Bipolars for low noise input - Alternatives to 2SC2240

[Blackjack8635]

Originally Posted by Salas

The sensitivity of your loudspeakers and your system's total voltage gain and listening distance plus background quietness will play a major role in that test. Noise in applications is always relative. Best engineering goal should be 'adequate' noise level in any specific application than absolute minimum IMHO.

Good point, I will reference against a real music input played at a suitable level and possibly played very loud. I should be then able to refer this back to an S/N figure, and as you suggest I may very well find that using a single input device is adequate and paralleling trannies and opamps in the quest for a lower noise MC input is a quest too far and that energy might be better expended else where. Which would be a good result

[ticknpop]

i believe I have have 3 or 4 each of the 2SD786 and 2SB737 Rohm around that I bought from Erno Borbely in the 1980s for his phono stage that I never finished. some were soldered in with trimmed leggs but never powered and some are spares - i'd sell them for $30 plus shippping. i'll count what I have if anyone's interested.

[Blackjack8635]

Lashed up a breadboard test on 2SA2547 MC preamp the other day and cct all working OK, but a lot of electrical pickup from lighting etc and limitations of test set meant that I need a special additional gain stage and some proper shielding to make a realistic noise measure.

Today built an input stage noise XL sheet and poking the 0.5nV /rtHz of 2SA1085 into it at the stated parameters back calculates to something of the order of rbb' 15R

[rayfutrell]

Not surprisingly, 0.5 nv/rt-Hz is approximatly the Johnson (thermal) noise of a 15 Ohm resistor.

[Blackjack8635]

Originally Posted by rayfutrell

Not surprisingly, 0.5 nv/rt-Hz is approximatly the Johnson (thermal) noise of a 15 Ohm resistor.

Well spotted, the test conditions on the data sheet specify an Rs of zero which nails current noise. Ic is specified at 10mA which reduces the shot noise component of en to an insignificant level which leaves the Johnson noise of rbb'. QED

Interestingly, I checked my Linn Troika at 2R ( 150, 200 uA? )and my Goldring ? at 9R (approx 500uA.)

Using required gain and real amplifiers, common emitter, common base ( cascode) and parallel op amps in spreadsheet noise calcs, what will work for one will not work for the other. I treated parallel transistors as a single device with rbb' simply divided by the number of parallel devices.

I have always thought that transistor matching is unimportant, and now looking analytically I believe that is correct ( tell me if I am wrong). and that attempts by Naim et al to balance collector currents was always misguided. Shot noise is not dependent on any individual device and so it does not matter whether all the current flows through one device or is shared. So long as each parallel transistor carries some current, even a tiny bit, then the rbb's are in parallel and so the Johnson noise calc stands.

In a cascode preamp with an Rs of 9R any collector current above say 2mA and the noise current component is dominating the figures.

Ideally the Rfb ( feedback resistor ) needs to be kept no larger than Rs and DC blocking capacitor is becoming HOOOOGE, or must be a avoided.

In this situation the best circuit I have is Nat Semi AN222, but once the currents and O/P gain are increased for a low O/P cartridge then distortion begins to suffer ( according to a SPICE simulation)

Against an Rs of 2R then the NF is going to be bad, ie. 9dB but that is still around the 0.5nV/rtHz.

So I come back to opamps. AD797's ( in the plural) 3 or 4 used in parallel even with Rfb of 4.7R or 10R will come out at a composite equivalent en of 0.5nV/rt Hz and so long as supply decoupling is correctly dealt with, no tantrums and vanishingly low distortion and no seriously bulky C's.

[hifidaddy]

I have tried THAT 300 array of four NPN transistors as well as MAT02FH in this circuit.

With the THAT 300 I experienced a high sensitivity against high frequency breakup and hum. It picked up a lot of noise. It took some half hour to cure this. In the end I installed a parallel combination of 330 Ohms with 15nF from each plus or minus input (remember both are floating) to ground. I also put the substrate of the THAT module to ground. As this device was unused, I gave it some hours before I was sure that the sound was not related to missing burn-in. The 2SC2546 I used before had been used years ago as well as the MAT02FH, but I will come to the MAT later.

The sound of the THAT was unpleasant, downward nasty. Lean, shallow bass, shallow soundstage, wirey treble, edgy midrange. Maybe there was still some HF leftover ... no wonder if one reads the datasheet. 350Mhz at 1mA is fast like hell.

Enter the MAT02FH. Immediately there was some relief. It sounds right OK from the start. It sounds more correct than the 2SC2546, one can notice that bass is tighter, but there was some idleness in the musical presentation, bordering the impression of boringness. Treble was less fine as with 2SC2546, and midrange less transparent. BTW, I had nearly the same impressions when I used them in discrete op amp style preamp modules a decade or one and a half ago.

Next to try: LM394, 2SD786, 2SC2240. Maybe I reverse the polarity for 2SB737.

Hartmut

[hifidaddy]

I just forgot to add: I tried to take meaningful measurements with ARTA software on my Laptop, but somehow I did not succeed.

Hartmut

[Onvinyl]

Hi Hartmut,
good thread, and I will follow it with interest.
But, at least for me, there is no way around jfets for MC-input. It is the only choice for a virtually DC-less input stage without compromises.

One might consider input transformers, but ... well ... show me a acoustically transparent one...

I did experiments with common base input stages myself, and could never guarantee not getting some DC offset at the input sooner or later. Plus, I found the sound of a complementary jfet input stage, both with a folded cascode or normal cascode topped with a current mirror (pass style) better in terms of bass firmness, resolution and magic (c) when used open loop. Even with feedback (syn08's HPS) could be made sounding quite good.

I support your findings vs. stiffness of working points, but not vs. mirroring.

However, I know you have a sharp ear, so your findings will be of interest for me!

Rüdiger

[masag1]

Hallo Hartmut,

If you need 2SA1316 and 2SC3329 for testing, please send me a PM.

With kind regards.

Sam

[hifidaddy]

Hello Ruediger,

I have many years ago tried SK170/240/146 and the like in common source configuration, with a resistor as load, no cascode at all ... among these usual suspects, I liked the 2SK97 best (see the 1995 Kaneda preamp), but what I missed with all these JFETs was naim style tightness, grip and PRAT. Since I had some lucky years with naim and naim copies in 80ies, I got used to what naim does right, and I am reluctant to pass on this (no pun intended!).

Though, as you mention bass tightness with your proposed configuration, I think I will give it a try this week. I still have a JFET moving coil stage here, have to add a cascode only.

Thanks a lot.

Hartmut