Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Build your own professional-grade audio amp on the sort-of cheap (2018) (ieee.org)
176 points by Tomte on Nov 7, 2020 | hide | past | favorite | 131 comments


The amp shown is nowhere near "professional grade".

Professional means it does the right thing in unusual circumstances - naughty inputs, naughty outputs, naughty mains power, running at full power on the grass under a pile of coats and scarves, pumping through old and fraying speaker leads lying on the ground, ...

The design is missing all sorts of safety features apart from the most basic (prevent electrocution from broken mains wires touching the chassis).

A 200VA transformer with a 350W amplifier is just asking for a transfomer meltdown (or fire, if the transformer doesn't have a functioning thermal fuse), unless there is a limiter in place to limit the output to 120W or less. (Fuses are also recommended, but in the professional environment you don't want them to blow, ever. Hence, a limiter is mandatory.)

The bridge rectifier is undersized.

In traditional mains-frequency rectification such as shown, the diodes only conduct for about 10% - 20% of the time, so the peak current in them is can be more than 10 times the average. At 36V and 10A output average, the rectifier current peaks are in the neighbourhood of 100A. A BR35 (35A 1000V) will provide a reasonable working life. At up to 36W dissipation, you can probably just bolt it to the chassis with some thermal compound.

The design is also missing speaker protection features and surge/spike protection/EMI prevention, and RF intereference filters on the inputs.

The design is also missing convenience features one expects of a professional amp, such as "power good" and "fault" lamps/circuitry. And handles.

It's nice that there are good inexpensive amplification modules now. But as with software, the difference between a toy example and professional grade is thought, time and money spent on reliability, safety, and usability.


I do think it's interesting, though, that you could conceivably add all those elements, and you'd still be well under $500 in parts.


I remember reading you should shoot for total BOM costs of about 30% sale price though... So that seems more reasonable to me


> The design is also missing speaker protection features and surge/spike protection/EMI prevention, and RF intereference filters on the inputs.

1.) TPA325x have built-in speaker protection circuits.

2.) Do balanced inputs need RF interference filters?

For this project: After blowing the bridge rectifier a few times (for the reason you mentioned above), I ended up replacing the linear supply with a SMPS instead.


If you want TPA3250 amps without having to build your own, they're widely available from the usual cheap vendors for less than $100.

The victory of class D is now total; extremely low distortion numbers can be achieved without heroic board layout or expensive passives made of unobtanium. All available in the tiny packages of AirPods, smartphones etc. And power levels in the tens of watts can be done with a single chip.


I built this project myself last spring. That 3e-Audio TPA3250 board isn't sold anymore, so I built my own custom variant amplifier using similar parts with a TP3251:

https://photos.app.goo.gl/gPH8VZiaGRiNYWfL6

It sounds MUCH better than my existing cheap vendor Fosi TPA3116 did -- the cheap vendor ones don't use quality components, especially for the power supply, so you're not going to see the expected performance.

That power supply board (diode bridge + filtering caps) was designed with a friend in KiCAD and fabricated by JLCPCB (I have a few boards leftover). The speakers are also DIY -- Hivi 3.1 with cherry wood veneer!

I showed article author Zorpette my work, and he gave it his blessing =)

EDIT: If you do try to build this, be VERY careful around the high-voltage power supply electronics.


There is no way DIY diode bridge and filtering caps on a separate board could make a slightest difference here, in fact high frequency noise from the power supply could only be filtered right near the chip. But leveling up an amplifier, to a 4 times more powerful as you did, would make a significant difference in sound.


In fact there is a way or two.

The extra resistance (2 to 10 milliohms probably) of the wire between the two boards, together with the smoothing capacitors on the downstream board, acts as a low pass RC filter.

Also the capacitors on the PS board may be low-inductance types, and filter HF noise more effectively than the caps on the amplifier module.

Having the diode bridge further away from the amp module's inputs means the antenna action of the rectifier's leads is less effective: the RF power of over-the-air rectification transients at the inputs is reduced by the square of the distance.

The layout of the amp as shown is good, except that I'd shorten and cable-tie wires to help prevent hum loops.


I would suspect imagine that putting a power supply like this in the same enclosure could introduce unpleasant noise. There will be a big current spike every half cycle when the rectifier starts conducting. The power factor will be awful, and I can easily imagine 120 Hz and its harmonics coupling somewhere you don’t want them. Not to mention that you might be able to hear the power supply buzzing.

Get a nice AC-DC converter brick and call it a day. If you want to be fancy, stick a linear regulator on the output.


If Apple or Dell sold 36V 10A power bricks, then maybe.

The ones that I've seen don't put out their claimed current, lose regulation badly at high load, and put a lot of hash on the DC rails. Seriously ugly. Opening them up, the circuit design often looks like it comes from the 1980s. Physical design usually doesn't meet standards for HV separation and creepage.

Rather than a linear regulator, lots of filtering (a few 100nF ceramics and a few millifarads of smoothing) would help. And a high-powered TVS or two for when the power brick dies, as it will.


Apple or Dell? I would start with Meanwell if I were trying this.


TPA3251 is the best on all performance


Very pretty! Thanks for showing it. What I'm going to say I hope will be taken as constructive, for those who are a bit more OCD about their electronics.

Safety: -

The mains safety earth (chassis bolt) should only ever have one lug: the mains input earth. Put a separate bolt an inch or so away from the mains safety earth bolt for the chassis connection of the bridge centre ground. (People have killed themselves doing repairs by ignoring this, and having faulty mains wiring.)

Fuses on the transformer secondaries are recommended if the transformer is expensive. If the bridge rectifier or an amp output stage fails, it'll almost certainly fail short, and the transfomer will heat up until its internal thermal fuse melts -- or if there isn't one, until it catches fire.

