Hi,
I'm designing an amp based on the LM3886 (I have a separate thread covering this over in Chip Amps), and with it a power supply for delivering +/- 28v-or-so.
I'm new to LTSpice, but thought the PSU circuit would be a good candidate for a first circuit to sim, so I've modelled it as per my schematic using 2 AC voltage sources running at 50Hz, around 29.5v to emulate the max output from a split secondary 18v toroidal transformer (UK mains).
The load circuit is borrowed from post 12 in this thread, as LTSpice doesn't have a LM3886 model included as standard, so this may skew things a bit!
It's full bridge rectified per rail, using 2x10,000uF caps for smoothing. According to the LTSpice sim, this gives about 600mV of ripple at just over 28V.
My question regards the current at C1/C7. The sim shows just under 80A immediately, quickly settling to around 11A spikes per half wave. Is this correct and normal, or will it quickly degrade the smoothing caps (at least C1 and C7)? The spikes last around 1.5ms. I do have an NTC thermistor pre-transformer for inrush limitation.
I notice that by removing the PI filter resistors, this current reduces to 8A, but the ripple waveform is much sharper. Will this matter?
The caps I've chosen are Nichicon UKW 35V, which have a ripple current rating of 3.7A. Mouser doesn't have anything resembling the UPW type I'd like to use (long life, low impedance, high temp) at 10,000uF 35v in the leaded type. What are your preferred caps for this kind of application? I can redesign my board a bit to cope with larger caps if necessary (currently it can fit 2x25mm diameter caps with the resistors in between).
Thanks, Christian
I'm designing an amp based on the LM3886 (I have a separate thread covering this over in Chip Amps), and with it a power supply for delivering +/- 28v-or-so.
I'm new to LTSpice, but thought the PSU circuit would be a good candidate for a first circuit to sim, so I've modelled it as per my schematic using 2 AC voltage sources running at 50Hz, around 29.5v to emulate the max output from a split secondary 18v toroidal transformer (UK mains).
The load circuit is borrowed from post 12 in this thread, as LTSpice doesn't have a LM3886 model included as standard, so this may skew things a bit!
It's full bridge rectified per rail, using 2x10,000uF caps for smoothing. According to the LTSpice sim, this gives about 600mV of ripple at just over 28V.
My question regards the current at C1/C7. The sim shows just under 80A immediately, quickly settling to around 11A spikes per half wave. Is this correct and normal, or will it quickly degrade the smoothing caps (at least C1 and C7)? The spikes last around 1.5ms. I do have an NTC thermistor pre-transformer for inrush limitation.
I notice that by removing the PI filter resistors, this current reduces to 8A, but the ripple waveform is much sharper. Will this matter?
The caps I've chosen are Nichicon UKW 35V, which have a ripple current rating of 3.7A. Mouser doesn't have anything resembling the UPW type I'd like to use (long life, low impedance, high temp) at 10,000uF 35v in the leaded type. What are your preferred caps for this kind of application? I can redesign my board a bit to cope with larger caps if necessary (currently it can fit 2x25mm diameter caps with the resistors in between).
Thanks, Christian
Attachments
80 A is of cause a large current but it seems to me all rectifiers work with similar transients and live a long life.
And you may add output resistanse to the AC voltage sourses to simulate real ones.
Output impedanse usually is a two or a bit more times larger then power transformer secondary winding DC resistance.
If your power transformer is relatively "large" power rated (for example, >100 VA), you may add a soft-start - it reduces peak cap charging current too.
And you may add output resistanse to the AC voltage sourses to simulate real ones.
Output impedanse usually is a two or a bit more times larger then power transformer secondary winding DC resistance.
If your power transformer is relatively "large" power rated (for example, >100 VA), you may add a soft-start - it reduces peak cap charging current too.
Last edited:
To start:
D1 is ok
D2 does nothing
D3 connects V1 supply to gnd on negative swing of V1
D4 both sides connected to gnd
D5 both sides connected to gnd
D6 connects V2 supply to gnd on positive swing of V2
D7 does nothing
D8 is ok
Just use one bridge ( 4 diodes ) and set V1 and V2 to be 180 degrees from each other.
You will find your ripple frequency will be 100Hz not 50Hz...just like reality
And what Vovk Z said
🙂
D1 is ok
D2 does nothing
D3 connects V1 supply to gnd on negative swing of V1
D4 both sides connected to gnd
D5 both sides connected to gnd
D6 connects V2 supply to gnd on positive swing of V2
D7 does nothing
D8 is ok
Just use one bridge ( 4 diodes ) and set V1 and V2 to be 180 degrees from each other.
You will find your ripple frequency will be 100Hz not 50Hz...just like reality
And what Vovk Z said
🙂
Last edited:
D'oh! Yes you're right - this is not wired up right. It is in my original schematic, just not properly in LTSPice. I've rectified that (pun intended) now and things look a lot healthier! Ripple current down to about 4A.
I'll try that. Practically, are you talking about wiring up the transformer as a split secondary (18-0-18), so commoning two of the secondary outputs as ground, with each rail sharing a single 4-diode rectifier? (Sorry for the newbie questions, for that's what I am!)
Thanks both, Christian
The ground loop in the OP schematic is a terrible mistake.
It jumped to my eye.
Center tap with 4 diodes, versus two independent windings with 8 diodes ?
IMO, the two independent windings with two bridges is better.
It jumped to my eye.
Center tap with 4 diodes, versus two independent windings with 8 diodes ?
IMO, the two independent windings with two bridges is better.
You do not seem to have included real-world parameters for the caps and voltage sources (=transformer).
Both the transformer and the caps have a finite series resistance, and the transformer also has a leakage inductance.
These parameters play a major in the shape/magnitude of the current spikes and shouldn't be neglected.
I think I have provided somewhere estimated formula's for transformers parameters, based on their size and construction.
I'll try to locate the post.
However, the best (most accurate) method is to measure them directly
To clarify the previous post: these aren't actually formula's, the values shown are transformer's constants, expressed in the correct units, like ohm*VA or mH*VA.
To find values for a given transformer, you need to divide the suitable constant by the VA rating: for example, a 100VA toroid would have a primary resistance of 4700/100=47ohm, and a leakage inductance of 250/100=2.5mH (for a 230V/50Hz system).
These are very crude approximations, but they are better than nothing if your are unable to measure the actual values.
To find values for a given transformer, you need to divide the suitable constant by the VA rating: for example, a 100VA toroid would have a primary resistance of 4700/100=47ohm, and a leakage inductance of 250/100=2.5mH (for a 230V/50Hz system).
These are very crude approximations, but they are better than nothing if your are unable to measure the actual values.
Remember that all the parasitic parameters are primary-referred: the secondary just has the magnetizing inductance required to achieve the transformation ratio, but is otherwise perfect.
Also try to include some esr for the caps: it will contribute to the realism of the sim
Also try to include some esr for the caps: it will contribute to the realism of the sim
Hi Elvee, thanks for the pointers. I finally got round to trying this out, simulating a transformer with the formulas you provided. It made quite a difference to the current draw of the first smoothing cap, which settles to something within spec of the real caps I intend to use.