Portable recorder noise specs

[jason]

This is from a company that sells gear for recording wildlife. Given
all the discussions here about flash-based recorders it might be
of interest.

http://www.avisoft.com/recordertests.htm

[[email protected]]

http://www.avisoft.com/recordertests.htm

they have some good recordings of animal / insects/ bird sounds as well...
thanks
Mark

[PStamler]

I note that some very popular recorders have exceedingly poor input noise results on these tests. Let the buyer beware! (Or buy a Sound Devices.)

Peace,
Paul

[Trevor]

"PStamler" > wrote in message
...
>I note that some very popular recorders have exceedingly poor input noise
>results on these tests.
>Let the buyer beware! (Or buy a Sound Devices.)

As it says on that page :
"The dynamic range at the maximum available gain setting alone should not be
interpreted as an indicator for the performance of a recorder because these
numbers will of course depend on the selected gain setting."

It's useless comparing the Dynamic Range at maximum gain setting when the
amount of possible gain varies, they should be compared at a standardised
gain setting with the input signal itself varied to give 0dBFS IMO. Clearly
a recorder with 70dB of possible gain will give a worse result on that table
than one with 40dB of gain.
I didn't check the sound files though, hopefully they provide real clues to
actual performance.

Trevor.

[Mike Rivers[_2_]]

On 12/21/2013 11:08 PM, Trevor wrote:
> It's useless comparing the Dynamic Range at maximum gain setting when the
> amount of possible gain varies, they should be compared at a standardised
> gain setting with the input signal itself varied to give 0dBFS IMO. Clearly
> a recorder with 70dB of possible gain will give a worse result on that table
> than one with 40dB of gain.

The problem with using "gain" as part of any measurement on an
integrated digital device is that there's no output that's in the same
units as the input. You have volts (or dBu) going in, and this becomes
dBFS, depending on the "gain." What you can compare is input sensitivity
for 0 dBFS at maximum (and minimum, if it's not zero) setting of the
input level control. There's no such thing as a standardized gain
setting other than "maximum."



--
For a good time, visit http://mikeriversaudio.wordpress.com

[Trevor]

"Mike Rivers" > wrote in message
...
> On 12/21/2013 11:08 PM, Trevor wrote:
>> It's useless comparing the Dynamic Range at maximum gain setting when the
>> amount of possible gain varies, they should be compared at a standardised
>> gain setting with the input signal itself varied to give 0dBFS IMO.
>> Clearly
>> a recorder with 70dB of possible gain will give a worse result on that
>> table than one with 40dB of gain.
>
> The problem with using "gain" as part of any measurement on an integrated
> digital device is that there's no output that's in the same units as the
> input. You have volts (or dBu) going in, and this becomes dBFS, depending
> on the "gain." What you can compare is input sensitivity for 0 dBFS at
> maximum (and minimum, if it's not zero) setting of the input level
> control. There's no such thing as a standardized gain setting other than
> "maximum."

Fair enough if you don't know and cant measure the voltage required for
0dBFS. But that still leaves the table as useless. You can at least
use a standard input voltage in that case, as the 0dBFS overload voltage
will probably be a lot closer than the maximum pre-amp gains, ie a lot less
than 30dB variation.
The pdf graphs show the problem, the real world noise performance is pretty
close for many of the devices, and it also shows the poor distortion
performance of the Sound Devices unit, making it's ranking pure speculation
on similar units not being faulty, and indicating an unsatisfactory level
of quality control at least.
As usual, you don't always get what you pay for. :-(

Trevor.

[Mike Rivers[_2_]]

On 12/23/2013 5:25 AM, Trevor wrote:

> "Mike Rivers" > wrote in message
>> The problem with using "gain" as part of any measurement on an integrated
>> digital device is that there's no output that's in the same units as the
>> input. You have volts (or dBu) going in, and this becomes dBFS, depending
>> on the "gain." What you can compare is input sensitivity for 0 dBFS

> Fair enough if you don't know and cant measure the voltage required for
> 0dBFS.

It's easy to measure. If you can't measure it, you probably shouldn't
worry about it. When I review an interface these days, I measure and
report the level required to reach 0 dBFS, but that really only helps
most readers when comparing one unit to another if input sensitivity is
important (as it is when recording wildlife sounds). I usually will
include an "in practice" statement such as "With the input gain turned
up full, speaking 1 foot away from an SM57 mic in a normal
conversational voice produces peaks around -12 dBFS." Of course this
doesn't reference any known standard (you don't know how loud I talk)
but at least it tells the reader that he may need to crank it up all the
way and/or get closer to the mic if he wants to see the meter hit full
scale.

