Mini12 - My First Amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
In my introduction post I said I wasn't going to post about the Mini12, but sharpi31 made me change my mind, and instead of posting about the amplifier itself, which is boringly simple, I decided to post about the construction of it, which I documented over at my blog: Building the Mini12.

Any feedback would be appreciated. I've also mentioned in the introduction that I've been playing around simulating discrete op-amps, etc. so here's a sneak peek (with a push-pull output stage) of what may be coming in the future (feedback about it is also greatly appreciated).
 
Last edited:
Neat little project, but I suspect high-frequency distortion of this circuit won't be pretty, especially at lower volumes. What I would suggest:

Add a ca. 100 ohm resistor from opamp output to amplifier output. (This keeps output stage gm from dropping off the map entirely at low signal levels.)
Add a ca. 330 ohm resistor between the two emitters of 3904/3906. One of about half value from each emitter to the output would be even better. (This gives them some idle current and output transistor base current suck-out capability. The latter option would also improve minimum gm further.)
Reserve a bit of space for a small ceramic capacitor (think double-digit pF, maybe low triple-digit pF) between opamp output and opamp inverting input. Stability in this design definitely is not ensured as-is.
(If board space is tight, you can always mount resistors vertically.)

I would also look into adding very minimal biasing at least, as the present setup is very taxing on slew-rate. Got some 20 mA red LEDs or ~2 V zeners floating around?
A CFP output instead of a Darlington may also be worth considering.

Of course you could also go the boring route of just using a chip amplifier - these tend to work decently out of the box. Not sure what would fit the criteria of +/-10.8 V (nom) split supply and moderate idle current draw though. LM1876 would be good if you can find one. Else, LM4755? (I know it says single supply, but I don't see why it wouldn't work with split supply with pin 6 tied to input ground. Input ground should still go to star ground somewhere, of course.) People in the Chip Amps section may be able to come up with some more options.

Speaking of chip amps, an NJM4556 (more oomph than the rather wimpy LM833) buffered with just a single EF buffer also comes to mind as a possible option. If that gives you enough space to add some biasing in the form of two small-signal diodes with a parallel electrolytic in parallel and a ~3.3k resistor to V- per channel, all the better. Then you'd need some emitter resistors though.
 
Last edited:
Neat little project, but I suspect high-frequency distortion of this circuit won't be pretty, especially at lower volumes. What I would suggest:

Add a ca. 100 ohm resistor from opamp output to amplifier output. (This keeps output stage gm from dropping off the map entirely at low signal levels.)
Add a ca. 330 ohm resistor between the two emitters of 3904/3906. One of about half value from each emitter to the output would be even better. (This gives them some idle current and output transistor base current suck-out capability. The latter option would also improve minimum gm further.)
Reserve a bit of space for a small ceramic capacitor (think double-digit pF, maybe low triple-digit pF) between opamp output and opamp inverting input. Stability in this design definitely is not ensured as-is.
(If board space is tight, you can always mount resistors vertically.)

I would also look into adding very minimal biasing at least, as the present setup is very taxing on slew-rate. Got some 20 mA red LEDs or ~2 V zeners floating around?
A CFP output instead of a Darlington may also be worth considering.

Thanks very much for the feedback. I've always seen these base to emitter to emitter resistors in datasheets for darlington pairs like the TIP120. So they are there to provide some idle current improving the high frequency response? What's the best way to choose the correct value for each resistor?

About the small cap between the op-amp output and the inverting input, I'll add either a 100pF or 47pF since I have a bunch of those and I can easily solder it on the under side of the board.

Also thanks very much for pointing out the CPF output. I'll try it out in my next amplifier for sure.


Of course you could also go the boring route of just using a chip amplifier - these tend to work decently out of the box. Not sure what would fit the criteria of +/-10.8 V (nom) split supply and moderate idle current draw though. LM1876 would be good if you can find one. Else, LM4755? (I know it says single supply, but I don't see why it wouldn't work with split supply with pin 6 tied to input ground. Input ground should still go to star ground somewhere, of course.) People in the Chip Amps section may be able to come up with some more options.

