I'm getting the major parts together for my active tri-amped speakers and I wanted to check with more experienced members of the forum if my plan for the power supplies are sufficient. There will be two of these supplies (L&R) each providing a recommended +/56Vdc to three mosfet amps each with a max power of 100w into 8ohms (though not all that will actually be used in my application).
I've attached a diagram; I'm currently intending 'Option1' and have most of the parts for that already. Its fairly conventional and aside from the transformer the power supplies are duplicated for each (low, mid & high) amp. I went with 80v caps (rather than say 63v) because I calculated 40Vac + 10% transformer regulation + 10% mains supply variation + rectification could mean nearly 70Vdc should be expected at times, even though it would nominally be 56Vdc. There are two 4,700uf caps per rail separated by a small resistor; their ripple rating is 4A each so a total of 8A - a rule of thumb seems to be that they should be rated to at least twice the max load current (which I make to be 2x 3.5A@100w/8ohm = 7A), so 8A is okay rather than superb, though my actual requirements of the amps are likely to be more like say 60W max (much less for the tweeter) so its not too tight in practice.
However, I've noticed some projects using considerably more capacitance, and have also seen some designs which share the first pair of capacitors after the rectifier ('option2'). I don't like this as much but with three amps sharing it would mean that I could double up my capacitors 'relatively' economically, if people think thats really worth doing?
Thoughts would be appreciated!
Cheers
Kev
I've attached a diagram; I'm currently intending 'Option1' and have most of the parts for that already. Its fairly conventional and aside from the transformer the power supplies are duplicated for each (low, mid & high) amp. I went with 80v caps (rather than say 63v) because I calculated 40Vac + 10% transformer regulation + 10% mains supply variation + rectification could mean nearly 70Vdc should be expected at times, even though it would nominally be 56Vdc. There are two 4,700uf caps per rail separated by a small resistor; their ripple rating is 4A each so a total of 8A - a rule of thumb seems to be that they should be rated to at least twice the max load current (which I make to be 2x 3.5A@100w/8ohm = 7A), so 8A is okay rather than superb, though my actual requirements of the amps are likely to be more like say 60W max (much less for the tweeter) so its not too tight in practice.
However, I've noticed some projects using considerably more capacitance, and have also seen some designs which share the first pair of capacitors after the rectifier ('option2'). I don't like this as much but with three amps sharing it would mean that I could double up my capacitors 'relatively' economically, if people think thats really worth doing?
Thoughts would be appreciated!
Cheers
Kev
Attachments
Thanks Richie - thats helpful; I can easily understand the effect of multiples of option 1 but I'm less sure of what happens when the amps start sharing things.
KatieandDad, its just 3x 100w mono amps per transformer - I'll be building two completely separate ones for stereo, as you suggest - though admittedly with a slightly smaller transformer than you'd like. I know what you mean about that, it does seem something many people favour to have at least twice the VA of the amp's power and I'm coming in a little beneath that. Though in practice I won't be using all 3x100W (the tweeter in particular will be much less), so I was hoping this would account for the slight short-fall - guess I should try to calculate energy lost by the the mosfets too though. I'd imagine/hope that with it being music rather than a continuous sine wave theres little chance of the transformer actually overheating, so presumably its the stiffness of the supply that I'd need to worry about.
Cheers
Kev
KatieandDad, its just 3x 100w mono amps per transformer - I'll be building two completely separate ones for stereo, as you suggest - though admittedly with a slightly smaller transformer than you'd like. I know what you mean about that, it does seem something many people favour to have at least twice the VA of the amp's power and I'm coming in a little beneath that. Though in practice I won't be using all 3x100W (the tweeter in particular will be much less), so I was hoping this would account for the slight short-fall - guess I should try to calculate energy lost by the the mosfets too though. I'd imagine/hope that with it being music rather than a continuous sine wave theres little chance of the transformer actually overheating, so presumably its the stiffness of the supply that I'd need to worry about.
Cheers
Kev
There is a lot to be said about lots of small guys will beat a bigger guy. That's also true for caps. You're going to need at least 20000uF per amp, better to use 20 x 1000uF than 1 x 20000uF. Even then bypass each bank with 100uF and then a 1uF Poly. Going for CRC or CLC will improve matters ever further.
Thats reassuring, thanks both. I've got 20,000uf per amp (well okay, 18,800uf) in my first/preferred option and am using a CRC configuration in it (CLC may be a bit advanced for me at the mo) so not too far off the mark then. I'm also going to be using some decent decoupling capacitors on/at the amps themselves, so hopefully that will be about on the money.
I think thats pretty much what I'm aiming for, really; I don't have the budget (or skill) to be making ultra-high end designs, but do I hope for something thats much better specified than any shop-bought amp I'm likely to buy at the same sort of price.
Cheers
Kev
I think thats pretty much what I'm aiming for, really; I don't have the budget (or skill) to be making ultra-high end designs, but do I hope for something thats much better specified than any shop-bought amp I'm likely to buy at the same sort of price.
Cheers
Kev
Last edited:
There's always lots of options to improve power supplies, but at least both suggested
one will work. I'd prefer option 1 too, and consider to add some more capacitance to
the bass amp.
