archang has quit [Remote host closed the connection]
archang has joined #qi-hardware
fengling has joined #qi-hardware
DocScrutinizer05 has quit [Disconnected by services]
DocScrutinizer05 has joined #qi-hardware
sandeepkr__ has joined #qi-hardware
GeorgeHahn has quit [Read error: Connection reset by peer]
xiangfu has joined #qi-hardware
FDCX has quit [Ping timeout: 244 seconds]
FDCX has joined #qi-hardware
icanicant_ has joined #qi-hardware
pcercuei has joined #qi-hardware
<DocScrutinizer05>
>>I measured the gate of the MOSFET with a scope and it gives me a very flat 24V.<< !!!!!
<DocScrutinizer05>
provide circuit diagram with testpoints you used for the measurement and I can tell you what exactly is wrong. So far I just can tell that either your probing or your circuit (or both) is flawed
<DocScrutinizer05>
for the FET we also need the particular type you used (either part number or at very least whether it's P or N, depletion or enrichment type)
<DocScrutinizer05>
shevek: ^^^
<DocScrutinizer05>
also please provide details about the PSUs you used
fengling has quit [Quit: WeeChat 1.4]
<DocScrutinizer05>
searching Mouser I hardly found FETs with Vgs >25V (and that's ABS MAX). Vgs_th (recommended) maximum I found is 12V, usual are up to maybe 5V
<DocScrutinizer05>
most are 0.6 to 2V though
<DocScrutinizer05>
I hope you didn't connect drain to +24V and load from source to GND
<DocScrutinizer05>
otherwise odds are your FET autoadjusts to Vsource = Vgate - Vgs_th (up to 3V maybe), so when you tie gate to source level (24V) you have a autoadjusting voltage of Vds_th between drain and source. multiply that with the current of 4A and you're in the 2figure Watts dissipated by FET
sandeepkr__ has quit [Read error: Connection reset by peer]
<DocScrutinizer05>
you see there's no need to go higher than 5-10V Vgs with the latter one, and when you have load in Source and tie Gate to Drain, you'll have a ~3V across the Drain-Source pins, no matter which current you draw
<DocScrutinizer05>
you also see that in a load-in-Source circuit, you need to drive the Gate with 4-10V *higher* voltage than Vdrain
<DocScrutinizer05>
yes, the last is Vdrain, not Vsource
<DocScrutinizer05>
there's honestly hardly any good reason to use load-in-Source circuit for on/off applications
<DocScrutinizer05>
the exact opposite applies for analog regulators etc, like Voltage regulators and audio amps, where you regularly use load-in-Source circuit - and you regularly expect the FET to get hot
<DocScrutinizer05>
another aspect: both of the quoted FET types would escape magic blue smoke with Vgs=24V, it's far beyond ABS MAX for both of them. No matter where the load
<DocScrutinizer05>
and even when you got your load in Source, in the moment you apply Vhigh (=24V) to Gate, your Source ist still 0V since the FET hasn't opened yet, so you'll blow the FET any way
<DocScrutinizer05>
to mitigate that effect, you need a series R in gate line, to get a sufficiently slow rising slope on gate's capacitance
<wpwrak>
yes, Vgs(max) tends to be quite limited. surprised me, too. in anelok, i have some FETs with the gate on VBUS. then i tried to vary them, just to notice that the new ones i was looking at had Vgs(max) = 5.0 V. not good :)
<wpwrak>
shevek: but in any case you should post a circuit diagram. hard to tell what's going on in your circuit. if the FET is intended to be either fully on or fully off, it shouldn't burn much power (if properly dimensioned). if it ends up somewhere in the resistive zone of operation, it may very well become an indian communication device (smoke signals)
<wpwrak>
also, how is the gate modulated ? just on/off at a low frequency ? PWM ? ...
<DocScrutinizer05>
yeah, another possible failure mode I considered (but rejected as too little plausible) is (PWM) modulation with very high frequency, several MHz
<DocScrutinizer05>
anyway I wonder how an arduino(?) is driving a 24V gate voltage
<DocScrutinizer05>
without schematics that's all fortune telling
<roh>
well.. either not properly -> smoke signals, or with atleast one more transistor as driver i guess ;)
<roh>
what i dont quite get: why not just use classic low side switching and a logic level fet?
<roh>
its only a heating coil, isnt it?