> You can at least
> use a standard input voltage in that case, as the 0dBFS overload voltage
> will probably be a lot closer than the maximum pre-amp gains, ie a lot less
> than 30dB variation.

This isn't really any more helpful than measuring the voltage for 0 dBFS
since most pepole don't know how many millivolts are coming out of their
microphone. You can find microphone sensitivity (usually in millivolts
per Pascal) on some spec sheets, but how many people know that 1 Pascal
= 94 dB SPL (pretty loud)?

> The pdf graphs show the problem, the real world noise performance is pretty
> close for many of the devices

I've found that to be pretty much true. The only thing that's really
impressed me lately has been the Focusrite Forte (I know that's not a
portable recorder) because with the gain set for about the same
quiescent noise as a more typical unit, 0 dBFS at about a 6 dB lower
level than what's more typical. But then "typical" covers a pretty wide
range.

The value of any measurement (or specification) is in how well you
understand and believe in the legitimacy of how the measurement was
made. When I measure quiescent noise of a device with an external mic
input, I terminate that input with a 150 ohm resistor, which
approximates the noise resulting from the source impedance of a typical
mic. But if there's no external mic input, since I don't have a
sufficiently good "quiet box", the best I can do is put the recorder
under a pile of pillows at a time when a neighbor isn't running a leaf
blower, chain saw, or lawn mower (night time is good).

> As usual, you don't always get what you pay for. :-(

Oh, you usually do, but sometimes you can't get everything you want no
matter how much you pay.


--
For a good time, visit http://mikeriversaudio.wordpress.com

[William Sommerwerck]

I have not been following this thread in great detail, but I'll stick my nose
in, anyhow.

If I were buying a digital recorder //with an integral microphone//, the noise
spec I'd like to see is the following...

For a given SPL at the mic (90dB, say), and with the record level at full
scale -- what is the residual noise when the recording is played back?

That wouldn't necessarily answer every question -- but it would be a good
start.

[Les Cargill[_4_]]

William Sommerwerck wrote:
> I have not been following this thread in great detail, but I'll stick my
> nose in, anyhow.
>
> If I were buying a digital recorder //with an integral microphone//, the
> noise spec I'd like to see is the following...
>
> For a given SPL at the mic (90dB, say), and with the record level at
> full scale -- what is the residual noise when the recording is played back?
>

Isn't "residual noise" the level of erased analog tape?

How do you calculate the residual noise under a 90dB sound source? And
wouldn't the answer be "whatever room tone happens to be"? Even an
anechoic chamber has some amount of room tone.

> That wouldn't necessarily answer every question -- but it would be a
> good start.

--
Les Cargill

[Les Cargill[_4_]]

William Sommerwerck wrote:
> "Les Cargill" wrote in message ...
> William Sommerwerck wrote:
>
>> I have not been following this thread in great detail, but I'll stick my
>> nose in, anyhow.
>
>> If I were buying a digital recorder //with an integral microphone//, the
>> noise spec I'd like to see is the following...
>
>> For a given SPL at the mic (90dB, say), and with the record level at
>> full scale -- what is the residual noise when the recording is played
>> back?
>
> Isn't "residual noise" the level of erased analog tape?
>
> How do you calculate the residual noise under a 90dB sound source? And
> wouldn't the answer be "whatever room tone happens to be"? Even an
> anechoic chamber has some amount of room tone.
>
>
> My heavens! The point I'm making is that this noise measurement is
> related to a real-world recording situation.


got that, but I still don't know very much about what the measurement is.

I expect that unless the measurement is made in a controlled space, the
answer will tell you nothing about what the recorder is capable of.

SFAIK, there's no microphone equivalent of a 600 ohm resistor
across an input in shunt.

--
Les Cargill

[Scott Dorsey]

Les Cargill > wrote:
>got that, but I still don't know very much about what the measurement is.
>
>I expect that unless the measurement is made in a controlled space, the
>answer will tell you nothing about what the recorder is capable of.

I think that is the point. If you're going to measure anything to do
with the microphone, you have to do it in a controlled space.

>SFAIK, there's no microphone equivalent of a 600 ohm resistor
>across an input in shunt.

That's true, so you're stuck having to use a more sensitive microphone
as a reference to verify how quiet your quiet room is.

Alternately it's possible to measure the electronics with the microphone
capsule disconnected or unpowered, and then determine the thermal noise
of the microphone capsule in isolation (in a controlled space with
calibrated electronics) and then add the two.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[William Sommerwerck]

"Scott Dorsey" wrote in message ...
Les Cargill > wrote:

>>> got that, but I still don't know very much about what the measurement is.

