Guitar/Audio amp poor quality

[Bren]

Guess the class D amplifiers only produce noise than, since it's basically a switching power supply as amplifier.

There's a major difference between power amplifier, and power supply.

The class-D amplifier your speaking of is another beast entirely, it works on a completely different principle than the amplifier your trying to build and requires much more intricate circuit design. If you'd like to learn how class-D amplifiers function, there are plenty of documents online, just do a simple google search.

If not, then in the simplest terms, Class-D amplifiers use a type of Pulse Width Modulation switching as an input source to the power amplifier stage, the output of the power amplifier is then filtered to "reconstruct" the reference waveform before being sent to the load.

This doesn't mean that they are sourced by non-rectified and unfiltered voltage. Class-D amplifiers require bus voltage filtering as much as any audio amplifier.
Also, keep in mind that the gain in a class-d amplifiers stage is directly proportional to the bus voltage. Swings in bus voltage due to excessive power supply pumping causes distortion on a class-D output at lower frequencies. There are many different causes of noise in class-D amplifiers, power supply voltage fluctuations due to output impedance and reactive power flowing through the DC bus, is only one source of this noise.

You are building a Hi-Fi amplifier? Regardless, Noise is Noise, although it can't be completely eliminated, reducing it, especially in a Hi-Fi audio amplifier is usually a good idea. In fact it's a good idea to reduce noise sources in any noise sensitive system, but I guess that goes without saying.

My general advice for someone who is grabbing any old schematic online and trying to reproduce it, is to stay away from unregulated unfiltered switching power supplies unless you already know that the designer took the necessarily steps for the supply, or you really know what your doing.

Many audio amplifiers don't even have a regulated power supply, they often run directly off a rectified power supply.

As an audio/electronics engineer this makes me uneasy,

In real circuits it is not too hard to get ripple levels that contribute inaudibly to the output noise.

This is why.

Even if the ripple itself is not heard through the output, it can contribute to output noise.

Also anyone who even spends $500 on a preamplifier is an audiophool, let alone $2000 which should be enough for a decent hifi with good speakers and more.

True for the average home user. I design pre-amplifiers for recording studio use, part of my design mythology is to reduce or eliminate noise at any possible source, leaving the cleanest sound with slight coloration from the electronic components.
Its a PITA to try an remix a track with noise, or poor audio coloration that shouldn't even be there all due to a badly designed, mediocre pre-amp.
So as you can see, my application is different.

[alm]

There's a major difference between power amplifier, and power supply.

The difference is fairly minor in my opinion. A power supply is basically a power amplifier with an internal reference connected to the input. Of course the implementation and application is quite different.

If not, then in the simplest terms, Class-D amplifiers use a type of Pulse Width Modulation switching as an input source to the power amplifier stage, the output of the power amplifier is then filtered to "reconstruct" the reference waveform before being sent to the load.

Which is similar to what an SMPS does. With proper filtering (actually easier since you can roll-off far below audio frequencies), you should get the same noise performance. Except that a class D amplifier doesn't have PSRR of the amplifier powered by the SMPS to reject some noise.

This doesn't mean that they are sourced by non-rectified and unfiltered voltage. Class-D amplifiers require bus voltage filtering as much as any audio amplifier.

I'm not suggesting an unregulated power supply, just that it's usually not necessary to go overboard thanks to the PSRR of the amplifier.

In fact it's a good idea to reduce noise sources in any noise sensitive system, but I guess that goes without saying.

Sure, less noise is better. It only makes sense though if it's actually the dominant noise source.

As an audio/electronics engineer this makes me uneasy,

For power amplifiers, building a linearly regulated power supply requires huge heat sinks. And power amplifiers (apart from the class D and similar types) aren't the most power efficient to begin with due to biasing. So the other option would be a switcher I guess .

True for the average home user. I design pre-amplifiers for recording studio use, part of my design mythology is to reduce or eliminate noise at any possible source, leaving the cleanest sound with slight coloration from the electronic components.

