EEVblog Electronics Community Forum

Electronics => Repair => Topic started by: Funkensteyn on June 01, 2016, 10:54:58 pm

Title: How to lower noise of active speakers?
Post by: Funkensteyn on June 01, 2016, 10:54:58 pm
I have a pair of active monitor speakers which i would like to improve. Overall sound is quite aceptable but I can hear a little noise.. So today i opened them up and had a look at the circuit. I did not trace out the complete schematic (yet) but this is how it looks:
For the power supply there is a toroidal transformer with 2 x 115 VAC input and 2 x 20 V output followed by a rectifier, filter caps and 7815/7915, looks like a basic bipolar supply from the datasheet.
There is a TL084CN op amp, one amp is set up as an amplifier whose gain is controlled by a pot on the back of the speaker, the remaining three amps form an active crossover network. The high and low parts each go to an LM3886TF audio amp.
As I said I want it to be less noisy. What could be sources of the noise? The TL084, the LM3886, the pot or something entirely different?
I need some advice on how to approach this.
kind regards,
Jan
Title: Re: How to lower noise of active speakers?
Post by: edavid on June 01, 2016, 11:33:58 pm
Is the noise still there if you short the input?

Does it vary with the gain setting?
Title: Re: How to lower noise of active speakers?
Post by: Funkensteyn on June 02, 2016, 12:38:20 am
Shortening the input doesn't change anything. The gain is huge, usually I (and my neighbours) can not stand it when the gain is higher than 1/4.. however if i turn it all the way up the noise only gets a tiny bit louder, the increase is barely noticeabel.
Thinking about it.. does this indicate that the main source of the noise does not originate from the crossover and the pot but from the 3886?
Title: Re: How to lower noise of active speakers?
Post by: edavid on June 02, 2016, 02:04:11 am
Well, it depends on where the pot is in the circuit... can you figure that out?

I would tend to suspect the TL084 more than the LM3886.

The noise could also be coming from outside the audio path... power supply or EMI.  Is it a hiss or a hum?
Title: Re: How to lower noise of active speakers?
Post by: Funkensteyn on June 02, 2016, 10:57:26 am
The noise is a hissing sound.

My initial guess on how the circuit works was wrong, but I think I figured it out. The speakers have balanced inputs, first the input is converted to a single-ended signal, then there is the pot wired as a simple voltage divider which feeds the filter sections.

I also noticed that the lm3886 is powered with the unregulated voltage from the rectifier, but according to the datasheet this should not be a problem.

Thank you for your help, it is much appreciated!
Title: Re: How to lower noise of active speakers?
Post by: Kleinstein on June 02, 2016, 04:04:02 pm
The noise seems to come from the part behind the pot. Without a clue on how the circuit looks like, it's hard to tell where the noise comes from. The TL084 are relatively noisy OPs, but it depends what they are used for.

Unregulated supply for the LM3886 should not be a problem.

An important factor is how much amplification is used after the pot. If the overall amplification is high enough, there might be a chance to reduce the amplification behind the pot. However the final amplifier might need special attention for to low amplification. 
Title: Re: How to lower noise of active speakers?
Post by: Anks on June 02, 2016, 07:53:24 pm
What speaker are they
Title: Re: How to lower noise of active speakers?
Post by: The Soulman on June 02, 2016, 09:36:24 pm
If you really like these speakers and plan to be using them for a long time I'd suggest replacing all power supply electrolytic capacitors (assuming it isn't using Nippon chemicon or Nichicon brand caps already) with some nice high spec low-esr ones, it will extend the lifetime of the speaker (or actually the amplifier module..)  and lower ripple voltage on your supply lines resulting in less noise from the speaker.
The tl084 (known to be noisee) could be replaced by a much better specked tle2074.

(Ducking and diving) After that you could add mkp capacitors between the ic's power pins and the local ground (plane) as nearby by as possible, bigger value is better.
Yes the opinions are spread across the spectrum about this and others will say bypass capacitors should be calculated for optimum functionality, in my opinion that is true for (cheap) tantalum capacitors in commercial designs that is going to be manufactured in large quantities.
For mkp Capacitors (with much more frequency independent specs, needed for audio) used for a one of it is usually not a concern if instead of a 5cent part a 25cent one is used.

