I do not see how the soft start circuit can be made into a slow charge for a large capacitor bank in a high powered class A amp, as either the load on switching is VERY high, or the dissipation in the bypass resistors is very high. without the class A load, it is no problem, with the load, it is extreme. I think one would have to also switch the load after a duration..... which means more relays....
If any interest, I can post an LTspice sim using inductors for transformers.....
If any interest, I can post an LTspice sim using inductors for transformers.....
In another thread, Mark Johnson pointed out the following document and made, amongst others, the following comments. See section 5.2 In-Rush Control. Perhaps this is instructive?
http://www.ti.com/lit/an/snaa057b/snaa057b.pdf
http://www.ti.com/lit/an/snaa057b/snaa057b.pdf
The most elegant and least complicated "soft start" design is to use a NTC device from Amatherm.
Inrush Current Limiters - Thermistors | Ametherm
Inrush Current Limiters - Thermistors | Ametherm
Using NTCs after the secondary windings will allow slow charging of the capacitor bank, they aren't just for the primary side. A convenient spot would between the rectifiers and the caps on the rail side. As Andrew suggests, you can bypass the NTCs after a few hundred ms. As long as it has the current capability, you could run that relay from the mains soft start as the required timing is similar.
That's where NTCs offer a big advantage.
A correctly selected NTC (or series of NTCs if the required is not available) will not overheat.
There is another advantage to NTC limited current.
As the charge voltage builds up, the NTC has heated during that short period. The resistance is now slightly less and in part compensates for the reduced charging voltage (charging voltage = transformer emf - voltage on the capacitor).
This gives closer to constant current charging and is the shortest time that the capacitor can be charged at that current limit.
A fixed resistor gives a non constant current charge. Theoretically the capacitor never actually achieves full charge. In practice >90% of full charge is achieved in ~5*RC time constant and in 10*RC time constant it is effectively fully charged.
An NTC does it faster than this.
A correctly selected NTC (or series of NTCs if the required is not available) will not overheat.
There is another advantage to NTC limited current.
As the charge voltage builds up, the NTC has heated during that short period. The resistance is now slightly less and in part compensates for the reduced charging voltage (charging voltage = transformer emf - voltage on the capacitor).
This gives closer to constant current charging and is the shortest time that the capacitor can be charged at that current limit.
A fixed resistor gives a non constant current charge. Theoretically the capacitor never actually achieves full charge. In practice >90% of full charge is achieved in ~5*RC time constant and in 10*RC time constant it is effectively fully charged.
An NTC does it faster than this.
Yes, the NTC will be faster. However, you need to take the NTC out of circuit so that it will be allowed to cool, which means what is equivalent to a form-C relay must be used. NB. LTspice simulation shows that the NTC will *still* conduct after the bypass relay has closed.
What if you power cycle the amp before the NTC has cooled? The inrush can blow the thermistor no?
The resistor has no such issue -- it is just important that the resistor is selected so that the power rating is respected. The Welwyn company provides a document that shows how much overpowered their resistors can be without failing, but it is just as simple to use a larger resistance to limit the inrush current to below the resistor's power rating.
This also increases turn on delay time.
IMO, it is worth the extra time at turn on ( and extra cost for resistors) to avoid any potential problems due to short cycling.
What if you power cycle the amp before the NTC has cooled? The inrush can blow the thermistor no?
The resistor has no such issue -- it is just important that the resistor is selected so that the power rating is respected. The Welwyn company provides a document that shows how much overpowered their resistors can be without failing, but it is just as simple to use a larger resistance to limit the inrush current to below the resistor's power rating.
This also increases turn on delay time.
IMO, it is worth the extra time at turn on ( and extra cost for resistors) to avoid any potential problems due to short cycling.
In a slow charge application, when power is cycled quickly the capacitor charge isn't likely to be depleted much (witness how long the amp plays after shutoff if you leave the source going) so the new inrush will be limited. I don't think you will have an issue burning out the NTC.
Also, good design practice in either case is to specify resistors and NTCs that can carry the full current in case the bypass relay fails to operate.
IMO, it boils down to personal preference. I like NTCs for getting to operating faster while still protecting the bridges and capacitors.
Also, good design practice in either case is to specify resistors and NTCs that can carry the full current in case the bypass relay fails to operate.
IMO, it boils down to personal preference. I like NTCs for getting to operating faster while still protecting the bridges and capacitors.
True, but that is more of an issue on the primary side, where the transformer may present essentially a short circuit if the magnetization is worst case. Then you wouldn't have time for the NTC to cool and present a significant resistance. But an amp with enough capacitance to require a slow charge (current discussion) will bleed off its rails very slowly even at fairly significant bias, so a short interruption isn't such a big deal
For example, my A75 with only 45K µF per rail and 2A bias will play for at least 20 seconds after removing power to the output stage. (separate transformer for the front end) In that time the NTC will have cooled significantly and recovered most of its cold state resistance. Not to mention that you've also got the primary side soft start helping to limit inrush currents.
For example, my A75 with only 45K µF per rail and 2A bias will play for at least 20 seconds after removing power to the output stage. (separate transformer for the front end) In that time the NTC will have cooled significantly and recovered most of its cold state resistance. Not to mention that you've also got the primary side soft start helping to limit inrush currents.
I've simmed a version where the rail voltage drops to below the deactivation voltage of the thermistor bypass relay. It takes around 14 -15 seconds.
Now, apply power, and the sim shows the thermistors passing a 40A peak.
I don't think waiting for them to cool is going to work all that well.
The thermistors must remain deenergized for a significant length of time before the power can be cycled. Resistors do not have this issue, but they won't fill the cap bank as the amp bias current drains the caps very quickly. ONLY if the bias current is switched can you assure that power cycling can be safe.
I honestly don't see an idiot proof method except for using lots of resistors and applying the bias current to the outputs ONLY when the caps have been filled. ie, place an NO relay between the power supply and the amp, or even between the front end and the output boards (thinking F5T here).
Is there really no better way?
Now, apply power, and the sim shows the thermistors passing a 40A peak.
I don't think waiting for them to cool is going to work all that well.
The thermistors must remain deenergized for a significant length of time before the power can be cycled. Resistors do not have this issue, but they won't fill the cap bank as the amp bias current drains the caps very quickly. ONLY if the bias current is switched can you assure that power cycling can be safe.
I honestly don't see an idiot proof method except for using lots of resistors and applying the bias current to the outputs ONLY when the caps have been filled. ie, place an NO relay between the power supply and the amp, or even between the front end and the output boards (thinking F5T here).
Is there really no better way?
the NTC starts cold and has the high resistance required to limit the inital charging current.
After a short charging time the NTC is warming and the resistance is dropping.
after a further time, say 200 mains cycles, the NTC current has fallen considerably below the initial peak charging current and the NTC dissipation has fallen (Intc * Vdropntc) considerably. The NTC cools before the smoothing capacitance has fully charged.
The charging rate in the later period of charging is quite slow, the ntc may have come back up to 20% to 50% of the cold value.
Finally the smoothing caps reach ~ 99.9% of their full charge voltage and the NTC has come back up to 50% to 90% of it's cold resistance value. NOW is the time to trigger the NTC bypass.
The NTC manufacturers make NTCs specifically to control the speed of charging capacitors.
They know their product.
They show you how to apply their product.
Do as they say and the slow charge is taken care of.
BUT !!!! then bypass the NTC to make the PSU ready to do it's job. To provide a low impedance source of current.
The PSU cannot do that job of "low impedance" if the NTC is still in circuit.