Granted, studio use requires higher standards, especially since the signal tends to go through multiple pieces of equipment before it reaches it destination (although much of this is probably in digital form these days). However, even when designing for low noise, it makes sense to determine the dominant noise source and go from there, as opposed to blindly throwing money improving a power supply.

[Bren]

Of course the implementation and application is quite different.

Well, at least you understood what I was referring to.

Which is similar to what an SMPS does. With proper filtering (actually easier since you can roll-off far below audio frequencies), you should get the same noise performance. Except that a class D amplifier doesn't have PSRR of the amplifier powered by the SMPS to reject some noise.

Similar, but as you stated before,

Of course the implementation and application is quite different.

And when the implementation and application is different......?
You usually take different precautions in the design process. 
It's like taking a jet engine and putting it on a horse and expecting the horse to run faster.

I'm not suggesting an unregulated power supply, just that it's usually not necessary to go overboard thanks to the PSRR of the amplifier.

True, amplifiers nowadays have great PSRR. In audio applications though (at least where I use them), its important to keep a few things in mine, one being circuit layout, and clean voltage supply. And why not? It seems like lazy engineering practice, as if to leave something half done and have an attitude of "well... it works good enough", especially when there's a possibility for poor circuit operation because of the half-assed design.
Then again you have the usability factor. If your design works good enough, and it meets specs, its what customers want, then consider it done!

Sure, less noise is better. It only makes sense though if it's actually the dominant noise source.

Depends on how sensitive to the noise the system is, where its being applied or used, and the requirements of the system. I wouldn't want to use an amplifier that has a 1.7Vp-p output and 30% of that signal is effected by noise generated equally in various places in the system. Sure the noise doesnt have a dominant source, but its still going to sound like crap.

For power amplifiers, building a linearly regulated power supply requires huge heat sinks. And power amplifiers (apart from the class D and similar types) aren't the most power efficient to begin with due to biasing.

Once again, usability. I keep bringing up this usability factor, but the truth is, it's what separates good engineers from bad engineers. Good engineers keep this in mind when they design products, they think about the system in terms of its electrical performance, but they also have to consider the user.
Sometimes sacrifices have to be made in electrical performance (efficiency and size being two examples) to accommodate the end user. Switchers definitely not the only option out there, some googling would open you up to a whole bunch of possibilities or even improvements to existing designs, try it, you'll be amazed.

blindly throwing money improving a power supply.

This is called blind engineering, or brute force engineering. It's a very unwise practice, no self respecting engineer would do this. When I design, I start with dominant noise sources around the system and work down from there until noise levels are at acceptable values.
I don't follow this half-assed "well it's good enough attitude" either, engineers that "do whatever it takes to get the job done" just aren't marketable in today's economy. Sometimes this requires a little extra time spent on a project, you'll see that it'll get you a long way in the industry.

[Zero999]

I don't follow this half-assed "well it's good enough attitude" either, engineers that "do whatever it takes to get the job done" just aren't marketable in today's economy. Sometimes this requires a little extra time spent on a project, you'll see that it'll get you a long way in the industry.

What I don't understand is the let's over-engineer audio amplifiers to give near zero noise, THD, a huge bandwidth etc. attitude of audiophoolery.

Nowhere else in engineering do I see people making things 10 times as expensive as they need to be for improvements which won't even be noticed unless state of the art test equipment is used. I mean why bother designing a power amplifier with a TDH of say 0.0001% when there will be more distortion in the best of the best speakers and room acoustics?

Nowadays it's possible to produce a solid state amplifier which is much better than good enough for audio fairly cheaply. I don't understand why people bother spending much more, even for studio usage.

[Bren]

Well to be honest, $500 for a studio grade pre-amp is already near the lower end of the spectrum as far as price goes. It's not all that uncommon to have a studio pre-amp around $2000. These are just standard industry prices. Expensive, maybe. But there is a lot more than just low noise, bandwidth and THD to consider when designing a pre-amp for studio use. Right down the the type of transformers used.

