EEVblog #1118 - Why Are Studio Monitors Noisy?

[EEVblog]

Why do most powered studio monitor speakers have a low level audible background noise? Time for some investigation on the KRK Rokit 6 studio monitor speakers to track down where the noise is coming from.

[johnlsenchak]

It's because the output transistors (inside the chip package ) are slightly biased on, causing the noise/ripple in the  sound floor  to increase

[EEVblog]

It's because the output transistors (inside the chip package ) are slightly biased on, causing the noise/ripple in the  sound floor  to increase

Probably. Bottom line, it's the amplifier chip.

[thmjpr]

This is much worse on Class D powered monitors, especially the cheap ones. Same sort of inherent noise floor, regardless of gain (if lower), supply source, voltage, etc. I returned a pair of Kanto monitors because of the unbearable hiss. And yet a smaller chinese amp brand has managed to work around the issue very well: auto shutdown is incredibly fast when no sound is playing, so of course if its shutdown there is zero hiss, same as Daves A/B amp in mute. It also helps that they used a bit better output amp (STA328). Unless you are playing something with incredibly low volume, you would never notice the hiss.

Its one thing that is not really covered well in reviews or specs, especially for powered monitors.

[laneboysrc]

Interesting video as always
Would it help to decrease the gain of the output amplifier, for example by changing the 12k feedback resistor to a 5k6 one?

[Raj]

They can digitally add anti-noise into 'audio signal in-line' to cancel it, can't they?
Someone, get on it and make it happen

[EEVblog]

This is much worse on Class D powered monitors, especially the cheap ones. Same sort of inherent noise floor, regardless of gain (if lower), supply source, voltage, etc. I returned a pair of Kanto monitors because of the unbearable hiss. And yet a smaller chinese amp brand has managed to work around the issue very well: auto shutdown is incredibly fast when no sound is playing, so of course if its shutdown there is zero hiss, same as Daves A/B amp in mute. It also helps that they used a bit better output amp (STA328). Unless you are playing something with incredibly low volume, you would never notice the hiss.

Its one thing that is not really covered well in reviews or specs, especially for powered monitors.

That's why many of the monitors have auto-mute, there is no other reason for it than to mask inherent background noise.
Yes, I'm surprised by the lack of mention of any of this in reviews, when the go into insane depth on all the other stuff.

[b_force]

Most of it all has to do with PCB design.
Very often power ground and signal ground are just directly connected together.
Sometimes even in the amplifier chip itself!
In other cases these chips have an input side (signal ground) and output side (with power ground).
Make sure you divide the two properly and DON'T put them over the same wire back to the PSU.
In fact it's better to make a good start ground with enough capacitive buffering as close to the power amps if possible.

In the old days when we developed our own Class-AB (B) amplifiers, we could see an huge different in noise when connecting the gain stage to actually the signal ground.
Sometimes even as big as 6dB or more!

This was a huge factor with high power and high sensitive compression drivers (especially in line-arrays.
Nowadays there are more than enough Class-D amplifiers that are as good or even WAY better in terms of noise (and all the rest as well).

In application were there isn't much power needed for tweeters you can just simply add a resistor attenuator.
You only do have to compensate for the gain and keep an eye on the frequency response.
(I saw this was also mentioned in the comments by someone else)

It the end you just want a gain as small as possible, which is a bit tricky because of stability of the circuit.
It surprised me (find it hard to believe) that the pre-amp section wasn't contributing that much.
The rail regulation is pretty poor with a bunch of 15V zeners (which are noisy) plus these opamps aren't the best at all (against what Dave said)

@Raj
Yes in theory you could add pre-noise (and pre-distortion), but you will need a DSP for these things.
Noise is going to be pretty difficult since it's difficult to make the random characteristics.

Its one thing that is not really covered well in reviews or specs, especially for powered monitors.

There are so many things not covered at all in reviews.
One of the most important things is called directivity, were even expensive well known brands fail horribly.
Next to that is non-lineair distortion (especially in the lower frequencies).

[David Hess]

That was a great job Dave.  I like that you found that the mute function was powering down the amplifier confusing the noise issue.