<DocScrutinizer05>
yeah
<DocScrutinizer05>
source to GND, drain to coil, gate to arduino, done
<DocScrutinizer05>
basically
<DocScrutinizer05>
maybe use a normal transistor to control the gate
<DocScrutinizer05>
depending on VDD of arduino, and Vgs_th of FET
<roh>
4A is 100W.. i guess one could even still switch that bipolar, with a tip121 or such
<roh>
i would always suggest using a base or gate resistor still...
<DocScrutinizer05>
sure, but a FET has better properties
<DocScrutinizer05>
Rds_on is lower
<DocScrutinizer05>
yes, base R is mandatory, particularly in common-emitter circuit. Gate R is never a bad idea too
<shevek>
So you're saying I need a resistor on the gate? I don't understand what would be its purpose. I thought there's never supposed to be any current going through the gate?
<shevek>
I was originally using PWM on them, and that worked without the parts getting overly hot. Later I switched to full on/off because I thought this could be causing the heat.
<shevek>
(They started getting hot without me making any changes AFAIK. So I wanted to solve that problem, but I have no idea why it wasn't happening originally.)
<shevek>
Is it possible that the lack of a resistor on the gate could break the FET? Is there a current through the gate that I wasn't told about?
<shevek>
Or maybe the limited inductance of the heater coil (it's wrapped around a steel pipe, so there is some) could cause the FET to break? Would a diode like in the schematic that was referenced fix it?
pcercuei has joined #qi-hardware
pcercuei has quit [Remote host closed the connection]
pcercuei has joined #qi-hardware
sandeepkr__ has quit [Read error: No route to host]
<DocScrutinizer05>
shevek: the first FET in source-to-GND circuit is absolutely OK. Recommended Vgs-on is 5V. You must make sure that arduino delivers an active 5V (or *at least* 3V) to gate for ON, and pulls gate to GND solidly for OFF
<shevek>
Yes, I'm using 5V power on the Arduino, it's at 4.8V on the digital output.
<DocScrutinizer05>
is the output a push-pull?
<shevek>
When it boots, the first thing I do is set the pin to output, and for off I set it low. The gate is never floating.
<DocScrutinizer05>
resp a internal pullup resistor configured to be enabled?
<shevek>
The internal pullup can only be enabled when it is set to input, and that never happens.
<DocScrutinizer05>
ouch
<shevek>
Is that a problem?
<DocScrutinizer05>
so how is 4.8V going to happen on gate?
<shevek>
The pin is set to output and I set it high.
<DocScrutinizer05>
is the output a pushpull?
<shevek>
I think so. It's an atmega328p, just like on an Arduino Uno.
<DocScrutinizer05>
for many GPIO on SoC you can configure if pushpull or open-collector
<DocScrutinizer05>
I'd add a 10k resistor from gate to +5V
<shevek>
Ah right. Yes, the atmega is pushpull in output, and either floating or pullup in input. I set it to output, so it is always pushpull.
<DocScrutinizer05>
when it's set to input, the gate is floating and FET in resistive mode
<DocScrutinizer05>
add a pullup or pulldown R
<shevek>
If the output is floating for some reason (because the chip is in reset), I'd prefer the channel to be closed, so pulling it to ground would seem better?
<DocScrutinizer05>
yes
<DocScrutinizer05>
this is also the explanation why it escaped magic smoke
<DocScrutinizer05>
keeping CPU in reset for a few seconds might already suffice
<shevek>
But it shouldn't ever have been floating. When 24V is applied, it is booting for a few milliseconds, after that the output is asserted.
<DocScrutinizer05>
^^^
<shevek>
It's nowhere near seconds.
<DocScrutinizer05>
add a pulldown and you're safe
<DocScrutinizer05>
100k
<DocScrutinizer05>
or 22k, depending on what the internal pullup rresistor of SoC
<DocScrutinizer05>
it should be internal R * 0.1
<DocScrutinizer05>
I dunno if arduino has weak or strong pullups
<shevek>
I'm not sure, but that pullup is never active.
<DocScrutinizer05>
ATMEGA328 has no PowerOn reset input?
<DocScrutinizer05>
aaa pin 29
<DocScrutinizer05>
so that'
<DocScrutinizer05>
s on P12, pin2
<DocScrutinizer05>
pull that to GND and see your FET go BBQ
<shevek>
Right; I don't do that. ;-)
<shevek>
But I am aware of that risk.
<DocScrutinizer05>
add a pulldown R to gate and you're safe
<shevek>
Yes, I'm planning on that for that reason.
<DocScrutinizer05>
best EE practice
<shevek>
But I don't understand why it blew up already; I don't think I did anything that was wrong.
