softstart circuit with relay: please critique

[exe]

Happy Friday!

I'm working on a 100W linear PSU and I decided to make a softstart circuit for it (though not sure if it is really needed). So here it is! What do you think about it? To me it looks a little bit over-engineered and probably many things could be thrown away. I'm not even sure we need LM317 at all. Will 555 circuit work from an unrectified voltage? What do you think? One assumption I made is that 555 (bjt version) can drive the coil directly (which is ~40-50mA to activate, ~20mA hold). May C7 do any harm at discharge?

The relay I'm going to use: 6V TE RT174006 http://docs-europe.electrocomponents.com/webdocs/1397/0900766b81397bc0.pdf

PS I also tried to find an appropriate SSR instead of relay, but most of them are based on SCRs and require some minimal current which is not that low. Hence here is a mechanical relay

[ajb]

Is this meant to go on the AC input side of the PSU (I would assume so given that you have a transformer across the input terminals)?  The time delay for the relay could be as simple as a transistor, resistor, and capacitor.  Or, since you apparently have an MCU in there, simply use the MCU to switch the relay on during the power up sequence.  If you want something more dumb and reliable, you can use your method or the simpler RC+transistor method and just use one of your existing secondary supplies to power it rather than an extra transformer and regulator.

Note that using an NTC as the soft start element means that with a rapid off-on-off-on cycle the last 'on' may not be very soft if the NTC hasn't had time to cool down.  Maybe not a concern with with your use case, but something to consider.  Since you're bypassing the NTC with a relay anyway, you could just use a fixed resistor in place of the NTC.

[prasimix]

If you're going to use MCU then soft-start could be done with one or two triac (depending how "soft" you wants to circuit starts). Also with MCU control a "stand-by" functionality and protection against multiple switch on-off cycles in short time can be implemented.

[exe]

Thanks!

Is this meant to go on the AC input side of the PSU (I would assume so given that you have a transformer across the input terminals)?  The time delay for the relay could be as simple as a transistor, resistor, and capacitor.  Or, since you apparently have an MCU in there, simply use the MCU to switch the relay on during the power up sequence. 

If you want something more dumb and reliable, you can use your method or the simpler RC+transistor method and just use one of your existing secondary supplies to power it rather than an extra transformer and regulator.

This is true, but I wanted to order 5 or so PCBs and use this project for "many" other projects, including power amp that does not have an MCU... Another issue here is that my PSU is going to be multiboard and I start with the power part first. The digital part will come much later, that "power ok" is for future provision.

But point taken, I'll buy the relay and try to make this also with BJT. Do you think I still need an LM317? I think so because these 2-5VA transformers have so poor regulation (+64% when no load)... But in this case there I can't limit the hold current, that is 20-30mA loss . Though 555 itself has comparable quiescent current.

Since you're bypassing the NTC with a relay anyway, you could just use a fixed resistor in place of the NTC.

It happens that my circuits fail and I want them to fail safely. I hope NTC will allow the fuse to blow. This is the only reason it is there instead of resistors.

[exe]

Hm, what if I use just an RC + schmitt trigger? They are almost dirt-cheap. Like 74LVC1G17 (up to 100mA output): http://www.farnell.com/datasheets/1648909.pdf

[exe]

The time delay for the relay could be as simple as a transistor, resistor, and capacitor.

The 555 timer is quite cheap, but also provides sharp output edge (though this is irrelevant here). In any case it's three elements: a resistor, capacitor and a bjt or 555. So the cost is essentially the same, right?

[Siwastaja]

Basics:

You don't generally use NTC when bypassing it with relay. You can do it, but it's not optimal.

The idea of using NTC is that it's self-contained, there's nothing else. It's poor, but it's simple and cheap. It's lossy all the time, and you have the tradeoff between losses and inrush current, and you have quite a bit of inrush anyway, so it only works when it's nearly not a problem anyway. It also has problem with repeated on/off cycles during about 10-20 seconds, but you rely on it being nearly unnecessary anyway.

When you have a bypass relay, then you use a fusible resistor, or a standard power resistor with properly dimensioned time-delay fuse in series (with the resistor only; the fuse is bypassed too), or a PTC, or a PTC thermally coupled with power resistor in series. The idea is that if your precharge drive side fails for any reason, the fusible resistor blows; or the fuse with standard resistor blows before the resistor overheats. PTC needs to have enough cold resistance so it works like the resistor, but almost stops conducting (and hence, precharging) if the driver is stuck.

Driving based on time is waste of design effort. Measure the rectified output voltage (the one with the large caps) with a simple comparator, for example, and drive the bypass relay when the precharge is finished, i.e., the voltage is near the operating voltage.

If you have an MCU there anyway, it's even better to use it to measure the voltage rise and drive the bypass relay. The MCU needs to be powered up first through a secondary supply. This way, you can implement timeout too, so that you rely on the fusible resistor (or PTC action) only in a severe case of MCU malfunction.

