5
\$\begingroup\$

I have always wanted to design and build my own audio amplifier. Is the following a reasonable design?

  1. The input stage takes in line level signals (say, up to +/-3 V, but typically more like 0.3 - 0.8 V).

  2. TI OPA 134 "audio op amp", in buffer configuration, to impedance match for...

  3. Low pass filter (ideally one I can turn on and off, so I can put the amp in "subwoofer mode")
  4. TI OPA 134 in non-inverting amplifier configuration, aiming for gain of 5-20
  5. A push-pull BJT amplifier stage, with enough gain to push 100 W through 8 ohms of impedance.

All of this powered by a, say, a 150 W power supply with high voltage rails.

I originally wrote +/- 15 here. I had op amps on the brain, since I am actively working on designing the line level circuit and want to figure out what gain factor I'll eventually need.

Obviously, there are a lot of details/features I'm not including or even designing at the moment. The BJT part is especially sketchy at the moment. I'll get to that, but I want to know if it's even "plausible" and reasonable.

Questions:

  1. Is this a reasonable first attempt at sketching out design?(This is actually version 0.0.4 of my design). Any additions/advice for a beginner?

  2. Given that I'll have a low pass filter, is the buffer op amp necessary or does modern audio equipment typically have a high enough output impedance at line level? Does it make sense to use an audio op amp in the buffer configuration?

  3. I'm guessing it makes more sense to do the low pass filtering on a line level signal as opposed to after the BJT amplifier stage, but that's just a guess. Is it right?

  4. How do I calculate the voltage I need across my 8 ohm load? If I do a DC approximation, it comes to about a 28 V difference (i.e., +/- 14 V). (Did I flub this math? P = I^2 R = 100 W / 8o = 12.5 A^2, so I is about 3.5 A, which means the voltage difference must be about 28 V give or take.) But presumably, AC imparts less power at +/- 14 V than DC at +/- 14 V. (But then again -- perhaps as a naive guess? -- a square wave swinging between +14v and -14v "should" intuitively push out as much power as +/- 14 V DC...) So what output voltage do I really need to get 100 "real" watts out of my speaker? What assumptions should I make about the input signal and it's spectrum? Square waves? White noise?

Or am I overthinking this, and should just aim for some rule of thumb?

Despite some of the comments, I am going to do this. Even if it takes me 50 practice circuits and and auditing a course on EE. I am not trying to kill myself or start a fire. I am trying to plan out a long term project -- to figure out how much time to invest in learning the line-level components and to figure out what kind of transistor technology and amplifier layout to learn about next, in order to accomplish my goal. Assuming this sketch of a design (barring the power supply issues mentioned) is even plausible. Which is why I asked in the first place. I am working on a roadmap, not trying to sit down and solder random parts together.

I'm totally green when it comes to building circuits, though I did learn a lot as a kid (Mims III) and vaguely remember some circuit design/analysis from college (passive components, Ohm's law, DC analysis, RCL analysis (though I don't remember the formulas right now and would have to look them up to solve for capacitances, etc.), transistor basics, and op amps, Fourier analysis for signals and in real analysis. I also have a math degree and took a bunch of physics classes.