Also carbon resistors can add a lot of noise (these are yellowy ones), metal-film (bleuish/greenish) is the way to go for audio and typically have better accuracy specs needed in cross-overs and such.
Title: Re: How to lower noise of active speakers?
Post by: Funkensteyn on June 02, 2016, 09:53:50 pm
Without a clue on how the circuit looks like, it's hard to tell where the noise comes from.
I am trying to draw a schematic based on the pcb, but I want to get it right before I post it. Although the pcb is small with not that many components this is quite hard for me.
What speaker are they
They are desktop monitors for mixing etc. The name is 'the box' :) it's the store brand from thomann.de (http://thomann.de), but they don't sell these anymore. Each monitor has a speaker for the low/midrange and a tweeter, each driven by an LM3886.
Title: Re: How to lower noise of active speakers?
Post by: richnormand on June 02, 2016, 09:57:14 pm
Is the noise from the supply, a defective component or just bad design? One or both channels, if stereo?

If one channel:When looking for noise in audio systems I used to insert a capacitor (like 0.1uf or so depending on the circuit. I used up to 1uf in some) to basically short out the signal to ground.
This does not affect DC levels. Works for tubes in real time just contacting the grid input, but with ICs or transistors it is safer to power down at each trial.
Power down the amp, insert cap (tack soldered on pcb) at the input of the stage, repower. Is the noise less? If not, move cap to next stage and repeat.
If there is a sudden drop in hiss or noise you have found the culprit. In some cases it was a bad IC in others a bad design using too much gain from a transistor/IC instead of using two gain stages using a better quality device.
If both channels: do like before to locate noisy stage. Most likely bad design. Look for low noise substitutes.

I assume your device is stereo and that both channels are affected? Is the noise really a hiss high frequency wide bandwidth or a leakage from an inverter supply (if using such)? But you say it uses "normal" regulators for the +-15V so that should be OK. In some instances a quick trial might be to put a good filter cap on the PSU line at the board and see if it gets better.

Finally at high gain proper layout on the chassis  (like holy grounds and such) plays a large factor on the quiescent noise level.
I would suspect cheap opamps.... There are several low noise units available.


 
Title: Re: How to lower noise of active speakers?
Post by: Yansi on June 02, 2016, 10:25:40 pm
Too much noise may indicate also a poor design. The amplifier might be oscillating.  Try to take some measurements with oscilloscope.
Title: Re: How to lower noise of active speakers?
Post by: Funkensteyn on June 02, 2016, 11:28:46 pm
Wow, thanks guys, a lot of helpful suggestions!

Both monitors have comparable noise levels, so I think this is a design/ component selection issue, there are mostly carbon resistors, jamicon electrolytics and some tantalum and ceramic caps. Serves me right for beeing a cheap ass when I bought them some years ago :)

I have to finish the schematic and be confident that it is correct before I start poking around. Then I will try to locate the noisy stage with the 'cap-trick'.

However if everything fails and this is all due to the components beeing used I would rather keep the transformer and LM3886 and build my own circuit instead of replacing every other component.

I had a look at the TLE207x, but at more than 20 times the price of TL08x it seems a bit overpriced when my main concern is a lower noise floor (11-17 for TLE2074 vs 15 typ. for TL084).
Title: Re: How to lower noise of active speakers?
Post by: macboy on June 03, 2016, 12:44:46 am
Definitely replace that TL084. Alternative FET input amps might be OPA4134 or OPA4132. Those are both highly regarded audio opamps with lower distortion and noise, and they are quads, which is hard enough to find.

