Designing an Output stage for an LM402

[dfdye]

I have been “playing around” with chip amps for a little bit now, and have recently gotten a couple of LM4702 voltage gain stages working, but I haven’t really gotten an output stage that I really like to go with this chip. For those of you who aren’t familiar with the LM4702, it is from the same National Semiconductor family as the infamous LM3875 of gainclone notoriety, but this chip is a high voltage (+/-85V), low current (~5mA max) opamp that acts as voltage gain stage for your favorite output stage.

The datasheet from National suggest a Darlington output transistor, but I am sure there has to be a better output stage for this chip. I have been trying to put together one, but but since I have never designed one from scratch before, any suggestions that some more experienced builders have to offer would be most appreciated.

Here are the design limitations: typical output current is 5mA, +/-65V rails available (which would be a little higher than what I would use, but let’s shoot for the stars! The chip can handle it!), able to drive 4 ohm speakers @ 100W comfortably. That power would take (by my back of the envelope math) 5A output, making the overall minimum required Hfe of the output stage 1000. Getting to that point is fairly straightforward with a traditional two stage emitter-follower for each rail, but I know there are snazzier ways to do it than that.

The problem I run into is that I am not really experienced in building/using multiple transistor stages. The basic idea of what I want to use is shown below, but I know I need to add in some biasing somewhere. For low powered, single transistor stuff, I know I would run a resistor between the collector and base and another between the emitter and base. The problem here is that I would then only bias T1 and T4 like this, and I would think I would still get switching issues from T2, 3, 5, and 6. If I bias these output transistors too, I would think the current required for biasing these would make the power requirements for the amp absurd!

Could anyone offer some insight? Where/how would you bias this circuit, or does anyone have an alternate suggestion?

Hopefully this will at least get the ball rolling on my education!

Thanks in advance!!
David

[jaycee]

You would connect the bias circuit between Source and Sink (the bases of the driver transistors T1 and T4 on your diagram) in the same way as National show for the compound darlingtons. You will need to adjust resistor values though.

You need a Vbe multiplier in pretty much any transistor output stage. You will see the Qmult, Rb1 and Rb2 arrangement on most amps.

For an output stage, my personal preference is the Triple Darlington. I think the main reason National recommend a compound Darlington is that they are targeting compact designs with a low part count, which would use transistors such as Sanken's SAP15 range.

[dfdye]

Originally posted by jaycee
I think the main reason National recommend a compound Darlington is that they are targeting compact designs with a low part count, which would use transistors such as Sanken's SAP15 range.

I completely agree that is what National was going after, but I am sure we can do better.

As for the output stage in the National circuit, I agree the Qmult will compensate for thermal issues, but it won't bias away switching voltages required to turn the transistors on and off, right? (I may be missing something here, but won't there be a goofy zone for +/-0.6V?) Looking back at schematics for some classic Class A (OK, just the Krell amps) amps, I tried to figure out what the tricks were to bias output transistors and sound good. I haven't found a great answer so far.

[famous last words] HOWEVER, I think that other thermal management techniques can make Qmult unnecessiary. As long as the temperature of the transistors is kept reasonable, and as long as you have reasonable values for the output resistors, I think Qmult is redundant. [/famous last words]

[djk]

Do something tried and true:

The tripple is advised as the LM chip doesn't have much output current.

3.6V bias is needed. Complete diagram:

http://users.ece.gatech.edu/~mleach/...aphics/ckt.pdf

"As long as the temperature of the transistors is kept reasonable, and as long as you have reasonable values for the output resistors, I think Qmult is redundant"

Buy an exra set of outputs and drivers, you'll need them. When you re-build the amp add the Vbe multiplier.

Another well designed ±65V amp with an IC front end (output and bias stage on P6, P7 is front end):

http://www.crestaudio.com/media/sche...ematic_set.pdf

[dfdye]

Originally posted by djk
The tripple is advised as the LM chip doesn't have much output current.

Thanks for the suggestion. I was looking for something along these lines. As a follow up question, I was under the impression that triple outputs had higher noise levels than dual stage outputs. Granted it may be necessary, but have you seen this to be true?

To extend the discussion of triple output stages, there is another simple triple stage that I was looking at (see attached, from Douglas Self's "Audio Power Amplifier Design Handbook"). The draw of this for me was that it looks quite similar to a complementary feedback pair that I like the idea of a lot.

Does anyone have an idea as to which would be better? Also, could you parallel the final transistors of the Self output stage to get a little more current?

Buy an exra set of outputs and drivers, you'll need them. When you re-build the amp add the Vbe multiplier.

Fair warning taken. Assuming that the initial transistors don't get too hot (in this case Q14 and 15) how would runaway occur? I’ll defer to those who have seen it before, but it doesn’t make sense to me right now. . .

Another well designed ±65V amp with an IC front end (output and bias stage on P6, P7 is front end):

http://www.crestaudio.com/media/sche...ematic_set.pdf

Thanks! I am sure it will take me a while to wrap my head around that one, but it looks like it will be chock full if information.

David

[djk]

Q14, 15 are the predrivers, they run class A and don't have a problem with thermal runaway. The Q16, 17 drivers are usually biased quite heavy (about 1/2 the required current for the output stage) and are not a a big problem either. The main problem is the Vbe of the outputs over temperature.

The CFP varients are hard to stabilize. Plagued with oscillations and shoot-though coming out of clipping, schematics as shown are best regarded as 'conceptual', and a working design will need a Baker clamp, a few more resistors, and maybe a cap or two.

If you have the time, energy, and resources, go for the CFP. If not, a tripple EF is tried and true. Tack a cap around 2µF across R36, and use two pair of TO-3 or three pair of plastic outputs and you should be fine.

[dfdye]

Originally posted by djk
If you have the time, energy, and resources, go for the CFP.

The first two are always problems, but the third is fortunately not. I have a great stash of transistors so I can afford to blow a few up trying to get it right. So I guess I will have to focus on getting the CFP up and working.

[djk]

Go to:

http://home.kimo.com.tw/skychutw/

and click on Ampzilla, then Ampzilla III. The type of CFP tripple is used by many high end amp companies. You would only need half of the bridge of course, and the bias circuit too