EEVblog #1128 - Studio Monitor Speaker Noise - Part 2

[EEVblog]

Follow-up to the Studio Monitor Noise video, replacing the mains power supply with a battery supply. Does the noise floor go away?

[johnlsenchak]

I thought we determined Dave that it's the small bias voltage that slightly turns on the output transistor in those chip packages In some amplifiers you can adjust the (turn on) DC bias current by way of a trim pot  I would think that if you found the right   resistor  on the board and replaced  it with a trimmer  pot then  possibly  you could reduce  the noise  even further

[EEVblog]

I thought we determined Dave that it's the small bias voltage that slightly turns on the output transistor in those chip packages In some amplifiers you can adjust the (turn on) DC bias current by way of a trim pot  I would think that if you found the right   resistor  on the board and replaced  it with a trimmer  pot then  possibly  you could reduce  the noise  even further

Nothing was determined to be the cause, someone just speculated that IIRC. The input noise is 3uV typical x 27 gain alone, so almost the voltage measured with a DC supply, and it's still quite audible.
There is no bias pin on this chip.

[BrianHG]

Why doe the Youtube video title say EEVBlog #1127?

[EEVblog]

Why doe the Youtube video title say EEVBlog #1127?

Because I screwed up

[Muttley Snickers]

I have a few sets of both active speakers and sub-woofers connected to different computers and was doing some recording and testing yesterday of a couple of stand alone pocket type recorders, on the Philips DVT-6010 recorder I had accidentally left the input set to microphone instead of a line level input with all the levels set to high so the speakers squealed like stuck pigs until I pulled the plug.

This turned out to be a bit of a blessing as it resulted in me discovering that the computer or its sound card is generating all sorts of noisy garbage. To sum up as I just discovered, you can have a super quiet set of active speakers but if the computer they are fed from is spitting out all sorts of noisy crap then it's all for nought, this may also be why some sets incorporate the automatic mute function as your monitors have.

[jh15]

Why not buy a real pair, and no crap electronics inside?

Were I studio near field speaker designer into real design, I may not be that good at audio circuit selection from marketeers to my wood and paper company.

[Herby]

The input noise is 3uV typical x 27 gain alone, so almost the voltage measured with a DC supply, and it's still quite audible.

You could try to set the gain to 1 instead. That gives an improvement of 27 at best.

If that works and is deemed useful, a low noise amplifier in front of the power amplifier could help (for example AD8597 or OPA1611 on the +/-15V supply). With a gain of 27 and a voltage noise density of 1.1nV/sqrt(Hz), those amplifiers should have about 4uV of noise at a bandwidth of 20kHz at their outputs. All in all, that gives 5uV of noise after the existing power amplifier (in theory, assuming no other noise sources). That's quite an improvement.

Edit: This is a horrible idea, because the power amplifier probably isn't unity gain stable (30dB gain minimum according to the datasheet)

[2N3055]

Like I said on a Youtube comment on a first video, PSRR on TDA2050 is bad, and also noise and distortion.
If you want integrated APA, LM3886 or some of the STK series are very good... I was surprised to see TD2050 in  there..
They are mediocre performers...
Good chips (like those mentioned) have very high PSRR exactly because of that so you can power them from unregulated power supplies..

Regards,

Siniša

[Kleinstein]

The TDA 2050 is an amplifier more on the cheap side. However the noise specs are not that bad, just a little more than the really good ones.
However as Dave already mentioned, other chips tend to have a different pin-out / case - so no simple replacement, but it would need more like a new amplifier board.

Reducing the gain of power amplifier might be an option, if one keeps an eye on stability. One would not be able (and would not want) to go all the way to unity gain, but getting from the factor of 27 to maybe 10 could already help and might be possible with moderate tweaks to the circuit.  A slightly reduced gain might be compensated from the filter circuit. There is still be the brute force method with an attenuator behind the amplifier  .

When hearing the direct comparison of mains powered and battery powered (around 9:40), one can hear some hum (e.g. 100 Hz). As the speaker is not very efficient at the low frequencies this hum can give quite some RMS amplitude and thus explain the drop from 400 µV to around 100 µV.
PSRR also depends on the circuit layout.

[blargg]

Decades from now part 100 of this series:

[2N3055]

TDA2050 has puny 45dB PSRR compared to LM3886 that has 85 dB.. Also it has twice the noise figures..
From TDA series  TDA729x DMOS series is much better sounding, and better overall.

There is no simple replacement, maybe one of the  power opamps could be replaced with a bit tweaking ..

[b_force]

You can get at least the noise down by 6dB by adding a resistor divider after the power amp.
6dB because of the baffle step.
In reality even more because the tweeter has a much higer sensitivity.