Likewise fuses on the DC rails will help protect the expensive smoothing capacitors from amp failure. (Yes, the TPA3251 has built-in protection - but that might also be damaged by the surge that caused the output stage to fail.)

I hope the polyester cap across the transformer secondary is X2 class.

Other points:-

A snubber across the mains switch will reduce arcing on switch-off and prolong the life of the switch. 10R-50R 3W in series with 100nF, X2 class, will do it. (The wattage rating is a proxy for both voltage rating and surge current rating.)

From the scale it looks like the transformer is a 120 or 160VA model. That's seriously undersized for an amp module rated at 350W (circa 550 VA with a traditional mains-frequency transformer). Transformer secondary fuses are strongly indicated.

OTOH, going up to a 500VA toroidal transformer means you'd need a soft-start circuit to prevent your house circuit breakers popping on start-up, from the initial transformer saturation inrush current.

I'm amazed you don't get RF interference with all that length of unshielded wire on the input side. I just have to leave one inch unshielded to be plagued with EMI - unless I use filters on the input terminals (100R / 220pF RC filter).

You don't have DC blocking on your volume control. I guess none of your sources have any DC offset. If not, changing the volume will be scratchy.

References: -

1. Bob Cordell, Designing Audio Power Amplifiers.

2. Doug Self, Audio Power Amplifier Design 6e.

3. Rod Elliott, sound-au.com. "Designing Linear Power Suppplies part 2" https://sound-au.com/power-supplies.htm

4. Idem, "Transformers - The Basics, part 2". https://sound-au.com/xfmr2.htm


This level of free electronics review is awesome, thanks from a software guy with more solder than sense.

Glad I'm not into power or audio electronics. :)


I hope I haven't put you off for life! :)

It can be fun and rewarding to buid your own.

Your amp will be fine as long as you use it for lounge-room listening. Just don't use it in pro situations, e.g. a wedding reception or large party (50+ people). Please!

(But if that cap across the transformer secondary is only DC rated, not AC, you can expect it to explode. Might take a couple of years, if you're lucky. I had one last five years before I learned about "X-class" mains rated capacitors.)


- There is only one lug, see photos.

- Polyester caps are AC rated

- Transformer is 300VA


I disagree that the "victory is total". First of all, distortion in the upper part of the audio range is not that stellar. Secondly - I personally can tell apart Class-D, Class-AB and Class-A. Class-D has grainy sound, even the best ones, $3000+ I tried; I get fatigued pretty quickly from them.


Interesting - I didn't realize that AirPods have class-D amps.


I'm pretty sure everything small and portable uses switching amplifiers, in fact pretty much all modern consumer electronics use class-D from your phone to your TV unless you specifically seek out separates from an audio shop. This is because class-D amps are extremely convenient, they are small, cheap, power efficient and they require little consideration for cooling - audio quality is an area of debate, there are certainly lots of very poor quality class-D amps our there with absurdly high THDs, but there are probably lots of good ones too.

AB would be more likely than A, but I can't imagine finding them even in a large flat screen TV since they are inherently less efficient and require larger power transformers which make them fairly heavy and bulky. You only really find non switching amps these days when people want to sacrifice some of that convenience for some nicer sound.


The trifecta of outstanding power efficiency, low cost, and nowadays very good performance of a class D amp virtually guaranties that any amplified sound you've head from a mobile device, or any consumer device really, was amplified by a class D.

There's not really a good reason to use non-class D in any mobile application today. As few as 10 years ago I thought Class D = garbage you settle for if you must, I've heard some incredibly good Ds that have made me completely change my mind


Just to throw an extra confounding factor in, I can guarantee that your hearing now isn't as good as when you first heard them. But that's fine, it's the subjective experience that counts.


Sound is weird and I agree it is a subjective experience.

I’ve done some unscientific blind tests with nice modern class Ds on people who avowedly think all digital amplification is trash. 9 times out of 10 the reaction has been something like “hmm, how lovely, now when are you going to show me the digital amp?” After I tell them what it is, it becomes “oh yeah, I can hear it now. It’s fine I guess for casual listeners but clearly lacks X.”

People consistently report higher levels of satisfaction with more expensive wine versus cheaper one, even if it’s just the same wine rebottled. They’re not lying, our perceptions and preconceived notions genuinely affect the we experience reality.


I tried many Class-Ds - and all of them caused fatigue. The initially sound even better than AB, but in 15 minutes I get tired.


Fatigue is an interesting factor as there is no physiological bias involved AFAICT - when you can't listen any more you just know, it's almost physical.

I have noticed different headphones and using a class AB headphone amp vs builtin switching amps can significantly impact how long I can listen for... But that anecdotal correlation is not enough for me to assign blame to switching amps, it might just be crap switching amps... or switching amps with not enough headroom. The problem is class D is always used for cheap low power devices, so most bad experiences are inevitably going to strongly correlate with class-D - whether or not it's attributed to the fundamental property of switching vs just being a bit ... shit.


Blinded or unblinded?


Both. I do not believe that non-blind tests are worthless. When you are dealing with subtle effects, knowing what to listen for actually helps in detecting the difference. However, I tried both blind and non-blind. Did not like both times. I really wanted to like Class-D though, it is not like I was prejudiced.


When my amp/receiver bought the farm a decade ago, I decided to try a 50-watt/channel 12-volt class-D amp after seeing the specs. It cost under $50, and I powered it with a hefty wall-wart (120v-1.5amp to 12v) left over from the 90s. It's been on 24/7 since, never used for over 10 watts/ch. The savings went into a decent pair of near-field speakers ... the most important part of great sound.