For example, mediocre transformers can round off the waveforms, introduce phase shift, saturate at low levels, and color the sound with an uneven frequency response.
The cost builds up in the components used and the construction. Labour costs are high, and the cost of parts are high. Then the question is for active components that do color the sound, what to use? Tubes, most often, or solid state designs? Passive components can also have a large impact on sonic quality, capacitors having the most effect. The difference between electrolytics, with their high dielectric absorption, poor temperature stability, wide tolerance and short life span, or film capacitors with dramatically less dielectric absorption than even the best electrolytic capacitors, which results in better low frequency realism, transient performance, and enhanced spatial accuracy.

It really has nothing to do with "over engineering". It has to do with the following:

The audio engineers of today have one thing in mind; that is, to take down the truest sound. Being able to do this allows them to do their job easier. Coupled with this are extra features of the pre-amp that the engineer might need. All of these things are taken into consideration when picking a studio grade pre-amplifier.

The same could be said about electrical engineers. "who would spend 100X the cost on an oscilloscope from Agilent or Tektronix when someone can pick up a no name Chinese USB oscilloscope that's good enough to take measurements"... Well, the engineer will need the features and the more accurate measurements of a more expensive scope. This will allow them to do their job easier.

It's really no different.

[ejeffrey]

There's a major difference between power amplifier, and power supply.

The difference is fairly minor in my opinion. A power supply is basically a power amplifier with an internal reference connected to the input. Of course the implementation and application is quite different.

Actually they are quite different even in principle.  Almost all power supplies have unipolar outputs and can only source current, while an amplifier must be able to sink as well.  It is the difference between a 4-quadrant supply and a 1-quadrant supply.  There do exist 2- and 4- quadrant power supplies for driving electromagnets and the like, and their internal topology is much like a power amplifier.  However, your standard bench supply is not.  Of course in the details almost all of their design characteristics are different, for instance power supplies rely heavily on filter to eliminate ripple and noise while amplifiers must do it all via negative feedback.

Also, keep in mind that the gain in a class-d amplifiers stage is directly proportional to the bus voltage. Swings in bus voltage due to excessive power supply pumping causes distortion on a class-D output at lower frequencies. There are many different causes of

Only the cheapest and nastiest class-D amplifiers use open-loop PWM.   Anything worth using has the gain set by feedback resistors like any other amplifier.

What I don't understand is the let's over-engineer audio amplifiers to give near zero noise, THD, a huge bandwidth etc. attitude of audiophoolery.

It is not even that.  It is easy to design a power amplifier with near zero THD+noise.  Any audio amplifier should have THD+noise far below audible levels not because it is necessary, but because it costs next to nothing and is simply a sign of lazy or incorrect engineering to make something that has sub-par performance in this respect.  The same is true of bandwidth to a limited extent.  Large full-power bandwidth is in a high power amplifier gets to be expensive.  However, an audio amplifier should *not* have excessively high bandwidth -- any ultrasonic noise on the input should be filtered out rather than amplified and provided to the speakers.  The high frequency roll-off should be set by component values not parasitics, and the amplifier should have plenty of open-loop small-signal bandwidth to make sure this happens.  Again, it is not expensive or difficult to build an amplifier with enough open-loop bandwidth to satisfy this.  The cost of an amplifier is almost completely dominated by the power rating.  Once you account for the power supply, heat sink, and output transistors, making the rest of amplifier essentially perfect is practically free.

The problem with most 'audiophile' stuff is that often it is not better at all, even inaudibly, than a less exotic amplifier.  It is just random gold plating and expensive changes that actually provide no measurable benefit to output performance at all.  Frequently they are even measurably worse due to 'interesting' design choices that both raise the price while lowering the quality.

[Bren]

Only the cheapest and nastiest class-D amplifiers use open-loop PWM.   Anything worth using has the gain set by feedback resistors like any other amplifier.

