minimal noise smd npn transistrors?

[D45T]

Greetings everyone. Does anybody know any good really low noise smd npn transistors that are still in production?

It all started when I had built my first carcosa fuzz and then found out that 2n508x transistors which are supposed to be used in this project are causing way too much noise despite "low noise" characteristics in their datasheets. I've managed to find some ztx1048a — these are causing much less noise but they are hard to find (mostly they are sold in big batches only). Now I want to build this fuzz again and I going to try it with SMD this time. So, I am looking for SMD npn transistors with ~500 hfe and noise level significantly lower compared to 2n5088/2n5089 (2-3 dB under IC = 100 Adc, VCE = 5.0 Vdc, RS = 1.0 kOhm, f = 1.0 kHz as written in datasheet).

Any recommendations? Thanks in advance.

[antonis]

For 100 Adc continous collector current, you need an extra noisy power n-p-n.. 

P.S.
I don't presume transistor noise in Fuzz pedals is the main "noise" issue..

[D45T]

I don't presume transistor noise in Fuzz pedals is the main "noise" issue..

This fuzz is four high-gain stages with passive treble cutting tonestack. I do not think that there are many other things to cause noise. Amount of gain is atrocious and overwhelming majority of people trying to build this pedal are getting high frequency whine and self-oscillation. I've made a good pcb with a lot of ground polygons, but it obviously didn't change noise of transistors. I used my audiointerface to record noise with different transistors, and ztx1048a are causing twice lower volume of noise, moreover the noise itself has audibly lower tone.

[antonis]

Amount of gain is atrocious and overwhelming majority of people trying to build this pedal are getting high frequency whine and self-oscillation.

Probably due to lack of 47-150 pF Miller caps..

Transistor noise and circuit design noise are totally different issues, so proceed to "noiseless" BJTs only in case of circuitry "quietness" measures are run-down..

[Ripthorn]

What schematic are you working from?

[D45T]

What schematic are you working from?

I took schematic from pedalpcb carcass fuzz.

[D45T]

Ok, if someone will be looking for the same information, BC817 is recommended on other forums and in few papers, and it has multiple subtypes with various hfe ranges.

[duck_arse]

BC817 is listed as switching/GP AF driver on all the datasheets I peered at - not one has any mention of noise figures. an odd choise for a low noise need, I'd think.

[R.G.]

This fuzz is four high-gain stages with passive treble cutting tonestack. I do not think that there are many other things to cause noise. Amount of gain is atrocious

Mother Nature has a reputation for not caring what humans think may or may not think. She just sets Her Rules and leaves us humans to figure them out.

As you've found out, merely having a low noise number printed in the data sheet doesn't guarantee that the circuit the transistor is used in will be quiet. That is because the transistors are not the only noise contributor. The design of the circuit itself often makes a huge difference in the noise that the transistors produce, and even the source impedance of the signal gets into the act. I took a look at the schematic. Just the first stage tells me that with an atrocious amount of gain, that thing is going to have noise problems.

In any low noise design, the added circuit noise will be dominated by the first stage's added noise. This is because any noise added to the signal in the first stage will be amplified by all the remaining stages. With four high gain stages this gets even more critical. For lowest noise, the general design requires that you (1) use enough gain in your first stage to get the signal level up well above the added noise of the next stage (2) use all the possible means of reducing added noise in that first active device, and (3) use an input device which minimizes the added voltage noise and current noise with that particular signal source impedance.

The first stage in this circuit is an emitter follower fed from 1M bias resistors and has a series 1K resistor from the input cap. that means there is no gain in the first stage. So all of the noise the first stage adds is merely tacked onto the signal and is then amplified by everything else. That is then followed by a voltage divider pot, which adds the pot's thermal noise to the noise already added by the first stage transistor and its biasing resistors. Then you finally hit a gain stage. The thermal noise of the pull-down resistor, the series 1K and both biasing resistors is fed into the base of the first transistor to be "amplified" - in this case, merely buffered down to a low source impedance - and the signal further lowered by that pot. I haven't looked at the source impedance issues here.

Fuzz circuits are inherently a bad case for hiss because they amplify the presumably small noise at the input, but the signal is clipped while the smaller noise is not. So the noise becomes more prominent (i.e. bigger compared to the signal) every time the signal is clipped. This circuit does that multiple times.

In a way, you're right - the noise of the buffer transistor and the first actual gain transistor do matter - their noise is further amplified.You might be happier just replacing those with the ZTXwhatevers. The later transistors will matter less. But rearranging the circuit itself would reduce noise a lot more than cherry-picking transistors.

[D45T]

BC817 is listed as switching/GP AF driver on all the datasheets I peered at - not one has any mention of noise figures. an odd choise for a low noise need, I'd think.

This article mentions 817 — http://www.janascard.cz/PDF/Design%20of%20ultra%20low%20noise%20amplifiers.pdf
What would you recommend for low noise smd?

In a way, you're right - the noise of the buffer transistor and the first actual gain transistor do matter - their noise is further amplified.You might be happier just replacing those with the ZTXwhatevers. The later transistors will matter less. But rearranging the circuit itself would reduce noise a lot more than cherry-picking transistors.

Thanks for such a detailed response.
Since BC817 is cheap and reported to have quite good noise performance I am going to buy a whole cup of those for hfe matching instead of buying one ztx transistor, then I'll try to take a look at schematic of that fuzz in ltspice and I'll try to improve it — the first two obvious steps will be adding Miller caps and making 4 parallel transistors input buffer like Way Huge are doing it.

[duck_arse]

BC817 isn't a low noise choice except in that circuit, optimised for BC817's. I'd use BC849's in a circuit, because I have them under the bench. but why am I building an smd ultra low noise amp? I've also got some very nice Motorola 2N5089, and some BC337.

[antonis]

but why am I building an smd ultra low noise amp?

Maybe 'cause you like to consider a four high-gain stages fuzz as ultra low noise amp..??
(just a guess..)