ultra-low noise transistors.

Status
Not open for further replies.
Jump to Latest
#28
According to their datasheets, some of the Zetex low saturation voltage transistors might be low noise, judging from what those specs tell about the base spreading resistance. See if you can get samples of ZXTP19020 and ZXTN19020 to play with. They're high gain too, hFE ~350.
 
Last edited:
#30
I rarely see Rbb specified on a datasheet.
Why is that?
Do they think we don't need that information, or do they believe it is misleading, or is it too variable and would be embarrassing, or are they just hiding behind the HF noise specs that regularly get published and don't apply to Audio & LF?
 
#32
AndrewT said:
I rarely see Rbb specified on a datasheet.
Why is that?
Do they think we don't need that information, or do they believe it is misleading, or is it too variable and would be embarrassing, or are they just hiding behind the HF noise specs that regularly get published and don't apply to Audio & LF?
Click to expand...

Hi Andrew,

I have been measuring transistor noise in order to get an estimate of RB for the SPICE models that I am putting on my web site as supplementary material for my book on designing power amplifiers (the models I currently have up there do not have refined RB numbers).

Most RB I have seen in vendor SPICE models for small-signal transistors is way off; I've often seen place-keeper numbers of 10 ohms for something like a 2N3904 or even a 2N5551. The truth is that these numbers are often surprizingly higher, in the hundreds. As you probably know, the 2N4401/4403 are very good in this regard. I've seen RB quoted as 40 ohms for these, but I've measured them in sensitive noise test circuits and they come in even better. My noise measurement is usually centered around 10 kHz. I usually make the measurement at 1 mA.

Cheers,
Bob
 
#33
AndrewT said:
I rarely see Rbb specified on a datasheet.
Why is that?
Click to expand...

My guess is very few people need it and its expensive to put on the datasheet. That's because spec-ing it means measuring it in production and test time costs money. When I first found out about these trannies, I looked up the Spice models which do give a value for Rbb. I think its 10ohms, but Spice models are no guarantee of reality. If it really were as bad as 10ohms with these parts I believe the base-emitter saturation voltages would be much greater.

Here's what they say - VBE typical 0.96V at 0.28A base current. Now if there's 10R in series with the base at 280mA, that's already 2.8V. So to meet that spec I think the Rbb must be in the region 1 - 2 ohms. Which is pretty impressive and its why I'm curious to know if they're really low noise.

Telstar said:
I don't have means to measure noise...
Click to expand...

Build an 80dB amplifier and send its output into a soundcard with the input shorted. Should be good enough to check - 1uV of noise will come out as 10mV.
 
Last edited:
#34
abraxalito said:
My guess is very few people need it and its expensive to put on the datasheet. That's because spec-ing it means measuring it in production and test time costs money. When I first found out about these trannies, I looked up the Spice models which do give a value for Rbb. I think its 10ohms, but Spice models are no guarantee of reality. If it really were as bad as 10ohms with these parts I believe the base-emitter saturation voltages would be much greater.

Here's what they say - VBE typical 0.96V at 0.28A base current. Now if there's 10R in series with the base at 280mA, that's already 2.8V. So to meet that spec I think the Rbb must be in the region 1 - 2 ohms. Which is pretty impressive and its why I'm curious to know if they're really low noise.



Build an 80dB amplifier and send its output into a soundcard with the input shorted. Should be good enough to check - 1uV of noise will come out as 10mV.
Click to expand...

I found out the hard way that RB cannot be reliably inferred from DC measurements on small-signal transistors by looking at excess Vbe. For all transistors, Vbe at high current does not tell the story about RB because of emitter crowing that reduces RB at high current.

BTW, when you build a high-gain amplifier for transistor noise measurements, you must limit its LF and HF bandwidth with known filtering to yield a known noise bandwidth. You may want an LF corner in the range of 100 Hz to 1 kHz to keep out LF garbage. You may want an HF bandwidth on the order of 100 kHz (but this is too high for a typical soundcard).

I then SPICE the actual noise amplifier circuit with the target transistor and adjust RB to get the same rms noise voltage integrated over the relevant bandwidth. By SPICEing the actual circuit, careful calculation of actual noise bandwidth of the filter becomes unnecessary.

Getting the right value for RB can also play a role in getting the right amount of excess phase for the transistor in simulation at high frequencies. An RB value of 10 ohms, when it should be 200 ohms, for example, may give optimistic high-frequency performance in simulation. RB against Cbe introduces a pole in the transconductance characteristic.

Cheers,
Bob
 
#35
Bob Cordell said:
Hi Andrew,

I have been measuring transistor noise in order to get an estimate of RB for the SPICE models that I am putting on my web site as supplementary material for my book on designing power amplifiers (the models I currently have up there do not have refined RB numbers).

Most RB I have seen in vendor SPICE models for small-signal transistors is way off; ..
Click to expand...

Hi Bob, thanks for chiming in.
Please let me know when you will have updated models for the 5551/5401. They sim quite good for noise (using your current models) and I'm pretty sure that the real noise is much higher.
 
#36
Bob Cordell said:
I found out the hard way that RB cannot be reliably inferred from DC measurements on small-signal transistors by looking at excess Vbe. For all transistors, Vbe at high current does not tell the story about RB because of emitter crowing that reduces RB at high current.
Click to expand...

That's very useful to know, appreciate the tip. I've never heard of 'emitter crowing' but Google thinks you might have meant 'crowding'😀

BTW, when you build a high-gain amplifier for transistor noise measurements, you must limit its LF and HF bandwidth with known filtering to yield a known noise bandwidth. You may want an LF corner in the range of 100 Hz to 1 kHz to keep out LF garbage. You may want an HF bandwidth on the order of 100 kHz (but this is too high for a typical soundcard).
Click to expand...

I was imagining in this case to do it with FFT software. So long as the number of bins is known, together with the sample rate the effective bin bandwidth can be calculated. Some software (I'm guessing the more expensive variety) even does the integration and produces a single figure. This all does rather assume that the anti-aliasing filter on the soundcard is going to be good enough - so for good measure it would be conservative practice to include one's own LP filter.
 
Status
Not open for further replies.
Top Bottom