I did the rough math and concluded that the power amplifier itself was contributing more noise than anything before it; I was expecting the input stage to be the problem.  While not being a particularly low noise circuit, whoever applied the RC4580 operational amplifiers used about the right impedance levels so they knew what they were doing.  The gain distribution could have been better, but with the TDA2052 contributing so much noise it did not matter.

Besides relatively poor input noise, the TDA2052 also only has power supply rejection of 40dB minimum and 50dB typical.  No detailed schematic is available but I suspect its high input noise comes from emitter degeneration of its differential input pair which is commonly done to reduce transconductance allowing for better frequency response and higher full power bandwidth which is required in an audio power amplifier.

I might consider adding a low noise voltage gain stage before the TDA2052 so it can operate at a lower voltage gain reducing output noise however the TDA2052 might not be stable at lower closed loop gains and the datasheet does not say.  It would be odd for it not to have a minimum stable gain because lower gains require more frequency compensation which would lower performance in its intended application.

[b_force]

BTW, could you share the schematic/service manual of the thing?

[Karel]

The power amp chip is poorly designed.
Power amps with discrete transitors usually don't suffer from this noise problem.

[David Hess]

The power amp chip is poorly designed.
Power amps with discrete transitors usually don't suffer from this noise problem.

I do not know about usually since everybody seems to use the same basic topology with emitter degeneration of an input differential pair but integrated amplifiers definitely have some disadvantages because the capacitors needed for complex frequency compensation are very expensive on an IC and economics means they use the least expensive power IC process available so they cannot take advantage of JFETs or high performance PNPs.  For instance no consumer device is going to use an expensive part from Apex Microtechnology.

[BrianHG]

I'm surprised.  (Talking about the power amp stage, not the puny input preamp noise) The cheapest discrete design power amp, when I say cheap, it usually means simple, but a proper design, would have low enough noise floor to make any hiss inaudible, even with a high power 200 watt design, and the power supply rejection ratio of the supply rails should be high enough that even if you have lemon dying caps on your power supply, the 100hz should also be inaudible when there is no source audio connected.

It's like someone put together ok effort on speakers, ok effort case, ok effort power transformer, ok effort front end, then, any amp IC that just happen to be available which had enough power and it just had to be a 2$ by the spec, no engineering solution.

[Nominal Animal]

Am I the only one who noticed a wisp of magic smoke escaping from that thing opened on the right side of the speaker at 9:27?  I swear I could smell something..

[b_force]

Am I the only one who noticed a wisp of magic smoke escaping from that thing opened on the right side of the speaker at 9:27?  I swear I could smell something..

That's reflection

[EEVblog]

Most of it all has to do with PCB design.
Very often power ground and signal ground are just directly connected together.

Not in this case. I shorted the input to ground and the layout looks good, same noise was still there. Seems inherent in the amplifier chip.

[EEVblog]

BTW, could you share the schematic/service manual of the thing?

Service manual and schematic:
http://eevblog.com/files/RP6G2schA1.pdf
It's excellent actually.

[EEVblog]

I'm surprised.  (Talking about the power amp stage, not the puny input preamp noise) The cheapest discrete design power amp, when I say cheap, it usually means simple, but a proper design, would have low enough noise floor to make any hiss inaudible, even with a high power 200 watt design, and the power supply rejection ratio of the supply rails should be high enough that even if you have lemon dying caps on your power supply, the 100hz should also be inaudible when there is no source audio connected.

It's like someone put together ok effort on speakers, ok effort case, ok effort power transformer, ok effort front end, then, any amp IC that just happen to be available which had enough power and it just had to be a 2$ by the spec, no engineering solution.

But it's not just KRK. Yamaha, JBL, Alesis and others all have it. And even someone in the youtube comments mentioned Adam as well.

[EEVblog]

That was a great job Dave.  I like that you found that the mute function was powering down the amplifier confusing the noise issue.

Many speaker have an auto-mute function to mask the noise, that's it's only function. I believe the KRK Gen 3 has this.