<DocScrutinizer05>
never build circuits with self destruct function ;-)
<DocScrutinizer05>
your crystal might be faulty so the CPU fails for a while (or forever) to generate clock
<DocScrutinizer05>
seen that with AT89c51 chip
<DocScrutinizer05>
the circuit needed a finger snip to start up, literally
<DocScrutinizer05>
knock the PCB, so crystal starts mechanical vibration and suddenly the CPU booted
<DocScrutinizer05>
reproducable
<wpwrak>
regarding gpio input = Z: careful with clamp diodes. first, don't burn them. second, they can mess up the voltage. e.g. it's perfectly safe to have a NPN transistor (or an n-FET) to drive a p-FET gate at, say, 24 V, with an MCU at 3.3 V or such.
<DocScrutinizer05>
hmm?
<wpwrak>
do the same without transistor in the middle, you either burn the clamp diode or worse, or you get at least Vg clamped to the MCU's VCC. so the gate may be at ~3.6 V instead of the desired 24 V, when the FET is "closed"
<DocScrutinizer05>
I don't see how a 24V get on gate there
<shevek>
No, that's the other board.
<DocScrutinizer05>
forget the other board
<shevek>
That's a chinese board with a controller on it; it takes a logic input and for some reason puts the high voltage on the gate.
<shevek>
Yes, let's forget about it.
<wpwrak>
the one with 24 V / FET looks quite right. low-side switching keeps this simple
<DocScrutinizer05>
yes
<DocScrutinizer05>
just lacks a pulldown for gate when GPIO = Z
<shevek>
The load is 48 windings of nichrome around a steel pipe of d=2cm. You think that could be a problem with inductance?
<DocScrutinizer05>
umm yes
<DocScrutinizer05>
add a snubber diode
<shevek>
Ok. If that's what went wrong, I can have some confidence that with a diode it won't happen again.
<DocScrutinizer05>
48 windings around a steel pipe sounds like a pretty decent inductivity
<shevek>
I'll also add a pulldown, of course.
<shevek>
But I'm pretty sure that wasn't the problem.
<DocScrutinizer05>
:nod:
<shevek>
I didn't quite understand what you were saying about a resistor on the gate. I understand it's not needed here, but I don't see how it would ever do anything. Isn't the gate supposed to be zero current at all times?
<DocScrutinizer05>
yes, except for charging the gate capacitor
<shevek>
Ah, ok. With a capacitor it makes sense indeed.
<shevek>
Then it's just a low pass filter.
<DocScrutinizer05>
a FET has parasitic gate capacitance
<DocScrutinizer05>
some few pF
<shevek>
Ok.
<DocScrutinizer05>
and yes, the R forms a low pass
<shevek>
I think I should be able to make this work now. Thanks!
<DocScrutinizer05>
I'm not used to such long timing constants we should use here :-)
<DocScrutinizer05>
make that 220kR
<DocScrutinizer05>
;-)
<DocScrutinizer05>
and 100nF
<DocScrutinizer05>
wow, that sounds odd
<shevek>
If the FET burns up, I think I'll throw away the board and make a new one, so protecting the controller isn't really a priority in that case. Not unless FETs burning up is supposed to be a regular thing, but in that case I want to make them easily swappable. ;-)
<DocScrutinizer05>
FET should work infinitely
<DocScrutinizer05>
wpwrak: could you please check the timing constant for 220kR -- 100nF ?
<DocScrutinizer05>
is it really 22ms?
<DocScrutinizer05>
I'm not used to that stuff in this magnitude anymore
<DocScrutinizer05>
hmm, well, a 22s for 100uF * 220k sounds right again, so yes, it's prolly 100nF and 220kR you should use
pcercuei has quit [Quit: leaving]
<DocScrutinizer05>
shevek: you can use this softstart method _only_ when you don't use ('high' frequency) PWM to regulate the heater
<shevek>
Ah yes, that's a good point. I do actually want that.
<DocScrutinizer05>
when you use PWM with a frequency >1Hz, you can't use that softstart and you need a diode that has 4A continuous
<shevek>
Ok. I'll do that then.
<DocScrutinizer05>
or you at least must use a softstart timing that's shorter than 1% of your PWM frequency. So when you use 10Hz PWM, softstart should have a timing for 1ms
<DocScrutinizer05>
IOW 100kR and 10nF
<shevek>
Right.
<DocScrutinizer05>
I suggest slow PWM for a inductive heater
<DocScrutinizer05>
fast PWM has no benefits but brings lots of hassle with such stuff
<shevek>
The induction is not its main feature; it's a resistive heater, that happens to have some induction.
