Want to build a SS PA

[R.G.]

I just bought two of them for LM3886 power amps. It's certainly A right one, if not THE right one.

[lowell]

Cool! Thanks. What supply circuit are you using for them?

[lowell]

Ok so that "120v to 42v" Transformer has a 21-0-21 secondary. That will give 21*1.4=29.4v when rectified correct? Or am I off here? Finally will the bipolar supply then be half of that, or apprx +-15v?

[R.G.]

When you put the two secondaries in series as 21-0-21, and full wave rectify, both sides are full wave rectified. You get about +/-30V, give or take a diode drop and ripple. It's a good compromise between the voltage for most power into 8 ohms and most into 4 ohms, and won't overvolt the chip when the AC line takes a spike up to 135Vac.

What you do is to get one of the 25A/200V or more integrated diode bridges, and run the ends of the secondary into the AC sides of the bridge. You connect the CT to the junction of your filter caps, the place where the + end of one is connected to the - end of the other. The + lug on the bridge goes to the free + end of one cap, the - lug on the other goes to the free - end of the other cap. The cap terminals then become +31V and -31Vdc. The connection of the two filter caps and the secondary center tap becomes the ground point in the middle of the bipolar supply.

[lowell]

Ok so there's no connection to "true" ground regarding the bridge rectifier and filter caps then right? Also, the B+ can be left at +-31vdc right? It will still be well within maximum (94v) supply ratings even with PT primary spikes. This will get me about 43watts right? Or somewhere in the low 40's. Finally, how does one go about choosing filter cap values in uf? Obviously the caps have to have a voltage rating of 50v or more.

[R.G.]

Ok so there's no connection to "true" ground regarding the bridge rectifier and filter caps then right?

If you mean AC power line safety ground, the correct answer is "not necessarily".

A transformer secondary floats independent of any other voltages. In fact, that's a major part of what transformers are used for. However, in building the amp, you will connect the incoming AC power line safety ground to the metallic chassis to protect you and others from the chassis becoming live with AC power if there is a flaw in the wiring. You will also connect the metal chassis to the signal ground to get the RF and hum shielding you can get from that.

So yet, there is a connection to AC power safety ground with the CT and filter caps, but it's done for other reasons. And it has no bad effect on the power supply because the transformer secondary truly floats free of any ground, so any point on the secondary can be connected to AC ground as long as it is the ONLY secondary point connected to AC ground.

Also, the B+ can be left at +-31vdc right?

Yes. That's the whole point - get a transformer that gives you the max voltage that's practical and usable for the power amp.

It will still be well within maximum (94v) supply ratings even with PT primary spikes.

And it is low enough to not make the LM3886 get too hot and have its internal thermal protection mute it erratically.

This will get me about 43watts right? Or somewhere in the low 40's.

I'd have to look at the chip specs again, but I think more like 50W for 8 ohms and maybe 60 for 4 ohm loads. It's just about the max the chip can do continuously.

Finally, how does one go about choosing filter cap values in uf? Obviously the caps have to have a voltage rating of 50v or more.

Power Supplies Basics at geofex has a lot of what you want.

The mathematical solution is that C = I*dt/dV, where C is the capacitance, I is the load current, dt is the time between AC power line peaks, and dV is the amount of ripple you want to let happen. You get to pick how big the ripple is just by picking the cap size.

As a practical matter, go here: for 18,000uF 50V caps for $2.56 each. I ... um... bought a bunch of these for just such a use. They also have these rectifier bridges which are major overkill, but handy to work with and will never burn out in this application, as well as being cheap.

[lowell]

RG thanks for that I'll be sure to check those links out. Where in the lm3886 app notes do I find the load current value for the filter caps?

[R.G.]

RG thanks for that I'll be sure to check those links out. Where in the lm3886 app notes do I find the load current value for the filter caps?

Oh, sure, call me on something I waffled and handwaved on...

It's difficult and not very clear even when you have the numbers. If you want to dig through it, I can help with that, but as a practical matter, it's not a clear-cut calculation. 4700uF to 10,000uF is a good value for amps of this size.

The reason it's not clear cut is the same reason that transformer sizing is not clear cut - the crest factor in audio. Audio signals have a peak-power to average-power ratio of about 20:1. That is, they spend very little time actually putting out their maximum power. It's only at that maximum power that the caps and transformers are working full-tilt. Federal laws on advertising require the advertisers to test their amps at full continuous sine wave power at some low distortion percent, and that's a very severe test of the power supply.

The power out of either the positive or negative side of the bipolar supply is (conceptually, for a pure sine of continuous full power) a series of half-wave pulses corresponding to the current pulled on the positive or negative half waves of the signal out; this is after all, a class B amplifier, and the positive outputs all come from the positive power supply, the negative half cycles all come from the negative half of the power supply.

