peranders said:
Hi Peranders. I am thinking of building, actually I think assembling is more accurate, a UCD180 amplifier. After looking at the soft start board that DIY Cable sells I decided to go that route. I plan on using a 800VA toroidal transformer with two 50v windings with a couple of IXYS bridges and some Nichicon Gold Tunes to build the supply. I can build a much better supply than the one DIY Cable is selling for $195 for a lot less money so I will be building that. Still trying to convince myself to go ahead and build a (gasp!) SS amp. Are the UCD modules as good as all the reviews say?
I had major problem with blowing 10A main fuses in my apartment when my tv and pc was on and i switched on my GC. Solution: Delay circuit for toroids
It works perfectly and now I never need to change my fuses
It works perfectly and now I never need to change my fuses
A single CL-type thermistor seems possibly problematical, unless the load current will be constant and known in advance.
I stopped using them in my power supplies and switched to using a MOSFET-based soft-start circuit. That way, I get a known (and very, very low) resistance, after the soft start is complete. (And the total cost is not much higher, although it takes more room. However, it also doesn't get nearly as HOT as a CL thermistor would get.)
The circuit is very simple and uses only six components. It goes between the diode bridge and the main filter capacitor(s). There is a hefty P-MOSFET (e.g. 55-Volt/80-Amp STP80PF55, or IRF4905) in parallel with a low-value power resistor, which are placed in series with the power rail (i.e. between the bridge's +DC output and the filter caps' "+" side). The mosfet's Source is connected to the rectifier bridge and its Drain is connected to the filter caps. Then components are added so that the mosfet (normally "off"/open circuit) turns on "gradually", eventually becoming a 0.018-Ohm resistance in parallel with the power resistor, effectively removing the power resistor from the circuit.
One that I recently designed was configured like this: 10uF in series with 33k Ohms, from mosfet's Source pin to ground (e.g. to "-" DC pin of bridge/neg terminal of filter cap), with the mosfet's Gate connected to the point between the 10uF and 33k (which set the time-constant/delay), plus a 1 Ohm 5W resistor from Source to Drain (used only during the startup-delay time). I also add a 10 Meg resistor in parallel with a 15v Zener, from Gate to Source (with Zener's cathode toward Source).
Some or all of those components' values and ratings would almost-certainly need to be changed, for different power supplies. The example above was for a small power supply (56VA transformer, 18VAC secondary), with a 12000uF filter cap (followed immediately, in this case, by a boost-mode SMPS with 38vdc output). The mosfet turns on gradually, over about 100ms, which keeps the input current spikes below 30A or so. In this supply, after startup, the mosfet only dissipates about 0.7 Watt, max.
It was a bit difficult to determine what power rating was needed for the 1 Ohm "bypass" power resistor. In the case above, the little 1 Ohm 5 Watt resistor can experience short periods of 10 to 20 Amp currents, i.e. up to several hundred watts of dissipation. Its initial (and maximum) stress was shown (in simulations) to be having its current go from zero to eighteen amps (332 Watts) and back to zero, in about six milliseconds. Over the first 90 milliseconds, its worst-case average power dissipation was 85.38 Watts, with an average current of 7.03A (9.24A RMS). So, you would probably want to use a resistor for which the manufacturer's datasheet gives data for "pulse-handling", or "surge" ratings of some sort. I ended up using part number 594-AC05W1R000J from http://www.mouser.com , which has a link to the datasheet for that series, which includes pulse-handling graphs that seem to indicate that a 3W 1 Ohm resistor would not quite be able to handle the 90ms of such surges, forcing me to make room for the 5W version. (I hope that I interpeted the graphs correctly... But, no flames or explosions, so far.)
To redesign with this topology for another power supply, the easiest way would probably be to download the free LTSpice (aka SwitcherCAD) from http://www.linear.com , and get the STP80PF55 spice model from the st.com website (or a model for whatever P-MOSFET you choose, from wherever), and simulate your power supply's startup with it, with worst-case loading. Then you could simply TRY different values for the power resistor, and different RC time constants for the mosfet's gate drive, and find something that might work for your particular power supply.