>> I expect that unless the measurement is made in a controlled space, the
>> answer will tell you nothing about what the recorder is capable of.

> I think that is the point. If you're going to measure anything to do
> with the microphone, you have to do it in a controlled space.

>> SFAIK, there's no microphone equivalent of a 600 ohm resistor
>> across an input in shunt.

> That's true, so you're stuck having to use a more sensitive microphone
> as a reference to verify how quiet your quiet room is.

> Alternately it's possible to measure the electronics with the microphone
> capsule disconnected or unpowered, and then determine the thermal noise
> of the microphone capsule in isolation (in a controlled space with
> calibrated electronics) and then add the two.


Golly, Mr Kent... Why is there confusion about this? Looking at these little
recorders //as closed systems//, we're not interested in the noise of the mic
capsules, or the preamp, or the ADA processing. All we want to know is how
audibly noisy the playback will be. The measurement I suggested is one
approach to determining that.

[Les Cargill[_4_]]

William Sommerwerck wrote:
> "Scott Dorsey" wrote in message ...
> Les Cargill > wrote:
>
>>>> got that, but I still don't know very much about what the
>>>> measurement is.
>
>>> I expect that unless the measurement is made in a controlled space, the
>>> answer will tell you nothing about what the recorder is capable of.
>
>> I think that is the point. If you're going to measure anything to do
>> with the microphone, you have to do it in a controlled space.
>
>>> SFAIK, there's no microphone equivalent of a 600 ohm resistor
>>> across an input in shunt.
>
>> That's true, so you're stuck having to use a more sensitive microphone
>> as a reference to verify how quiet your quiet room is.
>
>> Alternately it's possible to measure the electronics with the microphone
>> capsule disconnected or unpowered, and then determine the thermal noise
>> of the microphone capsule in isolation (in a controlled space with
>> calibrated electronics) and then add the two.
>
>
> Golly, Mr Kent... Why is there confusion about this? Looking at these
> little recorders //as closed systems//, we're not interested in the
> noise of the mic capsules, or the preamp, or the ADA processing. All we
> want to know is how audibly noisy the playback will be. The measurement
> I suggested is one approach to determining that.

There is a hole in your story.

The overall EIN of the recorder/preamps/mics could be... say, -90dB and
room tone could easily be 40 dB above that.

--
Les Cargill

[William Sommerwerck]

"Les Cargill" wrote in message ...
William Sommerwerck wrote:

>> Golly, Mr Kent... Why is there confusion about this? Looking at these
>> little recorders //as closed systems//, we're not interested in the
>> noise of the mic capsules, or the preamp, or the ADA processing. All we
>> want to know is how audibly noisy the playback will be. The measurement
>> I suggested is one approach to determining that.

> There is a hole in your story.
> The overall EIN of the recorder/preamps/mics could be... say,
> -90dB and room tone could easily be 40 dB above that.

Okay. Then you measure the unit in an anechoic chamber. I assume Zoom can
afford to make that measurement.

Let me spell this out, because I know that what I've just said will be
deliberately misinterpreted.

You take the recorder and a sound-level meter to a concert, and note that for
a 90dB sound level, the volume control on the recorder has to be set to, say,
7.5 to get a full-scale recording. Okay?

Then you walk into the anechoic chamber and set the recorder to 7.5 and make a
recording of nothing. This should give a close approximation of where the
recorder's noise floor -- considered as a system -- is, with respect to a 90dB
sound level.

[Les Cargill[_4_]]

William Sommerwerck wrote:
> "Les Cargill" wrote in message ...
> William Sommerwerck wrote:
>
>>> Golly, Mr Kent... Why is there confusion about this? Looking at these
>>> little recorders //as closed systems//, we're not interested in the
>>> noise of the mic capsules, or the preamp, or the ADA processing. All we
>>> want to know is how audibly noisy the playback will be. The measurement
>>> I suggested is one approach to determining that.
>
>> There is a hole in your story.
>> The overall EIN of the recorder/preamps/mics could be... say,
>> -90dB and room tone could easily be 40 dB above that.
>
> Okay. Then you measure the unit in an anechoic chamber. I assume Zoom
> can afford to make that measurement.
>
> Let me spell this out, because I know that what I've just said will be
> deliberately misinterpreted.
>
> You take the recorder and a sound-level meter to a concert, and note
> that for a 90dB sound level, the volume control on the recorder has to
> be set to, say, 7.5 to get a full-scale recording. Okay?
>
> Then you walk into the anechoic chamber and set the recorder to 7.5 and
> make a recording of nothing. This should give a close approximation of
> where the recorder's noise floor -- considered as a system -- is, with
> respect to a 90dB sound level.


