• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DSP Measurements and Rising Noise Floor

sfdoddsy

Active Member
Joined
Jun 6, 2019
Messages
291
Likes
429
I've been using Digital EQ for many years, starting the now venerable Behringer DCX2496, moving on to better pro units such as dbx driveracks, then various MiniDSP units and auto EQ systems.

Measurements clearly showed the early units added noise, and in the case of those with analog inputs so did listening unless really careful attention was paid to gain structure.

However (for me) since the MiniDSP onwards (and again paying attention to gain) noise has ceased to be an issue.

I'm sure it is measurable. But I can't hear it and the benefit of DSP EQ is so obvious that even if I could hear noise with my ear against the driver I wouldn't care.

The only noise that bugs me these days is that from poorly designed amps in active speakers.
 

Sokel

Major Contributor
Joined
Sep 8, 2021
Messages
3,287
Likes
2,923
The only noise that bugs me these days is that from poorly designed amps in active speakers.
Yep,we still remember the diyaudio thread some 5-6 years ago who someone dismantled them and their measurements put them in 60-70 SINAD area.
On the other hand the end result is inaudible (apart from the hiss which is intorelable in some of them).So...
 

sfdoddsy

Active Member
Joined
Jun 6, 2019
Messages
291
Likes
429
Well, that's the SINAD problem isn't it. Given that the noise level of a room is vastly worse than the SINAD of even the worst electronics, does it matter?
 
Last edited:

jensgk

Active Member
Forum Donor
Joined
Mar 21, 2020
Messages
253
Likes
555
Location
Denmark
You do often see 96dB dynamic range stated for undithered 16bit audio (you said 98dB but Wikipedia says 96dB like just about everywhere else) but that doesn't really make sense if you think about it. 96dB comes from the calculation 6dB x 16 = 96dB, i.e. every bit gives additional 6dB SNR. However, with audio we're always talking about AC signals and one of the 16bits represents the polarity of the signal. The smallest AC signal you could encode would be oscillating from -1 to +1 (1 least significant bit or LSB). And the largest would be +32767 to -32767. The amplitude difference between these is 20*log10(32767) = 90.3dB.
I don't understand that. If you have a signed 16 bit integer, it has 65536 different values, from −32768 to 32767.
As far as I can see, the sign does not change the precision. The precision does not change by subtracting a constant, - in this case 32768
 

Sokel

Major Contributor
Joined
Sep 8, 2021
Messages
3,287
Likes
2,923
Well, that's the SINAB problem isn't it. Given that the noise level of a room is vastly worse than the SINAD of even the worst electronics, does it matter?
My dedicated is 250 m³ with a noise floor 33dbA at it's best.
I listen at 3 meters.
Klipell distortion hearing test puts me in the -70db area I think (translated,I don't remember their numbers,they were different).
Do I need to say any more?
I know my (personal) SINAD chase is just a vice after all that,just for the fun of it.
 

jackocleebrown

Member
Technical Expert
Audio Company
Joined
Jan 30, 2020
Messages
54
Likes
712
Location
UK
I don't understand that. If you have a signed 16 bit integer, it has 65536 different values, from −32768 to 32767.
As far as I can see, the sign does not change the precision. The precision does not change by subtracting a constant, - in this case 32768
You're quite right. It was clarified in this post. If you allow 0.5LSB of DC then you can record a signal of 1LSB peak to peak amplitude without dither.
 

Trell

Major Contributor
Joined
May 13, 2021
Messages
2,532
Likes
2,921
Well, that's the SINAD problem isn't it. Given that the noise level of a room is vastly worse than the SINAD of even the worst electronics, does it matter?

Hiss can be an issue with some active speakers, especially near-field like in a desktop setup.
 
OP
M

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,027
Likes
2,713
Location
Detroit, MI
Well, that's the SINAD problem isn't it. Given that the noise level of a room is vastly worse than the SINAD of even the worst electronics, does it matter?

With poor gain structure it is easy to combine relatively high SINAD components and end up with hiss, especially if you are using digital volume control. The important issue here is residual noise at the amplifier output coupled with speaker sensitivity.

From experience an audio interface like a Focusrite 18i20 2nd gen (106 dB SNR at 4.86 V) coupled with a Hypex NC252MP amplifier (100 dB SNR at 5 W in to 4 ohm, 25.6 dB gain) will result in audible hiss with an unpadded highish sensitivity (92 dB) tweeter. Residual noise at the amplifier output in such a setup is > 450 uV.