\$\endgroup\$
  • 5
    \$\begingroup\$ You forgot to add the block diagram. You ask questions that suggest you don't know Ohm's Law or Watt's Law...how "vaguely" do you remember these? \$\endgroup\$ – Elliot Alderson Dec 28 '19 at 21:25
  • 2
    \$\begingroup\$ nomen, Do you seriously want to engage a \$150\:\text{W}\$ audio amplifier with the background you've mentioned? I can't say this strongly enough in words without having my backside thrown out of here, but you really need to think about something powered off of a 9 V battery first. Like maybe something around half a watt, instead?? I mean, seriously, if you can get a good half watt out of a 9V battery, then you can consider it a great success. And you'll learn the important things. The power you say brings in so much more. Too much more, in my opinion. Set sights lower. \$\endgroup\$ – jonk Dec 28 '19 at 21:29
  • 1
    \$\begingroup\$ @nomen Also... note that your suggested rails aren't going to support the power you specified. Do you know how to calculate the rails you'll need for 150 W? I think you are like so.... way.... out of your depth. Think smaller. Take this in steps. Please. \$\endgroup\$ – jonk Dec 28 '19 at 21:31
  • 2
    \$\begingroup\$ @nomen - jonk is a lot closer to correct than you are, I'm sorry to say. What you need to keep in mind is not the Fourier decomposition - it's the peak signal in the time domain. After all, a speaker does not work on a set of Fourier harmonics. It just gets a signal in and tries to drive an air mover proportionally. So, find the peak power you'll encounter if you're in the mood for precision. Otherwise, just specify an RMS power, multiply by 2 and call it P. Then for power supply voltages +/- V, P = (V - H)^2 /R, where H is the headroom required for the output stage. \$\endgroup\$ – WhatRoughBeast Dec 29 '19 at 3:03
  • 2
    \$\begingroup\$ The factor of two comes about because the peak value of a sine wave is 1.414 times the RMS value. Also, you need to be careful how you produce your supply voltages, since a simple transformer/rectifier/capacitor setup will "sag" during high-power episodes, but you're likely to have measured it with no load. Another good reason to take it gradually is to learn about heat sinks on your output stage. Dealing with or minimizing the waste heat from a linear transistor stage is not something which will come easily (or cheaply) at the 150 watt level. \$\endgroup\$ – WhatRoughBeast Dec 29 '19 at 3:07
6
\$\begingroup\$

To get 150Wrms on an 8 ohm load you'll need 35xsqrt(2) volts on your DC rails... Plus and minus 50V. Plus a little bit for losses etc. 150 watts rms is a lot of power for home use. 150 watt amp will melt the voice coils of most home HiFi speakers. Use P=V^2/R, and remember Vpeak is sqrt(2)xVrms

Don't bother with toroids bridge rec etc... Making your own psu is a mugs game unless you're a professional psu manufacturer. Buy a pair of isolated switch mode ones. Don't believe the audiophile nonsense about switch modes being noisy etc... Technology has moved on in leaps and bounds since they all formed their opinions.

Likewise, the leach amp is a relic of a bygone era. Lm3886 is an excellent beginner choice (I've built 2 of them myself, with 2x 24V switch mode PSUs) the sound is subjectively and objectively excellent if done right-but beware of fake chips) and irs2092 is more complicated to do yourself, definitely not beginner friendly, but a nice solution.

You don't normally need to buffer modern outputs from e.g your mobile phone, but older kit may have higher output impedance so suffer voltage drop and distortion without a buffer. The buffer amp is useful if you have a weak signal or your filter is liable to create harmonics on your input line (filters impedance changes with frequency so presents a different load to the input at different frequency).

You should absolutely do the filtering at line level. Filtering at power level is expensive since it requires large components and more likely to not work with the reactive load of the speakers/resonate/...

For building and testing, get yourself some power resistors rather than using speakers at first, an oscilloscope if you can afford it, and a power supply with a current limit and variable voltage (start low and turn up the volts once it is working).

Edit: you said 100W not 150W... For which you'll need 28sqrt(2)V, +/-40V rails. This is still a pretty monsterous amp for home HiFi.