Try reducing the gain of the 3886, but not below 10, the minimum for stability. Reducing gain here reduces the amplification of noise (and signal of course but you admit there's plenty or too  much signal gain). It also reduces distortion by increasing feedback.
Title: Re: How to lower noise of active speakers?
Post by: Yansi on June 03, 2016, 09:21:21 am
There's nothing wrong with an TL084 for line level audio. If it produces excessive noise, you have issues in your circuit design an the issue is really not the TL084.  ;)
Title: Re: How to lower noise of active speakers?
Post by: macboy on June 03, 2016, 12:44:08 pm
There's nothing wrong with an TL084 for line level audio. If it produces excessive noise, you have issues in your circuit design an the issue is really not the TL084.  ;)
No, there is nothing "wrong" with it but nothing "right" about it either. It is a general purpose op-amp that people used 30+ years for audio because it was cheap and available and wasn't terrible. Far better audio op-amps are available very inexpensively today so there is not reason to tolerate the mediocre noise and distortion performance of the TL084 or similar op-amps.

Nonetheless, you are probably right. Although the TL084 isn't super low noise, it isn't horrible, and even through 4 stages of them there should not be excessive noise. This is probably an issue with poorly thought out inter-stage gain and/or too-high impedances.

I'd start with the LM3886. The gain is set by two resistors so it is easy to find. If higher than 10, reduce it to (as low as) 10. Test for stability and for adequate gain. The noise will be reduced, since this stage amplifies everything at its input including noise produced by the filter stages.
Title: Re: How to lower noise of active speakers?
Post by: Yansi on June 03, 2016, 01:40:00 pm
If your circuit design is fucked up, better opamp will *NOT* solve noise issues. Howgh.  TL084/074 noise is low enough for line level audio applications. So you better figure out the circuit around, to be designed correctly. (especially, when there is a tone stack (hi-mid-lo) present - these designs are usually messed up)
Title: Re: How to lower noise of active speakers?
Post by: acbern on June 03, 2016, 02:29:43 pm
In any design, there are different sources of noise. Semiconductors generate noise (various forms), and resistors do. The noise generated by these is amplified, the noise of the semiconductors (opamps) adds.  Also, carbon resistors are known to add noise beyond what one would expect doing a simple resistor noise calculation (noise by laws of physics depends on resistor value and temperature; I wont go into details, but there are resistor noise calculators available online).
So this is what should be done (described in a very simplistic manner):
-check circuitry and analyze the noise sources and noise amplifiers (high resistance source followed by high gain amplifier=high noise) that contribute most; check what can be done here; e.g., use lower values of resistors
-reduce gain (if you cannot use the total gain anyway, why not reduce it and also amplify noise less)
-check to replace parts (carbon resistors, opamps; opamps being the least important issue imo)
Title: Re: How to lower noise of active speakers?
Post by: f5r5e5d on June 03, 2016, 03:48:54 pm
reducing the gain of a LM3886 is not so simple - it really will oscillate if feedback gain setting is too low

the answer then is "noise gain compensation" - which could let you move the 3886 signal gain down even below the min stable data sheet number

the noise of the 3886 would be amplified by the lower signal gain over audio with a RC noise gain compensation leg

the noise gain increases above audio to meet the 3886 stability requirement (preferably with some margin since the corner/pole of the shelving noise gain costs some additional loop phase shift)
Title: Re: How to lower noise of active speakers?
Post by: Kleinstein on June 03, 2016, 05:54:35 pm
If the OPs are a problem depends on the circuit and the place they are used. Also not all resistors are critical. In many places carbon resistors don't matter - though there is not much to save here. The same is true for the electrolytic caps. Low ESR caps are not better in all places. The old style classical ones usually have lower leakage and longer lifetime (if not overloaded by ripple). Also to low ESR might cause problems in some cases (e.g. higher peak currents). Usually the caps are not responsible for noise.

There is not need to use expensive MKP caps  - especially in decoupling. Here the low ESR might even be a problem, causing resonances. Even for inter stage coupling MKS is perfectly fine - the slightly higher loss factor mainly shifts cross over frequencies or filter quality factors by small fraction of 1% - hardly measurable and way below audible.

Using many OPs and separate power stages suggests they use active filters. Often the filter is behind the volume pot - so the OPs in the filter stages might be important noise wise.