Haha, yeah I could have mentioned that.

The cost of an amplifier is almost completely dominated by the power rating.

This is definitely true in power amplifiers, I cant say the same applies for pre-amps.

The problem with most 'audiophile' stuff is that often it is not better at all, even inaudibly, than a less exotic amplifier.  It is just random gold plating and expensive changes that actually provide no measurable benefit to output performance at all.

 

True for all devices.

Frequently they are even measurably worse due to 'interesting' design choices that both raise the price while lowering the quality.

Also true for all devices. Lets just compare Tektronix 2000 series scopes with their blazing fast waveform refresh rate, and their deep 2.5k memory (... sarcasm) and Agilent DSOX2000 scopes. Agilent has the obvious upper hand, but Tektronix scopes are still selling for twice the cost. I know what you're gonna say; the Tek2000 series scopes are a little dated (which means they should probably be dropping the cost on them a couple of grand). Anything by Tek now that is comparable to a Agilent DSOX is nearly twice the cost.
I'll be interested to see if the amazing new Tektronix revolution is comparable, at the same costs.

People just have to be careful with the things that they purchase. They have to do the research. It's like buying a luxury car and realizing it just doesn't perform like that ford tempo you had when you were a teen.

[IanB]

Nowhere else in engineering do I see people making things 10 times as expensive as they need to be for improvements which won't even be noticed unless state of the art test equipment is used.

I think it's worth noting that good engineers in all branches of industry will start by laying out a design using good design principles that is as near to ideal (or perfect) as possible. Thinking about the problem correctly in this way is pure professionalism and doing a job right. Once the ideal design is sketched out, the good engineer will go on to make considered simplifications for reasons of cost or practicality, in each case considering the impact on the design using appropriate cost/benefit criteria. The outcome should then be a practical and economic design fit that is fit for purpose, based on sound underlying principles.

The bad "engineer" will merely bodge something together using woolly thinking and random cheap components and hope for the best.

[alm]

Actually they are quite different even in principle.  Almost all power supplies have unipolar outputs and can only source current, while an amplifier must be able to sink as well.  It is the difference between a 4-quadrant supply and a 1-quadrant supply.  There do exist 2- and 4- quadrant power supplies for driving electromagnets and the like, and their internal topology is much like a power amplifier.  However, your standard bench supply is not.  Of course in the details almost all of their design characteristics are different, for instance power supplies rely heavily on filter to eliminate ripple and noise while amplifiers must do it all via negative feedback.

You're correct, but from a noise point of view, a power supply shouldn't be worse than an amplifier, since it is easier to design as you point out. A good amplifier should make a fairly decent power supply (though stability will likely be worse). A power supply for amplifying audio... not so much. Although I'm sure someone built an amplifier from something like an LM317.

[Zero999]

For example, mediocre transformers can round off the waveforms, introduce phase shift, saturate at low levels, and color the sound with an uneven frequency response.

What do you mean by colour the sound? Is it just a fancy way of saying distort the sound?

And why on earth would you put the analogue signal through a transformer? That's a very bad idea and should be avoided where ever possible. There are normally better ways of avoiding ground loops or just galvinically isolating a signal than a horrible transformer.

The cost builds up in the components used and the construction. Labour costs are high, and the cost of parts are high.

What labour costs?

What component costs?

An excellent quality amplifier can be made from inexpensive op-amp ICs and assembled by machines.

Then the question is for active components that do color the sound, what to use? Tubes, most often, or solid state designs?

Valves/tubes are just shit and a complete waste of money and solid state op-amps hardly colour the sound at all, assuming that means causing all sorts of distortion.

Here are a couple of fine op-amps which are more than good enough for audio and are relatively inexpensive:
AD797
LM4562

Passive components can also have a large impact on sonic quality, capacitors having the most effect. The difference between electrolytics, with their high dielectric absorption, poor temperature stability, wide tolerance and short life span, or film capacitors with dramatically less dielectric absorption than even the best electrolytic capacitors, which results in better low frequency realism, transient performance, and enhanced spatial accuracy.

