discrete, portable "toy" amp

Hello everyone.  I am new here, but have learned a lot by reading posts for awhile.  I have been playing with the Meier cross-feeds, so I need an amp between it and my Grado SR-60s.  I have been having a lot of fun on the breadboard with the following amp:

I haven't been able to find anything else like it in this forum so I thought I'd post it.  I think it sounds quite good, although I realize that some folks don't like output caps.  Two channels draw about 13 mA with 9V source, so it is easily portable.  I first built with 1N914 diodes instead of T7 and T8; the transistors simulate lower THD for low output voltage (I listen quiet!) but I cannot hear the difference.  The transistors are all general purpose, chosen solely due to their inclusion in the Radio Shack bags-o-transistors.  In any case I learned a lot by playing with this, and guess that some others may like to play with it, even if they don't solder it up.

I saw the basic design in a book by Soclof - changes I made are basically trivial.  It is a basic differential pair with an active load to increase gain, coupled directly to a class AB output.  No level shifting; no other gain stages, nothing fancy.  Gain is 1.5, which is more than enough for me.  I doubt this could ever be made to work with DC coupling.

I was wondering if anyone had any suggestions for modifications that would improve it without turning this into something of much larger scope (and current drain).  I realize it is a simply "toy" compared to the other discrete non-portable amps folks around here can design, but it is fun.

Thanks,

jason

Gain can be adjusted, and I have played with gain.  In the schematic, the gain is the usual 1 + R2/R1, but R2 and R1 also set the DC level of the output (but before Cout, of course), which we want near the middle of the voltage rails.  When I increased gain I added a resistor and yet another capacitor.  Here is an example with a gain of about 2.5:

The Gain is basically 1 + R2 / (R1 || R6).  C1 is chosen so that it isn't the limiting factor in the low frequency rolloff.  

edit: was playing with TINA (TI's free spice program) and find that increasing the gain does not increase the available output voltage before distortion creeps up.  For 32 Ohm loads the amp should be able to output 1.9 volts peak (1.3 rms). 

I have found that when I change R3 to 1k the amp seems to sound the same, and the current drawn of course falls by a factor of 2.  So 6.5 mA for two channels, which I really (!) like for portable applications.  According to TINA this reduces the maximum output to about 1V rms, which of course is way more than I ever use, and in theory is 30 mW into 32 ohms.  By the way, I use TINAs THD prediction to estimate the max voltage - I assume the max THD I want is 0.1%.  I do not know the "as-built" THD - presumably is higher than the prediction. 

By the way, I didn't do a stellar job matching transistors.  The bags of npn transistors Radio Shack sells have 5 each of 2N2222, 2N3904, 2N4401.  Likewise the pnp bag has 5 each of 2N2907, 2N3906, 2N4403.  So I had 5 2N4401s to make two "matched" pairs.  I am not an expert on electronics (not by any stretch!), but I just biased them forward active, measured the terminal voltages, and made two "best" pairs.  Perhaps that is part of why the DC offset at the output is the order of 10% off from what I expect ....  Likewise, I attempted to "match" the 2N4403's for the active loads, again making two pairs from 5.  I made no attempt to match anything in the output stages.

I may solder this up with a crossfeed attached.  Eventually I'd like to learn enough to do a pencil and paper analysis of this to predict open loop gain, etc.   I know, it would take a lot of hard work on my part, but a nice goal I think!

jason