NE5534 as mic pre without input transformer

;D

NO...

op amp input noise consists of multiple terms...

By using a step up transformer of 7x the 150 ohm mic, looks more like  7.3k ohm mic but with 7x the output voltage...

Now the noise voltage and noise current of a 553x op amp makes a much better match for delivering minimal total noise.  (when direct connected the noise current pretty much doesn't matter but signal to noise isn't as good). Too much voltage step up makes the noise current contribution dominate the noise voltage because it is current time impedance.

I am too lazy to repeat the math, but last time I calculated a 553x connected directly to a low z mic has a noise figure around 9 dB, With a step up transformer more like 2 dB,  (noise figure is noise vs a perfect noiseless amp).

Of course for a hi-Z  (600 ohm) mic the math is different.

JR

You want to look at things in terms of impedance.  To minimize noise you use the optimum source impedance for the particular amplifier.  The transformer is chosen to get this optimum.  This helps explain it

http://www.jensen-transformers.com/transformers/mic-input/

But you are correct in thinking the transformer allows you to use a non inverting opamp.  This is the topology of the classic API style mic pre.

If you look at the graph, at the 10K resistance the noise is 2uV. If you used a 1:10 mic transformer is will look like 100 ohms at the input and the EIN will be 20dB lower or 0.2uV. Looking at the same graph, the 100 ohm point has 0.4uV of noise so the transformer  is 6dB quieter. 150 ohms should be very similar.

Cheers

Ian

Typical termination for a microphone is "bridging" so 10x the mic source impedance.

With a transformer the impedance is changed by the turns ratio squared.

To realize 1.5k impedance at the primary requires 150k at secondary of 10x transformer.

JR

user 37518 said:

Do you mean a 10K resistor on the secondary reflected back to the primary at 100 ohms?

Click to expand...

No, I mean 100 ohms source at the primary reflected to the secondary as 10K.  This is the main noise resistance seen by the op amp. The secondary would probably have a 100K resistor across it so when this is reflected to the primary it looks like 1K so as not to load the 100 ohm source.

I know you are thinking the secondary now has a noise resistance of 100K but it doesn't. The 100K resistor is in parallel with the 10K reflected from the primary so the noise resistance is just under 10K.

Cheers

Ian

ruffrecords said:

No, I mean 100 ohms source at the primary reflected to the secondary as 10K.  This is the main noise resistance seen by the op amp. The secondary would probably have a 100K resistor across it so when this is reflected to the primary it looks like 1K so as not to load the 100 ohm source.

I know you are thinking the secondary now has a noise resistance of 100K but it doesn't. The 100K resistor is in parallel with the 10K reflected from the primary so the noise resistance is just under 10K.

Cheers

Ian

Click to expand...

Excellent explanation - it's difficult to articulate it without drawings (for myself anyway) even when you know the concept 'on paper'.

user 37518 said:

Ok I get, what I dont understand is, how if a 10K resistor is reflected to the secondary, how does it reduce the noise as you mentioned? or is it that input of the transformer (the primary) should be considered as the input resistance to which EIN is refered?

Click to expand...

The transformer gets you 20dB of 'free' gain, so the signal is now louder, as well as the noise from the microphone. So, by comparison, the 5534's high noise is now proportionally less noisy when compared to the stepped up source signal and its noise. The transformer steps up the impedance, and places the source in a better spot on the noise figure vs. source impedance curve.

A noise figure vs. source impedance curve will look somewhat like a bathtub. It's high at low impedances, because as the source impedance goes down, the fixed amplifier noise becomes proportionally greater than the noise of the source. As the source impedance increases, its noise also increases, making the amp's contribution to the total noise proportionally less. At still higher impedances, the amp's input current noise can dominate (especially with bipolar input amplifiers), since this current works into the source impedance to generate a voltage, which is then added to the noise of the source and the amp's input voltage noise. So, the sweet spot for a 5534 is around 1-2kΩ, and it's worse at higher and lower impedances. Transforming 150Ω to 15K gets it closer to the sweet spot than 150Ω alone.

Read the Jensen 990 paper.
http://www.technicalaudio.com/pdf/Jensen_Transformers/Jensen_OpAmps_990_and_related/Jensen_JE-990_opamp_JAES_reprint_1980.pdf

Fig 3 shows a very-good transistor at several currents.

The 990 runs its devices at 1.6mA. At 1K source it gives 0.5dB noise figure.

If a perfect impedance transformer is available, Fig 5 plots both the resistance hiss and the amplifier hiss for various transformed resistances.

The '5532 IIRC runs its devices at ~~0.2mA. The OSI is more like 3K-10K.

user 37518 said:

Makes a lot of sense, do you happen to know where can I find the 5534 graph which displays what you are saying, in the data sheet theres only the one I posted in my first post and doesnt follow the parabolic response you are refering to.

Click to expand...

NF curves are generally published for discrete (low noise) transistors. Op amps generally publish ein noise voltage and noise current specs. 

When using a discrete bipolar input device in a design you can vary the current density (operating current) to optimize for expected source impedance. In general higher current density delivers lower noise voltage but higher noise current making it a better match for lower  source resistance.

Op amps operating at fixed operating current are generally optimal for higher nominal source impedance. This is changing with some modern uber op amps with lower input noise.  But old school op amps worked well with 7x or so mic transformers, perhaps a bifet op amp could rock a 10x transformer, but there is no free lunch from higher step up.

Jensen was known for transformer design and good overall performance.

JR

Note: the 990 NF curves are lifted from the LM394 transistor data...