3
\$\begingroup\$

enter image description here

I'm trying to drive a speaker from the 47u capacitor , it seemed to be working fine and the volume scales well but when it comes to human voice , it adds this weird bathroom effect with some poor performance when it comes to high frequency/low volume . There is also some distortion

It weirdly enough , works fine when it comes to instruments (that or I have pleb ears )

What could be the cause of this ? Sorry if there's any glaring mistake but this has been bugging me for quite some time .

Edit : I'm audio I'm playing is from my phone's mp3 jack.

\$\endgroup\$
  • 1
    \$\begingroup\$ When you say "human voice", are you talking about a recorded voice, or a live signal from a microphone? In the case of the latter, you might be experiencing some acoustic feedback that is complicating matters. \$\endgroup\$ – Dave Tweed Oct 21 '16 at 10:28
  • \$\begingroup\$ @DaveTweed, the input is from my phone's mp3 jack, I'm playing some music through it \$\endgroup\$ – Vrisk Oct 21 '16 at 10:29
  • 2
    \$\begingroup\$ The small value of C1 creates a high-pass pole at around 500 Hz. This will definitely make voices sound "tinny" -- like through a telephone. Try making this capacitor at least 10x larger. \$\endgroup\$ – Dave Tweed Oct 21 '16 at 10:35
  • 2
    \$\begingroup\$ @AnirudhGanesh Consider this answer I wrote elsewhere here: electronics.stackexchange.com/questions/262274/… \$\endgroup\$ – jonk Oct 21 '16 at 11:36
  • 2
    \$\begingroup\$ @AnirudhGanesh Read through that link I mentioned just a moment ago and see if it makes any sense to you. It will help me decide if I can add anything useful. \$\endgroup\$ – jonk Oct 21 '16 at 11:46
4
\$\begingroup\$

You are going to have significant problems driving a low impedance speaker such as 8 ohms. To produce (say) 4 Vp-p across it requires that the peak current is 250 mA. This has to be delivered by the 2N2222 transistor on one half cycle of the waveform and by the 330R resistor (R5) on the other half cycle. Clearly, trying to take 250 mA through a 330R resistor needs a forcing voltage of at least 83 volts but you only have a 10 volt supply! So, my recommendation is to massively reduce R5 or preferably, use a push-pull output stage.

Given that the current gain of the output transistor might be 100, you will need to put at least 2.5 mA peak into the base but, taking 2.5 mA through R3 is not going to work because 8.25 volts will be dropped by it. This a similar story to that of R5 mentioned above.

Because of this you could set Q2 to have a much increased quiescent current of maybe 25 mA - this means reducing R4 considerably to something like 22 ohms AND reducing R3 to something like 220 ohms.

Now, when 2.5 mA is asked to be drawn through R3 it will drop about 0.55 volts and much more easily drive the final transistor Q1.

For the same reasons as above, you will need to reduce R1 and R2 because Q2 requires a quiescent current of 25 mA and, assuming it has a current gain of 100, its base will need 250 uA. The generally accepted rule here is that if the base needs "X" amount of current then the bias resistors ought to take 10 times this current. So choose R1 and R2 such that about 2.5 mA flows though them.

The other problem with distortion is C1 across emitter resistor R4 - it will make the output shape (if fed a sine wave) asymmetrical. This might catch you out later so please be aware of this.

I would also strongly urge you to use LTSpice (a free sim tool) to develop your ideas because its analysis of this simple problem will be more accurate than what I can do with pen and paper (no access to a sim today).

\$\endgroup\$
  • 1
    \$\begingroup\$ R5 is a serious problem. When Q1 cuts off, and it will long before the situation you describe above occurs, R5 has to sink too much current and it can't. The whole thing needs work. \$\endgroup\$ – jonk Oct 21 '16 at 10:48
  • \$\begingroup\$ How can you assume the gain? \$\endgroup\$ – Vrisk Oct 21 '16 at 10:49
  • \$\begingroup\$ @AnirudhGanesh - are you talking to me? If so then use the "@" symbol followed by my alias \$\endgroup\$ – Andy aka Oct 21 '16 at 10:53
  • \$\begingroup\$ @Andyaka, yes, did you assume the gain as hundred? It would actually vary highly, wouldn't it? \$\endgroup\$ – Vrisk Oct 21 '16 at 10:55
  • \$\begingroup\$ When driving a significant current the hFE falls off rapidly - look at the range of DC current gain in the data sheet. At Ic = 150 mA the gain is only 50 to 100. At 10mA Ic the gain might be 75. Also, if the answer I left yesterday explained things for you please consider formally accepting it or explain where you needed more information. \$\endgroup\$ – Andy aka Oct 21 '16 at 11:00
9
\$\begingroup\$

The 47 µF cap (C3) in series with a 8 Ω speaker forms a high pass filter with a rolloff off 420 Hz. That's the point at which the volume is already down by 2, and it gets lower inversely proportional to frequency below that. The reason "instruments" sound better than voice is because voice has more frequencies below this point.

However, that's only one problem of many with this circuit:

  1. Q1 will drive the output up with lower impedance than R5 will drive it down. The result will be assymetrical output shape to a symmetrical input. In other words, distortion.

  2. There is no cap across the power supply. When the output is driven high, it takes more current, which sags the supply. This feeds back into the first stage making a mess, possibly even causing oscillations depending on the impedance of the battery.

    There should be a decent size cap across the power supply, and the supply to the early stages filtered to protect them from noise on the supply anyway.

  3. 2N2222 is inappropriate for driving a low-impedance load like a 8 Ω speaker. If you fixed the output drive, then the next problem would probably be that Q1 is getting fried.

  4. The DC bias is quite dependent on the specific supply voltage. It looks like it will be close enough for the 10 V supply shown. You want the output of Q2 to be about in the middle of its possible range.

  5. C1 causes a high-pass rolloff of about 480 Hz. Below that, the gain is roughly 10. Above that it goes up with frequency. For audio, you want a flat gain across the intended frequency range. That frequency range is 20 Hz to 20 kHz for "HiFi" audio.

    This is now the second overall high pass filter, both together making "voice" sound a lot worse than "instrument".

Overall, if you really want to drive a speaker, ditch this mess and use a circuit intended for that. Driving a speaker isn't as simple as slapping a cap on the output of a low power amp, like what seems to have been done here.

\$\endgroup\$
  • \$\begingroup\$ Made the 47 cap 100 u and the 1u to 10 u, quality, well, it's still bad. \$\endgroup\$ – Vrisk Oct 22 '16 at 7:56
  • \$\begingroup\$ @olinlathrope, also tried with 470u, sound quality is waay worse, the voices are not fixed \$\endgroup\$ – Vrisk Oct 22 '16 at 8:05
  • \$\begingroup\$ @Ani: All that means is that you haven't bothered to listen to or understand what I said. \$\endgroup\$ – Olin Lathrop Oct 22 '16 at 16:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.