The LTSpice simulator will give plots of any or all currents, voltages, power dissipations, etc, versus time, just by clicking the mouse "probe" wherever you want on the schematic. And when a simulation run has stopped, you can also get the average and/or integral of any plotted quantity, by holding down CTRL and clicking on the plot label. (Power dissipation for any device can be plotted by holding ALT and clicking on the device, on the schematic).
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
I stopped using them in my power supplies and switched to using a MOSFET-based soft-start circuit. That way, I get a known (and very, very low) resistance, after the soft start is complete. (And the total cost is not much higher, although it takes more room. However, it also doesn't get nearly as HOT as a CL thermistor would get.)
The circuit is very simple and uses only six components. It goes between the diode bridge and the main filter capacitor(s). There is a hefty P-MOSFET (e.g. 55-Volt/80-Amp STP80PF55, or IRF4905) in parallel with a low-value power resistor, which are placed in series with the power rail (i.e. between the bridge's +DC output and the filter caps' "+" side). The mosfet's Source is connected to the rectifier bridge and its Drain is connected to the filter caps. Then components are added so that the mosfet (normally "off"/open circuit) turns on "gradually", eventually becoming a 0.018-Ohm resistance in parallel with the power resistor, effectively removing the power resistor from the circuit.
One that I recently designed was configured like this: 10uF in series with 33k Ohms, from mosfet's Source pin to ground (e.g. to "-" DC pin of bridge/neg terminal of filter cap), with the mosfet's Gate connected to the point between the 10uF and 33k (which set the time-constant/delay), plus a 1 Ohm 5W resistor from Source to Drain (used only during the startup-delay time). I also add a 10 Meg resistor in parallel with a 15v Zener, from Gate to Source (with Zener's cathode toward Source).
Some or all of those components' values and ratings would almost-certainly need to be changed, for different power supplies. The example above was for a small power supply (56VA transformer, 18VAC secondary), with a 12000uF filter cap (followed immediately, in this case, by a boost-mode SMPS with 38vdc output). The mosfet turns on gradually, over about 100ms, which keeps the input current spikes below 30A or so. In this supply, after startup, the mosfet only dissipates about 0.7 Watt, max.
It was a bit difficult to determine what power rating was needed for the 1 Ohm "bypass" power resistor. In the case above, the little 1 Ohm 5 Watt resistor can experience short periods of 10 to 20 Amp currents, i.e. up to several hundred watts of dissipation. Its initial (and maximum) stress was shown (in simulations) to be having its current go from zero to eighteen amps (332 Watts) and back to zero, in about six milliseconds. Over the first 90 milliseconds, its worst-case average power dissipation was 85.38 Watts, with an average current of 7.03A (9.24A RMS). So, you would probably want to use a resistor for which the manufacturer's datasheet gives data for "pulse-handling", or "surge" ratings of some sort. I ended up using part number 594-AC05W1R000J from http://www.mouser.com , which has a link to the datasheet for that series, which includes pulse-handling graphs that seem to indicate that a 3W 1 Ohm resistor would not quite be able to handle the 90ms of such surges, forcing me to make room for the 5W version. (I hope that I interpeted the graphs correctly... But, no flames or explosions, so far.)
To redesign with this topology for another power supply, the easiest way would probably be to download the free LTSpice (aka SwitcherCAD) from http://www.linear.com , and get the STP80PF55 spice model from the st.com website (or a model for whatever P-MOSFET you choose, from wherever), and simulate your power supply's startup with it, with worst-case loading. Then you could simply TRY different values for the power resistor, and different RC time constants for the mosfet's gate drive, and find something that might work for your particular power supply.