I understand that paragraph until you mentioned enhanced special accuracy?

Yes electrolytic capacitors are not good for distortion and nether are ceramic which can be microphonic and cause oscillation if used in a preamplifier. Decent film capacitors are not that expensive though.

Another thing I've never understood is why some audiophools swear by horrible carbon composition resistors when thin metal film resistors are much cheaper, less noisy and are also non-inductive.

It really has nothing to do with "over engineering". It has to do with the following:

The audio engineers of today have one thing in mind; that is, to take down the truest sound. Being able to do this allows them to do their job easier. Coupled with this are extra features of the pre-amp that the engineer might need. All of these things are taken into consideration when picking a studio grade pre-amplifier.

What extra features should a studio pre-amplifier have and why do they cost the earth?

The same could be said about electrical engineers. "who would spend 100X the cost on an oscilloscope from Agilent or Tektronix when someone can pick up a no name Chinese USB oscilloscope that's good enough to take measurements"... Well, the engineer will need the features and the more accurate measurements of a more expensive scope. This will allow them to do their job easier.

It's really no different.

I disagree.

There's always a need for a faster oscilloscope with more memory, channels higher precision, faster update rate etc. but why on earth would anyone want to spend ten times as much on an audio amplifier because it has a THD of 0.0001% rather than a cheaper one with a THD of 0.0005%? It's not like they can actually tell the different without using expensive test equipment.

[DrGeoff]

For example, mediocre transformers can round off the waveforms, introduce phase shift, saturate at low levels, and color the sound with an uneven frequency response.

What do you mean by colour the sound? Is it just a fancy way of saying distort the sound?

And why on earth would you put the analogue signal through a transformer? That's a very bad idea and should be avoided where ever possible. There are normally better ways of avoiding ground loops or just galvinically isolating a signal than a horrible transformer.

Analogue signals are often put through transformers in many professional situations. Look up 1176 compressor, one of the most widely used in the industry. It has an output transformer. It also has options for an input transformer, or op-amp input stage. Passive DI boxes use a transformer. Neve consoles are full of them on the mic input stages.

Transformers are used for impedance matching and balancing. Particularly useful for high impedance, low output sources (such as guitar pickups) where a 1:5 transformer can be used to match it to a lower impedance balanced input stage. A good audio transformer is not cheap though, often $80-$120 for a mic input or line output transformer with good specs. It will not add much colour to the sound until it is driven hard, acting a bit like a limiter when the transfer curve becomes non-linear.

[Bren]

OK, OK, I agree with you in the fact that it's stupid for someone to pay $200 for cables, these are as you say, "audiophools", through and through, no doubt. However, I'm not disagreeing with you that's there is no such thing as "audiophools". These people are just the type that don't know enough about the things they buy, and pay a premium for them. Is that to say that all things that are $500 or more are a waste of money? Of course it doesn't.

Valves/tubes are just shit and a complete waste of money and solid state op-amps hardly colour the sound at all, assuming that means causing all sorts of distortion.

Comments like this just make me roll on the floor with laughter.

I've seen this being argued by so many closed minded people time and time again. I definitely wont go into a debate about solid-state vs tubes, each are just as desirable as the the next, each have their place. I like to use both for different situations, sometimes a mixture of the two. If someone were to debate about that though, they should open a new thread, because guaranteed it'll be close to a billion pages long in less than a week. 

I still find it funny and somehow amazing how typical it is of people with limited knowledge in the audio industry to so quickly bash "audiophools" (sarcasm- by that I mean people who pay a premium and know exactly what their getting). I could go as far as to say that the majority of them also lack the understanding as to how electronic components change or color the sound of an audio signal, and whether or not that color(sound) is desirable.

