The quietest audio op-amp with ultra-low THD

[Eeyore]

National's LME49710/20/40 amd LM4562 seem to have the best figures at
2.5nV/sqrt Hz but in a practical circuit I found the venerable NE5534
with its 3.5 nV/sqrt Hz was quieter, presumably due to lower input noise
current in that particular configuration.

The LMEs also have insanely low THD
RL = 600ohms 0.00003% (typ)

http://www.national.com/pf/LM/LME49710.html
http://www.national.com/pf/LM/LM4562.html

Does anyone know of anything better without sacrificing unity gain
stability or THD ?

Graham

 

[Eeyore]

What resistance values are you assuming? Not one month ago, in this
very boutique, I posted a link to a new part with some really cute
specs--the ADA4898-1.

http://www.analog.com/static/imported-files/Data_Sheets/ADA4898-1.pdf

I gather it's probably intended for lower source resistances than yours,
but its combination of low eN and low Cin is amazing.

Yes, my source resistance may dominate. It's a new module for SSL 4000E/G
professional sound mixing desks to convert the 'channel buckets' into a
monitor desk for Digital Audio Workstations. I was shocked at how high their
mix resistor values were (15k). Crikey, even at Studiomaster in the latter
days I was using 4k7 !

It's certainly amazingly quiet though.

I'm not familiar with the THD being quoted in dBc though. Can you assist there
?

Graham

 

[Martin Griffith]

Not got round to using it yet but the ADA4899 looks tasty. 1nV, 0.7ppm,
600MHz.

Open the window, more shit flies in.
600MHz is not required for audio, only audiophools

martin

 

[Martin Griffith]

Yes.
I was thinking more of it allowing a gain of say x5000 to 100kHz. The chip's
low Vos allowing this in one hit.

A good audio mic at 100KHz and 70dB+ gain, what on earth are you are
you recording?

martin

 

[Eeyore]

15K? Sounds like jfet turf.

Depends on whose jfet !

The existing bus amp uses an LM394 and then destroys the noise performance by
following it with a -6dB 5534 stage.

The worst of all worlds. The bus mix amp will clip before its buffer and the noise
figure goes to hell.

SSL weren't noted for the best audio electronics.

Graham

 

[Eeyore]

Open the window, more shit flies in.
600MHz is not required for audio, only audiophools

Absolutely. The LME's 60 or so MHz is more than enough.

Graham

 

[Eeyore]

"Martin Griffith" wrote

Yes.
I was thinking more of it allowing a gain of say x5000 to 100kHz. The chip's
low Vos allowing this in one hit.

Max gain required would be about 24.

Graham

 

[Eeyore]

dBc = 'decibels with respect to the carrier', i.e. the undistorted output.

Unfortunately that doesn't specify a specific voltage or power level AFAICS.

Graham

 

[Les Matthew]

Yes, my source resistance may dominate. It's a new module for SSL 4000E/G
professional sound mixing desks to convert the 'channel buckets' into a
monitor desk for Digital Audio Workstations. I was shocked at how high their
mix resistor values were (15k). Crikey, even at Studiomaster in the latter
days I was using 4k7 !

At Midas Audio in the 80's we used 12k.


les...

 

[[email protected]]

15K? Sounds like jfet turf.

John- Hide quoted text -

- Show quoted text -

15K? Sounds like jfet turf.

Yes, Aren't you going to be dominated by the current noise? Speaking
of jfet opamps, Do you have a low noise favorite? I like the OPA134
from Burr-Brown (TI). But am always looking for something better.
George

 

[Eeyore]

At Midas Audio in the 80's we used 12k.

Yes, but you were going downhill them. KT group bought out Midas in 88 when I was
there.

Dave Dearden and I were dumbstruck (and not in a good way) just with the mic input
stage.

Graham

 

[Eeyore]

Yes, Aren't you going to be dominated by the current noise? Speaking
of jfet opamps, Do you have a low noise favorite? I like the OPA134
from Burr-Brown (TI). But am always looking for something better.
George

That depends how many inputs are assigned. As these are going only to be monitoring
desks, the number of inputs will be 24 or 36 typically.

So the suming node will see ~ 400-600 ohms. Current noise shouldn't be a problem
here, in fact I'm now thinking again of a discrete front end where en can be as low
as 0.5nv/sqrt Hz with a certain relatively inexpensive device I know, widely used in
direct coupled low-Z mic preamps. Time to do the sums and a balancing act !

