[sTaTIx]

I currently have unstacked BUF634's running in normal bandwidth mode in a single 9V amp. Here are some questions:

1) I know the procedure for stacking the Elantecs is to place their pins in a corresponding manner and solder them together. Is it the same for the BUF634? What about the 5-pin TO-220 package?

2) Is it possible to stack them without connecting them in high-bandwidth mode? If so, would I just connect the unconnected BW pins from one buffer to the next?

3) Would I benefit more from running them in intermediate high bandwidth mode, or simply stacking two of them per output in parallel with the BW pin disconnected? I'm driving a 32 ohm load.

 

[Voodoochile]

Which amp is running in front of the BUFs?

You would likely do better with a wider bandwidth than not connected at all. If it's a really fast OPA up front, then full-BW (direct connection to V-). If it's one of the slower OPAs, you could throttle the BW somewhere in between with a resistor.

As for stacking, yep, pin-for-pin, including the BW pin.
Whether your in wide BW mode or not.

They do draw quite a bit of juice when in wide BW mode, however.

 

[sTaTIx]

Quote:

Originally posted by Voodoochile You would likely do better with a wider bandwidth than not connected at all.

So non-stacked with BW connected with resistor > stacked normal-bandwidth? I'm not going to do both.

Quote:

If it's a really fast OPA up front, then full-BW (direct connection to V-). If it's one of the slower OPAs, you could throttle the BW somewhere in between with a resistor.

Even if it was a fast OPA, a full-BW connection wouldn't be feasible for a battery operated amp, would it?

Thanks for the help.

 

[tangent]

Stacking BUF634s works just like stacking Elantecs. This doesn't work for all buffers, but it does work for these two. Just solder pin 1 to pin 1, etc. Even the bandwidth pin.

As for the TO-220 package, you could stack that one the same way as the DIP package, but I personally wouldn't. The only purpose of going with the TO-220 package is for its higher current dissipation ability, and if you're going to depend on that you need to add heat sinks, which puts the pins too far apart to do standard stacking. You could connect the pins with PCB traces, of course, but that isn't quite the same thing as "stacking".

If you run a stack of 2 BUF634s per channel in low-quiescent mode, that adds up to 6 mA just for the buffers, a significant amount of current for a battery powered amp. If you put them in wide bandwidth mode, the current draw jumps to 60mA, just for the buffers. Heaven help you have more buffers on the vground and a hungry op-amp as well.

If you're using alkalines, this current level is far too high to be efficient -- alkalines can only handle 30mA or so before their battery life starts to drop faster than linearly with respect to load. In other words, going over 30mA is a good way to kill your batteries quick. Rechargeables can handle this level of current draw without trouble, but this still amounts to an awfully high current level for a portable amp. Do the math to see how often you'd have to change your batteries before deciding if this is worth the sound improvement to you.

The concern of bandwidth when it comes to the op-amp matters when you're using the buffer as a plain voltage follower. In that case, the buffer needs to be significantly faster than the op-amp to be stable. But if you've got the two in a Jung multiloop configuration (like on the META42) you can just raise the inner loop gain of the amp if you have instability when using a fast op-amp with a slower buffer.

 

[Voodoochile]

That was a much more educational response, Tangent!

That's what I was driving at, though, regarding current consumption. The Buf643s are very hungry little chips. Just the amount of heat they dissipate is a good indicator of how much power they are converting to heat compared to the elantecs.

Not really a good choice for a battery amp. It seems that the AD823 with a pair of 2001s following it is about as efficient as it comes for a two-stage amp. Things like the TLE railsplitter instead of a resistor divider help, too.