I recently ventured into the Campervan world. A friend gave me a couple of older nice waterproof Polk audio speakers. Mounted them in the back of the van on a custom rubber shock mount I build from random pieces at the hardware store.

I wanted a way to power the speakers and send music via bluetooth 5. All powered off my 12v solar system. Looked around at all the 'commercial' offerings and found nothing attractive. Everything was over priced or just didn't have modern features.

So... I searched around on Amazon for the right keywords. Ended up getting one of these little class D 100w+100w amps with treble and bass knobs off Amazon for $30 [1]. Got a 12v->24v step up for $16 [2] to give the amp a bit more power and less distortion.

Works amazingly well for a fraction of the price of anything else.

[1] https://www.amazon.com/gp/product/B089KVH9J1

[2] https://www.amazon.com/gp/product/B01EFUHFW6


But it looks like there was a commercial offering. You say it yourself that you could find a solution on amazon. Or I’m missunderstand what the quotes around ‘commercial’ mean?


When I mean 'commercial', I'm talking about something specifically made for car audio by a large name brand manufacturer... something like this...

https://www.amazon.com/Sony-XM-GS6DSP-Channel-Bluetooth-Ampl...

$400 and not even bluetooth 5 (it first came out in 2018 and hasn't been updated since).


Since this isn't explicitly clarified in the article, it's worth mentioning that THD (or THD+N for that matter) is not always a helpful spec to compare two amps. You can easily have two amps where the one with the worse THD is actually the superior one because its distortion is in low-order even harmonics, while the nominally lower THD amp has more distortion in odd or higher-order harmonics. The distribution of distortion over harmonics is as important as the raw sum. (For the layperson who doesn't care about the math: this is why tube amps sound good even when their nominal THD spec may be higher than low-end solid-state amps.)


This is kind of right, but...

Speaker distortion is in the 1% to 5% range.

If your amplifier's distortion is less than 0.1%, you probably won't hear it, whatever it is.

The main reason people like the sound of tube amps (the "warmth") is due to the fact that most of them can't amplify high audio frequencies very well.

For reference, the frequency range of human hearing is taken to be 20 Hz to 20,000 Hz. Hearing is logarithmic in frequency, i.e. the perceived difference in pitch between 100 Hz and 200 Hz is the same as between 10,000 Hz and 20,000 Hz - an octave.

Tube amps have output transformers. It's difficult to design wideband power transformers (heck, designing any transformer is difficult), so most of them have reduced output in the top octave. (AKA "high frequency rolloff".)

One reason for the bad reputation of transistor amps is this: -

Many early transistor amps (and some still today) were prone to oscillation at MHz frequencies under some circumstances. The oscillation itself is inaudible but it affects the linearity control of the amplifier (negative feedback). Amplifiers were oscillating at MHz frequencies for a millisecond or two and losing control, which had a noticeable effect on the sound.

The oscillation wasn't picked up by the designers (of the early transistor amplifiers), because who needs a 100MHz oscilloscope for designing audio circuits that only go up to 20kHz? That's just silly. (Until, of course, it isn't.) And in those days, 100MHz scopes were expensive.

There are other reasons. For example, tube amps withstand some overload; transistor ones don't. Exceed the output transistors' second breakdown rating for a tenth of a second, and suddenly you have no sound. Your new speakers you just brought home from the store killed your amp. "But they're 8 ohms, same as the old ones!" (Another long story there.)

It took a while to figure out all this stuff and how to protect against it, and in the meantime, reputations were made and lost. Tube good, transistor bad.


The article covers some of the alternatives below, so take this rant with a grain of salt.

I don't understand the need to connect a massive linear power supply to a class D power amplifier. That seems completely backwards to me. If you're going class D, then why not eliminate what is often one of the biggest and most obvious expenses and replace the linear power supply with a switching power supply?

Typically in these projects you spend some chunk of the BOM cost on a few nice transistors or an amp chip, another big chunk on the transformer, smaller chunks of money on capacitors, then miscellaneous ICs/amps/hardware/etc. There's some variation here between components, but it just seems bonkers to pay for a big hunk of metal for a 1960s era solution for your power supply, when you are already using super fancy high-speed switching circuits in the amp.

There is no reason that these devices should have any appreciable mass beyond the heat sink. If you've got a massive transformer, you're building an amp like you've traveled in time back to the 1980s, before class D amps existed (well, before they were feasible for these applications). Nothing wrong with that, it's just that the linear power supply adds a lot of cost and mass for no reason.


At my last company, we tested tons of combinations of class D modules with various power supplies, and the best sounding results tended to be with linear power supplies. Don't get me wrong, there's tons of tricky stuff when it comes to the subjective side of audio, so it may be at least partially placebo effect. But we pretty consistently found that off the shelf switching PSUs sounded worse subjectively, even if they measured the same on the bench.

I haven't been doing audio for a while, so I don't remember more of the details. But audio signals tend to need all their power in short bursts all at once, when a low frequency bass note hits for example, and so the transient current tends to be much more important than a stable voltage rail. A lot of switching supplies do not optimize for this. Well designed versions can both sound great, but a cheap linear supply is going to sound much better than a cheap switcher. Audio is stuck with 1960s technology for the most part anyway though.


> so it may be at least partially placebo effect

That really sounds like a kind of thing that should be tested blinded.


We did a lot of blind testing, but blind testing takes more effort and we were always swamped. Usually the engineers would just test things ourselves, and then I would give my boss a blind test for the final approval.


You could add the massive capacitors if you needed the bursts of energy, it just seems like the mains transformer itself is out of place.


>I don't understand the need to connect a massive linear power supply to a class D power amplifier.