[exe]

I think PTC increases chances for a malfunctioning circuit to remain in steady and broken state. I would like to have a fuse so it will just blow (though I don't leave my devices unattended because I know what kind of an engineer I am).

Driving based on time is waste of design effort. Measure the rectified output voltage (the one with the large caps) with a simple comparator, for example, and drive the bypass relay when the precharge is finished, i.e., the voltage is near the operating voltage.

I will have multiple isolated channels in my PSU. This way using feedback and comparators... Too much engineering. On the other side, a delay circuit is just three cheap components. It is also self-sufficient and robust.

If you have an MCU there anyway, it's even better to use it to measure the voltage rise and drive the bypass relay. The MCU needs to be powered up first through a secondary supply. This way, you can implement timeout too, so that you rely on the fusible resistor (or PTC action) only in a severe case of MCU malfunction.

Can be an option, but this is quite a bit of engineering, isn't it? And with multiple isolated channels... What's wrong with dumb delay circuit, again please?

[Siwastaja]

I don't see much complexity difference between a timer and a voltage comparator and a reference. Both can be done with about two or three components for about 10 cents. But maybe I'm not following your full circuit correctly.

BTW, you need to implement similar delay to the power consumer side so that outputs cannot be enabled during the soft-start. That would cause very high dissipation in the power resistor, possibly blow it up (after which the soft-start just disappears), and in any case, make the inrush happen suddenly after 100ms because the precharge could not charge the caps due to the load. With voltage-based precharge, these kind of faults can be detected.

But, IMHO, 100W linear supply doesn't need to get this fancy. I'm 99.9% sure it doesn't need any kind of inrush protection, and if it, for some reason, does, the crappy NTC only approach is probably more than enough. Linear supply is going to be very inefficient anyway, so the NTC loss is not that meaningful.

What we are discussing would make a lot of sense in the 500-1000W range, i.e., something that is considerably more than 1/10 of the house wiring fuse rating.

[exe]

I don't see much complexity difference between a timer and a voltage comparator and a reference.

This needs tuning for every individual power application, but I don't see clear benefits in my case comparing to the delay in my case. So, yes, this is superior to what I propose, but I'm more concerned about reliability, rather than making super-cool soft startr. This looks cool, but, probably, too cool for what I'm doing. Please see below why.

BTW, you need to implement similar delay to the power consumer side so that outputs cannot be enabled during the soft-start. That would cause very high dissipation in the power resistor, possibly blow it up (after which the soft-start just disappears), and in any case, make the inrush happen suddenly after 100ms because the precharge could not charge the caps due to the load. With voltage-based precharge, these kind of faults can be detected.

This is what "power ok" is for -- the MCU will listen to it. Anyway, I don't mind mild inrushes if the relay is already closed the contacts so the resistors/contacts are safe. My main concerns are the fuse (and may be power switch) and may be EMI? It's not that big load to create much harm. Let's say I'm doing an exercise of thoughtful robust design. So I can show off

But, IMHO, 100W linear supply doesn't need to get this fancy.

Absolutely, it's more like a toy project. I just don't know what else to assemble that would be simple, inexpensive and useful for future application. So I'm doing it mostly because I can.

[sarepairman2]

a 555 timer for a soft start circuit is a good choice in my opinion.

I think though you might be able to go cheaper. Since you are shorting out a circuit that has little voltage drop across it (theoretically) you could just use a RC delay.

You mainly want to drive a relay hard if its to prevent arc damage to the contact. If the potential is almost equal you can probobly get away without using it.

Why you would do this? I have no idea. If you want to avoid using a transformer to get some logic level, you can probobly make a delay circuit using triac/sidac/diac

What I don't get is that if you are designing a circuit in your own time that you can easily buy online for cheap, why would you worry about over engineering or expense? You are already spending a fortune in design time with or without it.
 
Theoretically its a good idea because it lets you select a lower value fast acting fuse and offers you more protection, not to mention extends the life time of the circuitry. If you do it you will know how to do it and if you decide to build something that runs nearer to the maximum specification of the components you are using then you will have experience.

[NiHaoMike]

Basics:

You don't generally use NTC when bypassing it with relay. You can do it, but it's not optimal.

The idea of using NTC is that it's self-contained, there's nothing else. It's poor, but it's simple and cheap. It's lossy all the time, and you have the tradeoff between losses and inrush current, and you have quite a bit of inrush anyway, so it only works when it's nearly not a problem anyway. It also has problem with repeated on/off cycles during about 10-20 seconds, but you rely on it being nearly unnecessary anyway.

When you have a bypass relay, then you use a fusible resistor, or a standard power resistor with properly dimensioned time-delay fuse in series (with the resistor only; the fuse is bypassed too), or a PTC, or a PTC thermally coupled with power resistor in series. The idea is that if your precharge drive side fails for any reason, the fusible resistor blows; or the fuse with standard resistor blows before the resistor overheats. PTC needs to have enough cold resistance so it works like the resistor, but almost stops conducting (and hence, precharging) if the driver is stuck.

