Transistor choice help

[Alex1478]

Hi,

First, I want to say that i'm learning english, so I'll try to explain as well as I can.

I'm an university student and I'm designing my first amp.
I just finished to design my amp but I forgot to give importance on a specific thing. I used a TO-92 transistor for my voltage amplifier stage. Now, my amp is working well (I think) on LTSpice, but those transistors consumes 800mW... That too much ! I want to know if someone know a good transistor that I can use instead of 2N5550(Q18) and 2N5401(Q15) with similars parameters and that can dissipate more power (TO-220 for exemple).

I did many research on internet and, unfortunately I didn't found what I need.
I just attached my amp simulation files on this dicussion thread

Your help is much appreciated !

Thank you.
Alex


EDIT: If you have any comments about my amp, feel free to discuss

[lineup]

TO-126 is smaller and faster than TO-220
and such TO-126 transistors are between TO-92 and TO-220

Most commonly used TO-126 are BD139/BD140 and MJE340/MJE350
These are used to drive bigger powertransistors
Sometimes used without heatsinks, but with small heatsink can take several watts.

800mW can one TO-126 take without any aluminium heatsink
Most TO-92 should not be used more than 300mW, when rated 500-600mW

[tiefbassuebertr]

check out this threads:
Transistor data with sorting
bipolar (bjt) transistor families for audio power output stages
High Voltage BjT's for Pre-driver stages of High Power Amplifier - Overview
ckeck out also the circuit topology from post #39 about
Your opinion about this schematics?
your topology gives no significantly better audible result, only the measurable results are much more better, especially if the power output stage runs between 20 and 100 mA idle current.

[ilimzn]

I think you have bigger underlying problems than just transistor choice.
1) is real life Vcc and Vee constant, and if not, are your front end tail currents constant?
2) What sets the DC current in Q13 and Q14?
3) What is the role of Q4 and Q6?

[Art M]

Originally Posted by Alex1478

I'm an university student and I'm designing my first amp.
I just finished to design my amp but I forgot to give importance on a specific thing. I used a TO-92 transistor for my voltage amplifier stage.


EDIT: If you have any comments about my amp, feel free to discuss

The amplifier looks almost exactly like a Randy Slone design.

I suggest you use the same transistors shown in his book

[infinia]

VAS selection link http://users.tpg.com.au/gerskine/gre...ransistors.htm

[Alex1478]

Originally Posted by ilimzn

I think you have bigger underlying problems than just transistor choice.
1) is real life Vcc and Vee constant, and if not, are your front end tail currents constant?
2) What sets the DC current in Q13 and Q14?
3) What is the role of Q4 and Q6?

1) I attached an image file of my current tail. I can say that it's pretty constant. I took that on the left transistor (emitter) of U9 (matched transistor).

2) & 3) The role of Q4 and Q6 is to put a DC constant current on the transistors (Q15-16-18-19-09). I tried a current mirror topology. I don't know if it a good idea to do like that. In the same way, it fix the constant DC current in Q4-Q6. R26 limits the current that flow in those transistors.

I don't know if I answered correctly to your interrogations. Tell me if my justification is wrong. I'm not very familiar with analog electronic yet.

Originally Posted by Art M

The amplifier looks almost exactly like a Randy Slone design.

I suggest you use the same transistors shown in his book

Yes, my design has been based on this book. But I want to make some variations from myself like using different components parts and use different constants current topologies.

Originally Posted by tiefbassuebertr

check out this threads

Thank you for those informations. I'll look that with a particular attention.

Originally Posted by lineup

TO-126 is smaller and faster than TO-220
and such TO-126 transistors are between TO-92 and TO-220

Most commonly used TO-126 are BD139/BD140 and MJE340/MJE350
These are used to drive bigger powertransistors
Sometimes used without heatsinks, but with small heatsink can take several watts.

Thank. I'll test those transistors

[ilimzn]

Originally Posted by Alex1478

1) I attached an image file of my current tail. I can say that it's pretty constant. I took that on the left transistor (emitter) of U9 (matched transistor).

