Hey Elvee ! Not sure to keep posting here or PM you.
Trouble with DIYaudio is that updated threads get on top of the list and I'm sure not many want to see or read it there.
Started on a dry run without the mosfets to get the timing right. (but diodes where the mosfet drain/source diode would be).
R5 100k and C1 10uF , is way to quick and the mosfets will turn on too soon.
I put 5x 33k =165k for R5 and 31uF for C1 , that seems better.
The R4 of 2k2 didn't have enough voltage over it to keep the Q1 off. So with 4k7 I
get -1,9 V BE. I also put 100 ohm from the collector to the "ground" , to limit
current to 90 mA max ( zener voltage 9,1V) .
As far as I can judge it takes much longer than 1 missing period to get Q1 conducting and the voltage to go below mosfet threshold , or the time it takes to drain 31uF with 100 ohm. I may use the BC807 , with lower than 100 ohm. It can take 500mA.
What I'm not entirely clear about is C2 . Is it there too to smooth the ripple at the
base.
You say in post 18 that 'as soon as a cycle is missing, it makes the transistor
conduct and discharge C1'
Doesn't the transistor only conduct after C2 is discharged until it is around the
needed 0,6 V for Vbe ? So the transistor will conduct and shut the mosfets down
after the 1uF x 33k goes from -1,9 V to 0,6V.
Lowering C2 will make the transistor conduct quicker ?
Trouble with DIYaudio is that updated threads get on top of the list and I'm sure not many want to see or read it there.
Started on a dry run without the mosfets to get the timing right. (but diodes where the mosfet drain/source diode would be).
R5 100k and C1 10uF , is way to quick and the mosfets will turn on too soon.
I put 5x 33k =165k for R5 and 31uF for C1 , that seems better.
The R4 of 2k2 didn't have enough voltage over it to keep the Q1 off. So with 4k7 I
get -1,9 V BE. I also put 100 ohm from the collector to the "ground" , to limit
current to 90 mA max ( zener voltage 9,1V) .
As far as I can judge it takes much longer than 1 missing period to get Q1 conducting and the voltage to go below mosfet threshold , or the time it takes to drain 31uF with 100 ohm. I may use the BC807 , with lower than 100 ohm. It can take 500mA.
What I'm not entirely clear about is C2 . Is it there too to smooth the ripple at the
base.
You say in post 18 that 'as soon as a cycle is missing, it makes the transistor
conduct and discharge C1'
Doesn't the transistor only conduct after C2 is discharged until it is around the
needed 0,6 V for Vbe ? So the transistor will conduct and shut the mosfets down
after the 1uF x 33k goes from -1,9 V to 0,6V.
Lowering C2 will make the transistor conduct quicker ?
Attachments
No problem: that's the place where it belongs, since all the details and relevant files are immediately available
From the info I posted earlier, I gather that the turn-on delay is ~3 mains cycles, around 60ms.
If you change one component value, the others need to be scaled in the same proportion:
Logic, but the 100 ohm is probably not necessary, as the base current is limited anyway.
It might impair the later stage of the discharge, and should probably be removed
A larger cap will need a larger discharge current, but everything is linked, and the first thing to do is to decide the turn-on delay value you want.
Once this is decided, the rest will flow naturally, but making spot-changes here and there will not work: the circuit is a whole, and all the aspects are interdependent.
If you need a large turn-on delay whilst keeping the one-cycle reset, it might be necessary to use another type of transistor, like a small PMOS.
It prevents the transistor from permanently discharging C1. If it is too small, C1 will be discharged even when there is no interruption, and if it is too large, it will increase the reset delay
Yes, but as I said above, making C2 too small will make it conduct at times when it shouldn't.
First state your requirements, mainly the turn-on delay, and we will see how the circuit can be adapted.
"Logic, but the 100 ohm is probably not necessary, as the base current is limited anyway.
It might impair the later stage of the discharge, and should probably be removed"
Yes but these transistors have a big Hfe , so even a small base current may not be enough to limit the collector surge current. Bipo's can't take short surges like mosfets can.
Indeed at less than around 1 V , it discharges much slower , it's more the 240k bleeder that takes over.