That makes perfect sense - thank you for clarifying it.

--
Les Cargill

[William Sommerwerck]

"Les Cargill" wrote in message ...

> That makes perfect sense -- thank you for clarifying it.

You're welcome.

[None]

"William Sommerwerck" > wrote in message
...
>> There is a hole in your story.
>> The overall EIN of the recorder/preamps/mics could be... say,
>> -90dB and room tone could easily be 40 dB above that.
>
> Okay. Then you measure the unit in an anechoic chamber. I assume
> Zoom can afford to make that measurement.
>
> Let me spell this out, because I know that what I've just said will
> be deliberately misinterpreted.

LOL! Yeah, try to blame everyone else for your inability to write a
clear and adequate description of an audio measurement. Yet another
example of your delusion about being always right, or usually right,
or whatever bull**** you claim and nobody believes. Your own failings
always seem to get blamed on everyone but you. High comedy, that!

[jason]

On Mon, 23 Dec 2013 18:19:15 -0600 "Les Cargill" >
wrote in article >
>
> William Sommerwerck wrote:
> > "Les Cargill" wrote in message ...
> > William Sommerwerck wrote:
> >
> >>> Golly, Mr Kent... Why is there confusion about this? Looking at these
> >>> little recorders //as closed systems//, we're not interested in the
> >>> noise of the mic capsules, or the preamp, or the ADA processing. All we
> >>> want to know is how audibly noisy the playback will be. The measurement
> >>> I suggested is one approach to determining that.
> >
> >> There is a hole in your story.
> >> The overall EIN of the recorder/preamps/mics could be... say,
> >> -90dB and room tone could easily be 40 dB above that.
> >
> > Okay. Then you measure the unit in an anechoic chamber. I assume Zoom
> > can afford to make that measurement.
> >
> > Let me spell this out, because I know that what I've just said will be
> > deliberately misinterpreted.
> >
> > You take the recorder and a sound-level meter to a concert, and note
> > that for a 90dB sound level, the volume control on the recorder has to
> > be set to, say, 7.5 to get a full-scale recording. Okay?
> >
> > Then you walk into the anechoic chamber and set the recorder to 7.5 and
> > make a recording of nothing. This should give a close approximation of
> > where the recorder's noise floor -- considered as a system -- is, with
> > respect to a 90dB sound level.
>
>
> That makes perfect sense - thank you for clarifying it.

+1

[PStamler]

Usual caveat: I'd want to know how the measurements were made. Was each recorder's input shunted by a 150 ohm resistor? That would be the right way to do it.

Okay, leave the gain and dBFS issues out of it for a moment. Think of it this way: the worst of these recorders has an equivalent input noise measuring -99dBu.

Now, an SM57, according to the spec sheet, has a sensitivity of -56dBV output (unloaded) when the SPL is 94dB-SPL (1Pa). That's about -54dBu (actually -53.8dBu, but let's round off to the nearest decibel.

If the recorder has EIN=-99dBu, that's 45dB below the mic's output at 1Pa. That means the mic-recorder combination has an equivalent room self-noise level of 94 - 45 = 49dB SPL.

I submit that that's godawful performance for any kind of serious recording of quieter instruments. Or voices (think interviews). Or anything. Especially when, according to the same source, there are much better-performing portable recorders out there.

Peace,
Paul

[William Sommerwerck]

"PStamler" wrote in message
...

> If the recorder has EIN=-99dBu, that's 45dB below the mic's output at 1Pa.
> That means the mic-recorder combination has an equivalent room self-noise
> level of 94 - 45 = 49dB SPL.

> I submit that that's godawful performance for any kind of serious recording
> of quieter instruments. Or voices (think interviews). Or anything.
> Especially
> when, according to the same source, there are much better-performing
> portable recorders out there.

I used to own some really nice mics from a company I won't name. (The company
still makes mics similar to these, more than 40 years later.) Their self-noise
was about -20dB (I forget the reference, but it was a commonly accepted one),
which was considered marginal for recording not-very-loud instruments.