Generally from what I understand, linear, unregulated (gasp) supplies are preferred over anything else. I think the reasoning is that the PSRR is great at lower frequencies so having a bit of 120Hz ripple isn't a big deal.

Though this is a bit of antiquated advice, IMO, since you can get amazing performance out of SMPS supplies with a bit of filtering.


Although they are conceptually related, it is harder to engineer switching power supplies than it is to engineer class-D amplifiers, so I could sort of understand a hobbyist drawing up a design that uses 1960s technology to make DC to go in to 1990s technology.


It's hard to engineer a good class D amplifier, but you don't have to do that, you just buy a chip and read the application notes. Same reason why you don't need to engineer an SMPS, you buy a module (if you are a hobbyist) or buy a chip and read the application notes (if you are making a product).

If you are just gluing modules together, then just get the SMPS module and save yourself the weight. If you think you need a linear power supply, then why do you need a linear power supply if you don't need a linear amplifier? Just seems odd.


If you are just gluing modules together... then buy an amplifier. ;)


Because gluing modules together you can spend less money than buying an amplifier with similar specs, and because it's your hobby.

In the past, when I've built amps, it might have cost me something like $300 in materials to make something which would have cost me $1000 to buy. Back when I was a high school student and didn't have much money, that was a good deal.


Perhaps because you'd have to synchronize the switching of the supply with the switching of the amp to prevent aliasing effects.


What I've been looking for is a not-outrageously priced home theatre system that supports all the latest standards, like DTS-X, Dolby Atmos, etc... and has proper, decent speakers, not tiny little tin cans or a soundbar or something.

However, nobody seems to make the "proper" speakers that have the up-firing part needed for Dolby Atmos and are also wireless so I don't have to run cables from the TV to the rear speakers.

The few that I've found so far are crazy expensive.

I think home audio needs an IKEA-type manufacturer to upend the traditional economics of the products.

The expensive part of home theatre audio is that proper wood speakers are big and expensive to ship.

Someone needs to figure out a way to flat-pack full size speakers that consumers can assemble at home, but without requiring soldering irons and power tools.


If you want good value for money, Dayton Audio is your best bet. They used to do some kits, IIRC, but their FBU speakers are reasonably priced for what they are.

For kits there's Madisound (high-end[1]), and in Australasia, The Loudspeaker Kit[2] for more moderately priced kits -- often using Dayton drivers.

You can get cabinet makers (kitchen designers and the like) to make up the boxes locally for you if you can supply them plans in a format they can use.

1. "High end": Buy components for a total of $1k, build your own box to the plans, and get the equivalent of a $10k pair of speakers.

2. http://www.theloudspeakerkit.com/

3. Dayton Audio: https://www.daytonaudio.com/


People had a lot of success buying cheap amps from AliExpress based on a known chipset, like the TPA3250 mentioned in the article, and change the few components that matter.

Usually, it is about changing the cheap knockoff caps by proper audio-grade ones.

Now you are left with the 90% of the work, that is getting a good power supply ;) The easy way is battery power, but then you need to charge the batteries. Linear power supplies are big, expensive and inefficient, but relatively easy to get right.

Switching power supplies are tricky as they tend to introduce all sorts of noise, in fact, they work almost in the same way as the amplifier itself, so you are basically powering an amplifier with an amplifier, both with their own feedback loops.


Counterargument: The power supply is overrated. You want to start with high-power but lousy switching power supply. You want to position it far and shielded from your amplifier. You want a decent ferrite bead on the cable, and a decent bypass filter in the box (both electrolytic and ceramic). The high-frequency stuff should mostly be gone. You can then use a linear regulator and a pretty basic filter to clean up the rest, and introduce an extra volt-or-so of drop on the output.

It sounds like a lot, but it's not.


Yes, I built a few headphone amplifiers using a similar setup: switching power supply - EMI filter - rail splitter - amplifier. No noise, just wonderful sound.


Could you pair a okay-ish power supply with a Jitterbug perhaps?

https://www.headfonia.com/review-audioquest-jitterbug-usb-ma...


I understand the DIY allure, I really do (I do some electronics DIY myself - guitar pedals, tube amps, that kind of thing), but if you want to get great sound and save money at the same time, then just get a used QSC amp from Craigslist for $400 or so. It's better than pretty much anything you'll be able to build on your own for "reasonable" money, and it doesn't look "DIY".


I used to be pretty into audio (I used to have bi-amped Vandersteen 2Cs). I've kind of wanted to build a new setup, speakers following the Tech Ingredients "World's Best Speakers" or "World's Second Best Speakers" at https://www.youtube.com/watch?v=CKIye4RZ-5k&t=18s

I had looked at building this amp to power it, but I realized I really want something that can do HDMI-CEC, where my TV can power the amp on and off and control the volume. You know, how my current Yamaha soundbar works. But I haven't for the life of me been able to find any boards that would allow me to add HDMI-CEC into the project. I did find USB modules that would allow a computer to send CEC commands, but no pre-amp or class-D board that could take CEC input.

Building speakers and a soundbar would be pretty fun though. Been really getting into DIY projects.


Tech Ingredients is in my top three favorite YouTube finds of 2020.


Would something like the SHARC work? https://thenaudio.com/


Looks like it should, it does say numerous times that it allows the TV remote to control volume and mute, but it also says line-level output. I had looked at some other similar ones on Amazon last year that wasn't clear could do volume control, but the SHARC sounds like a good option. Thanks!


Came here looking to see who would post about that YouTube Videos (Tech Ingredients). Those two videos are brilliant !


How good are those DLM speakers? I can't find much about it, but the idea of having very good, completely flat $100 speakers is very appealing.