Timing without a digital oscilloscope , is difficult to see , I just look at the bargraph on my fairly fast Fluke. Although the goal is to avoid the very short inrush current , I think a delay of around 0,25 to 0,5 sec is best but it goes fast past the 2,5-3 V mosfet gate treshold and then more slowly to the 9 V.
The 1 missing cycle discharge is not necessary , but no more than 10 = 200ms .
Like it is now with these components , it seems more and changing the transistor to BC807 with only 22 ohm is not going to change much.
Maybe putting the gates not on the full 9 V but with a resistor divider on a lower voltage.
Have you noticed SMPS's have a 1 sec delay between plugging it in and having a voltage on the output ? While the "normal" transformer type immediately has a slower rising output voltage.
It might impair the later stage of the discharge, and should probably be removed"
Yes but these transistors have a big Hfe , so even a small base current may not be enough to limit the collector surge current. Bipo's can't take short surges like mosfets can.
Indeed at less than around 1 V , it discharges much slower , it's more the 240k bleeder that takes over.
Timing without a digital oscilloscope , is difficult to see , I just look at the bargraph on my fairly fast Fluke. Although the goal is to avoid the very short inrush current , I think a delay of around 0,25 to 0,5 sec is best but it goes fast past the 2,5-3 V mosfet gate treshold and then more slowly to the 9 V.
The 1 missing cycle discharge is not necessary , but no more than 10 = 200ms .
Like it is now with these components , it seems more and changing the transistor to BC807 with only 22 ohm is not going to change much.
Maybe putting the gates not on the full 9 V but with a resistor divider on a lower voltage.
Have you noticed SMPS's have a 1 sec delay between plugging it in and having a voltage on the output ? While the "normal" transformer type immediately has a slower rising output voltage.
This is the circuit adapted for a longer delay.
The problem now, is that the gate voltage is rising more slowly, resulting in a "ramping" conduction for a few cycles.
If the SOA of the MOS is not sufficient, they will be blown.
One would need a snap-up action somewhere, but this would complicate the circuit.
No discharge resistor is required, because there is an implicit R1/Hfe resistor in the circuit
The problem now, is that the gate voltage is rising more slowly, resulting in a "ramping" conduction for a few cycles.
If the SOA of the MOS is not sufficient, they will be blown.
One would need a snap-up action somewhere, but this would complicate the circuit.
No discharge resistor is required, because there is an implicit R1/Hfe resistor in the circuit
Attachments
Not much difference . You went for higher current and a bigger 47uF.
I'm going with less than 3k for Z , so the voltage will be higher than 50 V and for
longer so the cap behind the rectifier in the SMPS will be fuller , those 3 periods of
ramping up with higher voltage should not be not a problem ... I think. But then
again the negative resistance of a SMPS input is tricky.
I have the Rohm R6030ENX , 2 for about 3 $ . Tested them , high treshold that
starts at 2,5 V , but really starts to matter at 2,8 V , so pretty high , not like the
1,25 V of a 7002. Data sheet doesn't show a SOA.
I connected them today but still without the Z bypass , and a subjective view of the
voltage rise on the bargraph , looks as if the delay maybe good enough and the
discharge too.
I do measure -1,9 V between the 9V line and the base but with a 1,2 V ac ripple.
Not keen on a higher value for the C2 . I have these short 0,5 to 2 sec power drops/cuts and I want to be sure the mosfets are off then . Your simulation shows the discharge in 1 period and I calculate with 1 uF and 33k and a voltage difference of about 2,5 V a 10 ms , so that is close.
Even if I had my frequency counter here that can measure time intervals between 2
inputs , it would still be impossible to connect it to the 230V mains to be sure of the
timing.
I guess the ultimate test will be trying it with a fuse and see if it blows because even
if there is no arc while plugging in , there still can be a big current.
Here's pics of the PCB , with and without the mosfets : (the 33K x5 are on the other side)
I'm going with less than 3k for Z , so the voltage will be higher than 50 V and for
longer so the cap behind the rectifier in the SMPS will be fuller , those 3 periods of
ramping up with higher voltage should not be not a problem ... I think. But then
again the negative resistance of a SMPS input is tricky.