The noise floor rise described in this thread as a result of applying filters in a DSP is much less audible as it is only present when high level signal is present which provides a lot of masking. When no or low level signal is applied the noise floor is much better with these DSPs.

Michael
 

Head_Unit

Major Contributor
Forum Donor
Joined
Aug 27, 2018
Messages
1,170
Likes
583
This showed that the elevated noise floor is level dependent.
Therefore with music signals, the noise floor will "pump" (modulate)? I remember a brief moment in time where someone was showing noise floor modulation of CD players, not sure why that measurement disappeared...
 

Head_Unit

Major Contributor
Forum Donor
Joined
Aug 27, 2018
Messages
1,170
Likes
583
From experience an audio interface like a Focusrite 18i20 2nd gen (106 dB SNR at 4.86 V) coupled with a Hypex NC252MP amplifier (100 dB SNR at 5 W in to 4 ohm, 25.6 dB gain) will result in audible hiss with an unpadded highish sensitivity (92 dB) tweeter. Residual noise at the amplifier output in such a setup is > 450 uV.
To put words in your mouth, it's not the residual noise of the amplifier itself causing hiss, but the amplification of hiss from earlier stages? That might explain why I see so many amps tested with SINAD too low to possibly have hiss, yet some people do complain about hiss. Now I will have to look at tests of integrated amps maybe and see how that data is...it's also why I would like to see a straight residual noise value without SINAD...
 
OP
M

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,027
Likes
2,713
Location
Detroit, MI
Therefore with music signals, the noise floor will "pump" (modulate)? I remember a brief moment in time where someone was showing noise floor modulation of CD players, not sure why that measurement disappeared...

Yes, it is some sort of noise floor modulation, although given the low absolute level it will be lower than the noise floor in the recording.

Michael
 
OP
M

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,027
Likes
2,713
Location
Detroit, MI
To put words in your mouth, it's not the residual noise of the amplifier itself causing hiss, but the amplification of hiss from earlier stages? That might explain why I see so many amps tested with SINAD too low to possibly have hiss, yet some people do complain about hiss. Now I will have to look at tests of integrated amps maybe and see how that data is...it's also why I would like to see a straight residual noise value without SINAD...

It is the combination of amplifier residual noise and amplified residual noise from early stages (including noise from the source music itself!). Which is the dominant factor really depends on the specific components. ASR reviews of DACs and amplifiers give you enough data to estimate the residual noise, for DACs you have SNR at 2 or 4 V output and for amplifiers you have the gain and SNR at 5 W in to 4 ohm.

An interesting experiment is to listen to setups with varying levels of residual noise and see where the audibility threshold is for your specific ears, speakers and listening environment. I did this here -> https://www.audiosciencereview.com/...topping-performance.24768/page-2#post-1008785 and now shoot for residual noise in the 100 uVs.

From what I've seen there are a few categories of people who complain about noise. 1) Folks using active monitors which have high noise DAC / amps and potentially tweeters with no padding. 2) Those with relatively high noise components on the front end of their system like miniDSP 2x4s followed by high (30+ dB) gain amplifiers. 3) Those using extremely high sensitivity speakers, especially if driven directly by an amplifier with typical gain.

Michael
 

spalmgre

Member
Joined
May 22, 2019
Messages
38
Likes
12
Location
Helsinki
Well, that's the SINAD problem isn't it. Given that the noise level of a room is vastly worse than the SINAD of even the worst electronics, does it matter?
It is completely wrong to say that you can not hear noise from the speakers if the level is below the noisfloor in the room.
 

