|
http://www.onsemi.com/pub_link/Collateral/2N3055-D.PDF
0. First of all EC is a dangerous place to learn about Amplifiers, do not trust anything you see or read (including this). None the less there are some gems. Get a text book and use google to validate, cause books have errors also.
1. Hfe on the datasheet can be down to 20 so I set the power devices so I can see how things will act at high power levels.
2. Quiescent current is the total current used to bias the amplifier without a load. Open the loads, notch and signal to see the bias levels in all stages. An amp datasheet should probably list the bias of each stage but they typically don't: https://apexanalog-public.sharepoint.com/Resources/PA01-73U.pdf
3. A Vbe multiplier allows the quiescent current to be trimmed, I have set it at 3.1V. However for Class B I would trim it down to 2.5V to turn off the power stage biasing current. When a transistor is considered off is a little fuzzy, and I would not really care but having worked on these things at an outfit that sold them as Class AB I am going to stick with there notion of biasing which was something like 150uA on a power device was off and 15mA was biased.
4. Most of the quiescent current should be in the power devices, so use the Vbe multiplier to steer the bias current where it will do some good. The bias is used to operate the devices in its linear range, for example the 2N3055 datasheet shows .1A to 10A on its hfe plots this is a clue that operating bellow 100mA will become increasingly non-linear and thus cause switching discontinuity (I may be pushing my luck at 15mA, and do not advise any less). Small signal transistors will also have datasheet that should show a range, e.g. something like 100uA to 100mA.
5. The darlington driver has over 1mA and since it is a medium power device this is likely enough to bias it within the linear range, but check the datasheet to be sure. This works out because the power devices have such poor gain that they end up biasing there own driver. So I will stress again that the point of bias is to get the device into its linear range, anything more is just wasteful.
6. biasing the Vbe multiplier stage with much more than a few mA is not helping, since 3mA is right in the linear range sweet spot. Also the op-amp will have to source/sink the current for that stage.
7. Class B biasing current is defined as off in the power stage, no one makes a commercial class B amp, but lots of outfits make class AB. The power stage I show has 30mA (15mA * 2) of bias which is enough to keep everything linear (BJT are not perfectly linear but are good). This keeps the high gain stages from causing rapid discontinues changes.
8. I added current limit sort of like one by @thebugger, but with some mods. It works like a Vbe multiplier that shunts drive current away from the darlington input which starves it. I like the looks of this method more than what I have seen at Apex Microtechnology (where I worked for a few years). The Vbe multiplier will also temperature track with the power transistors, which is why we don't want a band gap reference in this type of circuit. The op-amp can screw up the current limit so it needs a fairly high output resistance.
THD looks low ... added a notch filter to measure. Search for "THD measurement" in EC. I see less than 3mVp-p on the notch with the dummy at 20kHz... so that's better than 0.0075% THD. But I wonder if these BJT models (that run on a cell phone) are very accurate, I am thinking that is just a fun number.
Should I make a circuit board for this? I have almost no use in it so why do I keep tinkering with it. I don't even listen to music, perhaps because I'm tone def and everything sounds wrong.
|