<DocScrutinizer05>
yeah, I know
<DocScrutinizer05>
sorry for the fuzzy wording
<shevek>
It's actually only 12 windings, I was mistaken before; it's 4 of those in parallel.
<DocScrutinizer05>
still on a magnetic steel rod
<shevek>
Yes.
<DocScrutinizer05>
12 windings air would be negligible. Steel core changes that completely
<DocScrutinizer05>
also take care for proper isolation of the NiCr wire on the steel ;-)
<shevek>
Yes, of course. :-)
<shevek>
It has a glass fiber coating.
<DocScrutinizer05>
good
<DocScrutinizer05>
if you want short circuit protection, add a 0.25R/4W resistor between FET source and GND
<DocScrutinizer05>
or even as small as 0.1R
<DocScrutinizer05>
which then only needs 2W
<DocScrutinizer05>
will greatly improve the reliability of your design, avoiding FET to break even on shorts
<DocScrutinizer05>
for 90%, yes. 22ms is for iirc 63%
<DocScrutinizer05>
or 66?
<wpwrak>
i wonder if adding massive filtering to the gate won't make the FET operate a lot as resistor, thus heating up. kinda not what you want with PWM.
<DocScrutinizer05>
as I said: max 1% of PWM freq
<DocScrutinizer05>
I strongly discourage to use high freq PWM with heaters of that type
<wpwrak>
put a cap across the heater ?
<DocScrutinizer05>
the system is very inert, and you get all sorts of trouble, from switching loss like the one we discuss here, to noise generated by the heater itself, wehn you go beyond 50 or 100Hz
<shevek>
DocScrutinizer05: I think you mean 63%, 1-1/e.
<DocScrutinizer05>
yep, 63
<DocScrutinizer05>
and yes, of course the FET gets a little warm with this
<DocScrutinizer05>
rough estimation: 100W * 1% * 2 (switch state changes) * 0.5 (very rough estimation for 'efficacy' between ramping FET and load) = 1W
<DocScrutinizer05>
probbaly the 0.5 are way too high
<DocScrutinizer05>
worst case is half of 24V * half of 4A
<DocScrutinizer05>
at 4A voltage is near zero, at 24V current is zero
<DocScrutinizer05>
so it's not 0.5 but rather the integral over one half sine * 0.5
<shevek>
Integral, or average?
<DocScrutinizer05>
plus the FET going from off to on not during capacitor charging up from 0 to 5V but rather between 2.2V and 2.6V or sth like that
<wpwrak>
hmm, not for steady state. what's roughly the DC resistance of the heater ?
<DocScrutinizer05>
24v@4A
GeorgeHahn has joined #qi-hardware
<shevek>
6Ω
<wpwrak>
60 Ohm ?
<shevek>
6
<wpwrak>
ah . 6 Ohm :)
<roh>
on heaters one ususally doesnt do 'fast' pwm... cylces with the length of up to a few seconds are normal
<DocScrutinizer05>
what I said
<wpwrak>
the Ω looks like a zero here :)
<shevek>
That makes sense; they don't respond fast anyway.
<roh>
anything faster does only heat the fet more, and not help stabilize the temperature much, since there is lots of thermal/mechanical integration happening
<DocScrutinizer05>
inertia I called it
<roh>
i think our pwm runs on something like a single digit frequency in hz ;)
<shevek>
1Hz seems reasonable, I think.
<DocScrutinizer05>
microwave ovens have a PWM at ~0.1Hz or less
<DocScrutinizer05>
;-)
<shevek>
They have a very slow startup.
<DocScrutinizer05>
yeah, sort of
<DocScrutinizer05>
the magnetron anode heating takes some time
<DocScrutinizer05>
err kathode
<wpwrak>
worst-case FET load would be around 2 A, where it would have to burn 24 W. so if the PWM uses the corresponding range, the smoke may return
<DocScrutinizer05>
huh?