*DML


I was under impression that class A amplifiers generally have the best amplifier characteristics (linearity, THD), and class D was mainly used in applications preferring efficiency/size/price over audio quality. Apparently class D has gotten lot better for the majority of applications.


Class A architecture's main advantage is it is very easy to achieve high objective performance, while trading away heat and power (and consequently size and mass). You can easily get 3rd harmonic below -80dB and you can eliminate even harmonics entirely. All you have to do it spend a lot on heat sinks and commit to a large power bill.

The problem is going to be reaching really stellar objective performance with the Class A design, because all those space age (not in a good way) power transistors are very slow and hard to drive, and that limits your error correction techniques. That's where these fancy designs are better. Basically they're using lightning-fast transistors and a DSP to bridge a power supply directly into a speaker. Objectively these designs have 3 orders of magnitude less total distortion, unmeasurable 2nd and 3rd harmonics, inaudible (-120dB or better) intermodulation products. The space age designs have simply no chance.


By slow do you mean high capacitance between base and emitter? I’m under the impression that even high power BJTs will be good out to 100 kHz. FETs are even better I believe. Adding more components to the signal path degrades the signal slightly at best and introduces additional non-linearities at worst.


I think signal path degradation is thinking that can only be applied to linear amplifiers, and Class D amps are anything but linear. They literally short the power supply into the load with nanosecond timing. The transfer function through all that lossy, distorting stuff in the output path is pre-computed and equalized out by the DSP.

If I had been more clear about what I was trying to say, I wouldn't have said slow and fast. What I was trying to emphasize was that the linear power amps are optimized around a linear ideal of a transistor that never existed. By contrast, the Class D amplifier exploits the characteristics of the transistors we actually make today: high dv/dt, high di/dt, Rds(on) close to zero. There are fifty companies out there trying to outdo each other on these stats and the closer they get to infinity/zero, the better the output of these new amps becomes. There are fewer people out there any more trying to make transistors more linear in the linear region.


You can always just use a big power op-amp. For some reason, audiophiles object to feedback, but any decent op-amp should give excellent audio performance.

Power consumption and heat dissipation will be pretty bad, though. So you should go with class D.


I tried designing a modest amplifier using op amps for the majority of the frontend.

It's basically impossible unless you add a lot of stuff to buffer the opamp (which is why I wanted to use an op amp instead of discrete stuff)...

big power op amps that have decent distortion exists, but not at their rated power. You get like 0.1% thd at typical levels needed to drive an output stage.

And also thermal distortion will absolutely be an issue at that point.


It's actually class B that has the best performance. A lot of people say class B is bad because crossover distortion, but what they are actually referring to is class C (the classic picture where a bit of the waveform is conducted by the top one and a bit of the waveform is conducted by the bottom one, with some flat spots in between) Class AB is bad because near the crossover, both top and bottom devices conduct and that actually causes an assload of crossover distortion. Properly biased class B precisely hands over conduction from one side to the other, making crossover distortion practically negligible.

Class A is theoretically the best but you can get good performance, and perhaps even better performance than A with a properly designed class B amplifier.


I'd always heard the problem with class AB was less that both devices were on, but that when it's switching off one side near the zero point, that action added noise.

There were a lot of branded tricks in the dying moments of "stereo as a status symbol" (early 1980s) which tried to mitigate this. (Technics New Class A, JVC Super-A, etc.)


think of it this way: if you apply a sinusoid into a power storage, when the device (or the top/bottom device on its own) conducts 100% of the time, it is class A. When it conducts <50% of the time, it is class C. When it conducts >50% of the time (i.e. near the middle, both devices conduct), it is class AB. When it exactly conducts 50% of the time, it is class B.

Turns out at the crossover point, when it's is class AB, you get something called gm doubling where the large signal gain goes up because both devices are conducting and therefore providing gain. When you're in class C (incorrectly called class B by noobs), no device conduct near the middle so you get crossover distortion in the opposite direction. In class B, you theoretically get near-perfect handoff which makes the crossover distortion minimal (at least compared to class A)


This is the terminology Douglas Self uses. No one apart of him call class B class C.


Dude's right though. If you use the naming scheme everyone else uses, there isn't a name for the "class B" biasing that he talks about in that book. It's like a weird in between between class AB and class B.


Yes, I believe the main thing that’s changed over the years is the switching speed. IIRC that’s due to geranium transistors?


Geranium is the plant, germanium is the transistor. But everything is silicon, and has improved for the usual Moore's law reasons.

A class D amp is a closed-loop control system. While it can be done entirely in analogue, having a microprocessor in there lets the designer algorithmically compensate for nonlinearity of the output.


Most commercial switchmode amps are fully analog. The large power amps with built in DSP tend to process the audio signal before it reaches the power amp circuit. Artifacts such as harmonics can be driven down way below any credible threshold of audibility, with conventional techniques.

Especially in big amps, the power amp already has some rather heavy responsibilities, just to be stable when driving odd loads, and generally not blowing up. This tends to discourage putting a lot of weird stuff inside the feedback loop.


Want to go audiophile?

Here's my Pre-Amp (apavel Dudek's PA03) and Power Amp (Nelson Pass DCB1) that I built a couple of years ago.

I even went the extra mile and used discrete OpAmps from Burson Audio.

https://www.bursonaudio.com/pa03-gainclone-power-amp-by-pave...

For those who are not afraid to go down a rabbit hole. -> https://www.diyaudio.com/


microphone -[XLR]-> interface -[USB]-> computer

(I don't know much about this stuff.) I've been wondering why "audio interfaces" are so expensive. Even the cheapest one is $100. Can you explain what makes these expensive? How much would it cost to DIY one of those?