The LTSpice simulator will give plots of any or all currents, voltages, power dissipations, etc, versus time, just by clicking the mouse "probe" wherever you want on the schematic. And when a simulation run has stopped, you can also get the average and/or integral of any plotted quantity, by holding down CTRL and clicking on the plot label. (Power dissipation for any device can be plotted by holding ALT and clicking on the device, on the schematic).
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Tom:
That circuit will do a good job slowly bringing up your caps, but it won't do anything about magnetizing current in a large transformer. Really, something line-side is needed for that, which a thermistor works very well for, at least in an amp with fairly constant current draw. Even some variation in load would be alright, as long as the continuous draw is high enough to keep the resistance near zero. I'm just expecting that these chips have a low enough idle current draw to allow the thermistor to cool. I guess I'll find out tonight when I hook that power supply up to my amp.
-Nick
That circuit will do a good job slowly bringing up your caps, but it won't do anything about magnetizing current in a large transformer. Really, something line-side is needed for that, which a thermistor works very well for, at least in an amp with fairly constant current draw. Even some variation in load would be alright, as long as the continuous draw is high enough to keep the resistance near zero. I'm just expecting that these chips have a low enough idle current draw to allow the thermistor to cool. I guess I'll find out tonight when I hook that power supply up to my amp.
-Nick
Dick West said:
Notice that this is an Elektor design which Promitheus of some peculiar reason has put his own copyright
, anyway could take a peak at my softstart softstart and check at the bottom of the page. You'll to change the voltage dividing caps. They must be bigger in value. I have done a simulation (LTSpice) of this circuit. Drop me a message if you want it.
another option
Using a thermister, such as a CL-60 is fine even for amps that draw uneven current if you shunt them from the line after a short delay. This is how mine works, the down side is that you need a small separate xformer to power the relay.
Using a thermister, such as a CL-60 is fine even for amps that draw uneven current if you shunt them from the line after a short delay. This is how mine works, the down side is that you need a small separate xformer to power the relay.
Arx said:
Nick,
You are right. Thanks. Sorry, I was only considering the secondary side of the transformer.
Thinking about the primary side, now, and having relatively little practical or detailed knowledge of transformers, I'm wondering what the relative magnitudes and time-behaviors are, of the magnetizing current versus any load-induced current, in the primary side, particularly during startup. I suppose I could try to measure them, while powering-up a transformer with and without a load. But, "off the top of my head", I'm not quite sure how I would go about doing that.
Does anyone know of a good reference or online information source, for learning more about the details of transformer behavior? Long ago, around 1978, I did take a course (EE311 at Purdue) called "Engineering Electromagnetics". But all I remember it covering was transmission lines, antennas, and Maxwell's Equations. I'll dig the textbook out and check for transformers.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
I think the magnetizing inrush is at a minimum when the power comes on at the peak voltage. ~170v in North America, twice that most other places. An NTC with a relay or something to bypass it once the caps are up to voltage would work well I think. I don't see why it couldn't be powered off the main transformer.
-Nick
-Nick
Arx said:
Nick,
OK. I'm just starting to do a little research on this. But I thought I'd pass along a couple of fairly-good "basic information" discussion-threads that I found, about transformer inrush phenomena:
http://groups.google.com/group/sci....hread/thread/68419eff354b0f1/e53036154633c009
http://groups.google.com/group/sci....read/thread/d739b8eb1095064f/076f7278c7055e45
If I understood (and remember) correctly, there are TWO simultaneous inrush phenomena that are occurring, in the primary transformer winding: 1) The secondary side's load, e.g. caps charging, is almost exactly reflected through to the primary (ratioed, of course). This effect is _worst_ if switched on at peak of input AC wave. 2) Magnetization current inrush: If first half cycle is same polarity as last half cycle was when last powered down, "serious" inrush can occur, if core is still magnetized enough in same direction (from previous power on, usage, then off) that it saturates. This effect is apparently _least_ when switched on at a peak of the AC input wave.
