Indeed. I've build several 'sequencers' that switch on a number of mains outlets in a predetermined order, and switches off in reverse order. So you are sure the preamp is before switching the power amp on, and vice versa
I am using a triac AND a relay. The triac switches on at zero-crossing, but has a small 'dead band' that can cause DC on the xformer. So after the triac is on, a relay shorts it. At switch off, the relay opens first and then the triac.
By timing the switch-on and the switch-off at the exact same point on the sine wave, there's zero remanent magnetism in the core and inrush is almost totally absent.
Once you decide to use a micro-controller, the sky is the limit!
Jan
Aha!! A testable hypothesis!
I configured my test board's jumpers to give Full Wave Bridge operation, installed a 1.5 ohm current sensing resistor (measured value on Keithley bench DMM = 1.44 ohms), hooked up the scope, and attached an electronic load to the output. My electronic load is just a current source + enormous fan, that lets you draw whatever load current you want, from a power supply. Schematic of my rig is shown in Figure 1 below. Please recall that the transformer is a 15VAC wall wart from the O2 Headphone Amplifier; it's pretty scrawny.
I snapped scope photos of it operating at four different values of load current flowing in "I1". The photos say this:
Code:
External Diode PulseWidth Diode
Load Current as % of Conduction
Current PulseWidth 8.33 msec Angle
0 mA 2.0 ms 24% 43 deg
54 mA 3.2 ms 38% 69 deg
158 mA 4.0 ms 48% 86 deg
400 mA 5.0 ms 60% 108 degEd: looks like the conduction pulse width isn't a fixed and constant 30%.
anatech: looks like the diode conduction angle isn't teeny tiny.
PH104 & Jan: see whether you can spot the oscillatory ringing part of the current waveforms. It's there, I promise; but very small. This is why I prefer to look at L*dI/dt instead.
Added later: notice how beautiful the scope traces are, when driven by a source impedance of 1.44 ohms! Noise doesn't stand a chance of polluting these little beauties. Averaging was not used when snapping these shots.
Attachments
Last edited:
A thought: the conduction angle of the diodes also depends on the 'softness' of the mains transformer. A smallish transformer will have more secondary voltage droop with increasing load current and thus will show larger conduction angle increase with load increase.
In your illustrative cases, when the transformer would have been a larger low-impedance one, the conduction angle wold not increase as much.
But of course, each supply will have a transformer that is more or less matched, in secondary impedance, to the load currents expected. So, maybe it would be interesting to look at conduction angle versus percentage of load from the maximum that is drawn. *Maybe* this is more or less constant for supplies small or large.
Posit: A large supply with a transformer sized for 10A will have the same conduction angle at 5A, as a supply sized for 1A will have at 500mA.
Or is that too far fetched?
Jan
In your illustrative cases, when the transformer would have been a larger low-impedance one, the conduction angle wold not increase as much.
But of course, each supply will have a transformer that is more or less matched, in secondary impedance, to the load currents expected. So, maybe it would be interesting to look at conduction angle versus percentage of load from the maximum that is drawn. *Maybe* this is more or less constant for supplies small or large.
Posit: A large supply with a transformer sized for 10A will have the same conduction angle at 5A, as a supply sized for 1A will have at 500mA.
Or is that too far fetched?
Jan
I dunno, Jan. This one's conduction angle increased pretty appreciably when going from zero (really: 2.9mA) to 54mA (really: 56.9mA). It's supposedly a 500 mA (AC RMS) transformer, so I would think pulling 54-57mA of DC current {after HUGE filter capacitors} would not be the least bit challenging. Besides, Ed Simon said "30%" not "30% with the following caveats".
Here's the transformer. At what kinds of load current would you expect it to start misbehaving? BTW the DC resistance of the secondary was about 3.5 to 4 ohms. So the 1.44 ohm current sense resistor did not change things all that much.
Here's the transformer. At what kinds of load current would you expect it to start misbehaving? BTW the DC resistance of the secondary was about 3.5 to 4 ohms. So the 1.44 ohm current sense resistor did not change things all that much.
Attachments
Last edited:
Mark,
I look at your data as a confirmation that conduction runs around 30% in a properly designed power supply. At 400 mA 15 V the wall wart is probably past legitimate long term power rating. Also I aim for 5% ripple maximum 10%. I suspect you were outside normal design parameters. Pretty sure at a high conduction angle you are not getting close to the maximum voltage.
BTY a transformer rated for 1 A RMS AC will not produce 1 A DC! (12 V AC does not give 12 V DC. A few other bits factor in so don't even get the 70% you would expect but closer to 50% depending on things like internal impedance.)
So I don't think we are in disagreement. (About is the key word!)
As to using a triac to switch a transformer I find the zero crossing dead zone puts amazing amounts of HF into the transformer. Relays really do a nice job, but do cost more. Of course pairing them does reduce the surge current on the contacts. When sizing a relay like that the issue becomes the surge current when the secondary of the power transformer is shorted.
I look at your data as a confirmation that conduction runs around 30% in a properly designed power supply. At 400 mA 15 V the wall wart is probably past legitimate long term power rating. Also I aim for 5% ripple maximum 10%. I suspect you were outside normal design parameters. Pretty sure at a high conduction angle you are not getting close to the maximum voltage.