\$\endgroup\$
  • 2
    \$\begingroup\$ It'll cost you more and work less well, and have associated chance of shorting, not being safe, allowing mains noise through... Etc. Professional psu makers are good at this fairly difficult task, the OP probably is not and stands to derive little value from it. \$\endgroup\$ – David Molony Dec 29 '19 at 19:10
  • 1
    \$\begingroup\$ @David: "A mug's game" is a British English colloquialism. Remember that you have an international audience. \$\endgroup\$ – Transistor Dec 29 '19 at 19:16
  • 1
    \$\begingroup\$ It does occur to me that the common RMS power terminology may be pedantically incorrect... \$\endgroup\$ – David Molony Dec 29 '19 at 19:29
  • 1
    \$\begingroup\$ I got a lot of awesome input, but this is the only answer that answered my original questions. Thanks everybody! I agree that the PSU is out of scope for my project. This is a long term project and I will make some detours into the op amp circuits you mentioned. This still helps me set up my line-level stages, which is awesome! :-) \$\endgroup\$ – nomen Dec 29 '19 at 22:06
  • 1
    \$\begingroup\$ Glad to have helped. All the best of luck on this project. It's a good electronics starting point - sufficiently hard to have a point but readily attainable, and has a nice, satisfying outcome! It's where I started out several years ago! \$\endgroup\$ – David Molony Dec 30 '19 at 9:25
9
\$\begingroup\$

All of this powered by a, say, 150W power supply with +/- 15v rails.

With 15V rails, a BJT output stage should output a peak voltage of about 13V. In a 8 ohms resistive load this is 21W peak, or 10W "RMS". Note "Watts RMS" really means "active power in a resistive load"...

So your amp needs to run on higher voltage rails, which means it can't use opamps to create all the gain, since the output voltage of opamps is limited to +/- 15V. So you have to use discrete, which means opening all the can of worms at the same time, from loop stability to thermal management.

So... If you're a beginner, I definitely wouldn't recommend designing a power amp, because you'd have to get everything right. It could be expensive and risky, and if it smokes there is a good chance one of the output transistors will fail short circuit, which means the amp will output one of the rail voltages continuously and burn the loudspeaker. Also the PCB layout is difficult to get right. Starting with a kit is a much better idea, you'll get all the info on how to choose the heat sink, hopefully a good pcb, etc.

If you want to build something, the line-level, low power part with the opamps is much less risky, and mistakes are less costly. Note a misbehaving preamp can still destroy your speakers if it oscillates, but that's uncommon.

Once I was probing inside a headphone amp ... that stuff isn't scary, right? The scope probe somehow shorted something, one output transistor blew, shorted, and burned the voice coil of the $200 Sennheisers I had forgotten to disconnect. OUCH! That really sucked. Always try the new amp on second hand garbage speakers...

If you want a DIY power amp for full range speakers, you can't go wrong with a LM3886 based kit. Search "LM3886 done right". It has full protection and is easy to use, and LM3886 sounds very good considering the price and simplicity. In fact a properly implemented LM3886 will probably sound better than most BJT amplifier kits, and it doesn't smoke if the output is shorted, which is always a plus...

For a subwoofer, a Chinese class-D module with IRS2092 would be a good option, at an unbeatable price/power ratio.

After you build that stuff, read these excellent books:

"Art of Electronics" by Horowitz & Hill

"Audio amplifier design handbook" by Douglas Self

"Designing audio power amplifiers" by Bob Cordell

The last one in particular covers every tricky detail that really make the difference between smoke and no smoke. I really recommend it, it's practical, well written and "user-friendly", it will answer all the questions you asked, and also the ones you didn't ask. I mean, designing a power amp is not simple, but definitely not rocket science either... having a book that explains all the mistakes you can make and how to avoid them sure saves a lot of time and blown transistors...

There's also something you have to do...

You say "I want a 150W amp", this is something people often say, but do you know how many watts you actually need?

You can measure this with a scope. Probe the output of your current amplifier, using your usual speakers, while playing a bit louder than usual on music you like.

Last time I did that, I measured voltage equivalent to 1W average, and 10-20 watts peak. So my 50W amp actually never clips. It's an interesting experiment to do. Subwoofers usually need a bit more power though, which is why I recommended a Class-D amp.

EDIT:

About 15W amps. You will save a few dollars on the heat sink.

If you use one pair of output transistors (one NPN, one PNP or the MOSFET equivalent) and they're the beefy kind like TO-247 packages then you can do 50-60W RMS. More power requires more power transistors in parallel which opens another can of worms: current sharing, they have to be matched, etc, plus a big heat sink, it gets expensive and annoying real fast, so you only do it if you need it.