I think you are right, that a current-limiting device on the primary side could be used, that is switched out of the circuit by a device that's powered by the main transformer (i.e. the transformer for which the inrush current is being limited), at least in the case where the switching device (relay, transistor, etc) is "normally open", so that power would need to be applied to close the switch, to bypass the current-limiting element, after the inrush has subsided. That would ensure that the current-limiting device was in already in place for the critical first half-cycle of mains voltage, which is apparently when the largest part of the portion of the inrush that's due to magnetization current can occur.
I guess I still don't have a good "feel" for what current or voltage conditions might be "best" to try to sense, to control the switch-out of the current-limiting element. A simple time-delay should be effective, under _normal_ startup conditions. I guess I'm just wondering what advantages there might be to trying to base the switch-out decision on other things, such as the voltage at the smoothing cap, and/or actual sensed inrush current into primary winding, et al, and whether there is any point to trying to use a semiconductor, for example, instead of a relay, to make the transition less abrupt. And, maybe a semiconductor-type of gradually/controlled switching device could also serve as the current-limiting element, eliminating the need for a power resistor, or thermistor (in which case, if I were going to bypass it after the inrush anyway, and assuming a power resistor is cheaper and not much larger than a thermistor, I'd use a resistor instead).
Sorry to have blathered-on about all of that, for so long. I'll do a little more research, see more of what others have typically done, and post here if I come up with anything good.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
I kludged together a little power supply
120V line -> CL50 thermistor -> 120:25-0-25 600VA Transformer -> Cheap 25A bridges -> 68000 uF per rail -> 1500uF local to the chip
meh... seems to work. 38VDC per rail around 12mV of ripple.
I'm not going to go through the trouble to get a difinitive measurement of the sag, but I played some really bassy music through a 100Hz digital lowpass filter at very high volume (Louder than I'd ever listen) while looking at the positive supply rail on my scope
I don't know a good way to see it since the sag is so slow I can only see it with the scope on DC, but the drop was very slight. I think probably well under a volt.
Keep in mind this is straight to high volume bass after 5 minutes or so of idle, so either the thermistor holds its heat very well, heats up very quickly, or some combination of the two.
To me it seems like the thermistor's contribution wouldn't be particularily harmful to the sound.
I haven't measured the rise time on the caps, though, so I'm not sure yet how effective it is in preventing inrush. Maybe I'll measure at some point.
Either way, limiting with just a thermistor is far from ideal. It'll take several minutes to cool down, so if you turn it off, and back on within a couple minutes (or the power fails for a few seconds before returning) the thermistor will still be hot and do very little.
A relay across a thermistor would be a slight improvement, as it would prevent nuisance trips and damage if power flicks off and on momentarily, but only if interruptions are spaced several minutes apart. A quick off, on, off, on would still be a problem.
Switch mode supplies are sounding nicer and nicer all the time.
-Nick
120V line -> CL50 thermistor -> 120:25-0-25 600VA Transformer -> Cheap 25A bridges -> 68000 uF per rail -> 1500uF local to the chip
meh... seems to work. 38VDC per rail around 12mV of ripple.
I'm not going to go through the trouble to get a difinitive measurement of the sag, but I played some really bassy music through a 100Hz digital lowpass filter at very high volume (Louder than I'd ever listen) while looking at the positive supply rail on my scope
I don't know a good way to see it since the sag is so slow I can only see it with the scope on DC, but the drop was very slight. I think probably well under a volt.
Keep in mind this is straight to high volume bass after 5 minutes or so of idle, so either the thermistor holds its heat very well, heats up very quickly, or some combination of the two.
To me it seems like the thermistor's contribution wouldn't be particularily harmful to the sound.
I haven't measured the rise time on the caps, though, so I'm not sure yet how effective it is in preventing inrush. Maybe I'll measure at some point.
Either way, limiting with just a thermistor is far from ideal. It'll take several minutes to cool down, so if you turn it off, and back on within a couple minutes (or the power fails for a few seconds before returning) the thermistor will still be hot and do very little.