<DocScrutinizer05>
[2016-03-15 Tue 19:25:37] <DocScrutinizer05> rough estimation: 100W * 1% * 2 (switch state changes) * 0.5 (very rough estimation for 'efficacy' between ramping FET and load) = 1W
<DocScrutinizer05>
[2016-03-15 Tue 18:49:27] <DocScrutinizer05> or you at least must use a softstart timing that's shorter than 1% of your PWM frequency. So when you use 10Hz PWM, softstart should have a timing for 1ms
<wpwrak>
that's if you quickly ramp through the resistive zone. if you switch fast enough (just a few Hz, it seems, with the large filter), you can hold the gate voltage such that the FET becomes a ~6 Ohm resistor
<wpwrak>
so that's what you want to avoid
<DocScrutinizer05>
^^^^^
<DocScrutinizer05>
I said 10Hz, _not_ 10kHz
<wpwrak>
your low-pass filter is around 7 Hz
<DocScrutinizer05>
*sigh*
<DocScrutinizer05>
[2016-03-15 Tue 18:49:58] <DocScrutinizer05> IOW 100kR and 10nF
<wpwrak>
ah, you made it smaller. good. now it's 160 Hz
<DocScrutinizer05>
[2016-03-15 Tue 19:25:37] <DocScrutinizer05> rough estimation: 100W * 1% * 2 (switch state changes) * 0.5 (very rough estimation for 'efficacy' between ramping FET and load) = 1W. probbaly the 0.5 are way too high. worst case is half of 24V * half of 4A
<DocScrutinizer05>
[2016-03-15 Tue 19:31:42] <DocScrutinizer05> plus the FET going from off to on not during capacitor charging up from 0 to 5V but rather between 2.2V and 2.6V or sth like that
<DocScrutinizer05>
that's significantly less than the 63% for one T
<DocScrutinizer05>
or even the 90% your calculator seems to use
<DocScrutinizer05>
so no smoke signals, the FET will just get a little warm
<wpwrak>
(my calculator) tr is now 2.3 ms
<DocScrutinizer05>
yes, for 90%
<DocScrutinizer05>
not for 2.2V to 2.8V, out of 0 to 5V
<wpwrak>
(smoke) what i mean that there is/was a fairly low frequency limit before Vg becomes "DC" and keeps the FET operating in resistive mode.
<DocScrutinizer05>
yet you avoid transients of 4A resp 24V @ <1us
<wpwrak>
with fc = 160 Hz and a switching frequency around 1 Hz, that would be fine, though
<DocScrutinizer05>
we're assuming frequency << timing constant of the low pass filter
<wpwrak>
yes. i was worried about the frequency you had before.
<DocScrutinizer05>
you should read the complete backscroll
<wpwrak>
it's sometimes hard to spot things in those walls of text ;-)
<DocScrutinizer05>
the point is that you must not 'spot things' since everxy part of info is equaly relevant
<wpwrak>
not if it supersedes previous information :)
<DocScrutinizer05>
in that case it's even more relevant
<wpwrak>
exactly. hence it's important to spot those bits. qed :)
<shevek>
All info is equally relevent. Some info is more equally relevant than others. :-p
<DocScrutinizer05>
no, since my and roh's suggestion is still to use sub-1Hz PWM frequency
<DocScrutinizer05>
thus the original suggestion still is valid
<DocScrutinizer05>
the 10Hz was an example why we need to take care, since even at 10Hz the filter timing gets relevant, then I gave an example how to calculate IF the PWM was at 10Hz
<wpwrak>
yes, but i was warning about a dangerously near "danger zone". it's not a "suggestion" but a fairly hard speed limit
<DocScrutinizer05>
ohmy, we all got that meanwhile
<DocScrutinizer05>
[2016-03-15 Tue 18:47:53] <DocScrutinizer05> when you use PWM with a frequency >1Hz, you can't use that softstart and you need a diode that has 4A continuous
<DocScrutinizer05>
[2016-03-15 Tue 18:48:23] <shevek> Ok. I'll do that then.
<DocScrutinizer05>
[2016-03-15 Tue 18:49:27] <DocScrutinizer05> or you at least must use a softstart timing that's shorter than 1% of your PWM frequency. So when you use 10Hz PWM, softstart should have a timing for 1ms
<DocScrutinizer05>
it's really not helping to replace a percieved wall of text with a 10 times larger wall of discussion text just to pick up the lost details in that first wall
<DocScrutinizer05>
let's rather evaluate which snubber diodes would work for a heavily inductive load at 24V@4A, for freq of <1Hz and for 50Hz
<DocScrutinizer05>
obviously a 4A diode with a Wattage of 4*0.7W should always suffice
<DocScrutinizer05>
but I guess we can go significantly 'cheaper'
<DocScrutinizer05>
particularly since our inductivity has a ESR of 6 Ohms
<DocScrutinizer05>
((would always suffice)) not true, for frequencies of several kHz and way above, you need a schottky (low paraistic capacitance)
sandeepkr__ has quit [Read error: Connection reset by peer]