To Jfet or not to Jfet

I'm playing around with a 1:5 mic transformer I wound and looking for a solid state circuit to  couple it with. Works well with a 5532 but the tweaker in me wants something a little different. I was looking at some of Mr. Groner's mic amp circuits and bread boarded a couple of them. I really like this 2SK170 circuit:

http://www.nanovolt.ch/resources/microphone_preamplifiers/pdf/E_r1.pdf

But those JFET's are hard to get. Actually tried a regular 2N5457 and it wasn't any noisier... But I would be pushing the voltage limit of it.

Also tried a little simpler design like this one:

http://www.nanovolt.ch/resources/microphone_preamplifiers/pdf/D_r1.pdf

With an OPA2604 it seemed to perform just as well noise and distortion wise (real measurements).  Might have not sounded as "thick" but I don't know for sure.

Is there any practical reason to use a discrete JFET input with a 1:5 step up TX????

bluebird said:

Is there any practical reason to use a discrete JFET input with a 1:5 step up TX????

Click to expand...

Reflected impedance of 5k brings us in the comfort zone for many BJT-based circuits (5534 , most DOA's).

E.G. 5534
5 kohm source Johnson noise ->1.25uV
Noise due to EIN voltage: 3.5nV.141->0.38uV
Noise due to EIN current into 5k: 0.4pA.5k.141->0.38uV
Noise due to opamp: quadratic summing of EIN voltage and EIN current contribution -> 0.57uV
After quadratic summing -> 1.375nV That is 1dB noise factor compared to Johnson noise of the source.
In order to equal or better that with a JFET circuit, you need an EIN voltage of less than 0.57/141-> 4 nV/sqrtHz, which rules out many FET opamps except AD743, but is achievable with discrete LN FET's (LSK170/389/489).
Now, the question is "how much of an improvement will this be?" In theory, with 1nV/sqrtHz EIN voltage, the contribution of the circuit would be 0.141uV, resulting in a total noise of 1.26uV for a NF of 0.05dB. No need to say the precautions needed to achieve this figure are just about impossible to ensure, but in the end the improvement may be imperceptible.
I don't mean that it's not worth trying it, but just don't expect too much.

Excellent breakdown Abbey, thank you.

Years ago I copied the following from the Jensen website.  Any thoughts on this rule of thumb?


TRANSFORMER SELECTION

To determine the proper ratio transformer for your application, you will need to know the Voltage Noise and Current Noise of the active amplifier device to which the secondary of the transformer will be connected. If you divide the Voltage Noise by the Current Noise of the device you will get a value in Ohms. This value is the Optimum Source Impedance for best noise performance using that particular amplifier device. Select the transformer whose secondary impedance most closely matches the calculated value.

Example: An AD797 has an input Noise Voltage of 1.2nV per root Hertz Max. @ 1kHz and an input Current Noise of 2.0pA per root Hertz. This gives an Optimum Source Impedance of 600 Ohms. The proper choice of transformers for this device would be either the JT-16-A or the JT-16-B.

Special Note for Vacuum Tube Applications

If the above calculation is performed for Vacuum Tubes, the value obtained will normally be much higher than the secondary impedance of even our JT-115K-E transformer. We do not manufacture Microphone Input Transformers with ratios higher than 1:10 due to the limited bandwidth that is possible in designs of this type. The JT-115K-E will work very well in most Vacuum Tube applications.

bluebird said:

Is there any practical reason to use a discrete JFET input with a 1:5 step up TX????

Click to expand...

no.

in fact you can get jfet input op amps with respectable ein now....

JR

bluebird said:

But those JFET's are hard to get.

Click to expand...

The Linear Systems version isn't hard to get hold of.  Not very cheap for A grade, but available.

BT

ruairioflaherty said:

  Any thoughts on this rule of thumb?

Click to expand...

It's not only a rule of thumb, it's a determining formula that must be scrupulously obeyed, at least for BJT's. FET's and tubes OSI is way much higher than feasible; for tubes, we have to live with the fact that we can hardly do better than 4-5 dB NF and for FET's, even when improperly matched, their performance can better BJT's.

Regarding opamps check http://www.onsemi.com/pub/Collateral/AN-6601.pdf.pdf
Very few JFET opamps meet the 5nV/sqrtHz EIN voltage density that would yield a figure similar to that of a 5534. Most are typically in the 8-10 nV range.

rackmonkey said:

The Linear Systems version isn't hard to get hold of.  Not very cheap for A grade, but available.

BT

Click to expand...

Cheapest I've seen is $62 for 8 pieces. Do you have a better source?

Thanks for the info everyone! ;D I think I will forget the discrete JFET and just go with an OPA2604 because I can do +/-24 volts with it. Headroom is more important to me than a tiny noise difference I won't hear...

Abby, I bought two LSK489's from here for $5.70 ea:
https://www.trendsetter.com/Transistor-p/lsk489%20sot-23%206l.htm

Its an smd part and hard to mess with when prototyping. I was able to solder leads on it for the breadboard. I paralleled the two jfets and again it wasn't any different noise wise to the 2SK170. Gate capacitance didn't effect freq response  paralleled.

I would just trust my AP measurements but bread boarding for high gain circuits is messy and I have a feeling I'm measuring noise and low level distortion from a not optimal layout.
Regardless, just listening with headphones to a medium level condenser and even a SM58, pretty much any op amp circuit I try is very quiet. I definitely hear ambient room noise louder than any hiss at 40db gain...

In my experience the biggest sound factor to a mic preamp is the input transformer, or the loading resistors in a transformerless input preamp.

Example being if I use a a Jensen JT-16-A  (1:2)with a 5534 or a TL701 or an AD797 or even a discrete  transistor amp, Its basically going to sound the same. One might have a bit more noise but unless your recording something super quiet its not going to be a factor.
But put the same ratio, say UTC A-20 wired 1:2, on the front and it will sound significantly different tone wise than switching out any of the amp typologies.

I'm pretty sure the Neve VR consoles basically had a simple input transformer and NE5534 mic amp. How many amazing albums were recorded on Neve VR's? ;D

abbey road d enfer said:

Cheapest I've seen is $62 for 8 pieces. Do you have a better source?

Click to expand...

Not any more. I was looking at the same price. I heard that DIYAudio Store is the only non-OEM distributor these days, but Trendsetter still has them for about $8 apiece in the metal cans.

https://www.trendsetter.com/Transistor-p/lsk389b%20to-71%206l.htm

BT

80hinhiding said:

From the few things I've read about N-Channel JFET characteristics/how it works, I got the impression it'd be similar to vacuum tubes behaviour.  Did some more reading and some people did compare it to the way tubes handle peak transients/softens edges.

Click to expand...

It's a very common subject in audiophool circles. It is true that a JFET can be used without NFB with good results, just like a tube, which is just about impossible with BJT's.

I find myself hearing unpleasant distortion now in some older recordings, like the engineers pushed a BJT based amp way too hard or something, either intentionally or unintentionally.. perhaps one that had little feedback.  Just guessing.

Click to expand...

Ther are so many probable causes of distortion in a recording chain it is often very difficult to identify the cause. OTOH, many 1960's and 70's Nashville recordings have examples of poorly DI'ed electric guitars, where the sound's nastiness can be attributed to BJT's rebelling against mistreatment. 

I ordered some TL072 opamps and some J112 JFETs to play around with.  Didn't know any others to try at this point, but they'll probably come in handy somewhere in my work.

Click to expand...

TL072 won't tell you much, since proper use of opamps imply a lot of NFB, which tends to eradicate whatever sonic character may be lurking here. JFET single stages will tell you much more.