Simpler PI...

[Aziz]

New favourite: BC140-10 (the only one I have found in my box).
It is achieving the best performance with one transistor!!!! Two would be better. I have to look further.
I have some more other transistors...
Aziz

[Aziz]

Hi all,

Yeah!, I have achieved the 0.64 nV/sqrt(Hz) noise density (measured)!!!

The other noise sources dominate now. The 4.7 Ohm and ESR resistance of the caps. And I haven't even increased the collector current for the two BD139 yet. I had to reduce some resistor values (the BD139 needs more biasing current). I can squeeze out more noise.

My noise figure (NF) of my amplifier is now 3.5 dB (comparable with transistor noise figure).


Cheers,
Aziz

[Aziz]

Hi all,

it is going to be more difficult practically going below 0.6 nV/sqrt(Hz).
Limitations:
- Collector current limitted to 1-2 mA for each transistor. The op-amp can't deliver enough drive current without distorting the amplified signal (causing harmonic distortion).
- Paralleling transistors limitted to 2-3 (maybe 4) quantity.
- Biasing the BJT's with more current. Input impedance of the amplifier gets low although it isn't an issue to my application. High power transistors have low hFE parameter and need more biasing and drive current.
- Other noise sources dominate (I have to choose more amplifier gain to reduce them overall - no go).
- More power consumption.
- Convenient transistors (there are so many out there... you have to try it all. The data sheet does not tell you. )
- Convenient and low ESR capacitors (*LOL* I don't have any at hand!)

Well, I tend to use small signal with high hFE to mid power signal BJTs. I hope, the transistors I haven't tried them yet, will work better. (maybe BF459 / MJE13007 ??? Who knows)
But I'm very very disappointed about the so-called "low-noise BC550C".
No, it can't be. I have been very likely conned.

New goal: 0.5 nV/sqrt(Hz)
The betting office has just opened again!
(I will know it next month, whether I can achieve the goal.)

My bet:
It's going to be extremely difficult. You wanna help me with tips?

Cheers,
Aziz

[PATCHES JUNIOR]

It's going to be extremely difficult. You wanna help me with tips?

Cheers,
Aziz

I would say the best help would be for you to post some pictures of the boards you are building. You know the old saying. "A picture is worth thousand words" So let the gang here see what you are up to.

[Davor]

You have not been conned with BC550 as it really is a low noise transistor at ... 1k input resistance
The BC550C datasheet clearly shows base spreading resistance (Rbb) being over 100ohm at10mA, so it is NOT a good transistor for real low noise. BC337 is much better as it's base spreading resistance is well below 50ohm, and a single transistor will go down to 1nV/sqrt(Hz) at 1mA.

[Aziz]

I would say the best help would be for you to post some pictures of the boards you are building. You know the old saying. "A picture is worth thousand words" So let the gang here see what you are up to.

PJ, you want to see some pictures? But my digital camera is kaput. I don't have a working cam at home.

But I have some FFT pictures of some measurements. There you are:








Prototype of the super-duper ultra low noise AC amplifier is confidential. Hey, it's a rocket science!
(Mr. Candy can't do it! )

Cheers,
Aziz

[Aziz]

You have not been conned with BC550 as it really is a low noise transistor at ... 1k input resistance
The BC550C datasheet clearly shows base spreading resistance (Rbb) being over 100ohm at10mA, so it is NOT a good transistor for real low noise. BC337 is much better as it's base spreading resistance is well below 50ohm, and a single transistor will go down to 1nV/sqrt(Hz) at 1mA.

Hi Davor,

you could be right. But the BJT spice model is modelled totally wrong.
I have seen many BJT models and a lot of them don't model the base resistance RB correctly. Even the BC337 isn't modelled correct.
Spice is spice but the real world is what counts. You have to buy and try a lot of BJT's.
Cheers,
Aziz

[Aziz]

1.65kOhm Resistor Noise

Ooops, forgot one picture:

The open input condition, where a 3.3k + 3.3k resistor divider at the input effectively generates a resistor noise of 3.3k/2 = 1.65 k Ohm. The impedance is the parallel resistance of 3.3k and 3.3k.
(BTW, I have changed the former 10 k into 3.3 k. )



Who has got the 1k resistor in the PI input stage? Keep dreaming of an ultra low noise amplifier.