I do find all theses questions somewhat humorous (not insultingly so). I mean, I definitely don't have the motivation to answer all of them because it quite frankly has just become ridiculous. If you feel so inclined you can find many of the answers to the questions you've asked in online documents and journals, you can also ask audio/video recording artists, producers, and musicians. The answers you get should also clear up a lot of these misconceptions. 

Look up 1176 compressor, one of the most widely used in the industry.

One of my all time favs, another one up there with it would be the LA-2A.

[DrGeoff]

It really has nothing to do with "over engineering". It has to do with the following:
The audio engineers of today have one thing in mind; that is, to take down the truest sound. Being able to do this allows them to do their job easier. Coupled with this are extra features of the pre-amp that the engineer might need. All of these things are taken into consideration when picking a studio grade pre-amplifier.

What extra features should a studio pre-amplifier have and why do they cost the earth?

Fully balanced inputs and outputs.
High quality XLR connectors.
+24dBm input handling capabilty.
Metering.
For microphone pre-amplifiers:
  +48 phantom power.
  Up to 70dB gain with very low noise.
  Usually a transformer input stage for good balancing and impedance matching.

[Zero999]

Transformers are used for impedance matching and balancing. Particularly useful for high impedance, low output sources (such as guitar pickups) where a 1:5 transformer can be used to match it to a lower impedance balanced input stage.

Surely a better engineering solution would be to integrate a tiny pre-amplifier into the pick-up?

Comments like this just make me roll on the floor with laughter.

I've seen this being argued by so many closed minded people time and time again. I definitely wont go into a debate about solid-state vs tubes, each are just as desirable as the the next, each have their place. I like to use both for different situations, sometimes a mixture of the two. If someone were to debate about that though, they should open a new thread, because guaranteed it'll be close to a billion pages long in less than a week. 

I still find it funny and somehow amazing how typical it is of people with limited knowledge in the audio industry to so quickly bash "audiophools" (sarcasm- by that I mean people who pay a premium and know exactly what their getting). I could go as far as to say that the majority of them also lack the understanding as to how electronic components change or color the sound of an audio signal, and whether or not that color(sound) is desirable.

I do find all theses questions somewhat humorous (not insultingly so). I mean, I definitely don't have the motivation to answer all of them because it quite frankly has just become ridiculous. If you feel so inclined you can find many of the answers to the questions you've asked in online documents and journals, you can also ask audio/video recording artists, producers, and musicians. The answers you get should also clear up a lot of these misconceptions. 

All right I'll do some Googling.

There's one question I have which I can't find an answer to: why are such ridiculously high bit rates mandated by some for digital sound?

I mean why sample at 24-bits per sample at 192kHz? It's way, way over the top and completely pointless. Fair enough, maybe it's a good idea to do some digital signal processing with a much greater bandwidth and sample rate to avoid rounding errors but there's no point in using such an ADC. You may think I'm crazy but it's common sense when thermal noise is taken into account.

For example, take a dynamic microphone with an impedance of 300R, at 300^oK the equivalent thermal noise voltage is 997nV_RMS over the audio bandwidth of 20KHz. Suppose the microphone is outputting a signal of 2mV, at 24-bits per sample the quantitisation level is 119.2*10^-3V which is way into the noise margin so totally overkill.

[ejeffrey]

Transformers are used for impedance matching and balancing. Particularly useful for high impedance, low output sources (such as guitar pickups) where a 1:5 transformer can be used to match it to a lower impedance balanced input stage.

Surely a better engineering solution would be to integrate a tiny pre-amplifier into the pick-up?

Powered by what?  This is common on microphones and the power is provided by the mixer via phantom power... which uses transformer coupling.

Transformers provide second-to-none impedance balance and common mode rejection while avoiding high (differential) input impedance and the associated noise and ESD risk.  Their main drawback is that they are expensive

There's one question I have which I can't find an answer to: why are such ridiculously high bit rates mandated by some for digital sound?

