LM4780 transformer question

[AndrewT]

the 4780 running @ 45W per side will deliver 90W into 8r0.

A single 3886 will deliver 68W into 8r0.

Your bridged amps have gained you ~1.2dB in extra output power. Is all that complication worth an extra 1.2dB?

[gbcomp]

No, it is not. I started thinking about doing the 4780 because I wanted more power but all of this back and forth is really making me change my mind. I have done a lot more research on the 3886 chip before too so I am much more confident working with it. As this is just a summer project for me between semesters I should probably keep it simple and leave the 4780 for version two. Thanks for all the input everyone. It helped me out a lot!

[ChristianThomas]

I'm sorry to say but I disagree with much of the advice Andrew is giving here. It seems he would have you believe that you can get more out of one chip than you can out of two in parallel. This is rubbish. The 4780 is just two 3886 dies on the device. It's worth using because a bit more care went into the number of bonds to the legs, especially on the grounds. For a bridged amp it is invaluable because you have near perfect temperature tracking, which is quite important.

If you use them in parallel you would use a higher voltage than you would singly. And if you use them bridged, which is perfectly valid, you would use a lower voltage than that single value.

Especially paralleled, rail voltages of 35V to 39V would be fine. You need to keep these up if you are going to get the power into 8 ohms. But the first thing I would do is check to see what the real impedance of your speakers is. Most nominal 8ohms are actually somewhere around 6 unless they are either v. traditional or quite old. That means you won't have to go quite so high in rail voltage, but there is no harm in having it.

There is slightly less good dissipation on the 4780, simply because it doesn't have twice the area of the 3886. Dissipation in a paralleled arrangement is not going to be a problem because you have half the current through each and only a slightly higher Vce than you would with a single. And if you stick to 35V, no higher at all, which means you are dissipating the same amount through a larger area.

I really wouldn't make too much of a meal of this, and there is no reason that anyone should be alarming you. National do an Excel spreadsheet which is quite useful for getting to know what your choices are, but it is a long way from being gospel.

If you doubt my word then have a look at Mark Hennessey's site. He has done a number of gainclones, one of which I think is switchable between single, parallel and bridged. I don't even remeber him using different voltages but I may be wrong. The link is Mark's Project Pages. He does know what he is doing.

With 8 ohm speakers, if that's what they really are, I have to say I would be very tempted to go for a bridged arrangement. Siegfried Linkwitz, I think, has done one using 4780s. It really will give you substantially more headroom than you'll get in any of the other arrangements, and that should pay dividends in clean dynamics. I would probably go for around 30V if I was doing it, but have a look at what Siegfried and Mark have chosen.

Finally, choose a heatsink which is thick near the chip, not one of these feeble things. Thermal mass can be as important as the dissipation.

Hope this isn't too late for you.

CT

[AndrewT]

Originally Posted by AndrewT

If you want to drive 8ohms speakers, then design each half of the 4780 as a 4ohm capable amplifier.
Take care with overheating.
The 4ohm capability requires more current from the chipamp and that requires more cooling.
BUT !!!!
the dual amp chip has very limited heat dissipation through to the heatsink. It is an extended To220 after all.

You might be better aiming for 40W to 50W into 4r0, rather than 60W into 4r0.

Originally Posted by AndrewT

the 4780 running @ 45W per side will deliver 90W into 8r0.

A single 3886 will deliver 68W into 8r0.

Your bridged amps have gained you ~1.2dB in extra output power. Is all that complication worth an extra 1.2dB?

Originally Posted by ChristianThomas

.........I disagree with much of the advice Andrew is giving here. It seems he would have you believe that you can get more out of one chip than you can out of two in parallel. This is rubbish.

Which phrase says that?
Which bit is rubbish?

My post4 did not offer any advice. It simply asks.

[pacificblue]

Originally Posted by ChristianThomas

National do an Excel spreadsheet which is quite useful for getting to know what your choices are, but it is a long way from being gospel.

An absolutely long way from being a gospel, because the recommended heatsink size is quite far off. That Excel sheet has three issues.

One is that they forgot temperature derating in the spreadsheet.

Two is that they did not take their own AN-898 into account which shows that from 125 °C junction upward the built-in protection will kick in, but the spreadsheet calculates the heatsink for 150 °C junction.

Three and the least of them is that the heat dissipation due to the idle current is not calculated correctly.

All three mean that the calculated heatsink is much smaller than what you really need.

Originally Posted by ChristianThomas

With 8 ohm speakers, if that's what they really are, I have to say I would be very tempted to go for a bridged arrangement.

In a BTL application you get even more trouble with heatsinking, because the dissipation is four times that of a SE amp into the same load. For that reason it is better to use two LM3886 instead of one LM4780 for a BTL amp. Just like Linkwitz did by the way.

Originally Posted by ChristianThomas

I would probably go for around 30V

30 V BTL into an 8 Ohm speaker correspond to more than 90 W of heat dissipation. You will need forced air cooling aka a fan, and in summer even that may not suffice.