Input transistor shoot out.

[TUBESMAN]

I am working on a MC preamp. If the collective you were to do the same, what transistor would be your input device? The choices are: THAT 300,320,340 or SSM2212,SSM2220. The idea is to use multiples in parallel for lower noise and place them in the feedback loop of some high performance op-amp. Or, use them to replace the input stage of an LM318, as shown in National AN-222.
I'm looking for low noise and low distortion with average low source resistance MC cartridges.
Or, does it really matter that much since vinyl is 20db noisier than any of those devices! Thanks for your suggestions!

 

[kevinkr]

IIRC The 2N4403 is another rather inexpensive device that when paralleled can provide very good low noise performance.

Look for devices with low rbe..

 

[blu_glo]

I believe back in the day they used medium power transistors as those tended to have low re

 

[MarcelvdG]

If you want more suggestions for low-noise bipolar transistors:

http://www.diyaudio.com/forums/parts/222485-lowest-noise-bjt-transistor.html

 

[TUBESMAN]

How important is High Hfe? If I use the discreet devices like the 2N440x, then I might need to match them if I use multiples in either an LTP or in parallel. But that might not be a big deal if they are really cheap.

 

[MarcelvdG]

For very low source impedances such as those of typical MC cartridges, hFE is not that important. hFE matters for noise optimization because the base current has an associated base current shot noise. For very low source impedances, base shot noise probably won't be the dominant noise source anyway at any reasonable collector bias current.

 

[TUBESMAN]

For very low source impedances such as those of typical MC cartridges, hFE is not that important. hFE matters for noise optimization because the base current has an associated base current shot noise. For very low source impedances, base shot noise probably won't be the dominant noise source anyway at any reasonable collector bias current.

OK, so I'll just buy a bunch of one of these singles and match them for Hfe, just so current sharing will be more or less uniform. Without emitter resistors.

 

[MarcelvdG]

I don't understand; if you connect them in parallel, current sharing depends on VBE matching rather than on hFE matching. Or do you want to give each one its own base bias resistor and input coupling cap?

About connecting them straight in parallel; a temperature difference of 1 K will lead to a VBE difference at a given current of about 2 mV, or equivalently a current difference of (2 mV/(kT/q))*100 % ~= 7 % to 8 % at equal VBE. If 7 % extra dissipation causes more than 1 K of temperature increase, you get thermal runaway.

Hence, you should choose the intended IC and VCE such that you have less than 13 K of self-heating.

 

[TUBESMAN]

Vbe matching is the criteria of concern. But I cant see much of a way to limit heating to the levels you describe because generally, I think, each transistor needs to run at several Ma of current for low noise. I will run the numbers to see what dissipation I get.

 

[AndrewT]

How important is High Hfe? If I use the discreet devices like the 2N440x, then I might need to match them if I use multiples in either an LTP or in parallel. But that might not be a big deal if they are really cheap.

hFE @ 100 will perform quite differently from hFE @ 1000

Somewhere in between is usually adopted for Power Amplifier input stage.
I tend to select around 400 to 600 for hFE in small signal stages.

 

[sesebe]

If I remember well, I see in the past some very low noise audio amplifier made with input block made from many transistors in paralel, and these transistors was transistors for high frequency antenna amplifier.

 

[TUBESMAN]

Their is a thread here someplace that mentions that High Hfe is associated with some form of lower noise. I didn't have time to read it yet but it seemed to suggest that it had some affect. I will pour over it today. If I can find it again.

 

[jackinnj]

If I remember well, I see in the past some very low noise audio amplifier made with input block made from many transistors in paralel, and these transistors was transistors for high frequency antenna amplifier.

That's a JFET -- the BF862 from NXP -- surface mount only, about $0.35 if you purchase 10 -- it's a very nice device. It was designed for low noise a.m. radio reception.

 

[MarcelvdG]

Vbe matching is the criteria of concern. But I cant see much of a way to limit heating to the levels you describe because generally, I think, each transistor needs to run at several Ma of current for low noise. I will run the numbers to see what dissipation I get.

That shouldn't be a problem when you cascode them. A 2N4403 has a junction to ambient thermal resistance of 200 K/W, according to the Fairchild datasheet. 13 K of self-heating then corresponds to 65 mW. At VCE = 2.5 V, more than enough to keep it out of saturation, Ic then needs to stay well below 26 mA.

By the way, the optimal total bias current for all paralleled transistors together is

Ic,tot ~= (kT/q)*sqrt(hFE)/(magnitude(Zs + rb/N))

with kT/q ~= 26 mV at room temperature
hFE DC current gain
Zs source impedance
rb/N base spreading resistance per transistor divided by the number of paralleled transistors

Or expressed per transistor:

Ic,each transistor ~= (kT/q)*sqrt(hFE)/(N*magnitude(Zs + rb/N))

 

[TUBESMAN]

Lets say you planned to buy 2N4403's. What company would you get them from? Fairchild, Central Semiconductor or Micro Commercial Components?

 

[MarcelvdG]

I don't know. Maybe kevinkr has a preference, he came up with the idea of using 2N4403s.

By the way, the equation that I gave only applies when there is no or little 1/f noise, otherwise the optimal bias current may be lower. You can find more on the effect of hFE in post 44 of http://www.diyaudio.com/forums/parts/222485-lowest-noise-bjt-transistor-5.html, keeping in mind that base resistance usually dominates at very low source impedances.

 

[kevinkr]

I'd try samples from several vendors to determine whether some performed better than others. They are inexpensive enough. No preference although I do tend to buy a lot of Fairchild parts because they are cheap - not necessarily a good reason here.

 

[TUBESMAN]

OK. I am starting a search for methods to measure base spreading resistance. I think I'd go with Fairchild too.

 

[TUBESMAN]

Biggest problem I see is that all the really good transistors are discontinued, or second sourced from questionable suppliers.

 

[AMV8]

2 months ago I bought some KSA 992 and KSC 1845 from Mouser which seemed to me to be the best currently available.
My first choice was for the Hitachi 2SA 872a but they are not around.
KSA992 were 0.016p each ( approx $0.01 each ) and KSC 1845 were 0.024p each ( approx $0.015 each ) All Fairchild transistors.
Don