The two main candidates for the main noise source are the OPs in the filter stages and the power amplifier, if configured for high gain.
Title: Re: How to lower noise of active speakers?
Post by: Yansi on June 04, 2016, 09:04:41 am
Sallen-Key filter with opamp should NEVER be placed behind a volume pot, without a buffer.  Respectively, anything which requires constant/low signal source impedance should not be placed behind a volume pot!
Title: Re: How to lower noise of active speakers?
Post by: Funkensteyn on June 05, 2016, 04:25:02 pm
Thanks again for all the replies. I finally finished the schematic of the input section and crossover, it's attached to this post. I studied it a bit and now have a rough understanding of how it works, but there are some things i am unsure about.

Starting with the input, there is an differential amp which converts the balanced input to a single-ended signal. However, this stage has some gain, 2.9 for the inverting input and 3.3 for the non-inverting input. This unbalance will introduce some dc offset, but i guess this doesn't matter since the output is coupled via C26? Just curious why it is not designed with unity-gain? Also, what's the purpose of the capacitors C1 and C2, are those added for stability?

Then there is the volume pot after which the signal is fed to the filters. I think both of them are 2nd order butterworth types.

What's the purpose of the R5, R6 and C3 network? Is this just a voltage divider or a low-pass or both? After buffering it is fed to the low-pass, but shouldn't R7 and R8 be equal?

Can someone point me to a source where it is explained how I can calculate the cutoff of the filters, I only found stuff for designing filters and don't understand how I can reverse this to get the cutoffs.

Edit: corrected a typo in the schematic.
Title: Re: How to lower noise of active speakers?
Post by: Kleinstein on June 05, 2016, 05:30:05 pm
R1 and R4 should be the same value to get a true difference amplification. The different values might be to compensate for different impedance at both input, but it's still strange as different value for C1,C2 would than be needed. C1,C2 add an upper band limit and also help a little with stability, though the amplifier should be stable without it. Anyway the input stage should not be the noise problem.

Some gain for this stage is good to have a high level at the following amplifier stage. The idea is to have as much gain there and as little as needed after the volume pot and filter. The best gain for the input stage depends on the input level. So a course adjustment here might be a good idea.

Its a little strange to have the tweeter filter not going from behind U1A.
Also C26 before and after the pot is strange - one of them should be enough.
Title: Re: How to lower noise of active speakers?
Post by: Yansi on June 05, 2016, 06:03:15 pm
It is also strange that the volume pot output is not buffered, or the filter inputs are not buffered respectively. This way the properties of the filters (like cutoff frequency, gain, ..) changes as changes the output impedance of the volume pot.
Title: Re: How to lower noise of active speakers?
Post by: janoc on June 05, 2016, 09:50:35 pm
If you really like these speakers and plan to be using them for a long time I'd suggest replacing all power supply electrolytic capacitors (assuming it isn't using Nippon chemicon or Nichicon brand caps already) with some nice high spec low-esr ones, it will extend the lifetime of the speaker (or actually the amplifier module..)  and lower ripple voltage on your supply lines resulting in less noise from the speaker.

Low ESR caps in a linear PSU? Sure, you could do it, but I doubt it would make much difference (apart from the OPs wallet). Also the amp is symmetrical, with dual rails, so common mode mains hum will be attenuated already. Moreover, he is complaining about a hiss, not a low frequency hum.

The tl084 (known to be noisee) could be replaced by a much better specked tle2074.

Could be also a red herring, depends on how is that opamp used.

Title: Re: How to lower noise of active speakers?
Post by: Funkensteyn on June 06, 2016, 10:15:26 am
This circuit is really strange. I took me so long to get the schematic done because I was always thinking I got an error somewhere. I did it two times and came up with the same circuit, this is really how it is wired up.. I guess it's just a bad design and may be the reason they don't sell these speakers anymore.

R1 and R4 should be the same value to get a true difference amplification. The different values might be to compensate for different impedance at both input, but it's still strange as different value for C1,C2 would than be needed. C1,C2 add an upper band limit and also help a little with stability, though the amplifier should be stable without it. Anyway the input stage should not be the noise problem.