When in practice the bass will draw max. 100 W, mid will be more like 50 W or less and
treble would be 25 W or so, that's less than 200 W pre channel. So I believe a 500 VA
transformer will be sufficient.
one will work. I'd prefer option 1 too, and consider to add some more capacitance to
the bass amp.
When in practice the bass will draw max. 100 W, mid will be more like 50 W or less and
treble would be 25 W or so, that's less than 200 W pre channel. So I believe a 500 VA
transformer will be sufficient.
Yes, thanks for the caution. The power amps will be built using the ESP project101 PCBs - I don't trust myself to design my own power amp PCBs yet so these seem like a good compromise between pre-built and full DIY at the mo, though I may customise them a little.
These are designed for two pairs of lateral mosfets with an absolute max rail voltage of +/-70Vdc, but can be built with just one pair for lower powers. With +/-56Vdc I'm at the max of what is recommended for a single pair, but I will still build them with two pairs anyway because my amps will be on for many hours a day and I want to reduce long-term stress on the fets.
(EDIT: the mosfets recommended are the Exicon ECX10N20/ECX10P20 or the Hitachi 2SK1058/2SJ162, which seem somewhat higher powered than is typical)
Cheers
Kev
These are designed for two pairs of lateral mosfets with an absolute max rail voltage of +/-70Vdc, but can be built with just one pair for lower powers. With +/-56Vdc I'm at the max of what is recommended for a single pair, but I will still build them with two pairs anyway because my amps will be on for many hours a day and I want to reduce long-term stress on the fets.
(EDIT: the mosfets recommended are the Exicon ECX10N20/ECX10P20 or the Hitachi 2SK1058/2SJ162, which seem somewhat higher powered than is typical)
Cheers
Kev
Last edited:
I agree with 500VA. With typical drivers, efficiency around 86 to 90 dB/W, you will not get even close to using 100W from any of the amps, unless your listening room is as big as a barn. I use four LM3886TF amps in an active three way (two bridged for woofer) and that is enough to blow the walls out, using, wait for it, a single 160VA transformer.
Many people vastly over estimate the transformer requirements, without giving any real thought to what the actual needs are. It is one of those audiophool "rules-of-thumb" that can be treated with a bag of salt.
Abs
Cheers for that, the toroids were reasonably expensive so I'm very glad you don't feel I under-specified them. I had believed that the difference between continual RMS and music material coupled with the lower power of (especially) the tweeter would make up for it, so I'm really glad I wasn't completely wrong. Thanks for the reassurance!
Thanks,
kev
Thanks,
kev
Last edited:
Why does a the treble amplifier need as much smoothing capacitance as the bass amplifier?
My usual way of designing the "system" is to determine the passband of the amplifier.
Determine the amplifier operating frequency range such that these all encompass the range of frequencies getting past the Passive input filters.
Determine the smoothing capacitance filter to suit the operating frequencies of the amplifier.
eg.
For a wide band amplifier, set the passive input filters to 2Hz to 200kHz.
check that the amplifier will operate at least just beyond that range.
Set the smoothing capacitance filter frequency to <1Hz, i.e. more than one octave below the pass band filter.
If you apply that principle to an active amplifier set, then the mid band and the treble band amplifiers do not need 1Hz capability from their PSUs. ~10Hz for mid band and ~100Hz for treble band may be closer to the requirement.
If you were to implement this way, then the bass smoothing could be +-40mF, the mid smoothing could be +-4mF and the treble smoothing would be +-400uF.
At these much smaller values, they can be placed very close to the amplifiers and higher quality components could be selected. Say using ultra low ESR and ultra low ESL and ultra low connection lengths (much like MF decoupling would require)
So an option 2 would have main smoothing say +-20mF, feeding into three paralleled sets of +-20mF//+-4mF//+-400uF, then feeding the three amplifiers.
To get this split supply to work with minimum grounding loops you must adopt dual rectifier fed from dual secondary transformer.
I can't see how a centre tapped transformer can avoid grounding loops, since by definition/design the centre tap MUST be linked to the Main Audio Ground of all three channels.
My usual way of designing the "system" is to determine the passband of the amplifier.
Determine the amplifier operating frequency range such that these all encompass the range of frequencies getting past the Passive input filters.
Determine the smoothing capacitance filter to suit the operating frequencies of the amplifier.
eg.
For a wide band amplifier, set the passive input filters to 2Hz to 200kHz.
check that the amplifier will operate at least just beyond that range.
Set the smoothing capacitance filter frequency to <1Hz, i.e. more than one octave below the pass band filter.
If you apply that principle to an active amplifier set, then the mid band and the treble band amplifiers do not need 1Hz capability from their PSUs. ~10Hz for mid band and ~100Hz for treble band may be closer to the requirement.
If you were to implement this way, then the bass smoothing could be +-40mF, the mid smoothing could be +-4mF and the treble smoothing would be +-400uF.
