Questions about reducing noise ina an audio design

[JAY X]

Hi All!!

Happy New year!! let's see how it does this year!! 8)

let's see.…I did not  know where to post the following questions because they are related with different topics. I hope they may be of interest for the diy community.

The long story short.. I have built a summing mixer with opamps and line drivers. Very low noise opamp LM4562.. Line driver: DRV134,  22k balanced summing inputs.(1/4w resistors 1%)
At unity gain I have a noise floor of –80.2db with one input routed, Reading with soundforge meter. But this may not mean nothing.

1) PARTS QUALITY

All these parts are of good quality… ok…they need care regarding supply decoupling… and here is maybe my first mistake…

I usually use two 100nf ceramic for each opamp…and go ahead.. maybe for NE5532 this can work, but after reading spec scheets of the parts I used, 100nf is not enough.

LM4562 ask for 47uf across each rail…
DRV 134 ask for 1uf across each rail… ¿does it make sense? maybe with these values i can reduce even more supply rail noise...??

2) PSU NOISE

let's see.… I have built a psu for my summing mixer design, and now  I try to tweak it to reduce the noise injected into the circuits…
The psu I built around LM317/ 337. CRC filter: 2200uf-2r2-2200uf- 2r2- 2200uf.

2A) ¿is this equivalent to a 6600uf filter?

2B) Maybe I can replace the 2r2 resistors by 10mh inductors…?? And have a PI filter?.

2C) Output capacitors: 100uf each rail +/-17v… maybe add another 470uf in parallel ?

2D)  Could not measure correctly the ripple with my  cheap multimeter…¿ In AC mode and  selecting microvolt range right? Then convert to dbu or dbv… I have electrodroid in my mobile…

3) MEASURING NOISE

Well, here is my bootleneck….got no picoscope… just a PC oscilloscope software working with my soundcard…. An old ESI PRO WT192M. (2003). Routing an output of the soundcard to its  input I read –70 db…  so any measure of mixer noise would be contaminated by soundcard noise…

Thank you very much for your comments,
JAY X

 

[ruffrecords]

You first need to get a reasonably reliable means of measuring noise. For now Soundforge is probably your best bet. The first thing to to is make usre what you are reading really is the noise from your summer. To do this you need to check the Soundforge meter with the input shorted. This will give you an idea of the lowest noisr you can measure. If it reads below -90dBu then it will do for now.

Secondly, it is hard to diagnose the likely source of noise without a schematic. Please post the compete schematic of your summer.

Cheers

Ian

 

[PRR]

Hiss or hum?

 

[JAY X]

Hi Ian, PRR,

I will try to short the inputs as you say...see what happens...failed to upload a picture...i'll try later .

The basic structure of the mixer: BAL SUM AMP(22k  inputs & 75R/ 1000uf outputs) -> the balanced SUM out forms the (INSERT SEND) --> INSERT RETURN (HALF NORMALLED) --> OPAMP 10K DIFF AMP( with RFI FILTER and bipolar caps)---> FADER AMP 10db-----> BAL OUT (DRV134) (+6db). all opamps LM4562.

PRR, the noise is kind hiss.. white noise... I suspect from the master fader amplifier. it has 10db gain, despite i only have to use 2 or 3 db to get unity gain..

There are three boards: left ch, right ch and psu board. Each channel is powered thru one IDC 10 cable, 2 pwr pins and the rest to reinforce ground.... but it may not be enough...

Besides, as power bypass at each LM 4562opamp i used 100nf, when the spec scheets of the parts ask for 1uf (drv134)and 47 uf...(Lm4562)..

JAY X

 

[ruffrecords]

We still need the schematic. How many inputs are there?

Cheers

Ian

 

[audiomixer]

depending on the choice of feedback resistors the LM 4562 is not the lowest noise option out there. it has suite a bit more current noise then say a NE5532. so with highish resistors you would might instroduce more noise then with say a NE5532. balanced inputs are not the lowest noise either, at least the generally used design. I believe you have two (input, send return) so that adds quite a bit of noise as well.

you might try to reduce impedance were possible.

- Michael

 

[JohnRoberts]

Everything makes noise. That's why they measure signal-to-noise ratios.

JR

 

[PRR]

> the noise is kind hiss.. white noise...

Then your first suspect will NOT be power supply or supply pin decoupling. Power supply hum/buzz is usually FAR worse than its hiss.

(Yes, a chip in ultrasonic howl *can* also hiss; but there's usually other clues that it ain't happy.)

I don't think -80dB is horrible. *Especially* since you then say you are testing with 7-8dB more gain than you need. -87dB is fantastic. Nobody but maniacs gets 96dB on the CD. Few performances touch 96dB. Few listening situations approach 96dB.

There's knobs on that thing, no? Change them. What does that do to the hiss?

A logical approach:

Everything at zero. What is the output hiss?

Raise the Master. What hiss at Full? What setting of the Master gives "slight" increase of hiss re: the full-down position?

Repeat for the Channel faders. Hiss at full, and hiss at "slight increase of output hiss".

With Master and a Channel at their "slight increase" positions, is the overall gain useful?

 

[JAY X]

Hi,

This morning i run a simple noise level test.
One input routed  (both left & Right). TRS jack
The rest of inputs are in db25 format. (not used in the test)

fader position        noise level (Left & Right ch)

0  (-inf)                      -80.8 db
1                                -80.8 db
2                                -80.8db
3  (unity gain)            -80.8 db (Unity Gain)
4                                -78.3 db
5                                -78.3 db
6                                -76.3 db
7                                -76.3db
8                                -73.4 db
9                                -73.4 db
10                              -72.2 db       

I think i can't improve it much more because of my layout. Layout is the key in an audio design.
Despite, i'm quite happy with the results achieved right now.

This is why most makers add at the end of user manual...:" the design would be improved without further notice..."  ;D

JAY X.

 

[ruffrecords]

You talk about inputs being 'routed'; does this mean there are routing or muting buttons?  We really do need a schematic to be able to help you.

Cheers

Ian

 

[abbey road d enfer]

Hi,

This morning i run a simple noise level test.
One input routed  (both left & Right). TRS jack
The rest of inputs are in db25 format. (not used in the test)

fader position        noise level (Left & Right ch)

0  (-inf)                      -80.8 db
1                                -80.8 db
2                                -80.8db
3  (unity gain)            -80.8 db (Unity Gain)
4                                -78.3 db
5                                -78.3 db
6                                -76.3 db
7                                -76.3db
8                                -73.4 db
9                                -73.4 db
10                              -72.2 db       

I think i can't improve it much more because of my layout. Layout is the key in an audio design.
Despite, i'm quite happy with the results achieved right now.

This is why most makers add at the end of user manual...:" the design would be improved without further notice..."  ;D

JAY X.

there are many unknowns there: I assume you run SoundForge at Single-Speed (44.1 or 48k). Running SF at DS would deteriorate the results by 3dB.
The meters in SF, I don't think they are RMS. Most meters in DAW's are by default Peak meters.
Peak meter indications are typically about 6-10dB above RMS on pink noise.

Your figures tend to indicate EIN of ca. -102dBu for the fader amp, which is good.

If your description of "One input routed" means all the others are left floating, the noise level of the mix amp seems to be about -94dBu, which is not bad considering the best-case calculated noise is about -100.