[S. King]

On Mon, 23 Dec 2013 18:59:04 -0800, PStamler wrote:

> Usual caveat: I'd want to know how the measurements were made. Was each
> recorder's input shunted by a 150 ohm resistor? That would be the right
> way to do it.
>
> Okay, leave the gain and dBFS issues out of it for a moment. Think of it
> this way: the worst of these recorders has an equivalent input noise
> measuring -99dBu.
>
> Now, an SM57, according to the spec sheet, has a sensitivity of -56dBV
> output (unloaded) when the SPL is 94dB-SPL (1Pa). That's about -54dBu
> (actually -53.8dBu, but let's round off to the nearest decibel.
>
> If the recorder has EIN=-99dBu, that's 45dB below the mic's output at
> 1Pa. That means the mic-recorder combination has an equivalent room
> self-noise level of 94 - 45 = 49dB SPL.
>
> I submit that that's godawful performance for any kind of serious
> recording of quieter instruments. Or voices (think interviews). Or
> anything. Especially when, according to the same source, there are much
> better-performing portable recorders out there.
>
> Peace,
> Paul

IIRC (and I probably don't since it has been 40 years or so) the EIN of a
Langevin 5116 (do I have that number tight?), a modular tube pre-amp, had
an advertised EIN of -122 dB with reference to an output of +16 dB into a
600 ohm load. I do remember measuring the S/N and finding that the
advertised specs were pretty accurate. We had 24 of these Langevin
modules in a home brew 4-track console. The modules could be strapped for
either line-amp or mic-amp gain. We thought we would select the quietest
modules for mic amps. Turned out there wasn't enough difference to
bother. -99 Db EIN isn't so impressive. On the other hand, I haven't
been in a studio in a long time with an ambient room tone that was quiet
enough to impress either.

Steve King

[Trevor]

"Mike Rivers" > wrote in message
...
>>> The problem with using "gain" as part of any measurement on an
>>> integrated
>>> digital device is that there's no output that's in the same units as the
>>> input. You have volts (or dBu) going in, and this becomes dBFS,
>>> depending
>>> on the "gain." What you can compare is input sensitivity for 0 dBFS
>
>> Fair enough if you don't know and cant measure the voltage required for
>> 0dBFS.
>
> It's easy to measure.

For maximum gain it is of course, and that takes us right back where we
started! The gain of these units varies considerably. Handy to know if you
need that extra gain, but pointless rating the noise performance based on
the maximum gain!


>If you can't measure it, you probably shouldn't worry about it. When I
>review an interface these days, I measure and report the level required to
>reach 0 dBFS, but that really only helps most readers when comparing one
>unit to another if input sensitivity is important (as it is when recording
>wildlife sounds).

Right, but not so for the fixed gain line inputs. I want to know if they
overload at +4, +10, +14dBu etc. so I know if I need pads or not.


>I usually will include an "in practice" statement such as "With the input
>gain turned up full, speaking 1 foot away from an SM57 mic in a normal
>conversational voice produces peaks around -12 dBFS." Of course this
>doesn't reference any known standard (you don't know how loud I talk)

Exactly, would tell me far more if you simply quoted the mV input level, or
gave the SPL figure, preferably both.


>> You can at least
>> use a standard input voltage in that case, as the 0dBFS overload voltage
>> will probably be a lot closer than the maximum pre-amp gains, ie a lot
>> less
>> than 30dB variation.
>
> This isn't really any more helpful than measuring the voltage for 0 dBFS
> since most pepole don't know how many millivolts are coming out of their
> microphone. You can find microphone sensitivity (usually in millivolts per
> Pascal) on some spec sheets, but how many people know that 1 Pascal = 94
> dB SPL (pretty loud)?

If they can't simply look it up, *any* measurements are probably pointless
to such people anyway.


>> The pdf graphs show the problem, the real world noise performance is
>> pretty
>> close for many of the devices
>
> I've found that to be pretty much true. The only thing that's really
> impressed me lately has been the Focusrite Forte (I know that's not a
> portable recorder) because with the gain set for about the same quiescent
> noise as a more typical unit, 0 dBFS at about a 6 dB lower level than
> what's more typical. But then "typical" covers a pretty wide range.
>
> The value of any measurement (or specification) is in how well you
> understand and believe in the legitimacy of how the measurement was made.
> When I measure quiescent noise of a device with an external mic input, I
> terminate that input with a 150 ohm resistor, which approximates the noise
> resulting from the source impedance of a typical mic. But if there's no
> external mic input, since I don't have a sufficiently good "quiet box",
> the best I can do is put the recorder under a pile of pillows at a time
> when a neighbor isn't running a leaf blower, chain saw, or lawn mower
> (night time is good).

Yes, but it's still no where near as good as using a terminating resistor,
so comparing measurements made with those different methods is pointless. I
assume you point that out.


>> As usual, you don't always get what you pay for. :-(
>
> Oh, you usually do, but sometimes you can't get everything you want no
> matter how much you pay.

No, my saying is, and has always been : "you always pay for what you get,
but you don't always get what you pay for'".
An extreme example of that is kilobuck "snake oil" cables, not that *I'd*
pay for those! :-)

Trevor.

