| Electronics > Projects, Designs, and Technical Stuff |
| What input protections to add and filtering across a transformer ? |
| (1/2) > >> |
| lordvader88:
I'm making a LV linear DC PSU with 2 step-down transformers, and neither had mains/earth, or any capacitors like X /Y or any chokes like on a SMPS. I know that on a SMPS it's meant to keep the switching noise off mains. But why don't old linear transformer power devices have common mode chokes and "stuff" ? I can add a fuse, and mains/earth wires to the PSU chasis and transformers. I don't know how to calculate resistor or cap values that might be needed across the various transformer wingdings to suppress ringing. I know some LCR circuit basics, but I don't know the construction of these transformers. I never actually did a short circuit test, but I know the rating of 1. I know calulating this stuff even with all the magnetic spec's etc, so harder. I have a $140 LCR/ESR meter, and scopes. So what values should I measure and actually trust ?. I saw a website that linked to a calculator for that, I'll look it up. Both these transformers have other smaller winding's I'm not using. So these aren't the best choice, but it home-made from salvage. Even if the primary winding acts about the same, should I leave the other wingdings open, or add a high value resistor ? |
| T3sl4co1l:
The transformer does a good job to begin with, although toroids not as much (good bandwidth / poor filtering). Split-bobbin (bank wound) types are the best at filtering, though also the poorest regulation. Shell type is inbetween. You may actually want a line filter, to block line noise coming in, if using an unshielded toroid and the circuit is low noise. Most any filter module would do, or you can make one from discrete parts of the same typical values (with some attention to layout and shielding, if the system is sensitive to RFI). A transformer with low interwinding capacitance, and preferably a shield, is also better, in whichever style. FWB doesn't generate much noise, some diode recovery at mains frequency if you're unlucky. An R+C of modest value (10R + 10n should be generic enough to work) can be used to help with that. Ideally, R = sqrt(L/C), where L is winding leakage and C is diode capacitance, and the series C is several times diode capacitance. Winding capacitance may have to be added to that as well. 10nF is a pretty safe overestimate, and not very substantial at line frequency; that leaves just R to optimize. Note that you need at least two RCs for a FWB (e.g., AC1 to DC-, DC- to AC2, and bypass cap from DC- to DC+). Or one for each diode in FWCT. Tim |
| coppercone2:
remember it comes down to a transfer function and you can do it a few different ways , the main reason why things are built the way they are with a hyper common device like a transformer is cost 'trying to study a hyper optimized mass manufactured design can drive you insane if you don't keep that mind and keep in mind the transfer function of impedance not only the pass voltage, and the distortion effects, and surge behavior |
| MosherIV:
--- Quote ---I can add a fuse, and mains/earth wires to the PSU chasis and transformers. --- End quote --- Fuse - Absolute bear minimum you should be doing with any mians based circuit (unless you like waiting for unknown periods of time before poof, smoke and fire). Again, with metal chassis or case - you must earth the case/chassis (unless you want to wait for unkown periods of time before electrocuting someone, maybe yourself) I do not know much about filtering transformer based linear psus. Sometime, with very big transformers a current inrush protection circuit is needed. From what you have said, I do not think you need one in your psu. |
| lordvader88:
--- Quote from: T3sl4co1l on June 21, 2019, 10:31:32 pm ---The transformer does a good job to begin with, although toroids not as much (good bandwidth / poor filtering). Split-bobbin (bank wound) types are the best at filtering, though also the poorest regulation. Shell type is inbetween. You may actually want a line filter, to block line noise coming in, if using an unshielded toroid and the circuit is low noise. Most any filter module would do, or you can make one from discrete parts of the same typical values (with some attention to layout and shielding, if the system is sensitive to RFI). A transformer with low interwinding capacitance, and preferably a shield, is also better, in whichever style. FWB doesn't generate much noise, some diode recovery at mains frequency if you're unlucky. An R+C of modest value (10R + 10n should be generic enough to work) can be used to help with that. Ideally, R = sqrt(L/C), where L is winding leakage and C is diode capacitance, and the series C is several times diode capacitance. Winding capacitance may have to be added to that as well. 10nF is a pretty safe overestimate, and not very substantial at line frequency; that leaves just R to optimize. Note that you need at least two RCs for a FWB (e.g., AC1 to DC-, DC- to AC2, and bypass cap from DC- to DC+). Or one for each diode in FWCT. Tim --- End quote --- I do have a line filter in a little can, from big Sony LCD TV, and SMPS. And there's plenty of SMPS plugged into things in the house for sure, so I'll use the filter, fuse and gnd'ing. The last time I made this PSU with smaller windings, and I had some funny oscillation, that went on for ages after being turned off. So I'm sure it was the transformer. Using the frequency of oscillation I just added a low Z cap across the transformer somewhere, iirc,, and I got rid of most of noise that was sometimes growing in 1 of the BJTs. But I also had wires going everywhere on the perf.board, and it's not in a metal case yet, I'm working on the re-made optimized perf.board now. Almost done. |
| Navigation |
| Message Index |
| Next page |