat5e pair.  The smallest wires used in Cat5e are 24AWG which is rated at 577 mA.  I'm using 170 mA/wire so it seems OK.

24 gauge copper wire is has 84.1976 milliohms per metre.  At 170 mA, one metre will have a voltage drop of 14.3 mV on the positive, and a voltage rise of 14.3 mV on the GND, for a total drop of 28.6 mV.  A 10 metre cable will see 0.286 V drop.  Make sure your gear can deal with that.

"Experience is what enables you to recognise a mistake the second time you make it."

"Good judgement comes from experience.  Experience comes from bad judgement."

"When you hear hoofbeats, think horses, not unicorns."

"Fast.  Cheap.  Good.  Pick two."

"Read a lot.  Write a lot."

"We see a lot of arses on handlebars around here." - [J Ekdahl]

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

My runs will be less than a foot.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Cosider ZXMS6004DT8TA...slightly more $$ but in stock now..I use it in my design...take careful note of the package (SM-8)  on the datasheet (does not compare to digikey website image!)

 

The other one will be back in stock at mouser June 11 & digikey june 3   Mouser has 60,000 coming in batches.

 

The internal fet circuit draw about 100uA , so probably acts as a "pull down"...but not needed anyhow, the processor output will be high or low (once out of reset & ports set up).

You might try one as a test--BEST bet!   I'd probably leave a spot for pulldown at GS.  I also like to have some series resistance to help prevent popping the micro should the fet blow.  Probably 1k for this part.

 

When in the dark remember-the future looks brighter than ever.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Just spoke to a Diodes FAE. No internal pull-down resistor, so I think I’ll stick with the dual 8LQ if it tests OK.

Pages