[Trevor]

"Jeff Henig" > wrote in message
...
> I've always wanted to record at FAME in Muscle Shoals, AL. When they fixed
> a noisy power line (IIRC), it turned out that the traffic/road noise was
> quite audible in their mics. Very disappointing.
>
> OTOH, it didn't seem to have hurt their production much.

Well in the days when 60dB DNR from a tape recorder was very impressive, it
didn't matter too much. And in these days when you're lucky to get more than
40DB DNR on a commercialy produced CD, it doesn't matter much either. :-(

Trevor.

[detha]

On Mon, 23 Dec 2013 08:15:13 -0500, Mike Rivers wrote:

> The value of any measurement (or specification) is in how well you
> understand and believe in the legitimacy of how the measurement was made.
> When I measure quiescent noise of a device with an external mic input, I
> terminate that input with a 150 ohm resistor, which approximates the noise

<pedantic>Carbon or metal film resistor ?</pedantic>


-d

[William Sommerwerck]

"detha" wrote in message ...
On Mon, 23 Dec 2013 08:15:13 -0500, Mike Rivers wrote:

>> The value of any measurement (or specification) is in how well you
>> understand and believe in the legitimacy of how the measurement was made.
>> When I measure quiescent noise of a device with an external mic input, I
>> terminate that input with a 150 ohm resistor, which approximates the noise

> <pedantic>Carbon or metal film resistor ?</pedantic>

I assuming you're joking.

All 150 ohm resistors produce the same noise (qv, Boltzman). Confusing can
occur when talking about excess noise, generated when current flows through a
resistor. Metal-film resistors produce less than other kinds.

[Mike Rivers[_2_]]

On 12/27/2013 6:11 AM, detha wrote:
>> When I measure quiescent noise of a device with an external mic input, I
>> >terminate that input with a 150 ohm resistor, which approximates the noise


> <pedantic>Carbon or metal film resistor ?</pedantic>

Metal film, but that's just because of what was handy. Maybe Paul
Stamler has an opinion as to whether there's a difference in this
application. He's made a lot of this sort of measurements.


--
For a good time, visit http://mikeriversaudio.wordpress.com

[William Sommerwerck]

"Mike Rivers" wrote in message ...

>> <pedantic>Carbon or metal film resistor ?</pedantic>

> Metal film, but that's just because of what was handy. Maybe Paul Stamler
> has an opinion as to whether there's a difference in this application. He's
> made a lot of this sort of measurements.

It makes no difference. See my preceding post.

A 150-ohm resistor doesn't generate that much thermal noise. I'm wondering
what effect it would have on preamp-generated noise. (I never did quite
understand "noise matching".)

[Trevor]

"Mike Rivers" > wrote in message
...
> On 12/27/2013 6:11 AM, detha wrote:
>>> When I measure quiescent noise of a device with an external mic input, I
>>> >terminate that input with a 150 ohm resistor, which approximates the
>>> >noise
>
>> <pedantic>Carbon or metal film resistor ?</pedantic>
>
> Metal film, but that's just because of what was handy. Maybe Paul Stamler
> has an opinion as to whether there's a difference in this application.
> He's made a lot of this sort of measurements.

Not really pedantic, carbon resistors are definitely inherently noisier than
metal film, and noisier than the thermally generated noise of a dynamic
microphone. A metal film is a far better choice regardless of what happens
to be "handy".
A carbon resistor might be a better match for a carbon microphone of course
:-)
(not really though since the noise of those varies far more than even a
carbon resistor does!)

Trevor.

[William Sommerwerck]

"Trevor" wrote in message ...

> Not really pedantic, carbon resistors are definitely inherently noisier than
> metal film, and noisier than the thermally generated noise of a dynamic
> microphone. A metal film is a far better choice regardless of what happens
> to be "handy".

As far as thermal noise is concerned, all 150-ohm resistors are equivalent. If
a dynamic mic has a resistance (not impedance) of 150 ohms, it generates the
same amount of noise as a cheap 150-ohm carbon-composition resistor.

Reactive elements do not generate thermal noise.

[detha]

On Fri, 27 Dec 2013 23:33:48 +1100, Trevor wrote:

>
> "Mike Rivers" > wrote in message
> ...
>> On 12/27/2013 6:11 AM, detha wrote:
>>>> When I measure quiescent noise of a device with an external mic input,
>>>> I
>>>> >terminate that input with a 150 ohm resistor, which approximates the
>>>> >noise
>>
>>> <pedantic>Carbon or metal film resistor ?</pedantic>
>>
>> Metal film, but that's just because of what was handy. Maybe Paul
>> Stamler has an opinion as to whether there's a difference in this
>> application. He's made a lot of this sort of measurements.
>
> Not really pedantic, carbon resistors are definitely inherently noisier
> than metal film, and noisier than the thermally generated noise of a
> dynamic microphone. A metal film is a far better choice regardless of what
> happens to be "handy".

