DIY Kits Signal to noise level

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

JAY X

Well-known member
Joined
Jan 9, 2009
Messages
683
Hi All!

When designing pcb dual layer or single layer....¿which is the signal to noise level to expect? or floor noise level...

Maybe this is a difficult question as Mic preamps can be noisier at higher gain levels... and compressors, mixers...?

¿are there any noise level figures list that can be a guide to follow?

For example: A friend, owner of a studio, has a neve8816 summing mixer, says it has -100db noise level...and, for example, dangerous music 2bus is -83db..

Any comments are wellcome!
Thank you very much!

JAY X
 
And on the measurement technique !

What is -100dB and what is -80dB  noise level ?

One man's method is not the same as the next man's.
 
alexc said:
One man's method is not the same as the next man's.

Unless the technique and equipment are documented.

If you are looking for a list of noise figures of well known equipment you'd do well to never again omit the dBu when quoting their specs. It's definitely there in both neve8816 and dangerous music 2bus specs and you should learn what that means.

Noise and it's implications to PCB design is a much larger topic. Larger than a single ME degree.
 
In theory a two-layer or more PCB offers the potential to self shield some high impedance traces. In practice noise is generally dominated by other factors.

JR
 
Hi john,

Other factors... you mean , like crosstalk and supply noise for example?

Yesterday i though i was on the right track...with a noise floor level of about -80db... and later i discovered that i mounted the volume pot pcb reverse side!! ... -74db  :p

All the levels i measured conneted to a pc soundcard, and monitored with soundforge.
 
JAY X said:
Hi john,

Other factors... you mean , like crosstalk and supply noise for example?

Yesterday i though i was on the right track...with a noise floor level of about -80db... and later i discovered that i mounted the volume pot pcb reverse side!! ... -74db  :p

All the levels i measured conneted to a pc soundcard, and monitored with soundforge.

For a summing mixer I expect the noise to be dominated by the number of stems being summed times the noise of the summing amp or makeup gain stage (even passive summers require make up gain). Self-noise of the resistors used to perform the sum will also contribute to the noise floor.

Finally signal path contamination due to poor layout or external noise sources is only a factor when they are present and larger than intrinsic circuit noise sources. I would not expect single-sided or double-sided boards to be a major concern compared to the rest of the design.

JR
 
Hi,

Well, in my summing mixer i used 22k summing resistors...maybe i can lower them to 15k or 10k...

But i think that, in my design the major noise contributor is the fader amplifier i placed before the DRV134 line driver...
It is build with 4k7/2k2 resistors in the feedback arm, and gives about 10db gain.  Maybe i can use a lower ratio resistors.. 2k2/1k or lower...680R/330R..

JAY X
 
JAY X said:
Hi,

Well, in my summing mixer i used 22k summing resistors...maybe i can lower them to 15k or 10k...

But i think that, in my design the major noise contributor is the fader amplifier i placed before the DRV134 line driver...
It is build with 4k7/2k2 resistors in the feedback arm, and gives about 10db gain.  Maybe i can use a lower ratio resistors.. 2k2/1k or lower...680R/330R..

JAY X

Resistors are rarely a significant contributor to noise. You can work out the noise in a resistor quite easily. Divide the resistor value by 3096 and take the square root of the answer. This value is the number of microvolts of noise the resistor contributes in a 20KHz bandwidth. To turn this into something meaningful just remember that one microvolt is -117.8dBu. 

Here's an example. A 22K resistor makes SQRT(22000/3096) = 2.67uV of noise.

20log(2.67) = 8.5dB above 1uV. So the noise from a 22K is 8.5 -117.8 = -109.3dBV

But, in a summer, all the 22Ks are effectively in parallel so if you have 16 of them the parallel combination is 22K/16. This effectively reduces the noise by four times or 12dB to -121.3. Definitely not worth worrying about.

The 4K7 and 2K2 resistor in your 10dB amp are also effectively in parallel so for noise purposes they look like 1K5 which produces a noise of 0.7uV or -120.9dBu. Again not worth worrying about.

The major sources of noise in almost any analogue audio circuit are the active amplifying devices and the gain the produce.