Aziz

[PATCHES JUNIOR]

PJ, you want to see some pictures? But my digital camera is kaput. I don't have a working cam at home.





Cheers,
Aziz

Too bad your camera is broke. But I knew that would be your answer. Maybe I could send you one free, shipping included. But I know you will say "no thanks'. Best of luck to ya.

[Aziz]

Too bad your camera is broke. But I knew that would be your answer. Maybe I could send you one free, shipping included. But I know you will say "no thanks'. Best of luck to ya.

Answer: "no thanks"
Why don't you build your own amplifier PJ? You could see it brilliantly working and could verify my measurements.

Cheers,
Aziz

[Davor]

you could be right. But the BJT spice model is modelled totally wrong.

Not all of them. You may check Rb value in a spice model, and if you wish you may change it there, or as a parameter. Some species of the same transistor model have Rbb close to reality and you may simply pick those. In Spice most models do not show real Rbb values as Rbb is seldom of interest for a transistor that is mainly used for switching purposes, and yet due to the low Rbb becomes increasingly popular among low noise enthusiasts.
BFxxx series transistors may prove even a better choice, but as you noticed before, Spice is not helping in modelling for low noise because the key parameter is seldom present in datasheets, let alone Spice models.

Because MC gramophone cartridges are fading into history, so do the fine transistors of the era, and soon will be made exclusively out of unobtainium. My approach is always-go-with-the-popular-part because those will tend to stay on the market.

[Aziz]

Hi Davor,

I think, that the BF459 should be more convenient for my application. It's a high voltage transistor (300 V), which should even survive high base currents as well (100 mA). So I can leave my current limitting resistor at the input. The RX coils resistance could be reduced down to 2 - 5 Ohm.
I should look at my old kaput TV/crt display, there could be some BF459's there.

At the same time, the gain setting resistor of 4.7 Ohm could be reduced down to 2.2 Ohm (or even to 1 Ohm) when using a dual high current op-amp (such as NJM 4556 or NJM 4558 ) . Hey, I could even drive mid/high power transistors, which tend to produce less noise.

Perhaps I could achieve the 0.5 nV/sqrt(Hz) this way. That's really a challenging goal.

For a good front-end however, everything between 0.7 nV to 1 nV/sqrt(Hz) would suffice by far. And this goal has already been achieved with simple & cheap means.

I'm really happy with my AC amplifier. I should design/solder a new detector controller prototype and make the world's best detector ever! (Oooooooaaah!, "It's the end of Greedlab!") *LOL*

BTW, the AC amplifier does consume approx. 8 mA at 8.5 V supply voltage.

Cheers,
Aziz

[Aziz]

Hi all,

I can measure and calculate the amplifier noise more accurately now.
The 2 x 2N2222 and NE5534 front-end AC amplifier at gain 206 has
e_n = 0.664 nV/sqrt(Hz)
input referred voltage noise density.

Not that bad I think!
It's the poor man's revenge attack! It's raining "blue" pills.

I'll be able to know, whether I can reach the 0.5 nV/sqrt(Hz) region now.

Cheers,
Aziz

[Aziz]

2 x BC337-40 and NE5534 at gain 206 and at 24 °C room temperature:
e_n = 0.886 nV/sqrt(Hz)
Good.

1 x BC140-10 and NE5534 at gain 206 and at 24 °C room temperature:
e_n = 0.66 nV/sqrt(Hz)
Very good. It's a single transistor version.

2 x BD139 and NE5534 at gain 206 and at 24 °C room temperature:
e_n = 0.623 nV/sqrt(Hz)
Nice. I'm even better.

1 x LM394 (both transistors parallel) and NE5534 at gain 206 and at 24 °C room temperature:
e_n = 0.742 nV/sqrt(Hz)
Nice, but expensive.

4212

[Davor]

It's raining "blue" pills.

Damn! I took the red one.

Regarding transistor current, if you don't go over ~6mA you are well within op amp supplies. However, once you reach some decent noise levels, the only upside of increasing currents is increasing the preamp frequency response. So in effect a transistor choice is a more determining factor than current.