The pedantic was more about specifying measurements - what resistor (e.g.
wire-wound can work as a coil and pick up EMI), at what temperature,
maximum wire length, etc. Thermal noise from the resistor should be
sufficient orders of magnitude below the measured noise to not matter (and
no I didn't do the math), unless it's kept at 1000 deg.K or so. And
hopefully there will not be going to enough current through the resistor
to introduce current noise - in that case something is /really/ wrong in
the setup.

-d

[Scott Dorsey]

detha > wrote:
>On Mon, 23 Dec 2013 08:15:13 -0500, Mike Rivers wrote:
>
>> The value of any measurement (or specification) is in how well you
>> understand and believe in the legitimacy of how the measurement was made.
>> When I measure quiescent noise of a device with an external mic input, I
>> terminate that input with a 150 ohm resistor, which approximates the noise
>
><pedantic>Carbon or metal film resistor ?</pedantic>

In practice, it doesn't matter since the current is so close to zero so the
excess noise is too.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Scott Dorsey]

William Sommerwerck > wrote:
>"Mike Rivers" wrote in message ...
>
>>> <pedantic>Carbon or metal film resistor ?</pedantic>
>
>> Metal film, but that's just because of what was handy. Maybe Paul Stamler
>> has an opinion as to whether there's a difference in this application. He's
>> made a lot of this sort of measurements.
>
>It makes no difference. See my preceding post.
>
>A 150-ohm resistor doesn't generate that much thermal noise. I'm wondering
>what effect it would have on preamp-generated noise. (I never did quite
>understand "noise matching".)

It has a dramatic effect... think of it as providing a path to ground for
noise generated at the base of the input transistor (for example due to
leakage). If the source impedance of the noise is 1500 ohms and the load
impedance presented by the microphone is 150 ohms, you just got 10 dB of
noise removed.

When you're thinking about noise, the input stage is the generator and
the microphone is the load, you have to twist everything around backwards
in your head.

My normal jackleg test for quiet chambers is that if you plug an SM-57
into a Millennia Media preamp, the noise level should go down and not up.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Mike Rivers[_2_]]

On 12/27/2013 7:20 AM, William Sommerwerck wrote:
> A 150-ohm resistor doesn't generate that much thermal noise. I'm
> wondering what effect it would have on preamp-generated noise. (I never
> did quite understand "noise matching".)

Right - we're not adding the trivial contribution of the noise from the
resistor in to the measurement. It's there so that the input current
characteristics of the preamp are similar to what it is when connected
to a microphone.

The difference in noise measured at the output between with the input
terminated by 150 ohms and with the input shorted is that the shorted
input produces 3 dB less noise. So with a little specsmanship -
connecting the input to something that doesn't represent a typical
microphone - can make it look 3 dB quieter than it is when actually
doing what it's intened to do.

--
For a good time, visit http://mikeriversaudio.wordpress.com

[PStamler]

As Scott says, using a 150 ohm dummy resistor should lower the measured noise.

Think of it this way: imagine a transformerless mic preamp with an input impedance that's created by a 1.5k resistor (typical). The input circuit has 1.5k going to ground, and will have a noise spec reflecting that.

Now plug a 150 ohm mic into the preamp's input. The input transistors now see 150 ohms in parallel with 1.5k, or about 136 ohms, and the noise measurements will reflect that.

Since the only way anyone is likely to use a mic preamp is with a mic preamp, the dummy plug simulates that more accurately then leaving the input unconnected will do.

And yes, I always make my dummy plugs from metal film resistors, simply because (a) that's what I normally have around, and (b) their behavior more closely resembles that of a pure resistance. See the data sheets for that last point.

Peace,
Paul

[PStamler]

On Friday, December 27, 2013 12:15:18 PM UTC-6, PStamler wrote:

> Since the only way anyone is likely to use a mic preamp is with a mic preamp, the dummy plug simulates that more accurately then leaving the input unconnected will do.


Er, make that "Since the only way anyone is likely to use a mic preamp is with a mic plugged into it..." Must have been Zen.

Peace,
Paul

[[email protected]]

>It makes no difference. See my preceding post.
>
>A 150-ohm resistor doesn't generate that much thermal noise. I'm wondering
>what effect it would have on preamp-generated noise. (I never did quite
>understand "noise matching".)

