Power pentodes as preamp tubes?

[Steve Conner]

Good idea Merlin. The transistors have to be surface mount, but we already use the FZT751 in other places, a surface mount version of the ZTX751. I'll try those out for noise.

It doesn't really address my original gripe, though: nobody makes a transistor actually specified for low-noise audio amplification any more. Diodes Inc. could decide to add some excess noise mechanism to the FZT751 tomorrow, it would still pass their quality control, and there would be nothing I could do about it. Not very convenient for mass production.

I wonder what makers of mixers and mic pres do nowadays.

[R.G.]

I wonder what makers of mixers and mic pres do nowadays.

In the popular jargon of the southern USA, they go fishin' and take pot luck.

The thing that helps is that on average, transistor processing is much better than it was in the 70s, which was pretty much the last hurrah of the discrete transistor, and the poor processing that made the worst of noise issues in discretes has been washed out with the semiconductor tide that's lifted all boats.

Still not a pointed design for low noise, but better in general.

[Merlinb]

I wonder what makers of mixers and mic pres do nowadays.

Douglas Self recommends the 2N4403, with rbb at 40 ohms.
This guy apparently measured the BC337 at 30 ohms http://www.janascard.cz/PDF/Design%2...amplifiers.pdf

Modern opamps are getting pretty darn good though, with noise specs that are approaching those of discrete transistors. Eventually there may be nothing to choose between them. (I recently picked up some free samples of the OP07D for a high source impedance application. It has some impressive specs without being absurdly expensive.)

[Alan0354]

BJT noise at LF is also dependent on base spreading resistance. Devices rated for high collector currents usually have smaller spreading resistance. Obviously there is a balance to be struck between minimising base current (implying high hFE) and in minimising spreading resistance (implying a high current rating and therefore probably low hFE). The ZTX653/753 and BC337/327 represent good examples, as they both have low spreading resistance and healthy hFE >200.

You know the two you point out are medium power transistor, I went and search the data sheets in digikey and there is no noise information on either one. They are not meant to be low noise front end application. Here is the data sheet of MPSA18:

http://www.onsemi.com/pub_link/Collateral/MPSA18-D.PDF

Look at the audio frequency performance, if you can find one better, I would love to use it.

There are three sources of noise, base spread is only one of the three. At audio frequency, 1/f noise is more important than others. There is no particular way to predict 1/f noise, you look at the data sheet on the referred input noise voltage and current to determine, not just by base spread resistance. Base spread only tell you about the thermal noise. High beta lower the base current and lower the shot noise. You have to look at your application to determine which is the best.

I look at 2N4403, again there is not noise spec. How come both of you keep looking at transistors that are not specialized for low noise application and keep talking about noise?

[Alan0354]

Good idea Merlin. The transistors have to be surface mount, but we already use the FZT751 in other places, a surface mount version of the ZTX751. I'll try those out for noise.

It doesn't really address my original gripe, though: nobody makes a transistor actually specified for low-noise audio amplification any more. Diodes Inc. could decide to add some excess noise mechanism to the FZT751 tomorrow, it would still pass their quality control, and there would be nothing I could do about it. Not very convenient for mass production.

I wonder what makers of mixers and mic pres do nowadays.

I think so. They might not say for audio, but you can look at the noise graph, you can get the information. Here is the data sheet of 2N5087. If you look at Fig. 1 to Fig.5, you have all the information you needed for design already. This is a through hole, you can easily find the surface mount equivalent.

http://www.onsemi.com/pub_link/Collateral/2N5087-D.PDF

This one has very good low frequency performance.

[Steve Conner]

Nice! I didn't know about the 2N5087. There is a surface mount version, the MMBT5087.

[J M Fahey]

Being in an "European Standards" Country I use BC559C. Cheap, available and *very* good.
The PNP has less noise than the equivalent NPN.

[Alan0354]

Nice! I didn't know about the 2N5087. There is a surface mount version, the MMBT5087.

http://www.onsemi.com/pub_link/Colla...T5087LT1-D.PDF

Go to Digikey, they are cheap. $0.22 ea.

http://www.digikey.com/product-detai...SCT-ND/1139821

[Merlinb]

You know the two you point out are medium power transistor, I went and search the data sheets in digikey and there is no noise information on either one. They are not meant to be low noise front end application.

Yes, there are very few transistors that provide good noise information. The types I mentioned have been found to be suitable for low noise circuits through the experience and research of various manufacturers, not because it says so on the data sheet. But there are lots of electronic devices that are suitable for applications other than suggested by their data sheets.

[Steve Conner]

I went through a load of transistor datasheets looking for noise data. I just didn't happen to find the 2N5087. (Thanks Alan! ) The 2N4403 used to have noise data, but it was removed in the latest version. I don't think the BCxxx series are available in surface mount, and I want to avoid through-hole for this project.