Speaking of chip amps, an NJM4556 (more oomph than the rather wimpy LM833) buffered with just a single EF buffer also comes to mind as a possible option. If that gives you enough space to add some biasing in the form of two small-signal diodes with a parallel electrolytic in parallel and a ~3.3k resistor to V- per channel, all the better. Then you'd need some emitter resistors though.

Yeah, I could use a chipamp, but it sucks all the fun out. What I'm really interested in the electronics hobby is to learn a lot about analog electronics and how they work (using a chipamp defeats this purpose) and ending with a nice sounding and personalized amplifier. :)
 
Thanks very much for the feedback. I've always seen these base to emitter to emitter resistors in datasheets for darlington pairs like the TIP120. So they are there to provide some idle current improving the high frequency response? What's the best way to choose the correct value for each resistor?
I would assume that the resistor in integrated Darlingtons is sized to provide quick-enough turn-off in switching applications and a decent and manageable current for the first transistor while turned on.

In hi-fi EF2 output stages you typically see values of 100-330 ohms per side, giving about 2..6 mA of driver idle current. In integrated Darlingtons you tend to find values in the kOhms instead, hence part of their bad reputation in audio circles.

Set up as in your amp, the first transistor of the Darlington idles at extremely low current and accordingly is very sloooooow. Ever looked at the datasheet of a typical transistor and noticed how much fT will vary with Ic? Now a 2N3904/3906 combo is a fair bit smaller than typical output drivers and probably will only require a fraction of the current (plus output power of your amp isn't so huge either), but I'd still give them 500 µA .. 1 mA. TIP31/32C are not exactly super fast, so those should get 10 mA or so, with emitter degeneration resistors of about 0.68-1 ohm.

Actually, I bet that one half of your output stage operates even in idle. The LM833 has a pretty substantial input bias current, and the large imbalance in DC resistance (almost 220 kOhms) would create an input DC offset of about 110 mV (nominal LM833 input bias current is 500 nA)... which is then amplified by the gain of 13, which is flat to DC. In other words, expected output DC offset would be almost 1.5 volts! :eek: You haven't checked that, haven't you?
With an 8 ohm load connected, this then proceeds to suck over 150 mA out of your positive supply (note DC resistance of 8 ohm speakers is closer to a bit of 6 ohms). Assuming the speakers are generously sized and their distortion does not degrade appreciably from this yet (a big IF), you'd find sound to be perfectly fine, as you've basically built a low-power single-ended Class A output and shifted things quite a bit away from the dreaded zero crossing. Unfortunately the corresponding output transistor has to dissipate over a watt at this point and would probably be seriously unhappy in the long run.

Maybe you have an On Semi LM833? These exhibit lower specified IIB, 300 nA typ. Then output DC offset would still be under 1 volt, and output transistor Pd would remain under 1 W (which these won't be happy about but at least they shouldn't blow up right away).

Hi-Fi amps typically balance DC resistance (and AC impedance) carefully, and often drop gain at DC to unity.

You wanted to build an amp the hard way, so you'll get all the associated fun. ;)
 
Yeah, I'm using a On Semi LM833. Sadly I completely forgot to check the DC offset when I was simulating the circuit.

You wanted to build an amp the hard way, so you'll get all the associated fun. ;)

Sounds like fun to me! ;)

I've been experimenting with your suggestions (specially the CFP output which is really interesting) on my next project. This is how it currently looks like:

fq2oUiP.png


I've also generated a nice THD graph (into a 8 ohm load) using my little THD plotting script in conjunction with ngspice:



This time I checked the DC offsets and while it's idling I get 32.8mV across the 8 ohm load. I've also included the LTSpice schematic if you want to check. :)
 

Attachments

  • Discrete Power Amplifier (Compound Output).asc
    4.7 KB · Views: 109
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.