It has a dramatic effect... think of it as providing a path to ground for
noise generated at the base of the input transistor (for example due to
leakage). If the source impedance of the noise is 1500 ohms and the load
impedance presented by the microphone is 150 ohms, you just got 10 dB of
noise removed.

10x voltage division is 20 dB but who's counting. :-)
Mark

[Trevor]

"detha" > wrote in message
...
> The pedantic was more about specifying measurements - what resistor (e.g.
> wire-wound can work as a coil and pick up EMI), at what temperature,
> maximum wire length, etc. Thermal noise from the resistor should be
> sufficient orders of magnitude below the measured noise to not matter (and
> no I didn't do the math),

Obviously, since there IS a difference between S/C, and 150ohm termination
at room temperature even in a fully shielded farady cage. The difference is
the thermally generated noise even for an ideal resistor. In fact the DNR of
microphone pre-amps these days has almost approached the ideal for their
thermally generated noise, that is why all the good ones have similar ein
specs now, and cannot improve much without cryogenic cooling of some sort.

Trevor.

[Trevor]

"PStamler" > wrote in message
...
> As Scott says, using a 150 ohm dummy resistor should lower the measured
> noise.

Of course, compared to an O/C input, which is why you never measure that
way. A S/C input will provide even lower measured noise if that's all you
are after. Some manufacturers do just that.

Trevor.

[[email protected]]

Ok let's be pedantic for a while.
When you connect a pure resistive dummy load to the input of a preamp, the output noise is from 3 sources
1 the thermal noise of the dummy load
2 the CURRENT noise from the amp input
3 the VOLTAGE noise from the amp input

When the dummy load is a short then all the noise is from #3
When the dummy load is an open then all the noise is from #2
In the general case it is a combination.
If the dummy load is is cold then #1 will be low in which case the noise will vary with the value of the dummy load depending on
The design of the amp which determines if 2 or 3 is dominant. The lowest noise usually occurs at the value of R where 2 and 3 are near equal for a particular design. FETs have very low current noise so work best with higher Z inputs. BJTs have lower voltage noise and work best with lower Z inputs.
These are all generalities of course.
Look up the app notes from TI and analog devices about noise.

Happy New Years
Mark

[Scott Dorsey]

> wrote:
>Ok let's be pedantic for a while.
>When you connect a pure resistive dummy load to the input of a preamp, the =
>output noise is from 3 sources
>1 the thermal noise of the dummy load
>2 the CURRENT noise from the amp input
>3 the VOLTAGE noise from the amp input
>
>When the dummy load is a short then all the noise is from #3
>When the dummy load is an open then all the noise is from #2
>In the general case it is a combination.

This is true, and that is the reason why the 150 ohm resistor is suggested;
it is about the impedance of a dynamic mike, and so the resulting noise
approximates that of a typical real-world application. The combination of
noise sources is arranged to approximate something useful.

>If the dummy load is is cold then #1 will be low in which case the noise wi=
>ll vary with the value of the dummy load depending on
>The design of the amp which determines if 2 or 3 is dominant. The lowest n=
>oise usually occurs at the value of R where 2 and 3 are near equal for a pa=
>rticular design. FETs have very low current noise so work best with higher =
>Z inputs. BJTs have lower voltage noise and work best with lower Z inputs.
>These are all generalities of course.=20

Yes, this is all true. But, what we want to know is how much noise we
will get when we plug in the microphone that we're going to be using,
so we care about both current and voltage noise more or less in a
uniform proportion.

In the real world, of course, the microphone may alter it; not all
microphones can be modelled as a 150 ohm resistive source. If you
wanted to know how a preamp would perform over a variety of sources
you might want to present numbers at 50, 150 and 600 ohm source impedances.
But 150 ohms is a good start.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[[email protected]]

Scott.
Yep I agree 100%
150 Ohms is a good approximation for what many mics look like and is a good test point.
My comments were directed to why an open or a short are not good testing alternatives and that a short is not necessarily the lowest noise case.
It depends on the relative contributions of current and voltage noise sources inside the pre.


Mark

[hank alrich]

PStamler > wrote:

> On Friday, December 27, 2013 12:15:18 PM UTC-6, PStamler wrote:
>
> > Since the only way anyone is likely to use a mic preamp is with a mic
>preamp, the dummy plug simulates that more accurately then leaving the
>input unconnected will do.
>
>
> Er, make that "Since the only way anyone is likely to use a mic preamp is
>with a mic plugged into it..." Must have been Zen.
>

Right. Whereas ziss iss Now.

--
shut up and play your guitar * HankAlrich.Com
HankandShaidriMusic.Com
YouTube.Com/WalkinayMusic