Head_Unit

Major Contributor
Forum Donor
Joined
Aug 27, 2018
Messages
1,170
Likes
583
Well, that's the SINAD problem isn't it. Given that the noise level of a room is vastly worse than the SINAD of even the worst electronics, does it matter?
It is completely wrong to say that you can not hear noise from the speakers if the level is below the noisfloor in the room.
Having worked with a friend who has been doing noise monitoring in National Parks for decades now, let me categorically say it is NOT so simple as comparing two numbers. Numbers ignore
- The spectrum of the noise. 70 dB of treble noise might easily be heard over 80 dB of bass noise or whatever
- Our ears have directional hearing, not captured in one noise number.
- Human brains can pick out periodicity for instance the whup-whup-whup of helicopter rotors over the white-ish noise of a waterfall or such.
It's also possible you might hear hiss with your ear next to a tweeter, and after that it always bothers you because you know it's there, even if it is not audible at the usual listening position.
 

spalmgre

Member
Joined
May 22, 2019
Messages
38
Likes
12
Location
Helsinki
It's also possible you might hear hiss with your ear next to a tweeter, and after that it always bothers you because you know it's there, even if it is not audible at the usual listening position.
So I have JBL compression driver horns and also JBL 10" 2123 mid transducer with reported 101db sensitivity, 1W, 1m. These I drive actively with separate power amp channels.

I have tried several D-class amplifier boards that do have noise that is clearly audible. So I have installed L-pads to get the noise (hiss) down.

Now I can still hear the noise close to the speaker but not in the listening position. The L-pad is damping about 15db.
I did several iterations before finally setting it for the 15db level. I was not satisfied until I could not hear any noise in the listening position.

So at least in this case, I don't hear the noise even though I know it's there as I can hear it close to the speaker.

But I don't say that we HiFi enthusiasts can't imagine and hear all kinds of things that are not there.
 

Attachments

  • IMG_3463.JPG
    IMG_3463.JPG
    599.3 KB · Views: 37
Last edited:

spalmgre

Member
Joined
May 22, 2019
Messages
38
Likes
12
Location
Helsinki
Is that with a source? Or even with no inputs connected?
No source connected and also tested with Najda DSP preamp with internal 8 channel DAC connected. Najda is entirely silent so same noise with or without a source.

Tested D-amp bords with chip 3116 and also Tripath TK2050 amplifiers. I want to point out that the noise I hear is with very sensitive +100db/1W speaker drivers. See attached picture where I use the 3116 board with a SEAS driver. This setup has no noise that I can hear.

I drive my main speakers actively 8 channels and like to stay on a budget as I think the DIY HiFi don't need to be expensive.

The needed power is very low but all quality D-class amps seem to seek hi power handling. Lesser power -amplifiers like Topping PA5 have a very unreliable reputation but would be interesting.

Now I have ordered Connex Electronic LM3886SMPS without IC (69.99$, they have problems locating enough amp chips) as I have been able to buy genuine LM3886 locally in Helsinki. There is the SMPS and stereo amplifier on a 10x10x3.6 cm board.

LM3886 datasheet states "typical noise floor 2.0uV so we shall see how that works out.

 

Attachments

  • 3116 amp.jpg
    3116 amp.jpg
    241.8 KB · Views: 20

spalmgre

Member
Joined
May 22, 2019
Messages
38
Likes
12
Location
Helsinki
Is that board a colab with neurochrome, or is it just coincidence that the Modulus-286 tested here used SMPS from Connex?
Not a coincidence but both companies have their own LM3886 boards. It might be that one has looked at the other's design. But so has everybody who has built an LM3886 amplifier.

Not a coincidence as the Neurocrome Modulus amplifier needed an SMPS and Connex Electronics has a reputation for delivering good boards for audio applications.

-----------------------------------

Connex Electronic LM3886SMPS is a Connecx Electronic produced LM3886 board.

Neurocrome is a company that also produces LM3886 boards that use the brand name "Modulus"

Connex Electronic does also produce SMPS Switching-mode Power Supplies. These SMPS boars can also be bought from the HiFiMeDIY webshop.
I have emailed and also ordered from both Connex Electronics and HiFiMeDIY and have got good answers and delivery. This is the amp board from hifimediy that I bought and it has worked well minus the noise with high efficient horn drivers (see earlier posts).

The amplifier that Audio Science Review did evaluate is a Neurocrome-produced Modulus board amplifier that uses Connex Electronics SMPS.
 
Last edited:

somebodyelse

Major Contributor
Joined
Dec 5, 2018
Messages
3,040
Likes
2,321
I was thinking more along the line of the "LM3886 Done Right" and associated articles - the Modulus is a whole other beast. IIRC Neurochrome worked on the MiniDSP Studio series so doing work for Connex seemed at least plausible, but it's a widely used amp chip so could just as easily be independent. I just wondered if you knew either way.
 
Top Bottom