I have the Rohm R6030ENX , 2 for about 3 $ . Tested them , high treshold that
starts at 2,5 V , but really starts to matter at 2,8 V , so pretty high , not like the
1,25 V of a 7002. Data sheet doesn't show a SOA.
I connected them today but still without the Z bypass , and a subjective view of the
voltage rise on the bargraph , looks as if the delay maybe good enough and the
discharge too.
I do measure -1,9 V between the 9V line and the base but with a 1,2 V ac ripple.
Not keen on a higher value for the C2 . I have these short 0,5 to 2 sec power drops/cuts and I want to be sure the mosfets are off then . Your simulation shows the discharge in 1 period and I calculate with 1 uF and 33k and a voltage difference of about 2,5 V a 10 ms , so that is close.
Even if I had my frequency counter here that can measure time intervals between 2
inputs , it would still be impossible to connect it to the 230V mains to be sure of the
timing.
I guess the ultimate test will be trying it with a fuse and see if it blows because even
if there is no arc while plugging in , there still can be a big current.
Here's pics of the PCB , with and without the mosfets : (the 33K x5 are on the other side)
Attachments
Last edited:
This type of circuit is difficult to design: in theory, because using a rigorous pen and paper method, you quickly stumble on transcendental equations, and even in sim, because the actual behavior of a pre-made power brick is difficult to model accurately and simply.
This leaves the experiment as the only reliable method to assess and refine such a circuit, but it doesn't mean you should throw yourself into deep waters for your first try: I suggest you use an incandescence lamp (100W for example) as a dummy load: it does have a substantial inrush current, but more gentle than a cap, and it is completely fail-safe. It also provides a visual feedback of what's happening.
Since the circuit is almost completely referenced to one of the rails, possibilities for serious mishaps are practically eliminated.
Once you are happy with the behavior on the dummy load, you can go to the next level, but maybe not the actual load yet.
As Italians say: chi va piano va sano
Proceed step by step, and learn from each step
This leaves the experiment as the only reliable method to assess and refine such a circuit, but it doesn't mean you should throw yourself into deep waters for your first try: I suggest you use an incandescence lamp (100W for example) as a dummy load: it does have a substantial inrush current, but more gentle than a cap, and it is completely fail-safe. It also provides a visual feedback of what's happening.
Since the circuit is almost completely referenced to one of the rails, possibilities for serious mishaps are practically eliminated.
Once you are happy with the behavior on the dummy load, you can go to the next level, but maybe not the actual load yet.
As Italians say: chi va piano va sano
Proceed step by step, and learn from each step
Now that is some good advice , why didn't I think of that ?
I was going to do tests , first with a phone charger , but an incandescence lamp is much better to try out. I have a 25 W one. Who still uses these nowadays ?
It is a bit nerve wracking playing around with 230 V.
I was going to do tests , first with a phone charger , but an incandescence lamp is much better to try out. I have a 25 W one. Who still uses these nowadays ?
It is a bit nerve wracking playing around with 230 V.
Indeed: for that precise reason, my lab has been equipped with an isolation transformer for many years.
You still need to remain on your guard, but it relieves you of the GND/PE issues.
If you often work with mains, I strongly suggest you acquire one
You still need to remain on your guard, but it relieves you of the GND/PE issues.
If you often work with mains, I strongly suggest you acquire one
No I don't play around much with the mains. Luckily !
Ok , today the tests with current limiters . First my idea is using a lossless resistor : a 1uF X2 cap.
25W lamp 70Vac just lighting up , no problem.
With a a small nokia phone charger no load (235V) and max load (500mA 6,3V) at
173Vac , no problem.
Then a WD (western digital) external HD supply 12 V 1,5 A , no load no problem ,
with a 12 ohm resitor and a 100mA fan , doesn't work. The ac voltage oscillates
between 50 and 200 V , and a low variable dc voltage on output . The 1 uF is
definitely out , unless you start de SMPS with very low or no load.
Next 1350 ohm with 3 x 450 ohm 5W resistors.
Lamp gets 118Vac , 38V over each resistor , so around 84mA= 3,2 W/resistor.
Nokia 203Vac , again with max load 500 mA at 6,3 V ( ~3 W) , ok.
WD with only the fan (100mA) 218Vac , again with 12 ohm + fan =1,1 A (around 13 W) no go , still oscillating between 50-230 V.