OK, I forgot 0)
0) If you are just starting with analog, then you have definitely chosen the wrong design to start with. Do a search on this forum about that A. Sloan design and you will find it is inherently flawed. Your problem seems to be that you do not understand the topology. If you cannot see WHY and HOW this works from a topology standpoint, there is little hope to ever make it work in the real world, despite what the simulator shows you, simply because you don't know what you should be looking at in the simulator.
Your answer above is a typical example: I have asked if Vcc and Vee are constant in the real world, and therefore, if tail currents in the input stage are constant. What you did is, conveniently skipped over the first question, and simply looked at the plot out of the simulator. This will of course tell you the currents are constant, because in the simulator you used ideal voltage sources as Vcc and Vee, and they, unlike real world power supplies ARE constant, by definition. In other words, you have completely missed the point of the question, just like your simulator results completely miss the expected real world behavior, because you set it up in non-realistic conditions, failing to model real world ones, simply because you do not understand how it works in the real world.
Specifically, if your tail current is generated by a current mirror, then it will mirror whatever current flows from it into the diode end, and that is being generated by a resistor from the power supply. A real world power supply, unless regulated, will have significant ripple, and so will your tail current.
Just to make things clear, this does not mean you need to regulate the power supply, it means you have incorrectly chosen the topology of the tail current source, because of failure to understand what it does and/or modeling for unrealistic conditions.
On the other hand, this is a minor problem for your design, so nothing terribly dramatic would happen (except maybe slightly degraded performance compared what could have been done using the same or less parts) but this was actually the SIMPLE question out of the 3 i posed, so not getting it should at least give you some reason to do more thinking about this design.

2) & 3) The role of Q4 and Q6 is to put a DC constant current on the transistors (Q15-16-18-19-09). I tried a current mirror topology. I don't know if it a good idea to do like that. In the same way, it fix the constant DC current in Q4-Q6. R26 limits the current that flow in those transistors.
I don't know if I answered correctly to your interrogations. Tell me if my justification is wrong. I'm not very familiar with analog electronic yet.

In fact this is some sort of a cross between a current mirror and a cascode, and the only reason it seems stable is that it puts the transistors into boundary modes of operation, and by that I mean boundary modes of the simulated model of the transistor. Transistor models are by their very nature simplified representations of real world components, and modeling of operating areas close or outside of the expected 'normal' biasing is even more inaccurately modeled. Simulator results can rarely if ever be relied upon in these conditions, and they should be avoided unless special models are used or the limitations of the model are well known in comparison to the actual real-world component.
I could almost bet my life on the real world amplifier behaving completely erratically as it is drawn.

The correct approach to this problem is initially wilting down the schematic to a 'skeleton' amp. Never mind for now that only one output pair will not be enough, or how exactly references for current sources are generated - draw a battery! - or even use ideal current sources. Then study the amp first under expected conditions, and then, under unexpected ones. Carefully inspect all elements to see if voltages and current make sense (using ideal current sources can be particularly problematic but unexpected results also show where problems might happen in the real world). Keep it as simple as you can initially. No cascodes, Wilson mirrors, etc. What you will find in the case of your design is that the VAS currents are not defined at all. Each differential pair will try to put any current up to +-tail into the base of it's corresponding VAS transistor in order to get their joined collectors to a zero potential. This will happen as long as currents in the top and bottom half of the circuit from the VAS to the output are equal - no matter if it's 10mA or 10kA. In the real wold there would be lots of smoke. Alternatively, if you construct the amp as in your schematic, the result will depend solely on the tolerances of the actual parts used, which is NEVER good practice - circuits must behave as much as possible completely opposite, be independent of component tolerances.
There are ways around the problem but it will make this already overly complex amplifier even more complex. If you still intend to keep a similar input topology, I would suggest replacing the current mirrors driving the VAS with a folded cascode VAS. Although this is not equivalent, it makes it possible to simply set up a well defined VAS current.
Alternatively, I could ask you to calculate the components needed to get 10mA current through the VAS in your design. When you manage it, go and change the operating temperature and see what happens. Or, change the type of transistor, perhaps just a different gain group.
Remember - a topology has to make sense at a very basic math level of ohms law, on paper. Dont put it in a simulator unless you have such a topology, i.e. one that you understand. If you don't do that, you will almost certainly end up with something that works ONLY in a simulator (and in fact, may not even work in a different simulator using nominally exactly the same components).
Once you have verified the simple version, which includes introducing all sorts of anomalies, like overly large input that provokes clipping, overly low load (to check what happens to currents and dissipation as well as the ability of the stages to drive each other), AC voltage superposition on power rails to check for power supply ripple rejection, etc. - then you start embellishing, adding actual reference sources, current sources, etc. The data you get from (mis)behavior of the simple version is then great help at choosing the right support circuits.