At these much smaller values, they can be placed very close to the amplifiers and higher quality components could be selected. Say using ultra low ESR and ultra low ESL and ultra low connection lengths (much like MF decoupling would require)
So an option 2 would have main smoothing say +-20mF, feeding into three paralleled sets of +-20mF//+-4mF//+-400uF, then feeding the three amplifiers.
To get this split supply to work with minimum grounding loops you must adopt dual rectifier fed from dual secondary transformer.
I can't see how a centre tapped transformer can avoid grounding loops, since by definition/design the centre tap MUST be linked to the Main Audio Ground of all three channels.
2 pairs solves the psu prob, but now you need twice the bias current, at 120ma or so a pair, thus each amp will idle at around 30W, so a lot of sinking is (or other cooling method) is needed for all six....
Hi,
Its highly dubious and very unlikely a 3 way active speaker needs
3 identical power amplifiers and identical supplies to those amplifiers.
Probably only the bass section needs double devices (and half
the impedance of mid and treble) for nominally double power.
Some might use cascaded C for bass RC for mid and a further
RC for treble, including series diodes to prevent reverse current.
rgds, sreten.
Its highly dubious and very unlikely a 3 way active speaker needs
3 identical power amplifiers and identical supplies to those amplifiers.
Probably only the bass section needs double devices (and half
the impedance of mid and treble) for nominally double power.
Some might use cascaded C for bass RC for mid and a further
RC for treble, including series diodes to prevent reverse current.
rgds, sreten.
The treble and mid amplifiers could have very similar currents demands as the bass amplifier.
The difference between them is the duration of the demand.
Tens to hundreds of milliseconds for the bass amplifier, but tens to hundreds of microseconds for the treble amplifier. That is why the capacitance can be so much smaller.
I will continue to contend that the ONLY reason we see and get away with small amplifiers on treble drivers is because the sensitivity of treble drivers can be much higher than typically found in low frequency bass drivers.
The difference between them is the duration of the demand.
Tens to hundreds of milliseconds for the bass amplifier, but tens to hundreds of microseconds for the treble amplifier. That is why the capacitance can be so much smaller.
I will continue to contend that the ONLY reason we see and get away with small amplifiers on treble drivers is because the sensitivity of treble drivers can be much higher than typically found in low frequency bass drivers.
Thanks chaps, thats enormously helpful!
I'd liked the idea of the amps all being similarly/fully specified modules, just for future flexibility. But I'll admit that I hadn't reasoned through the extra cost of that and/or the compromises for the actual build I'm intending them for.... On reflection that would have been a poor design choice; thanks for pointing it out!
It makes a lot of sense to re-assign the larger capacitances to beef up the bass amp and get some decent quality smaller ones for the treble instead. I'll admit that my budget doesn't run to 'special' capacitors for all this (so I just went for decent ripple, esr and lifetime), but maybe it could stretch to some good smaller ones just for the treble amps. I decided not to go with full class A amps for the treble, so perhaps its the least I should do.
I'd also not considered just how much more idle power would result from doubling up the mosfets. I guess for each fet the bias current is no different and it would reduce stress on them when drawing higher powers, But I suppose doubling the bias current would make the mosfets run warmer on any given heatsink, and its not ideal for the power supply or my bills to be wasting this much energy. In theory (and assuming my mosfets are genuine ones) then I'm within the recommended spec to just use the one pair even if it were for full-range, so yes I can see that I'd be going over the top to use two pairs - for the mid and treble in particular. I understand that I shouldn't skimp on headroom/current-capability for the treble amp, but if I'm understanding correctly the 'average' power needed is quite low so the mosfets shouldn't be stressed very much at all.
Cheers
kev
I'd liked the idea of the amps all being similarly/fully specified modules, just for future flexibility. But I'll admit that I hadn't reasoned through the extra cost of that and/or the compromises for the actual build I'm intending them for.... On reflection that would have been a poor design choice; thanks for pointing it out!
It makes a lot of sense to re-assign the larger capacitances to beef up the bass amp and get some decent quality smaller ones for the treble instead. I'll admit that my budget doesn't run to 'special' capacitors for all this (so I just went for decent ripple, esr and lifetime), but maybe it could stretch to some good smaller ones just for the treble amps. I decided not to go with full class A amps for the treble, so perhaps its the least I should do.
I'd also not considered just how much more idle power would result from doubling up the mosfets. I guess for each fet the bias current is no different and it would reduce stress on them when drawing higher powers, But I suppose doubling the bias current would make the mosfets run warmer on any given heatsink, and its not ideal for the power supply or my bills to be wasting this much energy. In theory (and assuming my mosfets are genuine ones) then I'm within the recommended spec to just use the one pair even if it were for full-range, so yes I can see that I'd be going over the top to use two pairs - for the mid and treble in particular. I understand that I shouldn't skimp on headroom/current-capability for the treble amp, but if I'm understanding correctly the 'average' power needed is quite low so the mosfets shouldn't be stressed very much at all.
Cheers
kev
Last edited:
You have read correctly. The average power fed at treble frequencies is very much less than the mid band.
But the peak voltages on the very short transients are just as high.
And it's on the fast changing transients that speakers demand the extra current. The bass drivers probably demand just 1.5times the current of the nominal bass driver impedance.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.