A relay across a thermistor would be a slight improvement, as it would prevent nuisance trips and damage if power flicks off and on momentarily, but only if interruptions are spaced several minutes apart. A quick off, on, off, on would still be a problem.
Switch mode supplies are sounding nicer and nicer all the time.
-Nick
Hi,
the theoretical maximum peak current on starting up a transformer can approach Vpk/Rprimary, under the worst set of conditions.
In the UK the peak voltage is about 340Vpk but can be as high as 360Vpk.
The resistance of a 600VA primary (2*120Vac in series) is between 1r0 and 2r0. The lower values for a low regulation transformer.
These worst case numbers (= a very rare ocurrence) result in a peak current of about 300Apk but for a very short time period. A 10A fuse is likely to survive repeated starts indefinitely.
An 8A fuse may blow once or twice a decade, but this is more likely to be due to fuse fatigue rather than peak current overload.
A 6A fuse may blow as often as once or twice a fortnight.
Lets put a 50r resistance in series with the primary.
The worst case start up current is reduced to about 6.6Apk, (a reduction of about 50times).
You will find that a 3.1A fuse will last almost indefinitely and allow a working power of about 700VA to pass through the transformer.
The numbers for the US are similar and again the benefit of cloase rated fusing can be obtained by fitted a current limiting resistor or thermistor in the primary circuit.
Assuming Vpk=170Vpk and that the parallel connection of the 120Vac primaries then Ipk<=170/0.28=600A. I wonder how big the fuse would be to survive indefinitely with surges approaching just half this value at a frequency of once in every 50 re-starts?
But, maybe the advantage of close rated fusing is not attractive to the user.
the theoretical maximum peak current on starting up a transformer can approach Vpk/Rprimary, under the worst set of conditions.
In the UK the peak voltage is about 340Vpk but can be as high as 360Vpk.
The resistance of a 600VA primary (2*120Vac in series) is between 1r0 and 2r0. The lower values for a low regulation transformer.
These worst case numbers (= a very rare ocurrence) result in a peak current of about 300Apk but for a very short time period. A 10A fuse is likely to survive repeated starts indefinitely.
An 8A fuse may blow once or twice a decade, but this is more likely to be due to fuse fatigue rather than peak current overload.
A 6A fuse may blow as often as once or twice a fortnight.
Lets put a 50r resistance in series with the primary.
The worst case start up current is reduced to about 6.6Apk, (a reduction of about 50times).
You will find that a 3.1A fuse will last almost indefinitely and allow a working power of about 700VA to pass through the transformer.
The numbers for the US are similar and again the benefit of cloase rated fusing can be obtained by fitted a current limiting resistor or thermistor in the primary circuit.
Assuming Vpk=170Vpk and that the parallel connection of the 120Vac primaries then Ipk<=170/0.28=600A. I wonder how big the fuse would be to survive indefinitely with surges approaching just half this value at a frequency of once in every 50 re-starts?
But, maybe the advantage of close rated fusing is not attractive to the user.
Hi
I did similar calculations to Andrew and then did some tests with one of my power amps. The power amp has a 1kva transformer with a primary resistance of 1.8ohm. I experimented with various levels of capacitance on the rails.
With no thermister and 15000uf on each of 4 rails the power amp needed a 2.0 amp primary fuse. ( I only tried a limited number of starts). With 60,000uf on each of 4 rails the power amp needed an 8 amp fuse. With a thermister added in the primary circuit a 2.5 amp primary fuse worked.
I think this supports your calculations Andrew.
I currently use a 10ohm 15 amp thermister form Rapid electronics on my power amps. I bought them some years ago but Rapid still sell them. When I first tried them I was concerned about the high operating temperature and did not use the thermisters for a while. However I found that others-including pass labs- are using them so I have started to use them. However the temperature does still bother me.
Does anyone know how long thermisters last?