[1] https://www.amazon.com/PreSonus-AudioBox-USB-Audio-Interface...


> I've been wondering why "audio interfaces" are so expensive.

They're a minority interest. Designing good DACs is a subtle art and the engineer costs are amortised over small production runs.

The design of the ancillary circuits (analogue preamps, headphone monitoring, phantom power, input and output protection), and mechanical design and integration are also time-consuming and tricky (=expensive) to make reliable and within design goals. And marketing costs are probably as large as engineering.

You certainly could do it yourself, but the component cost is the least of your worries. You'll need a fair bit of bench gear - oscilloscope, function generator, various power supplies, DVM, ... but even that's not the problem. You need a lot of knowledge about component selection and circuit design and layout, and a lot of experience with the tools of the trade.

Making something like the Presonus (but probably six or ten times as big, because I can't do an all-in-one circuit board with all those functions on it and have it work in any reasonable timeframe), getting it to a state where I would let an unknown mic or computer be plugged into it, or let someone else use it, would easily take me more than three hundred hours. If you're new to this, multiply by n (n >=3).


Two reasons, first, those who want XLR inputs (and outputs in many cases) want quality sounds so they are willing to pay for something a little more high end. If you there are cheap USB audio interfaces that are pretty good, but the next level up is really several levels up in quality (though I'm not sure if you can objectively tell). Second, the volumes sold need high prices to pay for the engineering.


The people who really want XLR are usually commercial, theatrical, etc. In those applications, you have a whole bunch of sources, a whole bunch of sinks, a whole bunch of devices and, critically, a lot of wire. RCA is single-ended, generally not isolated, and, no matter how much you gold-plate it and how much money you throw at silly cables, it will couple to 60 Hz AC, AM radio, and pretty much anything else. The effect in a large system isn’t some mystical loss of “airiness” — its obvious horrible garbage. I have personally listened to AM radio by accident — a theater setup had a little nonlinearity somewhere, and AM radio was being pickup up on a microphone wire, inadvertently demodulated, amplified, and it came out quite clearly from the speakers.


That is a much higher price point. XLR is also used in tiny studios. The guy recording his guitar in the basement doesn't use RCA even though odds are very good none of the problems you cite would happen to him.

Not that you are wrong, but you have moved up several more levels. In price to get all those inputs and long cables...


I think for your application, you'll just need to separate the concerns out a bit

You have the mic preamp, which can definitely be diy'able

Then you have the AD/DA converter - this will convert the output from your diy'ed mic preamp into digital signals - this can't be done easily DIY, but you can purchase chips that does this : the ESS Sabre is one of the most popular AD/DA converters that you can probably source from aliexpress, etc - you might need to buy in bulk though otherwise the vendor might not talk to you.

Then you have the usb interface itself, which comes with all the issues about certification, interface types and properly engineering/building a usb interface that works with consumer ports across different platforms is non-trivial for DIY'ers. You might be able to do something like "make a usb work for my specific computer, with a specific motherboard/platform" but for example, making something that adheres to the usb2.0 or 3.0 spec for both mac + pc + anything else that uses usb (for example, your cell phone, etc) - that's a totally different exercise that most diy'ers won't be able to easily do.

Again, you can buy off-the-shelf chips to do this, but at that level, you're doing digital circuitboard design, and I'm not sure how well bread boarding works with things like usb. You might be able to layout a PCB and have someone etch it for you, etc.

After that you'll need to build the output portions of the interface (after coming out of the DA chip). That can be DIY'ed.

If you want to DIY an in-and-out audio interface you'll need to take on all these different aspects all at the same time.

That's why you don't see many DIY projects of this kind.

Rather, the DIY projects being talked about in this thread are one aspect of this chain: you want to drive some speakers, ok! we can build an amp diy. You want to do AD/DA ok you can do that one thing with a homebuilt board, with a chip, etc.

But having all aspects being integrated together is non-trivial for a DIY. You might be able to use a Raspberry Pi zero or something like that for driving the AD/DA - USB portions, etc and that could be a way to go.

In fact, I think using an RPi as a DAC would be a great way to DIY a usb audio interface, at least you've got all the chips integrated already, and you have an audio jack for getting inputs, etc, and you have the power source and usb OTG port[1] already designed, etc. And that's where you start to see $100 isn't all that bad for a usb audio interface.

[1] https://gist.github.com/gbaman/50b6cca61dd1c3f88f41


This. And if you're talking about audio cards, from a music production perspective, one substandard component in the chain makes all the rest kinda pointless. They all need to be up to snuff.


Another correction: Pre Amp is based on Nelson Pass's design, and Power Amp on Pavel Dudek's


should be: Pavel Dudek

Apologies for the typo


It seems like a lot of money is spent on the power supply. Small Bluetooth speakers draw power from lithium ion batteries which are, in turn, topped up with a cheap external power source. I understand that you can draw very high current from these batteries even if only briefly. Why don’t high end amps use the same trick instead of these giant toroidal coils and smoothing capacitors?


A dark secret of power amp design is that the power supply is half the battle. Note that this design uses a conventional line frequency transformer rather than a switchmode power supply. Making an all-switchmode amp is actually a bit of a challenge. I've been there. My home stereo is a cheap switchmode power amp board, but the power supply is line frequency, as a concession after I discovered that powering it from a switchmode power supply makes it put out all sort of weird noises.

An extremely nice product are the "IcePower" modules made by Bang and Olufsen, which incorporate both power supply and power amp into a single board. So they've worked out the bugs of making the parts play nicely together. These are now the jellybean power amps in smaller musical instrument amplifiers, of which I own a couple.