Cheers

Ian
 
Hi Ian!,

Wow!,  fast masterclass on noise! hmmm... i see more clear....but  ¿ why do i hear noise as i turn up the volume???. :eek:..  noise gain...in my volume scale of 1 to 10.... up to 3 or 4, noise is not much noticeable...-80 db...then starts to increase...up to -74db... :p

I used bipolar opamps NE5532 and LM833,  i try with fet opamp OP275,  and reduced a bit the noise...

So, if noise is inherent to active devices....there must be a limit to noise where we can't go lower!!.. due to pcb layout and supply noise...and this what i was trying to ask at the beggining of this thread: bottom and top limits to noise levels that are acceptable or tolerable in an analogue design... ???

¡Thanks Ian!
JAY X
 
JAY X said:
Hi Ian!,

Wow!,  fast masterclass on noise! hmmm... i see more clear....but  ¿ why do i hear noise as i turn up the volume???. :eek:..  noise gain...in my volume scale of 1 to 10.... up to 3 or 4, noise is not much noticeable...-80 db...then starts to increase...up to -74db... :p

I used bipolar opamps NE5532 and LM833,  i try with fet opamp OP275,  and reduced a bit the noise...

So, if noise is inherent to active devices....there must be a limit to noise where we can't go lower!!.. due to pcb layout and supply noise...and this what i was trying to ask at the beggining of this thread: bottom and top limits to noise levels that are acceptable or tolerable in an analogue design... ???

¡Thanks Ian!
JAY X

As I said in the last sentence of my last post "The major sources of noise in almost any analogue audio circuit are the active amplifying devices and the gain they produce."

In a mic pre you make strenuous efforts to reduce the amount of noise produced by the active devices but the gain is your enemy.

Suppose you have a mic with a source resistance of 150. This produces SQRT(150/3096) = 0.22uV  = -130.9dBu of noise. Seems pretty low but if you amplify the mic signal by 80dB you also amplify its noise so -130.9dBu at the mic pre input becomes -50.9dBu at the output and you hear this as hiss. This is inherent in the physics and there is nothing you can do about it. The above assumes a perfect amplifier that creates no additional noise at all. Real mic pre amplifiers are very good and often only add a couple of dB to the noise. The best way to reduce noise in this example is to reduce the gain. Drop it to 40dB and the noise will drop to around -90dBu. However, this usually means you need a more sensitive microphone that produces more output for a given sound level and that usually means a condenser type. They can easily be 20dB more sensitive than a ribbon mic.

So to answer your question, circuit layout and power supply noise can all be minimised to the point where they contribute little to the overall noise at the output. However, there is no getting away from the noise produced by the resistance of the source and that is what really defines the limit.

Cheers

Ian
 
Ian:

Have you ever done the "trick" of paralleling active devices to play off the correlated signal vs. uncorrelated noise for extra s/n ratio? The technique is often mentioned but I can't really say I know of anyone doing it....
 
Hi,

Yes paralleling active devices works for reducing noise, but in my case the pcb is already built...and now i'm trying to spot the noise origins... i have a friend that has an oscilloscope... so i hope we will find out somewhat...

On the other hand, after reading other threads, i realized how important is IDC wiring, and bus bar grounding..

In my case there are three boards : left ch, right ch, and psu board. Both channels are powered with 20 way idc cable.... 10 ways for relays ( 2 pins for power and the rest for ground) and other 10 ways for opamps ( 6 pins for power and 4 pins for ground)... so it is clear i have to reinforce the ground connections from the boards to the psu...maybe connecting both boards with 50way idc cable to their ground plane...and then to the psu... someway... besides, i don't know how a bus bar comes into the play.... i have to investigate!!! :)

JAY X
 
tubegeek said:
Ian:

Have you ever done the "trick" of paralleling active devices to play off the correlated signal vs. uncorrelated noise for extra s/n ratio? The technique is often mentioned but I can't really say I know of anyone doing it....

Yes and no. It is a well known trick. There is a semiconductor mic pre, can't remember the manufacturer, that uses about six transistors in parallel in the first stage. I also know someone on this group has built a point to point version of my Eurocard tube mic pre. Since he built just one channel there was half of a 12AX7 input tube going spare so I did some sims and gave him some components changes for a version where the two triodes in the 12AX7 were paralleled. In theory that should reduce the tube noise by 3dB.