[boB]

I was hoping you would go one step further and reduce or short the 12K power amp feedback resistor to see if reducing the gain would reduce the noise. 

If so, then you'd have to makeup the gain in front of it IF it could even be done quietly.

boB

[BrianHG]

I'm surprised.  (Talking about the power amp stage, not the puny input preamp noise) The cheapest discrete design power amp, when I say cheap, it usually means simple, but a proper design, would have low enough noise floor to make any hiss inaudible, even with a high power 200 watt design, and the power supply rejection ratio of the supply rails should be high enough that even if you have lemon dying caps on your power supply, the 100hz should also be inaudible when there is no source audio connected.

It's like someone put together ok effort on speakers, ok effort case, ok effort power transformer, ok effort front end, then, any amp IC that just happen to be available which had enough power and it just had to be a 2$ by the spec, no engineering solution.

But it's not just KRK. Yamaha, JBL, Alesis and others all have it. And even someone in the youtube comments mentioned Adam as well.

It is not possible to argue with selecting from what's available & the design choices.  I'm using a pair of Paradigm Monitors, but, originally fed with a homemade and amp which used OP27 opamps for voltage gain, feeding a NPN/PNP darlington emitter follower 75 watt amp, and today, a new external 200 watt amp both tied directly to the headphone output of my laptop and there is 0 hiss or hum.  So, technically, a proper amp can be designed, it's just that KRK, Yamaha, JBL, Alesis just picked a noisy IC audio amp off the shelf.

I personally do not blame them.  In a self powered monitor speaker system, they have a number of criteria and features to meet.  For the electronics, they went with published off the shelf reference designs and components as they do not have a specialist in each specific field and making an elaborate discrete audio amplifier (which takes many more components and engineering know how), or, choosing a quality premade 25$-200$ amp module was just out of the question.

If I were to re-equipt today with a cheap budget in mind, I would still choose a set of discrete monitor speakers similar to my Paradigms, but, maybe a 4 ohm version.  I would then either choose a higher quality Alibaba type Class D amp with a single 48v telecom switching supply, or, an Ebay used quality high end amp.  I do not think I can tolerate the performance bar of KRK, Yamaha, JBL, Alesis because at times, I need to crank up the volume and I cant tolerate any weakness or distortion, and at low volume, hiss and hum is something that should be obsolete in today's digital age.  Though, I still cannot faulter that you don't need an external amp with the KRK, Yamaha, JBL, Alesis.

[b_force]

Most of it all has to do with PCB design.
Very often power ground and signal ground are just directly connected together.

Not in this case. I shorted the input to ground and the layout looks good, same noise was still there. Seems inherent in the amplifier chip.

What makes you think that shorting the input bypasses the PCB layout?

I was hoping you would go one step further and reduce or short the 12K power amp feedback resistor to see if reducing the gain would reduce the noise. 

If so, then you'd have to makeup the gain in front of it IF it could even be done quietly.

boB

The datasheet says that closed loop gain needs to be at least 30dB for stability reasons.

[SilverSolder]

It's interesting that the HF noise viewed in the FFT application seems to start to drop off at around the crossover frequency?

Perhaps the power amp noise is spat back into the input, and is actually attenuated by the preceding stage...  but only above the crossover frequency?

[David Hess]

The datasheet says that closed loop gain needs to be at least 30dB for stability reasons.

So it does.  I looked for that in the datasheet but missed it and assumed it would be the case.

There are ways to lower the closed loop gain on a decompensated amplifier however they involve raising the noise gain so that is no solution.

[b_force]

The datasheet says that closed loop gain needs to be at least 30dB for stability reasons.

So it does.  I looked for that in the datasheet but missed it and assumed it would be the case.

There are ways to lower the closed loop gain on a decompensated amplifier however they involve raising the noise gain so that is no solution.

It's somewhere in the side notes.
But basically all these power opamps have a minimum gain to keep them stable.

There are ways to add extra compensation and do other tricks.
From manufacture point of view those are mostly not very interesting because you will find yourself in unknown territory.
Which means spending a lot of time (money) on research.