Because then you have to charge the battery.

Terminal voltage matters too; higher voltage is required to deliver more power. 3.7V into 4 ohm limits you to under a watt. So many systems have a boost converter in there too - which is a potential source of electrical noise.

You could certainly have a stack of 18650 cells do the job if you want a portable boombox...


And, importantly, discharge it. Keeping a battery healthy is a moderately involved task, and the circuitry to do it is another-thing-to-go-wrong that's avoidable complexity with just about any other option. Besides which, no matter what you do the batteries will fail out within a single-digit number of years, even under low usage. It's a good option if you need portability, and pretty terrible by almost any other metric.


I guess because batteries while being great for high current operations, are not the most environmentally friendly solution. My DIY power amp requires a +/- 20-40V power supply. That would be quite a lot of batteries. Toroidal transformers are the next best choice, since they are smaller than comparable regular transformers.

Large caps are not only for smoothing but also act as power bank for instantious large current needs.


Forgive me, I'm not an audiophile, but am curious about it.

What is the constraint(s) that makes good audio equipment expensive?

Is it about components that can handle enough power? Is it avoiding distortion? Or both of the above (or other factors) for the price that components are available on the market?

Looking at some of the photos of components in the amplifier, it looks way dumber and less sophisticated than a Raspberry Pi. Why has the cost of an amplifier not come down to say, $100?

And maybe related, why does a "good" pair of noise cancelling headphones cost $300 and not $50? Is there something related?


I design audio electronics circuits. I think what makes it is mostly:

* high bandwidth.

* over a high dynamic range.

* with a high sensitivity over directionality and pitch changes.

Our ears are exceptional things, and as social animals they are far more important to decoding extremely subtle cues in our surrounding voices and the voice of those talking to us than we probably are aware of.

The thing with audio is, that in the end you still have to move air. And you have to do so without having your vibrating thing breaking after a week. Mechanical reliability is hard while still maintaining these specs. Audio electronics is sensitive to inaccurate power layouts, trace layouts etc.


Some of it is the need for precision. Cheap resistors, capacitors etc. can be outside stuff spec by +/- 10%. Which is fine for all kinds of applications, but audio is really sensitive to that kind of variance, so to get the one or two orders of magnitude more precise components you need for it to sound good, you pay a premium.

At least this is what I learned trying to save money by building my own modular synth stuff. You'll save some money (labour costs), but it's still gonna easily cost hundreds for a handful of components.


First the usual consideration of mechanical construction and production runs; stopping the plastic from rattling and making headphones comfortable adds to the cost.

Then there is bulk; the transformer+caps arrangement is large and heavy, incurring cost all the way along the supply chain.

Then you get into arguments about how much distortion is OK; you can never get to zero but you can asymptotically approach it at increasing cost.

Personally I bought a Sabaj A3 for my TV, which is about $120 and has TOSLINK and Bluetooth. No audiophile will ever use bluetooth, but for playing music in the lounge it's completely fine.


> No audiophile will ever use bluetooth

That's because audiophiles are more about the equipment and less about the actual music.


Well, sure there are audiophiles who are gear fetishists to a stupid extreme. But bluetooth, ugh.. Sketchy, bad latency, audio distortion, weird time jitter. Companies are making things like bluetooth midi controllers, but I can't imagine ever, ever, ever.. bringing bluetooth on stage for a gig.


No, it's because audiophiles know you can't apply digital compression and still retain musical accuracy. The initial bluetooth specs didn't have enough bandwidth for uncompressed audio, so they mandated a non-standard compression.


Audiophiles may know that, but audiophiles are amazing at knowing things that aren’t true. A good modern codec at a moderate rate is indistinguishable from uncompressed audio. Bluetooth, sadly, isn’t one of these.


I work in film post production/audio as a freelance mixing engineer. While I have a certain amazement at stuff audiophiles will claim to hear, hearing the difference between compressed and uncompressed is certainly not impossible, but it depends on the signal you use to compare.

I am a cables person, but for me this is more about reliability and ease of use than sound. When I disconnect the cable stuff it ends. When I connect it and I don't use cheap or broken cables they just work. And they also work with somebody elses gear.

There is a reason professional wireless receiver/transmitter bodypacks cost upwards 500€ per pair.


I used to be able to reliably distinguish 128kbps MP3s encoded with some mediocre encoder from the original audio (blinded, but I never tried double-blinded). I doubt I could pull off the same stunt with a top-of-the-line encoder, let alone with a better codec at an appropriate bit rate. A good codec can be made “transparent” such that no one call hear the difference except perhaps with deliberately chosen music that abuses that codec.

Now if only the world could settle on actually using good codecs...


I agree. If you use OPUS for example, it becomes nearly impossible to tell.

But one thing we should not forget: these codecs are used for distribution or streaming. If you ever record something that has to be edited, filtered, treated with an EQ, denoised etc. then go for at least 24 Bit Wav. Compressed audio is great for when the thing already sounds the way it should, but it starts to fall apart very quickly when you manipulate it heavily.


No, it's because audiophiles know that bluetooth is terrible.



For a few dollars more there is Nelson Pass's(firstwatt) Do-It-Yourself amp camp. Very nice analog amp. https://diyaudiostore.com/products/amp-camp-amp-kit?variant=...


If you're the kind of person who equates harmonic distortion with "warmth" you're in for a big treat with these Nelson Pass designs. On objective measurements these are just about the worst amps you can buy (or build). We're talking about 2nd harmonics at only -50dB. Warm AF.


Bias current of 1.45A (not ma!) means it'll be pretty warm to the touch too.

The only advantage is simplicity for teaching purposes. Doesn't get simpler than class A.