So 1350 ohm is still too high.
Tried 900 ohm , WD with 12 ohm load , no , the same oscillating.
450 ohm : WD with 12 ohm and fan : 197 Vac and 11,87 Vdc = the same as
without the 450 . So this works for 13 W output.
Time to put the mosfets back in , and here it really paid off to use the lamp . I can
clearly see the the < 0,5 sec step between the 450 ohm or the open mosfets.
Then the WD , that is a real sparker/arcer , tried many times no sparks , no arcs ,
not that typical sound .
I'm happy with the circuit , but the neg resistance of SMPS is a bother.
This was only the WD SMPS delivering 13 W on its output , that is still far below
the 30-36 W that I intend to buy. Lowering the 450 ohm will dramatically increase
the watts over it , and currents can already arc at a couple of A , but it will still be
harmless. Like 2x 450 = 225 ohm at max 350 V peak , worst case 1,5 A that's 544 W , 272 W each for an instant , can a 5 W resistor survive this ?
My SMPS will start with minimal load , so I probably take the 450 ohm , but made
with 4 of them to spread the wattage .
Still some caveats : With brown outs , this is periods of lower voltage , the mosfets
may not get enough Vgs to conduct or only partial , and the 450 (and maybe the
mosfets) will have to handle not just the higher current because of the lower ac
voltage but also the voltage drop over it . This can be a problem , luckily brown outs
are rare , not like little or big powercuts. Maybe a thermal sensor to cut off the load ?
And if for some reason the mosfets fail to conduct , the 450 will get the full load ,
with the negative resistance of the SMPS wanting more an more current , until it
fails and dissipation of the 450 will be to much . Again a sensor ?
If there is something wrong with the Z (here 450 ohm) , like break down because
of too much dissipation , the mosfets can get the full inrush current. These R6030
can take it though.
Would I try this with my 135 W laptop power brick ( and no charging battery in ) .... uh , no I wouldn't.
For the few that follow this and hate these inrush currents too , there is no 1 easy
universal circuit here for all . It all depends on how much watt you're going to
switch on and there are the costs of 2 capable mosfets , the big cement resistors and some
small but cheap little components and of course , 2 TO220 Mosfets and big 5 or 10 W resistors take in a lot of space.
For me it's worth it , and I am going to use it. It will power a LED light with remote controled dimmer . The DC side of the SMPS will get an inrush current limiter too , to limit the current to the E-caps , and the LED's wouldn't conduct below certain voltage , so these things will make the use of the primary inrush limiter easier.
And thanks Elvee. Your idea with only 1 transistor is much better than mine with an optocoupler. Getting the cut off time right , would be harder too.
Ok , today the tests with current limiters . First my idea is using a lossless resistor : a 1uF X2 cap.
25W lamp 70Vac just lighting up , no problem.
With a a small nokia phone charger no load (235V) and max load (500mA 6,3V) at
173Vac , no problem.
Then a WD (western digital) external HD supply 12 V 1,5 A , no load no problem ,
with a 12 ohm resitor and a 100mA fan , doesn't work. The ac voltage oscillates
between 50 and 200 V , and a low variable dc voltage on output . The 1 uF is
definitely out , unless you start de SMPS with very low or no load.
Next 1350 ohm with 3 x 450 ohm 5W resistors.
Lamp gets 118Vac , 38V over each resistor , so around 84mA= 3,2 W/resistor.
Nokia 203Vac , again with max load 500 mA at 6,3 V ( ~3 W) , ok.
WD with only the fan (100mA) 218Vac , again with 12 ohm + fan =1,1 A (around 13 W) no go , still oscillating between 50-230 V.
So 1350 ohm is still too high.
Tried 900 ohm , WD with 12 ohm load , no , the same oscillating.
450 ohm : WD with 12 ohm and fan : 197 Vac and 11,87 Vdc = the same as
without the 450 . So this works for 13 W output.
Time to put the mosfets back in , and here it really paid off to use the lamp . I can
clearly see the the < 0,5 sec step between the 450 ohm or the open mosfets.
Then the WD , that is a real sparker/arcer , tried many times no sparks , no arcs ,
not that typical sound .