Cheers

Ian
 
JAY X said:
In my case there are three boards : left ch, right ch, and psu board. Both channels are powered with 20 way idc cable.... 10 ways for relays ( 2 pins for power and the rest for ground) and other 10 ways for opamps ( 6 pins for power and 4 pins for ground)... so it is clear i have to reinforce the ground connections from the boards to the psu...maybe connecting both boards with 50way idc cable to their ground plane...and then to the psu... someway... besides, i don't know how a bus bar comes into the play.... i have to investigate!!! :)
Ground resistance is crucial regarding longitudinal noise (noises that come from "bad" voltage developing between different stages). They appear as hum/buzz (most of the time) and spurious (RFI/EMI). But it seems you're complaining about hiss...?
 
Hi abbey!

Yes!...hiissss is my worry!.. :p...and if a little can be done about it...
Despite i'm happier than last  version i built...so i achieved an improvement, but still there is work to do..

JAY X
 
tubegeek said:
Ian:

Have you ever done the "trick" of paralleling active devices to play off the correlated signal vs. uncorrelated noise for extra s/n ratio? The technique is often mentioned but I can't really say I know of anyone doing it....

This made sense before really low noise devices were available. In the late '70s early '80s I came across some low noise transistors (2sb737/2sd786)  that were designed for moving coil phono head amps, so optimized for source impedance down around 10 ohms. For perspective those transistors were roughly 20dB less noise voltage than ne5534. While the noise if a bare part is never the final result.

Another issue with paralleling devices is the capacitance goes up too so HF response can suffer in the extreme when paralleling devices.

If you look inside the old LM394 low noise dual transistor it used a physical base-emitter structure that consisted of many parallel fingers, not unlike parallel devices. 

JR

PS: There are many opamps quieter than 5534 these days.
 
I know this is going to sound pretty obvious, but ensure the quality of your incoming signals/cables are good. If you're compromising your feeds into the summing amp, then it's going to be noise + noise + noise +... = way too much noise.

If the PCB is already made, it will be tricky to adhere to newly learned design principles.  :-\
 
JAY X said:
I used bipolar opamps NE5532 and LM833,  i try with fet opamp OP275,  and reduced a bit the noise...
You may consider using VLN opamps, that are designed for low source impedance operation (about 500-1000 ohms) like the AD8599 or LT1115 (unfortunately a single so you would need an adapter). They may reduce noise by as much as 12dB compared to a 5532.
You say your measurements have been done with a souncard. Was the sampling rate 44.1k? If you do the measurements at a different sampling rate, the noise bandwidth is higher (24kHz at 48k SR, 44kHz at 88.2k SR, and so on), so the measurement is worse.
If we had a schemo, we could help you more.
You say you measured -74.
With 10dB gain in the fader amp, that's -84 for the mixing amp. If it has 16 inputs, that means a noise gain of about 24dB, which computes at -108 for the opamp, which seems about correct for me.
The Neve summing amp uses discrete components that give a noise of about -130dBu, some 20dB improvement over a 5532. Now I bet their measurement is done at unity gain (which is the only correct way of doing).
In your case, without the 10dB gain, the -84 figure is just about what one can expect from a 5532/LM833/OP275-based circuit.
Correct me if I'm wrong about the measurement conditions (10dB gain, 16 inputs, soundcard SR...)
 
Hi,

Today i performed an audio test of my summing mixer design.

I run a 16 bits wave file 0 db level, 1 khz test tone thru the mixer and adjusted the gain of the output till the recording meter in soundforge read -0.3 db. All the faders in my soundcard all way up 100%. Winamp player also 100% at full volume.

After passing the test tone thru the mixer and recording the result in soundforge, i played it again to listen about the noise levels. At the silent passages (floor noise only)  i run the statistics in soundforge, and these were the results:

Min sample value: -3 -80,77 db L & R
Max sample value: 2 -84,29 db L & R
RMS POWER: -90,36 db L , -90,23 db R

I know this is not a full compliant lab audio test, but may be useful to say if this unit is on the right track, regarding noise.
I would like to post some screen captures, but i don't know how to post an image.. sorry! :-[

Thank you for your comments,

JAY X
 

Latest posts

Back
Top