A 50-60W RMS amp is never going to clip anyway at normal listening levels with reasonably efficient speakers. So do you need to pay extra for 100W? Only if you determine, by measuring the voltage at the output of your amp, that you need the extra watts! Could be the case for inefficient speakers, a sub, or if you like it REALLY loud.

However a 15W amps will clip, and if it uses discrete components its schematic will be basically the same as the 60W amp, only the heat sink will be smaller. So the 15W amp has no advantage, unless you do it Class A and have super efficient speakers.

\$\endgroup\$
  • \$\begingroup\$ Thank you for all your input and advice. Maybe 100W is too ambitious for someone without a lot of practical experience. I understand that, with a high power circuit, stakes are high so and so I will do as you say and start smaller. Would you say a 15-20W amp is too ambitious for a beginner to design? On the cusp? (I can do cusps well). \$\endgroup\$ – nomen Dec 28 '19 at 22:39
  • 2
    \$\begingroup\$ Well I was also wondering where does the fixation for 100W RMS power into 8 ohm load really comes from. Is this some arbitrary value or you have a really big room and it is really based on some actual calculations? As a math person this might help you calculate what you really need. A normal speaker puts out about 80 dB of SPL when fed with 1 Watt and measured at 1 meter. That is approximately 1 lawnmower. Each time you double the power, you get one lawnmower or +3dB SPL more, and each time you tenfold the power, you get ten lawnmowers or +10dB SPL more. How much dB or Watts you need? \$\endgroup\$ – Justme Dec 29 '19 at 0:29
  • 1
    \$\begingroup\$ @nomen I've added stuff to the answer. Also you can find Cordell's book as pdf on the net if you search for it. \$\endgroup\$ – peufeu Dec 29 '19 at 0:50
  • 2
    \$\begingroup\$ @Justme peak power isn't average power! 1W average was really loud on my speakers. Usually I listen lower than that. But peaks 20dB above average power are common and do not sound loud, they sound natural if they are, well, just peaks. \$\endgroup\$ – peufeu Dec 29 '19 at 0:53
  • \$\begingroup\$ @justme: thanks for your suggestion. My 100W goal comes from my original plan, which was to build the "Leach amp" (which is 100W). If 40W or 10W is enough, I'd be happy with that. I just want my subwoofer to sound loud enough to match my main stereo pair (which is on 2x125W denon circuits). But as peufeu says, I'm probably wasting money on specs I don't need. Also, I do want to learn about this stuff. It was always a goal of mine to build cool stuff. \$\endgroup\$ – nomen Dec 29 '19 at 0:58
5
\$\begingroup\$

A suggestion. Rather than approach this as one enormous project, approach it as a project with subprojects and learn along the way:

  • First buy off-the-shelf modules for a pre-amp, a power amp, pre-built power supplies for each (perhaps wall-warts), a case, volume controls, and make it work. Given availability of (normally Chinese) imports on Amazon etc., this will be ridiculously cheap (cheaper than buying the components yourself), and has a much higher percentage chance of working first time than assembling it yourself.
  • Next build a single power supply for both (e.g. make the pre-amp supply work of the higher voltage power amp supply, or - if confident with mains electricity - buy a torroid, a couple of bridge rectifiers, some electrolytics, and some regulators etc.)
  • Next replaces the pre-amp stage with something you designed and built yourself
  • Next replace the power amp sage with something designed yourself around an existing chip (LM3886 or similar)
  • Next (if you must) progress a home-built MOSFET based output stage or whatever

At each stage you will have a working audio amp. When you get to the end you'll have one working, home designed and built, and in effect the bits for another. I bet this will cost less in destroyed components or misdesigns than trying to do it all in one go. And you can break the project into manageable stages, take a breather and read-up on the next stage after each.

You might grumble "but my pre-amp, power amp, and power supplies will be on separate boards!" This is not a bad thing for a beginner, and indeed not a bad thing in general even if you are not a beginner.

\$\endgroup\$

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.