This is just a quibble, but most switchmode amplifier chips are analog. Granted the term "digital" has been adopted by the audio community for switchmode amplifiers, but it still raises eyebrows.


On a side note has anyone tried the Devialet Phantom speaker system?

https://www.devialet.com/en-eu/phantom-speaker/phantom-premi...


Class D amps are pretty amazing. the Fender class D bass amps are ridiculously good and light for the price now. like 20% of the weight of their predecessors light. Which, if you've lugged a tube 4x10 around, is pretty awesome.


Can anyone recommend a DIY speaker in the same vein as this amp? (Aside from the panel speaker linked elsewhere in these comments)


I’d be interested in this too but get lost down the rabbit hole every time I try to research speakers. I think the problem is an ideal amplifier is easily described as “a straight wire with gain” while a practical speaker is a much more complicated problem with many tradeoffs based on physical properties of vibrating materials and distances etc.


I tried making my own but it's like 99% woodworking and 1% actual audio engineering (i.e. doing the whole Tiele-Small stuff). Be sure you know what you're getting into haha


A few years ago I built the lx-mini speaker system and it was a fun project. Foe details see Linkwitzlab.com


At 84€ just for the plans I will give this a miss, but if you happen to have images and other opinions of the ones you made yourself I'd be very curious

http://www.linkwitzlab.com/LXmini/Introduction.htm


Pardon my newbeeness here, but is it possible to build a home cinema amp (say 5.1 hdmi) in the same spirit ?


It's effectively impossible to DIY anything with HDMI but if you can get an adequate 6-channel analog output from an HDMI line-level converter then at that point all you need is three 2-channel amps or design one with all 6 channels in one box.


Agreed. Aliexpress has some digital-to-6-channel decoders that work pretty well (and even support the more advanced codecs that are usually licensed by regular amps). Some of them even run Linux under the hood. :)

I would suggest using TOSLINK rather than HDMI as you can almost always extract TOSLINK from HDMI. Technically HDMI can carry more audio bandwidth, but I'm not sure where that might get used (if at all).

EDIT: I haven't specifically tested this one, but it's similar to one I've played with: https://www.aliexpress.com/item/1005001491403444.html?spm=a2...


Yeah, but I can't really recommend building D/A converters any more, except as a learning exercise. Even the cheapest off-the-shelf gear now has performance that will be difficult to beat for the DIYer.

This is somewhat disappointing for me as I have built many D/A converters over the years.


I think we're on the same page here. Plus with the advanced digital codecs, you should probably just outsource it to the all-in-one chip packages.


HDMI audio is a huge pain. Optical Toslink should be doable; you might be able to find the right board. But there's a lot more involved in doing the digital 5.1 decode and having more channels.


Summarizing what jeffbee says in sibling comments: -

Yes, it is certainly possible. But you need to acquire the prerequisite seven to ten years' experience first for a project of that complexity. When you've got that experience...you'll just buy one.


What about non-audiophile? Are there any similar resources for professional-grade guitar amps?


Take a look at Rod Elliott's web site, sound-au.com.[1]

Rod has many years' pro audio experience and his web site is a treasure trove of audio DIY information. He has at least one guitar amp project, and several articles about design traps and pitfalls he has seen... and others about how to do it right.

1. https://sound-au.com/index.html


Back in 1980 my Science Fair project was a class-D stereo modulation system.... ah, memories.


I wish there was something free/oss from NwAvGuy - i love his Objective2...


How do class D amps sound? In my limited amplifier build history, tubes still win over transistor based amps every time. Mosfet amps sound great in the mids and highs and bipolar sounds great in the lows, but neither of them beat nice tube designs.


An amplifier shouldn't "sound" like anything - the best possible amplifier would be a wire with gain. Tube amps do have their own sound, primarily due to higher harmonic distortion and the use of output transformers in the signal pasth. Whether this "sound" is agreeable to you varies by the individual, but there is certainly no universal agreement that a tube amplifier sounds "best".


I have built multiple amps of different types and topologies. Each one of them 'sounds' completely different. This is the reality of amplification. There is no perfect distortionless amp at all frequencies for all speakers. The interactions between the circuit and the impedance presented across all audible frequencies by any random speaker are infinitely variable. The sound of an amp is the combination of all factors in the system, which includes physical interactions and their response to the variations in presented impedance.

So yes, an amplifier shouldn't sound like anything, but they always do in reality. Within that reality, we can make judgments about the general qualities of certain types of amplifiers. If you want to comment on my question please do. Otherwise, kindly go away, and keep your comments about your magical fairy tail land where all amplifiers are perfect to yourself.


What sounds neutral also seems to be highly subjective.

I think "colour" of sound is a fitting analogy though - in the same way that our can eyes adjust to color temperature within a certain range i've noticed on the scale of hours/days I can become accustomed to speakers/amps with quite different frequency responses.


Silly misread—-deleted.


Class A amplifiers are "high-end, audiophile quality", class D is about as low as you can go


Sorry, this is completely incorrect. A, B, A/B, and D all refer to the amplifier topology, not to the quality.


Sorry you are incorrect, Only in Class A is the whole of the signal contained in the operating area of the transistor, thus only class A amplifiers transfer the whole of the original signal on to each successive stage.

Class B C and D chop more and more out of the crossover point out of the signal as it passes through each stage.


That used to be true, but it seems class D has got quite good. This prompted me to look up the old LM12 but I cant find it anywhere but Ebay. Must be out of production.


There's plenty of extremely high quality class D amps around now that exceed the performance of Class A, e.g. the Benchmark AHB2 ( https://www.audiosciencereview.com/forum/index.php?threads/r... )




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: