TDA7315 chip parameters interpretation

[Frarun]

Hi,
I analyze the datasheet for the TDA7315 chip, and have some doubts concerning noise.

First one is "Figure 4: Noise vs. Volume/Gain Settings".
It looks like under the 0dB the noise is below 5uV (typical). At +20dB noise going up to
30uV. 20dB is like 10times amplification, so can we say the input noise of internal op-amps
is like 3-5uV/rtHz. (I would say if we 'model' the whole internal audio line as op-amp).

In the NE5532 datasheet there is a 5nV/rthz 'Equivalent input noise voltage' (1khz). Is it
mean, if this op-amp closed loop with amplification 1, the noise on the output will be 5nV ?

Regards.

 

[Frarun]

Maybe I will clarify/simplify a bit my question. How this is possible this popular op-amp NE5532, has a 1000 times less noise then an Integrated Circuit?

 

[Chris Hornbeck]

I'm not certain how you're getting your numbers, but maybe you're conflating "uV" and "uV/ rtHz", quite different numbers. Full bandwidth noise can also be weighted in various ways, so an apples-to-apples comparison requires very strict definitions.

All good fortune,
Chris

 

[FoMoCo]

In the NE5532 datasheet there is a 5nV/rthz 'Equivalent input noise voltage' (1khz). Is it
mean, if this op-amp closed loop with amplification 1, the noise on the output will be 5nV ?

No. It'll be roughly 5nV times the square root of the bandwidth over which the noise is measured. There's also a factor for roll-off rate, but I've ignored it for simplicity. For a 20KHz bandwidth that's about a microvolt.

 

[Frarun]

Thank you for responses.

I see something is wrong with my approach to this topic. I am still digging the knowledge.
I found the "Chapter 10. Op Amp Noise Theory and Applications" by Texas Instruments. There is a section "10.2.5 Noise Units"
describing the noise calculation for op-amp. I have tried it for NE5532 and gain 1 (0dB):
rt(20000-20) = 141.35
5* 141.35 = 706.75 nV (EIN)
706.75 * 1 = 706.75 nV (on the output)

Discrete elements are omit with this calculation, so noise will be somehow higher then calculated.

Nevertheless the 706.75 nV looks much more close to 5uV for TDA7315.

Full bandwidth noise can also be weighted in various ways

Is it why the TDA7315 has a noise values given for 20-20kHz and for some "A CURVE"? One is typical 5uV and second 3uV. What is "A CURVE" anyway?

For a 20KHz bandwidth that's about a microvolt.

Hope I correctly calculated above.

Thanks.

 

[FoMoCo]

Hope I correctly calculated above.

Yes, you did. There's also that factor for the roll-off rate, but I don't have that textbook open and have forgotten the exact value for each situation.

 

[Chris Hornbeck]

Is it why the TDA7315 has a noise values given for 20-20kHz and for some "A CURVE"? One is typical 5uV and second 3uV. What is "A CURVE" anyway?

The A curve follows human hearing sensitivity, peaking importance is given to stuff in the upper midrange.

The factor that FoMoCo mentioned is a correction for real-world gentle rolloffs instead of perfect brick-wall filters. It's a theoretical 15.7KHz in place of 20KHz. Probably easiest to research details in your first language; I'm not good at explaining without a diagram anyway.

All good fortune,
Chris

 

[Frarun]

Ahh. The A CURVE is probably the same as described here.
A-weighting - Wikipedia, the free encyclopedia
And the roll-off looks like this:
Roll-off - Wikipedia, the free encyclopedia
(in my native language we call this like a 'characteristic sliding', but the dB/dec or dB/8ve is the same all over the world I think .

This makes perfect sense.

I will ignore a human ear characteristic and roll-off with this analyze. It is too much for one topic. This is interesting what commercial producers of hi-fi equipment uses for parameters inside manuals (not always giving the details)...

Thanks.

 

[Frarun]

I am digging this IC deeper. There is no details how the TDA7315 looks inside. Instead the TDA7300 gives me some news. I can see there are a couple of elements with op-amps.

I am just curious why the THD+N is so huge at -20db set in TDA7315 (like 0.1%) And again I want to compare this to NE5532 op-amp with gain -20dB, but don't know yet how to calculate this one.

 

[Frarun]

It seems everything is OK with 0.1% THD+N at volume set to -20dB. In such case the signal is weak, and the noise part of THD+N parameter goes higher and higher... more important is a value at the 0dB volume set which is like 0.005% which can be assumed is like THD. Still don't know how to compare this to one op-amp, but I saw the amplifiers with THD in this range.

What If we can build the electronic module with the same functionality, but from a discrete elements (op-amps, res, capacitors)? Could we achieve better THD & noise ? Which parameter of this chip is really weak? Do you have any suggestions?