I mean why sample at 24-bits per sample at 192kHz? It's way, way over the top and completely pointless. Fair enough, maybe it's a good idea to do some digital signal processing with a much greater bandwidth and sample rate to avoid rounding errors but there's no point in using such an ADC. You may think I'm crazy but it's common sense when thermal noise is taken into account.

Mostly they are marketing bits.  A high quality 24-bit audio ADC has somewhere in the neighborhood of 18 real bits and 6 marketing bits.  However, once you have a sigma-delta ADC it costs you almost nothing to round up to 24 bits.  They might even help a bit...  Sigma-delta converters actually require a fair bit of dither to perform well, and by making the LSB deep below the noise floor you ensure that you always have enough dither to avoid quantization artifacts.  The ADC could keep track of a 24 bit word for its internal FIR and only report the significant bits, but in reality they are going to need to be rounded to 16 or 24 for storage in a computer and 24 gives you a bit more leeway in your recording gain as well as some handy dynamic range for signal processing.

The bit rate is a bit more puzzling.  In the bad old days you would win by increasing the sample rate from 48 kHz to 96 or higher because you relaxed the requirements on the analog reconstruction filter.  Even if your source was 16 bit / 48 kHz it might be advantageous to upsample to 96 kHz, apply a digital reconstruction filter to eliminate aliasing, and then use a lower order or lower Q analog filter.  Now that sigma-delta DACs are universal, this is no longer necessary: they do this implicitly.  They use digital noise shaping to shove all the noise up to the MHz region where it is trivial to filter.

That said, it isn't expensive to do, and if you aren't starved for bandwidth or storage space, why not?  Every audio CODEC I have seen allows you to select the bit rate, so you can run it at 48 kHz if you are pressed for bandwidth.

[Zero999]

Transformers are used for impedance matching and balancing. Particularly useful for high impedance, low output sources (such as guitar pickups) where a 1:5 transformer can be used to match it to a lower impedance balanced input stage.

Surely a better engineering solution would be to integrate a tiny pre-amplifier into the pick-up?

Powered by what?

Batteries, a small power jack from the PA or mixer.

This is common on microphones and the power is provided by the mixer via phantom power... which uses transformer coupling.

Transformers provide second-to-none impedance balance and common mode rejection while avoiding high (differential) input impedance and the associated noise and ESD risk.  Their main drawback is that they are expensive

I suppose my point is if the signal is amplified as close to the source as possible then there should be no need for balancing or impedance matching because the signal is now large enough that hum shouldn't be a problem.

There's one question I have which I can't find an answer to: why are such ridiculously high bit rates mandated by some for digital sound?

I mean why sample at 24-bits per sample at 192kHz? It's way, way over the top and completely pointless. Fair enough, maybe it's a good idea to do some digital signal processing with a much greater bandwidth and sample rate to avoid rounding errors but there's no point in using such an ADC. You may think I'm crazy but it's common sense when thermal noise is taken into account.

Mostly they are marketing bits.  A high quality 24-bit audio ADC has somewhere in the neighborhood of 18 real bits and 6 marketing bits.  However, once you have a sigma-delta ADC it costs you almost nothing to round up to 24 bits.  They might even help a bit...  Sigma-delta converters actually require a fair bit of dither to perform well, and by making the LSB deep below the noise floor you ensure that you always have enough dither to avoid quantization artifacts.  The ADC could keep track of a 24 bit word for its internal FIR and only report the significant bits, but in reality they are going to need to be rounded to 16 or 24 for storage in a computer and 24 gives you a bit more leeway in your recording gain as well as some handy dynamic range for signal processing.

The bit rate is a bit more puzzling.  In the bad old days you would win by increasing the sample rate from 48 kHz to 96 or higher because you relaxed the requirements on the analog reconstruction filter.  Even if your source was 16 bit / 48 kHz it might be advantageous to upsample to 96 kHz, apply a digital reconstruction filter to eliminate aliasing, and then use a lower order or lower Q analog filter.  Now that sigma-delta DACs are universal, this is no longer necessary: they do this implicitly.  They use digital noise shaping to shove all the noise up to the MHz region where it is trivial to filter.