BTY a transformer rated for 1 A RMS AC will not produce 1 A DC! (12 V AC does not give 12 V DC. A few other bits factor in so don't even get the 70% you would expect but closer to 50% depending on things like internal impedance.)
So I don't think we are in disagreement. (About is the key word!)
As to using a triac to switch a transformer I find the zero crossing dead zone puts amazing amounts of HF into the transformer. Relays really do a nice job, but do cost more. Of course pairing them does reduce the surge current on the contacts. When sizing a relay like that the issue becomes the surge current when the secondary of the power transformer is shorted.
Last edited:
That size transformer is usually "impedance" protected" as in no fuse. You can short the output and it won't fry. It also means the peak current is effectively limited. Similar for a Sola constant voltage transformer, which operate in saturation.
Does that scope support waveform math? Try V*A waveform to see the peak power in. Also look at the phase relationship V to I as the load is changed.
Does that scope support waveform math? Try V*A waveform to see the peak power in. Also look at the phase relationship V to I as the load is changed.
I have been using a resistor and a relay to manage in- rush on 1kVA and 2kVA torroids under MCU control with great success - 10 years in one case with no issues.
I've gone over to a NTC ( high energy handling and small size) and a relay under MCU control for my latest PSU's (500 watt and 1.2 kVA) and it works very well.
In rush and cap charging currents on a big PSU top 100 Amps - I've seen some literature from a few years ago showing peaks of 200A for the first cycle. Soft start is a must.
I've gone over to a NTC ( high energy handling and small size) and a relay under MCU control for my latest PSU's (500 watt and 1.2 kVA) and it works very well.
In rush and cap charging currents on a big PSU top 100 Amps - I've seen some literature from a few years ago showing peaks of 200A for the first cycle. Soft start is a must.
Last edited:
Ringing is present in the current waveform but it is very small, and scopes often have a hard time displaying it cleanly. So I prefer to view the ringing voltage instead of the ringing current.
Remember this image is from terrible, awful diodes (1N4003s), the absolutely worst axial-lead diodes tested. Also using a transformer with unusually high leakage inductance. Imagine how tiny the ringing current would be, with half decent diodes and a better than average transformer. Note to self: display the ringing voltage, not the current.
Attachments
Last edited:
Ed, once the triac is conduction, there is only a small voltage across it, and if you time the relay closure such that there is no secondary load current, the surge current through the relay is almost negligible. As I said, if you know u-controllers and Ohm's law, you can do anything!
Jan
A (very) short digression: we talked a few pages back about current hogging due to temp differences on a MOSFET die. See attached - I thought this was a nice picture which I didn't want you guys to deprive off!
A 10:1 difference in T in degree centigrade across the die!
Courtesy an Infineon paper unearthed by my friend Ian Hegglun.
Jan
A 10:1 difference in T in degree centigrade across the die!
Courtesy an Infineon paper unearthed by my friend Ian Hegglun.
Jan
Attachments
Jan,
Once the relay is closed the triac stops conducting. If you have an output transistor fail the relay will see far higher current than normal. It is that surge that it must be able to handle. I have seen 20 A relays cook on a short circuit even with a 20 A circuit breaker.
Mark,
I think the significance of the turn off ringing is voltage dominant is that the transformer secondary is abruptly unloaded. If you can do the measurement with just a resistor across the secondary drawing 10% power it would be interesting to compare that to a full snubber. Of course the straight resistor approach is only useful on low power stuff.
I also like to put a capacitor across each diode equal to the diode's capacitance. That should halve any stored energy spike. Larger capacitors would offer some improvement of course but could never reach zero. The downside of any diode bypass capacitor is of course the increase in line noise.
Once the relay is closed the triac stops conducting. If you have an output transistor fail the relay will see far higher current than normal. It is that surge that it must be able to handle. I have seen 20 A relays cook on a short circuit even with a 20 A circuit breaker.
Mark,
I think the significance of the turn off ringing is voltage dominant is that the transformer secondary is abruptly unloaded. If you can do the measurement with just a resistor across the secondary drawing 10% power it would be interesting to compare that to a full snubber. Of course the straight resistor approach is only useful on low power stuff.
I also like to put a capacitor across each diode equal to the diode's capacitance. That should halve any stored energy spike. Larger capacitors would offer some improvement of course but could never reach zero. The downside of any diode bypass capacitor is of course the increase in line noise.
That's not the usual use of surge; that's the running current; not the switching current. At any rate, sizing the relay on the primary side is a routine engineering solution.
Nice, thanks.
The lowest temp is around 230 C, and the highest in the ...near 500's?
Bet the pattern matches the die attach integrity. Did they happen to do an ultrasonic scan of that?
ps..you did mean 2 to 1 difference, no? That's another of those graphs where labels were horribly done. I've never seen a thermal profile that listed temps in the color bar that weren't on the display..
Sigh.. what are they teaching kids in the schools today??
John.
Last edited:
Jan,
When I mean turn on surge that is what I call it. There are many other cases. Same reason I use loudspeaker not speaker.
I was looking at the colored squares which seem to indicate a 10:1 variation.
And yes they say the cold regions are the 'bond attach shadows'.
I'll send yo the paper by mail - has a very nice thermal Spice model developed.
Jan
- Status
- Not open for further replies.