Graham

 

[Joerg]

Graham,

What resistance values are you assuming? Not one month ago, in this
very boutique, I posted a link to a new part with some really cute
specs--the ADA4898-1.

http://www.analog.com/static/imported-files/Data_Sheets/ADA4898-1.pdf

I gather it's probably intended for lower source resistances than yours,
but its combination of low eN and low Cin is amazing.

But mind the knee which is a bit highish, see figure 13:
http://www.analog.com/static/imported-files/Data_Sheets/ADA4899-1.pdf

Since Graham does audio stuff this amp might not quite be the ticket.

How about a discrete design? Maybe with tubes ...?

<ducking>

 

[Joerg]

Open the window, more shit flies in.
600MHz is not required for audio, only audiophools

Yeah but: Think what that can do in advertising: Gold plated super
electrons that are made extra agile and can turn themselves around in
under 2nsec. Electrons spend less time in the muddier regions in the
device, therefore get less dirty and don't wear out so fast. Much less
electron fatigue. Makes for extremely crisp crescendos that could never,
ever be heard with an amp that is anywhere less than 500MHz

 

[Eeyore]

But mind the knee which is a bit highish, see figure 13:
http://www.analog.com/static/imported-files/Data_Sheets/ADA4899-1.pdf

Since Graham does audio stuff this amp might not quite be the ticket.

How about a discrete design? Maybe with tubes ...?

<ducking>

LOL !

Actually, now I know a little more about its practical intended function, I'm
looking at a discrete differential front end using 2SA1085s in front of the
LME49710.

Should be good for 0.7nV/sqrt Hz.

The manufacturer's original design which is not dissimilar in concept except
they got the gain structure wrong used 2 parallel pairs of LM394s !

Graham

 

[[email protected]]

Yeah but: Think what that can do in advertising: Gold plated super
electrons that are made extra agile and can turn themselves around in
under 2nsec. Electrons spend less time in the muddier regions in the
device, therefore get less dirty and don't wear out so fast.

Speaking of dirty... I'm waiting for the marketing types to sell some
super-expensive cans of compressed air to clean out heat sink dust.

Improve heat transfer! Make your heat sinks sparkling clean with
AudioPhool(tm) Compressed Air! Crisper highs! Tighter Bass! Make
your power transistors happy! Give your electrons a bath! We blow
away the competition! (in cubic feet per minute that is)

Michael

 

[Martin Griffith]

Martin Griffith wrote: snip

Yeah but: Think what that can do in advertising: Gold plated super
electrons that are made extra agile and can turn themselves around in
under 2nsec. Electrons spend less time in the muddier regions in the
device, therefore get less dirty and don't wear out so fast. Much less
electron fatigue. Makes for extremely crisp crescendos that could never,
ever be heard with an amp that is anywhere less than 500MHz

I've just done a deal with the LHC at Cern, I'm buying the stripped
off electrons from the ultra fast protons that they use in round
thing.

How should I package them?

martin

 

[Les Matthew]

Yes, but you were going downhill them. KT group bought out Midas in 88 when I was
there.

Dave Dearden and I were dumbstruck (and not in a good way) just with the mic input
stage.

Graham

Dave Dearden, that name rings a bell. I jumped ship to Brit Row around
87. It was mental at Midas then, bank holding the purse strings, full
order book but no cash flow. Pro40 was crazy, so much more electronics
crammed into the same old Pro4 frame, Geoff just wouldn't contemplate
having a larger frame. They got seriously hot.

les...

 

[Martin Griffith]

Dave Dearden, that name rings a bell. I jumped ship to Brit Row around
87. It was mental at Midas then, bank holding the purse strings, full
order book but no cash flow. Pro40 was crazy, so much more electronics
crammed into the same old Pro4 frame, Geoff just wouldn't contemplate
having a larger frame. They got seriously hot.

les...

Dave Dearden... Many Confused Individuals (MCI consoles). them where
the days, when SMPTE timecode was new

martin

 

[Joerg]

I've just done a deal with the LHC at Cern, I'm buying the stripped
off electrons from the ultra fast protons that they use in round
thing.

How should I package them?

Those need to be transported in a humidor