I'm happy with the circuit , but the neg resistance of SMPS is a bother.
This was only the WD SMPS delivering 13 W on its output , that is still far below
the 30-36 W that I intend to buy. Lowering the 450 ohm will dramatically increase
the watts over it , and currents can already arc at a couple of A , but it will still be
harmless. Like 2x 450 = 225 ohm at max 350 V peak , worst case 1,5 A that's 544 W , 272 W each for an instant , can a 5 W resistor survive this ?
My SMPS will start with minimal load , so I probably take the 450 ohm , but made
with 4 of them to spread the wattage .
Still some caveats : With brown outs , this is periods of lower voltage , the mosfets
may not get enough Vgs to conduct or only partial , and the 450 (and maybe the
mosfets) will have to handle not just the higher current because of the lower ac
voltage but also the voltage drop over it . This can be a problem , luckily brown outs
are rare , not like little or big powercuts. Maybe a thermal sensor to cut off the load ?
And if for some reason the mosfets fail to conduct , the 450 will get the full load ,
with the negative resistance of the SMPS wanting more an more current , until it
fails and dissipation of the 450 will be to much . Again a sensor ?
If there is something wrong with the Z (here 450 ohm) , like break down because
of too much dissipation , the mosfets can get the full inrush current. These R6030
can take it though.
Would I try this with my 135 W laptop power brick ( and no charging battery in ) .... uh , no I wouldn't.
For the few that follow this and hate these inrush currents too , there is no 1 easy
universal circuit here for all . It all depends on how much watt you're going to
switch on and there are the costs of 2 capable mosfets , the big cement resistors and some
small but cheap little components and of course , 2 TO220 Mosfets and big 5 or 10 W resistors take in a lot of space.
For me it's worth it , and I am going to use it. It will power a LED light with remote controled dimmer . The DC side of the SMPS will get an inrush current limiter too , to limit the current to the E-caps , and the LED's wouldn't conduct below certain voltage , so these things will make the use of the primary inrush limiter easier.
And thanks Elvee. Your idea with only 1 transistor is much better than mine with an optocoupler. Getting the cut off time right , would be harder too.
Last edited:
I remember a large 1500W switching power supply from the late eitghties and early nineties which used a mains inductor as an inrush current limiter. The promotional argument was, that an inductor does not distort the mains line, hence, the inrush current is more well behaved and looks like a lightly distorted sine shape. It worked by having a triac turn on to short the inductor after a few seconds. The inductor was in series with the mains feeding the power supply.
Since, an unloaded mains transformer is much like an inductor, whose nature is to resist current changes, I do not think transformers by themselves, are an issue of large inrush currents. The problem of large inrush currents has to do with large uncharged capacitors and stationary motor rotors.
In my opinion, the simplest and best solution is to use a thermistor with a negative temperature coefficient. With a few additional components this can be used to turn on/off a thyristor or an IGBT.
Since, an unloaded mains transformer is much like an inductor, whose nature is to resist current changes, I do not think transformers by themselves, are an issue of large inrush currents. The problem of large inrush currents has to do with large uncharged capacitors and stationary motor rotors.
In my opinion, the simplest and best solution is to use a thermistor with a negative temperature coefficient. With a few additional components this can be used to turn on/off a thyristor or an IGBT.
The Mean Well SMPS I want to use , says 45A for about 440 us . Can a thermistor react that fast ? The puls will be over by the time its resistance caught up to limit it.
I don't know much about inductors , but at 50 Hz , it's going to be big and heavy.
When there is a powercut isn't there a peak reverse voltage generated , like cutting the power to to a relay coil ?
I don't know much about inductors , but at 50 Hz , it's going to be big and heavy.
When there is a powercut isn't there a peak reverse voltage generated , like cutting the power to to a relay coil ?
Thermistors do not need to react fast: they start in a cold state (high-resistance), and warm-up depending on the power they dissipate.
They would be ideal, since they are simple, low-cost and jam-proof, but if you make a restart seconds after a power-down, they remain in a low-resistance state and cannot erase the current peak.
Rick, you use very high bypass impedances, like 450 ohm. There is no need to go that high: you just want to limit the inrush to a few amps, maybe 10A.
To achieve that, something like 20~30 ohm would be OK.