[k4rlhp]

I propose a test: change the R2 and R3 of the tda2052 in the negative feedback loop say for 260 and 10k ohm values  (currently 560 and 22k) This will preserve gain of 32db but should reduce johnson noise into the negative feedback. I did not check the datasheet for what the lower limit might be but these values should be ok to test and if that reduces noise predictably then smaller values and solution suitability could be checked. if you take the AD797 as an example, the nfb resistor values are critical design consideration for a lna application.

[Kleinstein]

TDA2050 has puny 45dB PSRR compared to LM3886 that has 85 dB.. Also it has twice the noise figures..
From TDA series  TDA729x DMOS series is much better sounding, and better overall.

There is no simple replacement, maybe one of the  power opamps could be replaced with a bit tweaking ..

The DS I have (e.g. ST/NS) show typical 2 µV noise (A weighted) for both the TDA2052 and LM3886. So about the same value, with even a higher worst case value for the LM3886. The TDA729x however seem to be better (1 µV). The weak point of the TDA2052 is more with distortion and PSRR.

My first try would be reducing the power amplifier gain, and have a higher gain setting in the preamplifier. Chances are one can still reduce the gain somewhat. The 30 dB seem to be the gain of the test setup, not the stability limit. With the internal speaker one avoids capacitive loading from long cables. Still it might be worth doing this on a separate circuit with a dummy load).

The 560 Ohms resistor in the current circuit should contribute some 3.3 nV/Sqrt(Hz) and thus around 300 nV A-weighted input referred. This is still way lower than the amps own noise (2 µV). So the resistors noise is not critical here.

[IanMacdonald]

It does seem that the TDA 2052 can't be expected to do much better than 90uV or so of hiss with a gain of 30. Possibly a lot worse if you get an upper-limits sample.

My thoughts on this are that separate bass and treble drive has its pros and cons. It may allow for a more accurate crossover filter than a simple passive arrangement, but if it requires the use of two consumer-grade power amps to stay within budget,  then maybe one good amp and a passive x-over is preferable.

Auto muting strikes me as a nasty bodge because in a studio you'd absolutely want to know if you'd got background noise on a channel. You could end up with a seriously red face over that one if it led to a production batch being rejected.  Probably would be noticed on a headphone sound check anyway, but still not a good situation.

[b_force]

It does seem that the TDA 2052 can't be expected to do much better than 90uV or so of hiss with a gain of 30. Possibly a lot worse if you get an upper-limits sample.

My thoughts on this are that separate bass and treble drive has its pros and cons. It may allow for a more accurate crossover filter than a simple passive arrangement, but if it requires the use of two consumer-grade power amps to stay within budget,  then maybe one good amp and a passive x-over is preferable.

Auto muting strikes me as a nasty bodge because in a studio you'd absolutely want to know if you'd got background noise on a channel. You could end up with a seriously red face over that one if it led to a production batch being rejected.  Probably would be noticed on a headphone sound check anyway, but still not a good situation.

You can still use a hybrid as a mentioned above.
And as Dave mentioned that effect is very minimal anyway

[SilverSolder]

...low noise amplifier in front of the power amplifier could help (for example AD8597 or OPA1611 on the +/-15V supply). With a gain of 27...

Is this what's called a "compound amplifier"?

Edit: This is a horrible idea, because the power amplifier probably isn't unity gain stable (30dB gain minimum according to the datasheet)

Does this still apply if it sits in a feedback loop with another, faster, better, amplifier?

Might still be a good idea!

[Kleinstein]

Edit: This is a horrible idea, because the power amplifier probably isn't unity gain stable (30dB gain minimum according to the datasheet)

Does this still apply if it sits in a feedback loop with another, faster, better, amplifier?

Might still be a good idea!

Having a loop with an inner and outer amplifier only works the simple way, if the inner amplifier is faster than the outer.  The inner amplifier would still need to be stable on it's own. There is another problem if the inner amplifier has a gain higher than the full amplifier gain. Already some gain makes things a little tricky.  It is possible to use a separate higher frequency feedback for the outer loop, but this is not easy. So adding an extra outer loop is not that simple, especially if fast.

It is usually easier to tweak the original amplifier circuit so that it is stable with a lower gain. In the DS I found a gain of 30 is the gain of the standard circuit, not as the minium stable gain, which I have not seen.

[thm_w]

Auto muting strikes me as a nasty bodge because in a studio you'd absolutely want to know if you'd got background noise on a channel. You could end up with a seriously red face over that one if it led to a production batch being rejected.  Probably would be noticed on a headphone sound check anyway, but still not a good situation.

In a studio absolutely, at home where 90% of these sort of active monitors are used its beneficial. I would hope the high end ones wouldn't do this or have a selection switch.