Like I said before the quickest and easiest way is just add a resistor divider after the power amplifier.
Also keep in mind that this noise isn't noticeable or only barely noticeable at the listeners position.
So it's a bit like solving a non-existing problem.

[Richard Crowley]

I posted the YouTube video over on the audio forum Gearslutz.com
https://www.gearslutz.com/board/geekslutz-forum/1229581-why-studio-monitors-noisy.html#post13503068

There was just a response from widely-respected design guru, Jim Williams (not the one from Linear Technology)
This Jim Williams is famous for modifying commercial audio gear to significantly improve performance.
He also consults for many companies that design and make commercial audio gear.
And he makes a boutique line of audio gear (at http://www.audioupgrades.com/)

Anyway, he said:

This is common on powered monitors as the amps are not well designed. Even the Meyer HD1's hiss. That is an issue for me. It masks low level details. My monitors are silent here, zero hiss 1/2" away from the center of the tweeters. The noise is the same whether the amp is on or off. It can be done.

[David Hess]

The datasheet says that closed loop gain needs to be at least 30dB for stability reasons.

So it does.  I looked for that in the datasheet but missed it and assumed it would be the case.

There are ways to lower the closed loop gain on a decompensated amplifier however they involve raising the noise gain so that is no solution.

It's somewhere in the side notes.
But basically all these power opamps have a minimum gain to keep them stable.

I am intimately familiar with the how and why leading to my surprise at not to find a minimum stable gain listed in the datasheet.  I just did not look closely enough as I am used to bold letters on the first page saying "Minimum Stable Gain of 32" or similar.

There are ways to add extra compensation and do other tricks.
From manufacture point of view those are mostly not very interesting because you will find yourself in unknown territory.
Which means spending a lot of time (money) on research.

The real restriction is not on minimum closed loop gain but on minimum *noise* gain.  It only takes a series RC circuit across the inputs to solve but raising the noise gain while lowering the closed loop gain produces the same output noise so nothing is improved.  Altering the internal compensation if possible would lower the full power bandwidth which I suspect is why they configured it this way in the first place; the full power bandwidth was insufficient without decompensation and a minimum stable gain is hardly an issue for an audio power amplifier.  It is too bad it suffers from such a high input noise but that is likely the result of transconductance reduction in the input stage being required as well.  Discrete audio amplifiers commonly employ both of these techniques.

Choosing such a noisy part seems dumb considering how much effort they put into the preamplifier circuits but it sure is economical for undiscriminating users.  Are all of the remaining integrated audio amplifiers noisy as well?  I have never heard of a quiet one.  Maybe there were no better choices short of a discrete implementation.

To solve the noise problem when using an integrated power amplifier likes this I might place it within the control loop of a lower noise amplifier but this is not something which could be easily retrofitted.

Also keep in mind that this noise isn't noticeable or only barely noticeable at the listeners position.  So it's a bit like solving a non-existing problem.

It drives me crazy when I come across it.  It used to be a more common problem when high efficiency speakers intended for vacuum tube amplifiers were used with solid state amplifiers.  Electronic headphones of various types all seem to have the same problem with excess noise.

[b_force]

The datasheet says that closed loop gain needs to be at least 30dB for stability reasons.

So it does.  I looked for that in the datasheet but missed it and assumed it would be the case.

There are ways to lower the closed loop gain on a decompensated amplifier however they involve raising the noise gain so that is no solution.

It's somewhere in the side notes.
But basically all these power opamps have a minimum gain to keep them stable.

I am intimately familiar with the how and why leading to my surprise at not to find a minimum stable gain listed in the datasheet.  I just did not look closely enough as I am used to bold letters on the first page saying "Minimum Stable Gain of 32" or similar.

There are ways to add extra compensation and do other tricks.
From manufacture point of view those are mostly not very interesting because you will find yourself in unknown territory.
Which means spending a lot of time (money) on research.