Don
I did similar calculations to Andrew and then did some tests with one of my power amps. The power amp has a 1kva transformer with a primary resistance of 1.8ohm. I experimented with various levels of capacitance on the rails.
With no thermister and 15000uf on each of 4 rails the power amp needed a 2.0 amp primary fuse. ( I only tried a limited number of starts). With 60,000uf on each of 4 rails the power amp needed an 8 amp fuse. With a thermister added in the primary circuit a 2.5 amp primary fuse worked.
I think this supports your calculations Andrew.
I currently use a 10ohm 15 amp thermister form Rapid electronics on my power amps. I bought them some years ago but Rapid still sell them. When I first tried them I was concerned about the high operating temperature and did not use the thermisters for a while. However I found that others-including pass labs- are using them so I have started to use them. However the temperature does still bother me.
Does anyone know how long thermisters last?
Don
Hi,
nice to see some actual numbers.
I am surprised that 2A (240Vac) allows a start-up into 1kVA.
I cannot get a 625VA to start on less than 5A (single primary) and the poorer one (dual primary) needs 6.2A. Both now have 50r//relay and 3A fuse.
Try adding a timer or voltage activated relay to let the power thermistor to cool down for the next re-start.
nice to see some actual numbers.
I am surprised that 2A (240Vac) allows a start-up into 1kVA.
I cannot get a 625VA to start on less than 5A (single primary) and the poorer one (dual primary) needs 6.2A. Both now have 50r//relay and 3A fuse.
Try adding a timer or voltage activated relay to let the power thermistor to cool down for the next re-start.
AMV8 said:
Well, I have been using them in a 2x 120 WPC bridged amp for about 3 years now with no failure.
I had been blowing MUR860's which have a peak rating of 100 A - this when I was using the transformer from an anodizing power supply.
There is some care and feeding of the GE Sensing inrush limiters -- they do need access to free air flow.
If you use a thermistor with too high an "on resistance" you can get some modulation of the power supply output. You have both an I^2R resistance loss plus variation of the resistance with temperature.
regarding magnetization inrush:
It does occur the strongest when turned on at 0v and weakest when turned on at peak.
As far as the caps on the secondary pulling the most current at Peak input voltage, that's not entirely accurate either. They'll draw the most at THEIR peak input voltage, but that's not going to be tied directly to the peak input voltage to the transformer. There will be phase shift through the transformer.
Here's a really good article on the subject:
http://relays.tycoelectronics.com/appnotes/app_pdfs/13c3206.pdf
Heh.. we're kind of off topic on the original question of the thread.
I suppose it's important to know which side of the transformer you'll be putting the thermistor on before someone tells you how to calculate it.
I can't be much help there.. I just grabbed on the looked to be in the right ballpark. Maximum power draw from your amp should leave you below the thermistor's max current rating at its minimum resistance, but at the same time, you want it close enough to the maximum that you're keeping it very hot so the resistance stays low and doesn't sag your voltage.
Without any extensive testing, a CL50 seems to work well enough on my amp.
-Nick
It does occur the strongest when turned on at 0v and weakest when turned on at peak.
As far as the caps on the secondary pulling the most current at Peak input voltage, that's not entirely accurate either. They'll draw the most at THEIR peak input voltage, but that's not going to be tied directly to the peak input voltage to the transformer. There will be phase shift through the transformer.
Here's a really good article on the subject:
http://relays.tycoelectronics.com/appnotes/app_pdfs/13c3206.pdf
Heh.. we're kind of off topic on the original question of the thread.
I suppose it's important to know which side of the transformer you'll be putting the thermistor on before someone tells you how to calculate it.
I can't be much help there.. I just grabbed on the looked to be in the right ballpark. Maximum power draw from your amp should leave you below the thermistor's max current rating at its minimum resistance, but at the same time, you want it close enough to the maximum that you're keeping it very hot so the resistance stays low and doesn't sag your voltage.
Without any extensive testing, a CL50 seems to work well enough on my amp.
-Nick
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.