[beaker353]

There seems to be an inherent misunderstanding of the context in which studio reference monitors are actually used. I'm sitting right now in front of two Genelec studio reference monitors that are just shy of $10kUS a pop. They have an inherent very low-level noise across all three drivers, and that's fine and very likely within spec. While it's nice to have a low noise output, it's far from the most important spec for a reference monitor. Frequency response range and flatness are the big ones, followed by phase consistency and time alignment across the entire range, dynamic linearity (in to out gain is linear across the entire dynamic and frequency range) is also an important spec though it's typically not a measured spec. It's not uncommon to have 20 or more hand-tuned filters per monitor to get the aforementioned specs as close to ideal as possible. And just like a multimeter needs recal from time to time, so do reference monitors in serious applications. So does a reference monitor need low noise? It's nice, but sound (mixing) engineers will gladly give up noise for other more important performance values.

-EM

[Brumby]

Good point.

I hadn't really thought bout it - but it makes sense.  I mean, when you are using these speakers for their purpose, the signal level is way above this noise level.  Add to that the fact that you aren't really listening for that level of noise - but the quality and balance of a much louder signal.

I think this might be considered more of a storm in a tea-cup than an issue of any significance.  Sure, it's not ideal - but then what is?

[David Hess]

The input noise is 3uV typical x 27 gain alone, so almost the voltage measured with a DC supply, and it's still quite audible.

You could try to set the gain to 1 instead. That gives an improvement of 27 at best.

If that works and is deemed useful, a low noise amplifier in front of the power amplifier could help (for example AD8597 or OPA1611 on the +/-15V supply). With a gain of 27 and a voltage noise density of 1.1nV/sqrt(Hz), those amplifiers should have about 4uV of noise at a bandwidth of 20kHz at their outputs. All in all, that gives 5uV of noise after the existing power amplifier (in theory, assuming no other noise sources). That's quite an improvement.

Edit: This is a horrible idea, because the power amplifier probably isn't unity gain stable (30dB gain minimum according to the datasheet)

You identified the problem.  Power amplifiers are usually not unity gain compensated because it compromises performance.  They also almost always have emitter/source degeneration for transconductance reduction in their differential input stages for improved AC performance which is what produces higher input noise.

What could work although it would be difficult is to enclose the integrated amplifier inside the feedback loop of a lower noise amplifier.

Reducing the gain of power amplifier might be an option, if one keeps an eye on stability. One would not be able (and would not want) to go all the way to unity gain, but getting from the factor of 27 to maybe 10 could already help and might be possible with moderate tweaks to the circuit.  A slightly reduced gain might be compensated from the filter circuit. There is still be the brute force method with an attenuator behind the amplifier  .

Reducing the gain (and noise gain) requires access to the internal compensation node which unfortunately is not provided.

[b_force]

There seems to be an inherent misunderstanding of the context in which studio reference monitors are actually used. I'm sitting right now in front of two Genelec studio reference monitors that are just shy of $10kUS a pop. They have an inherent very low-level noise across all three drivers, and that's fine and very likely within spec. While it's nice to have a low noise output, it's far from the most important spec for a reference monitor. Frequency response range and flatness are the big ones, followed by phase consistency and time alignment across the entire range, dynamic linearity (in to out gain is linear across the entire dynamic and frequency range) is also an important spec though it's typically not a measured spec. It's not uncommon to have 20 or more hand-tuned filters per monitor to get the aforementioned specs as close to ideal as possible. And just like a multimeter needs recal from time to time, so do reference monitors in serious applications. So does a reference monitor need low noise? It's nice, but sound (mixing) engineers will gladly give up noise for other more important performance values.

-EM

Than you're actually missing the most important of them all, which is a constant directivity.
That's the mean reason why Genelecs perform so well.

And as I already mentioned now twice before, there is a very easy trick to get rid of the noise if it bothers someone.

[Icchan]

It's absolutely true, that this is an issue in very high end monitor speakers as well. I  have Genelec 8020B monitors and Genelec is regarded as a cream of the crop, these were back then their smallest model and they were 350€/piece.

They have completely custom power-supply and amplifier design and they're really low noise, but you can still put your ear to it and hear hiss. It's negligible though.

One of my friends has larger versions and they're many times more powerful so the hiss is a bit more pronounced so they tend to keep them off when not in use. I couldn't hear it in a quiet room, so it's a bit subjective.

Genelec does recommend, that when you buy the monitors, you turn them on, and you'll never turn them off ever again.
This is basically to increase the lifetime of the components when avoiding unnecessary surges and to allow everything to settle so that they'll sound the same, every day, all day.

But anyway.

---Now a general rant about quality---

KRK isn't very high end when it comes to studio monitors. Every decent monitor speaker can be bought one at a time, not as a pair. Genelec is bought one at a time because they can guarantee that any speaker coming from the production line is so close in tolerance that it can be paired up with any other from their production line without sacrificing anything.

My friend works at Genelec at this time, and he told me that every single speaker can be paired up with the serial number to the measurements done to it before packaging. And they do store eve3ry single measurement done to their speakers. Thus if I was so inclined, they could perhaps give me the measured frequency response for each of my speakers and I could in a sense use them as a decently calibrated sound sources for acoustical measurements.

That's quality control for you!