This thing can only supply maybe 28V peak into 8 ohms, so the peak current at signal frequency is 28/8 = 3.5A. The average current of that half-wave sine signal is 3.5/3.14 = 1.11A.

So the capacitor you need is C = I*dt/dv, or C =1.11A*0.017S/dV = 0.0185/dV.

The way to read that is that the capacitor required to suppress ripple to any particular voltage is 18.5millifarads per volt of ripple; an 18,500uF cap will give 1V of ripple at full load continuously. If you want to make the ripple be only 0.1V, you need a capacitor of 185,000uF.

A 4700uF cap will give ripple of dV= 0.0185/4.7mF = 3.9V, which is a bit over 10% ripple. A 10,000 uF cap will give ripple of dV = 0.0185/0.010 = 1.85V.

Those look bad at first glance, but as a practical matter, musical peaks are not continous, and 5% to 10% ripple doesn't sound all that bad in a musical amplifier; and the cost to suppress it further with capacitors can get quite large. It's this "how bad does that sound?" and "What's the actual crest factor of the music you're playing RIGHT NOW?" that make this question foggy.

I'm surprised at the answer coming out at 18,500uF in the example for the amp; I did not fudge the numbers to make that 18kuF cap look good. It just happened.

So that's how you do the numbers: you calculate to three decimal places, then guess what's good enough since your numbers show it can't be perfect. It's how engineering is done.

[lowell]

If you mean AC power line safety ground, the correct answer is "not necessarily".

A transformer secondary floats independent of any other voltages. In fact, that's a major part of what transformers are used for. However, in building the amp, you will connect the incoming AC power line safety ground to the metallic chassis to protect you and others from the chassis becoming live with AC power if there is a flaw in the wiring. You will also connect the metal chassis to the signal ground to get the RF and hum shielding you can get from that.

Ok so I've got all my stuff in line here and I'm ready to assemble my prototype hybrid amp! Very excited... I have that all-to0-familiar "can't sleep" excitment. We audio geeks can be pretty obsessive. But it's so much damned fun!! Anywho RG I'm rereading this thread again and I'm still unclear... I appologize.

Is the +PGND/-PGND grounded to the chassis or not? I know it IS connected to the LM3886 ground, but are these DC connected to the chassis? Thanks for dealing with my questions!

Also, how does one go about rating a bridge rectifier package as far as V and I are concerned? I get diodes and their ratings, just not the packages like the MUR860.

And I'm thinking if I run this amp with 42v rails and get 60watts of power, I'll need to rate my mains fuse for the toroid at: 60w/120v=500ma fast blo mains fuse. Correct?

Finally, is it necessary to use an IEC receptacle that has an EMI filter?

[Steve Conner]

Hey lowell,

That power supply is more complicated than it needs to be:


Now there is only one ground terminal, so no worries about whether the two ground terminals should be connected together. (In case you're wondering the answer was yes)

Generally all the grounds should be connected together, then connected to the chassis in one place: Douglas Self recommends at the input jack. The rules for grounding are the same as for tube amps, and if you get it wrong, it'll hum. (Be careful with that connection from the PT secondary center tap to the junction of the two filter capacitors.)

Bridge rectifiers: Even a 600V, 25 amp one only costs a few bucks, comparable to the cost of shipping. So I usually just get a bigger one than needed. Bolt them to the chassis for heatsinking.

IEC receptacle with EMI filter: Not necessary, but a nice touch.

RG: I got my 5 legged transistors working, and the thermal stability seems pretty good.

[kg]

Yeah, think about it dude. There's no winding on a Champ PT that would supply a SS power amp.

am i allowed to use an output transformer?

[lowell]

Chia Hu Co Ltd at Apex Jr. Are you suggesting that I turn the 2 non-center-tapped secondaries as 1 center-tapped secondary? So connect the lower end of the top primary winding with the top of the lower winding, thus making a center-tap? I'm a bit unclear here.

I would think that my tube preamp power supply should be grounded at the input jack. Then ground the ss PA power supply to the 3rd prong grounding point, and more near the speaker jack. ??

[lowell]

Chia Yu Co Ltd at Apex Jr is a dual secondary. Are you suggesting that I connect the 2 non-center-tapped secondaries as 1 center-tapped secondary? So connect the lower end of the top primary winding with the top of the lower winding, thus making a center-tap? I'm a bit unclear here.

I would think that my tube preamp power supply should be grounded at the input jack. Then ground the ss PA power supply to the 3rd prong grounding point, and more near the speaker jack. ??

[Steve Conner]

Yes, if you have a two winding transformer, that's how to hook it up.

If you have a preamp and a power amp, you have to ground them in the same place or you can get hum. You may need to experiment with the grounding a little.

[lowell]

Great thanks for the help Steve. I'll be able to post some audio of the amp when its done.