Some gain for this stage is good to have a high level at the following amplifier stage. The idea is to have as much gain there and as little as needed after the volume pot and filter. The best gain for the input stage depends on the input level. So a course adjustment here might be a good idea.
I usually have the pot turned to 1/4 so this roughly cancels the gain of the input stage. So for my listening habits I could make all the resistors equal and get rid of the pot. I understand that one does want less gain in the following stages, but first amplifying the signal just to turn it down in the next stage seems unneccesary.
Although this might not be source of the noise I am trying to understand the designer's thinking. What could lead to impedance differences in the balanced inputs?

Its a little strange to have the tweeter filter not going from behind U1A.
It is also strange that the volume pot output is not buffered, or the filter inputs are not buffered respectively. This way the properties of the filters (like cutoff frequency, gain, ..) changes as changes the output impedance of the volume pot.
That's what I was thinking, too. A better way would be to put a buffer at each filters input? Would the two filter sections influence each other if they were buffered by only one amp after the pot? As only one amp was still available on the TL084 maybe the designer decided to prefer the midrange over the tweeter.
Also C26 before and after the pot is strange - one of them should be enough.
I made a typo in the capacitor labels going to and from the pot, one should be C25, so strangely they are both there.
Title: Re: How to lower noise of active speakers?
Post by: David Hess on June 06, 2016, 03:17:44 pm
R1 or R4 is just wrong.  They should both be the same value and if they are different, common mode noise at the input is going to end up at the output.  I would use 1% metal film resistors for R1 through R4 and 1% capacitors for C1 and C2 to raise the common mode rejection.

The impedance levels are all reasonable for the TL084.  I do not think noise from the TL084 is a factor here.

As a test, I would short the inputs to the LM3886s (I assume there are two) and see if the noise remains.  If it does, then the source is the LM3886 circuits.  I am suspicious of the LM3886; the datasheet does not list input referred voltage and current noise but its input bias current makes me suspect it is intended to operate with low impedance sources like an LM833.  The low input impedance used for the noise specification supports this as do the circuit examples.
Title: Re: How to lower noise of active speakers?
Post by: edavid on June 06, 2016, 03:59:47 pm
It looks like there is no DC bias path for U1A  :-//
Title: Re: How to lower noise of active speakers?
Post by: Kleinstein on June 06, 2016, 05:41:09 pm
The capacitor behind the pot should not be there - this would also avoid the problem of missing DC path for U1A.

Having one buffer OP for both filter sections is ok, there is only minimal interaction between the filters - even than its a question if it's better or worse to have worse filters at one OP.

Having the same value for R1/R4 would not change much with the gain, only a factor of 2/3.

Different input impedance is possible if the signal source is a one sided source - going to ground with the one side, while the other input sees the output impedance of the source. Anyway if needed adjustment would be better done with R2,R3. There is no really need for 1% precision für C1,C2 ,  5% or 10% should be good enough. Currently R1/R4 are 50% off.

Having amplification and than attenuation again makes some sense, to get a higher signal level and thus less noise influence. The trouble could be too much gain behind the filters.
Title: Re: How to lower noise of active speakers?
Post by: Funkensteyn on June 07, 2016, 09:45:16 am
There was an typo in the labeling again, the buffer and lowpass were both labeled UA, sorry about that.

It looks like there is no DC bias path for U1A  :-//

That's UB now, just that we are on the same page. Assuming an input bias current of 10nA the drift would be 10nA/10uF=1mV/s=60mV/min, right? I never noticed any malfuntion even after hours of operation.

Different input impedance is possible if the signal source is a one sided source - going to ground with the one side, while the other input sees the output impedance of the source.
Sure, didn't think about single-ended input.

Later today I have some time to poke around and will report back if shortening the lm3886s will improve on the noise situation.

Thanks for all your help, I already learned quite a lot!
Title: Re: How to lower noise of active speakers?
Post by: macboy on June 07, 2016, 04:14:23 pm
Good job reversing the schematics.

One way to determine the filter response is to simulate the circuit using e.g. LTSpice, then run an AC analysis. This will plot the gain/phase over a specified frequency range.

You didn't include the power amplifier section in the schematics. Did you have a chance to at least determine the gain of the LM3886?