Driving based on time is waste of design effort. Measure the rectified output voltage (the one with the large caps) with a simple comparator, for example, and drive the bypass relay when the precharge is finished, i.e., the voltage is near the operating voltage.

If you have an MCU there anyway, it's even better to use it to measure the voltage rise and drive the bypass relay. The MCU needs to be powered up first through a secondary supply. This way, you can implement timeout too, so that you rely on the fusible resistor (or PTC action) only in a severe case of MCU malfunction.

Many modern PC PSUs use NTCs bypassed with relays.

[exe]

I think though you might be able to go cheaper. Since you are shorting out a circuit that has little voltage drop across it (theoretically) you could just use a RC delay.

Could you please elaborate on RC delay? Is it a resistor in series and a capacitor in parallel to the coil?

Concerning the cost of BOM. Definitely it is possible to make it cheaper, but, as you already mentioned, for a one-off device the time is the biggest concern. +$3 to the total cost is not a problem . This 555 circuit makes me feeling like it is really rugged and I can rely on it. But I'll consider all the suggestions in this thread. May be I'll try several approaches.

What I don't get is that if you are designing a circuit in your own time that you can easily buy online for cheap, why would you worry about over engineering or expense? You are already spending a fortune in design time with or without it.

It's my hobby. Also designing something new is a good way to learn electronics. In terms of money it's definitely waay cheaper to buy something from ali express for less than 10 dollars.

[sarepairman2]

I think though you might be able to go cheaper. Since you are shorting out a circuit that has little voltage drop across it (theoretically) you could just use a RC delay.

Could you please elaborate on RC delay? Is it a resistor in series and a capacitor in parallel to the coil?

Concerning the cost of BOM. Definitely it is possible to make it cheaper, but, as you already mentioned, for a one-off device the time is the biggest concern. +$3 to the total cost is not a problem . This 555 circuit makes me feeling like it is really rugged and I can rely on it. But I'll consider all the suggestions in this thread. May be I'll try several approaches.

What I don't get is that if you are designing a circuit in your own time that you can easily buy online for cheap, why would you worry about over engineering or expense? You are already spending a fortune in design time with or without it.

It's my hobby. Also designing something new is a good way to learn electronics. In terms of money it's definitely waay cheaper to buy something from ali express for less than 10 dollars.

i think you misunderstood what I meant. I mean that you are spending many many hours designing this thing. If you got paid for your work it would come out to like at least 30$ an hour for electrical design. The idea of trying to save a few dollars on a circuit you design for yourself that does not have a chance of seeing any kind of mass production is just ludicrous. Its like recycling toilet paper.

save that kind of thinking for dealing with your manager at work.... if your are gonna get stressed out about 4$ then have someone pay you a wage lol

I mean unless you just get off on building cheap shit... sometimes I see things on forums that boggle my mind. People spend hours researching, writing forum posts... then they get into some 3 page argument about saving 3$ on something they clearly worked 100 hours on. 

It's like that scene in office space, where the guy gets another job and his old manager is still there to haunt him. It's like the bean counters at peoples jobs get into someones head... brain slugs... Maybe they are just scared their wife is reading their forum posts IDK.

[Siwastaja]

I mean that you are spending many many hours designing this thing. If you got paid for your work it would come out to like at least 30$ an hour for electrical design. The idea of trying to save a few dollars on a circuit you design for yourself that does not have a chance of seeing any kind of mass production is just ludicrous.

Completely wrong, it's exactly the opposite. What you are describing is a small cost saving (in labor cost, which doesn't even apply in a hobby project!).

When designing a small-volume product (say, less than tens of thousands), then you can't spend the time you are paid for like that. I understand that very well. You use quick solutions, avoid overanalyzing, etc.

But when you design it for yourself, the idea is to analyze the solutions as much as you ever want to, because it is the learning point. Design space exploration for minimum BOM cost is an extremely good learning point.

Sure, you could make it work more quickly, but what you learn will be usable later, in other projects.

As a result, if this ever becomes a job, by self-employment or not, you are experienced and knowledgeable, and able to make important decisions more easily as you have more perspective.

Learning by doing is everything. It's easy to just pour more money at the problem (and usually it's something like $3, so very easy to do), but it takes the imagination away.

[exe]

Thank you all very much, I considered every opinion in the thread.

I built a test circuit and it seems the biggest issue is the capacitors that cause triggering immediately after a quick powercycle. I have yet to figure out how to fight the problem

[Siwastaja]

I built a test circuit and it seems the biggest issue is the capacitors that cause triggering immediately after a quick powercycle. I have yet to figure out how to fight the problem

That's exactly one of the reasons I suggested softstart based on measuring the DC link cap voltage with a comparator to a reference. Not too many components (integrated reference+comparators exist, add a resistor divider, relay and a relay driving transistor). This will skip the soft start only if the output caps are already (mostly) charged up, in which case the inrush is a lot smaller, mostly the inrush caused by the transformer only + the inrush caused by the DC link caps "topping off" from the reference level which cannot be too close to the actual operating voltage.