That said, it isn't expensive to do, and if you aren't starved for bandwidth or storage space, why not?  Every audio CODEC I have seen allows you to select the bit rate, so you can run it at 48 kHz if you are pressed for bandwidth.

Thanks, that makes sense.

[DrGeoff]

Transformers are used for impedance matching and balancing. Particularly useful for high impedance, low output sources (such as guitar pickups) where a 1:5 transformer can be used to match it to a lower impedance balanced input stage.

Surely a better engineering solution would be to integrate a tiny pre-amplifier into the pick-up?

Maybe it is, and there are some that have active pickups. However the majority don't and, way back in the days when electric guitars and basses were invented, there wasn't any way to do it that would be able to be fitted into the instrument. As already pointed out, condenser microphones have a preamp in them, powered by a 48V phantom power supply.

There's one question I have which I can't find an answer to: why are such ridiculously high bit rates mandated by some for digital sound?

I mean why sample at 24-bits per sample at 192kHz? It's way, way over the top and completely pointless. Fair enough, maybe it's a good idea to do some digital signal processing with a much greater bandwidth and sample rate to avoid rounding errors but there's no point in using such an ADC. You may think I'm crazy but it's common sense when thermal noise is taken into account.

For example, take a dynamic microphone with an impedance of 300R, at 300^oK the equivalent thermal noise voltage is 997nV_RMS over the audio bandwidth of 20KHz. Suppose the microphone is outputting a signal of 2mV, at 24-bits per sample the quantitisation level is 119.2*10^-3V which is way into the noise margin so totally overkill.

I think the same thing was said when 16 bits was chosen for CD's. There's now a push to re-release a lot of music at 24-bits (as downloads).

24 bits in not way over the top when dealing with a lot of processing on the signal. The samples are usually converted to 32 bit floating point internally and these values used through the algorithms for the processing through the DAW. Generally when recording, however, you aren't using 24 bits, since you never record at 0dBFS. I'd usually peak at -12dBFS during recording, which is 2 bits less.

Sample rate is a different story. I don't know anyone who is using 192kHz, most use 44.1, 48 or 96kHz. If I'm recording for a pop/rock/jazz CD release, I'll still use 44.1kHz at 24 bit rather than having to do a conversion later. For video, 48kHz is usually sufficient. It uses a lot less disc space too, which can easily grow quite large when dealing with an ever increasing number of tracks.

[DrGeoff]

I suppose my point is if the signal is amplified as close to the source as possible then there should be no need for balancing or impedance matching because the signal is now large enough that hum shouldn't be a problem.

There is always a need for balancing. There's a lot of noise around on stage with lighting rigs, amps etc. Balancing helps remove a lot of it easily. These days cable runs are shorter though, with wireless packs for guitarists and digital consoles running a fibre or cat5 cable between the stage and the FOH console and back to the amp racks.

[Bren]

Quote from: Hero999 on Today at 02:24:18 PM
I suppose my point is if the signal is amplified as close to the source as possible then there should be no need for balancing or impedance matching because the signal is now large enough that hum shouldn't be a problem.

There is always a need for balancing. There's a lot of noise around on stage with lighting rigs, amps etc. Balancing helps remove a lot of it easily. These days cable runs are shorter though, with wireless packs for guitarists and digital consoles running a fibre or cat5 cable between the stage and the FOH console and back to the amp racks.

This is especially true for live performances, but not just... Professional recording studios are an architectural specialty, and due to multiple noise isolated rooms or booths, the signal is almost never close to the source. You often have mic's plugged into patch panels and patch panels running cable through walls to the control room, and from the control room wall to the mixing desk. You could easily have a 30foot + length for the signal to travel before it meets the source.