Maker Pro
Maker Pro

Dodgy Power Amp Design

Stuck another one on Q4, they both got a bit warm, and distracted me while Q2 silently roasted to death.

I give up. I've ****ing had it with amplifiers. I hate the things!
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Try the trick with the globes. Pick a 12W globe (or similar). In quiescent conditions the bulb should not light up. If it does, something is wrong, but at least your transistors are (relatively) safe.

edit: It might also be a good idea to insert some small emitter resistors in Q1 and Q3. Otherwise there is not much to limit the quiescent current of your output stage. (I'm no expert in how this is actually set, but you seem to have no way to set it)
 
Last edited:
It seems to work and be stable now (thanks everyone by the way). More resistance seemed to do the trick.

There's alot of RF interference when I plug it into my PC (which I'm using as the sound source). Most of it seems to be coming from the computer but I did pick up a radio station at one point. I'm using proper audio cable. Is there any way to filter that?

Here's how it looks now. Note that I've also moved the op amp output to the centre:

pamp3.jpg
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
R8 and R9 are going to seriously compromise your output voltage swing, and thus your output power.

I was suggesting you place them in the emitter of Q1 and Q3. Those resistors will help ensure you don't blow the output transistors, and you could leave them in there for the short term while you get everything else sorted out.

The trick is to set the quiescent operating conditions so that Q2 and Q4 are only very slightly on, passing perhaps 50mA (I'm not sure what a suitable value would be, but it would be in the order of mA to tens of mA).

Most power amplifiers have some means to set this quiescent current, although I'd resort to Googling to find out how it's normally done.

Figure 6 on this page describes what I'm suggesting for your drivers. The current through the drivers will be amplified through the output, so the emitter resistors of Q1 and Q3 need to set their quiescent current to something quite small (100's of uA?) to get appropriate current through your output transistors without the need for additional resistance.
 

KrisBlueNZ

Sadly passed away in 2015
R7 should be a much higher value, especially if you want a reasonably flat low-frequency response. I would use 100k for R7.

You can simplify the circuit slightly by doing what I suggested in post #11. That is, (1) delete R7; (2) disconnect the junction of R1 and R2 from C2; (3) connect the junction of R1 and R2 to the "+" input of the op-amp.

As Steve says, R8 and R9 will greatly limit the output power of the amplifier. If the output transistors are overheating without those resistors, that is a separate problem that began in post #16 after you failed to connect the output transistors properly. Have you replaced the output transistors?

Connecting the op-amp to different places in relation to D1 and D2 will have almost no effect. Connecting it to the centre point will cause slightly more distortion because it is less tightly coupled to the bases of the output transistors. I recommend leaving it connected to Q1 base.

This amplifier uses fixed bias, set by D1 and D2. A better method is to replace D1 and D2 with an NPN transistor, with a resistor (e.g. 6k8) from collector to base, and a trimpot of slightly higher value (e.g. 10k) from base to emitter. This transistor must be coupled to the output heatsink to provide thermal feedback.

To set the output stage quiescent current, connect a current meter in series with the output stage (e.g. in series with Q2 emitter), set the trimpot to maximum resistance, power up, and adjust the trimpot for around 20~50 mA quiescent current. Leave the amp to warm up for 10 minutes, readjust the trimpot, and repeat a few times until the temperature and current stabilise.

That output stage, using Sziklai pairs, is probably not the best design. It's more usual to operate the output transistors as emitter followers.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
R7 should be a much higher value.

How much difference does input offset current make in an audio amplifier like this Kris? (to optimise this R7 = R3 || R6)

I was concerned about the choice of 1k for R7 mostly due to the high pass effect it creates with C1

Your suggestion ties the input to the center of the voltage divider (which has an impedance of 110k), but by connecting the feedback loop's ground reference there are you not causing a similar problem by essentially shorting out the input to a virtual ground? I'm not suggesting it won't work, but I can't understand how it does.
 
Moved the resistors to the driver emitters and it it didnt blow up - cool :p
Same RF interference though - and a low pass filter doesnt do much except kill the sound.
As for R7 - looks like Steve is right on that one - tried a 47K and it was very bass heavy and distorted.
 

KrisBlueNZ

Sadly passed away in 2015
How much difference does input offset current make in an audio amplifier like this Kris? (to optimise this R7 = R3 || R6)
No significant effect. The closed-loop gain is not high, and the signal levels are very high compared to the input offset voltage or any offset caused by input current.
I was concerned about the choice of 1k for R7 mostly due to the high pass effect it creates with C1
Me too.
Your suggestion ties the input to the center of the voltage divider (which has an impedance of 110k), but by connecting the feedback loop's ground reference there are you not causing a similar problem by essentially shorting out the input to a virtual ground? I'm not suggesting it won't work, but I can't understand how it does.
I wasn't suggesting returning the feedback network to that point. The feedback network consists of R6 from the output to the inverting input, and R3 and C2 in series to 0V. The voltage divider at the non-inverting input sets the DC bias. The voltage at the top of C2 will follow. Effectively, the top end of C2 is a virtual ground, with its DC voltage set by the DC voltage on the non-inverting input.

As for R7 - looks like Steve is right on that one - tried a 47K and it was very bass heavy and distorted.
You must have done something wrong. Make only, and exactly, the changes I listed, compared to the most recent diagram you posted.
 
Last edited:
Kris's way works now (R7 removed, + to divider, R3 to C1 only) - must have been doing other stuff wrong before when I tried that.

Interference is still a problem though. Most of it is noise form the computer. How do pc speakers get round this?
 
Last edited:

KrisBlueNZ

Sadly passed away in 2015
Is the noise coming in on the power supply or the signal? High-frequency noise on either one can be removed with filtering - simple R-C filtering for the signal, where some signal loss is acceptable, and L-C filtering on the power rail.

Show us a diagram of how it's all connected. How is the amplifier powered? Is there an earth loop somewhere?
 
Power is from a cheap ATX supply on my desk, audio is from headphone output (shielded cable) - definitely looks like its from the 0V rail of the supply - with the power off the noise is still audible with headphones until thats disconnected
 

KrisBlueNZ

Sadly passed away in 2015
Assuming that "With the power off ..." refers to the laptop power...

If you need to use an ATX supply, you should add L-C filtering in both the negative and the positive rails. That is, break the positive rail and connect an inductor across the gap, ditto for the negative rail, then connect some capacitors between the rails on the amplifier side of the inductors.

The inductors need to be rated for the current that your amplifier draws at full output. For example:
http://www.digikey.com/product-detail/en/5900-681-RC/M8331-ND/774871
http://www.digikey.com/product-detail/en/5900-391-RC/M8329-ND/774869
http://www.digikey.com/product-detail/en/1447423C/811-1333-ND/1924820
http://www.digikey.com/product-detail/en/1422441C/811-1339-ND/1924832
http://www.digikey.com/product-detail/en/1130-271K-RC/1130-271K-RC-ND/2534438

For capacitors, use a good quality ceramic cap such as
http://www.digikey.com/product-detail/en/K474Z20Y5VF5TH5/BC1167CT-ND/286789
http://www.digikey.com/product-detail/en/C430C474M5U5TA7200/399-4513-2-ND/818288

... in parallel with a good-quality low-ESR electrolytic such as
http://www.digikey.com/product-detail/en/EEU-FM1V471/P12415-ND/613776
http://www.digikey.com/product-detail/en/UHW1H471MHD6/493-6951-ND/3664391
 
Cool - I might already have some of that stuff salvaged from an old power supply. I'm also planning to run the final version with a pre amp off a battery (depending on current draw). The plan is to attach the transducer to the body of a guitar (that is using it as the output on an acoustic guitar with a coil pickup - in theory it shouldnt feed back or it would at least take alot of energy to do so)
 
Last edited:
Top