I justhad a look at a power brick including a soft-start, and the series resistor was 11ohm.
You do not need to go that low, but 20ohm seems more reasonable
They would be ideal, since they are simple, low-cost and jam-proof, but if you make a restart seconds after a power-down, they remain in a low-resistance state and cannot erase the current peak.
Rick, you use very high bypass impedances, like 450 ohm. There is no need to go that high: you just want to limit the inrush to a few amps, maybe 10A.
To achieve that, something like 20~30 ohm would be OK.
I justhad a look at a power brick including a soft-start, and the series resistor was 11ohm.
You do not need to go that low, but 20ohm seems more reasonable
Last edited:
If NTC's were so good and easy , why wouldn't every SMPS be fitted with one ? With only 1 low cost component , inrush current would be a thing of the past . Nearly all electronic appliances these days use SMPS. The very low power phone or mp3 player ones , don't arc , or barely , maybe they have a limiting resistor already ?
Elvee :
Say I use a 35 ohm and it's just at the moment at 350 V , that is a 10A peak = 3500 W , yes maybe only for 0,5 ms . Can a 5 or 10 W resistor take that much ? Even at a lower voltage , it is still a lot . And even a couple of Amps causes arcs , although harmless. I can make 1 A on my car battery arc.
I'm buying the 35W APV-35 , but I'll only use maybe ~26 W , it is good to keep some extra and there is practically no price difference between 25 or 35 W version , and dimensions are the same. If 450 ohm can switch the 13 W of the WD , maybe 225 ohm can switch the APV-35 at 26 W output .
A single 450 ohm at peak voltage , can get >250 W , still a lot for a 5 or even 10 W R.
I am an unlucky switcher , 9 times out of 10 I get an arc plugging in the WD of laptop power brick !
Elvee :
Say I use a 35 ohm and it's just at the moment at 350 V , that is a 10A peak = 3500 W , yes maybe only for 0,5 ms . Can a 5 or 10 W resistor take that much ? Even at a lower voltage , it is still a lot . And even a couple of Amps causes arcs , although harmless. I can make 1 A on my car battery arc.
I'm buying the 35W APV-35 , but I'll only use maybe ~26 W , it is good to keep some extra and there is practically no price difference between 25 or 35 W version , and dimensions are the same. If 450 ohm can switch the 13 W of the WD , maybe 225 ohm can switch the APV-35 at 26 W output .
A single 450 ohm at peak voltage , can get >250 W , still a lot for a 5 or even 10 W R.
I am an unlucky switcher , 9 times out of 10 I get an arc plugging in the WD of laptop power brick !
The PFC system normally keeps an eye on the current draw, most PFC controller ICs work with the switching controller to control current. Should too much current be requested over N number of ticks/mains cycles then the system locks out the PFC switching to prevent overload.
Normally an SMPS safeguard will kick in if the SMPS is given a short at startup. Perhaps the better option is to have an NTC after the SMPS to limit inrush that is then switched out, an alternative is to have a current limiter such as a mosfet load side to limit the incoming inrush to something sensible.
SMPS is like putting your phone on the inrush freeway at the peak commute period. Only real way is to limit the load.
Yes
You have to see the problem from the energy angle: most of the inrush current is used to charge capacitors. The energy is 0.5*C*U², and an equal energy is lost in the charging resistance.
The amount of energy doesn't vary with the charging time. Your 5 or 10W resistor will have to swallow the same quantity of Joules, whether it is 450 ohm or 45 ohm, and will end up with the same temperature rise.
In fact, if the startup is very long, the resistor will also have to dissipate the additional power consumed by the supply on top of the charging energy.
Yes there is an inrush limiter on the output because of the big E-caps after the SMPS.
Very interesting !
So you are sure that an 5 W resistor can take a very short very high dissipation burst ? I can't find anything about that in its datasheet.
You could swap banks in and out, each with their own NTC. Naturally you don't want to swap at max current so perhaps at 3RC or 4RC. Then switch to the next, so by the time the swap comes back or completes the NTC will have cooled enough and then continue until you have a fully charged bank at which point you can short out the charger and carry on.
As long as the load isn't inductive you shouldn't get boost style spikes of voltage.