Merlin, my point is that if you use it for something that's not on the datasheet, you can't complain to the manufacturer when it stops working. For instance, every transistor I ever tried, the base-emitter junction makes a nice 7.5V Zener. But if I used them for that purpose in production, I'd soon be in trouble.

Anyway, enough of our little three-legged fuses. Erm, friends rather.

[Alan0354]

Yes, there are very few transistors that provide good noise information. The types I mentioned have been found to be suitable for low noise circuits through the experience and research of various manufacturers, not because it says so on the data sheet. But there are lots of electronic devices that are suitable for applications other than suggested by their data sheets.

The concern is at audio frequency the major problem is not thermal noise from the base spread resistance, nor even the shot noise from the current. It's the 1/f noise that is not predictable. It does not show up on every device. For example, the carbon comp resistors. The thermal noise is EXACTLY the same as the metal film. What make the carbon comp noisy is the 1/f noise. It is unpredictable, you can test a 100 device and don't see any noise and then you design into the circuit. Then it'll come and bite you late. You can pick and find individual carbon comp resistor as quiet and metal film, but you don't design calling carbon comp on your design that require low noise.

You look at the general noise graphs, they rise at the low end, that's 1/f noise. Unless the data sheet specified for that frequency, don't take for granted even you did characterization on a few devices. Yes, if you are willing to hand pick device, it will work, if it does not have 1/f noise, then that device is good. But short of hand picking, don't count on it.

There are plenty of transistor providing all the noise data you need to do all the designs. The two I provide have everything you need to design the low noise front end. What yours are equivalent to 12AT7 or 12AU7 in the tubes, they are medium power transistors that are not meant for low noise front end.

It is not obvious which one to pick, it all depends on the operating conditions, your input impedance, power requirement and all. That's what the noise figure contour graph is for, showing you the noise figure at different impedance.

If it is not because of 1/f noise, all the FETs and the group III and V advanced semi-conductor will out perform BJT hands down, it's not even close. For one, FET has very low shot noise for very obvious reason, they don't draw gate current!!! Problem is their 1/f noise is so high that it is hard to use it in audio application. The Group III and V type has 1/f noise well into MHz. But it works for RF because you use small coupling cap to block the low frequency noise. As 1/f stated, noise is inverse proportional to frequency, noisiest at low frequency.

[Alan0354]

I went through a load of transistor datasheets looking for noise data. I just didn't happen to find the 2N5087. (Thanks Alan! ) The 2N4403 used to have noise data, but it was removed in the latest version. I don't think the BCxxx series are available in surface mount, and I want to avoid through-hole for this project.

Merlin, my point is that if you use it for something that's not on the datasheet, you can't complain to the manufacturer when it stops working. For instance, every transistor I ever tried, the base-emitter junction makes a nice 7.5V Zener. But if I used them for that purpose in production, I'd soon be in trouble.

Anyway, enough of our little three-legged fuses. Erm, friends rather.

The BC560 that JM Fashey might be interesting. It does not provide graphs, but it specified 2dB from 30Hz to high frequency. The thing I like is the beta stay at 100 even down to Ic=0.1mA.

http://www.fairchildsemi.com/ds/BC/BC556.pdf

Transistors sure are more friendly to me, don't you notice the tubes are kicking my behind right now!!!?

I agree with you 1000% even on the post before that unless it is specified in the data sheet, don't take for granted unless you are willing to hand pick device.

[J M Fahey]

BC559/560 *are* very good and alive and kicking.
The exact same die packaged in SOT23 is BC859/860
Go to Datasheet & application note database, pdf, circuits, cross reference, data sheet, datasheets | Datasheet Archive and in the search window paste "bc859"
Out of the many datasheets offered, a good one is:
>>>
BC859 Fairchild Semiconductor PNP Epitaxial Silicon Transistor
ri

5 pages, 58.94 Kb
Original
<<<
Covers both, 859 and 860.
I use them to build potted mini custom gain blocks ("hybrid ICs" if you wish)

[Alan0354]

BC559/560 *are* very good and alive and kicking.
The exact same die packaged in SOT23 is BC859/860
Go to Datasheet & application note database, pdf, circuits, cross reference, data sheet, datasheets | Datasheet Archive and in the search window paste "bc859"
Out of the many datasheets offered, a good one is:
>>>
BC859 Fairchild Semiconductor PNP Epitaxial Silicon Transistor
ri

5 pages, 58.94 Kb
Original
<<<
Covers both, 859 and 860.
I use them to build potted mini custom gain blocks ("hybrid ICs" if you wish)

I think the complementary part is BC550 which is used in EC boost. I have used this and I don't hear any difference in noise performance compare to MPSA18 in the circuit I designed. BC560 and 2N5087 is going to have a good shoot out!!!

[Merlinb]

You look at the general noise graphs, they rise at the low end, that's 1/f noise.

Thank you Alan, I am familiar with noise theory. I was just adding to the bulk of suggestions, not trying to out-do yours.