The real restriction is not on minimum closed loop gain but on minimum *noise* gain.  It only takes a series RC circuit across the inputs to solve but raising the noise gain while lowering the closed loop gain produces the same output noise so nothing is improved.  Altering the internal compensation if possible would lower the full power bandwidth which I suspect is why they configured it this way in the first place; the full power bandwidth was insufficient without decompensation and a minimum stable gain is hardly an issue for an audio power amplifier.  It is too bad it suffers from such a high input noise but that is likely the result of transconductance reduction in the input stage being required as well.  Discrete audio amplifiers commonly employ both of these techniques.

Choosing such a noisy part seems dumb considering how much effort they put into the preamplifier circuits but it sure is economical for undiscriminating users.  Are all of the remaining integrated audio amplifiers noisy as well?  I have never heard of a quiet one.  Maybe there were no better choices short of a discrete implementation.

To solve the noise problem when using an integrated power amplifier likes this I might place it within the control loop of a lower noise amplifier but this is not something which could be easily retrofitted.

Also keep in mind that this noise isn't noticeable or only barely noticeable at the listeners position.  So it's a bit like solving a non-existing problem.

It drives me crazy when I come across it.  It used to be a more common problem when high efficiency speakers intended for vacuum tube amplifiers were used with solid state amplifiers.  Electronic headphones of various types all seem to have the same problem with excess noise.

You're talking about a composite or nested amplifier I think?
That is actually a perfect way to get not only rid of the noise, but also improve the total performance significantly.
Totally forgot about these lol.

Yes, with high efficient compression drivers it can be an issue.
I have worked on huge line-array systems myself with 12 compression drivers in it, even the best amplifiers struggle with that.
Luckily these are being used pretty noisy environments.
In these situations you definitely can hear what a proper PCB layout and good parts (using NP0 instead of X7R etc) can do.
For home audio it's basically like Dave showed. You have to put your ear almost into the tweeter to hear anything at all.
It does depend what kind of speakers and how the filter is made.
Like I said, just adding a passive attenuation circuit can reduce the noise by 6dB minimum (because of the bafflestep compensation)
Since most tweeters are also much more sensitive than woofers, that number can easily go to about 12dB.
Of course that will cost a bit more power, but home hifi tweeters need around 10-15W anyway

The latest Class-D amplifier IC's are in terms of noise pretty much the same or better.
Professional modules are a lot better actually.

[boB]

My Mackie HR-824s (not made anymore) are virtually dead quiet.  I can hear slight hiss if I put my ear right up to the tweeter but it is very quiet.  These used discrete amplifiers and not a module like most of the monitors today use.

I hear that Genelec monitors are also very quiet.   Anything made to be cheap and usually Chinese just aren't going to be everything you could ask for.

That's why I buy used on Ebay.

boB

[Bassman59]

I hear that Genelec monitors are also very quiet.

I have a pair of Genelec 1031A which I bought new in 1996. They are dead quiet. The output amplifiers are discrete. Input stage is nothing fancy, just a differential amplifier made with a 5532.

[Kleinstein]

The residual noise problem is no specific to monitor speaker. It also sometimes happen with normal Hifi equipment if high efficiency speakers are used. I had that with an otherwise good amplifier and self build speakers too. Many of the old chip amplifiers are made with low efficiency speakers in mind. Speakers can be quite different in how much sound they make from a given electrical power.

Those monitor speakers tend to prefer high efficiency speakers as they could save on the amplifiers power this way.

[b_force]

The residual noise problem is no specific to monitor speaker. It also sometimes happen with normal Hifi equipment if high efficiency speakers are used. I had that with an otherwise good amplifier and self build speakers too. Many of the old chip amplifiers are made with low efficiency speakers in mind. Speakers can be quite different in how much sound they make from a given electrical power.

Those monitor speakers tend to prefer high efficiency speakers as they could save on the amplifiers power this way.

To my professional experience and knowledge I have never seen or come across significant differences in the sensitivity between home hifi and studio monitors.
The biggest difference is that studio monitors are mostly active speakers while home hifi speakers use a passive filter.
This passive filter already reduces the noise by at least 6dB because of the baffle step correction.
Since most tweeters have an higher sensitivity this attenuation can be easily around 9-12dB.