Discrete Opamp Open Design

[RNMarsh]

Originally Posted by EUVL

> This is not correct: 2x2SK117=2x2SK209=2SK2145=2SK3320

Was a typo from me in a rush. Sorry for my blunder.

2SK117 = 2SK209 = 2SK880 = 1/2*2SK2145 = 1/2*2SK3320
2SK246 = 2SK208 = 2SK879
2SJ103 = 2SJ106


Patrick

What is the short list of good complimentary jFETs??

Thx - RNM

[Wavebourn]

Originally Posted by elektroj

Wavebourn, do you mean by that circuit similar to Widlar Op-Amp Current Source?

I am not familiar with such terminology. I mean, recording head should be driven my current source, not voltage source.

[john curl]

Yes, Widlar current source is optimum.

[john curl]

My load was 1mH.

[john curl]

I take back my opinion, although the Widlar current source is OK, I used something better do current drive my record heads. I don't know the exact origin of the circuit, but it 'stumped' R.G.Meyer, when I described it to him and he drew it on the blackboard in class.

[scott wurcer]

Originally Posted by RNMarsh

Id set at the zero temp - they work fairly well.

From the datasheet that's about 500uA. With 1K between the sources you would only have an average of .25V Vgs, so you can't use a lot of the distribution.

[bcarso]

Originally Posted by RNMarsh

What is the short list of good complimentary jFETs??

Thx - RNM

I'd say there are about ... none. The only P channel devices still in production that I know of are from two rather-low-performance processes, that from which the 5460 is drawn, and that from which the 5114 is pulled. Low gm, highish noise, low breakdown voltages.

There is of course stock here and there of the 2SJ74, the nominal complement to the 2SK170, and the even rarer P channel dual the 2SJ109 (complement to the discontinued 2SK389 dual). And at that the complementarity was focused on the transconductance, at the expense of capacitances and undoubtedly an even more extravagant die area. It is no surprise that Toshiba found them unprofitable to maintain.

On the other hand, partly because of the Benjamin Franklin direction of current flow there continues to be substantial development and availability of P channel DMOS with a reasonable range of properties. As remarked, there are no guarantees on noise, particularly at low frequencies, so in general such parts and their N channel cousins are not recommended for low noise early stages where the noise gain is high. As cascoding parts however, since the preceding transistor is usually possessed of a fairly high output Z, they can function reasonably well, except for the requirement of a gate bias for the enhancement-mode devices. Since the d.c. current at the DMOS gate is negligible this bias source can be a photovoltaic one or a battery, although this will hardly be very compact. I'm not sure if there are any depletion-mode PMOS devices like the NMOS DN2450 and similar parts, although if one were to order a million pieces maybe Supertex could be persuaded to try to make them

EDIT: Of course a folded cascode gets rid of the PMOS biasing voltage problem, but the pullup resistor that supplies current to the input device and the PMOS device has to be large enough to make the voltage noise of the PMOS have a small effect. This costs voltage and power dissipation. The resistor can be replaced with a current source, although still some volts are needed and the noise contribution from that source will be more than that of a resistor spanning the same available voltage.

[scott wurcer]

Originally Posted by scott wurcer

From the datasheet that's about 500uA. With 1K between the sources you would only have an average of .25V Vgs, so you can't use a lot of the distribution.

Sorry Dick I misread the schematic (I read 51.1 Ohms on the penciled in values), you were aiming at 800uA but does not look too close.

[scott wurcer]

Originally Posted by bcarso

I'd say there are about ... none. The only P channel devices still in production that I know of are from two rather-low-performance processes, that from which the 5460 is drawn, and that from which the 5114 is pulled. Low gm, highish noise, low breakdown voltages.

The Fairchild PFET's are 'very' noisey.

[bcarso]

Originally Posted by scott wurcer

The Fairchild PFET's are 'very' noisey.

Indeed --- I used to wonder if the old databook values could possibly be that bad. Unfortunately, they appear to be about right. Ouch!

The attached is from an ancient National Semi databook, and the process 89 is from which the 2N5460-61-62 parts are sourced. The somewhat larger die from Process 88 is a good deal better, but both fall woefully short of the noise performance predicted by Van der Ziel, even at 100kHz, where the curves are beginning to level off.

In light of this, the 2SJ74 etc. was quite an achievement.