Do ULN FET's have more noise current?

I just bought a FetHead http://www.tritonaudio.com/index.php?option=com_content&task=category&sectionid=4&id=17&Itemid=33 and made some noise measurements on it (all done on AP system 1, rms detector, 22-22k)).
To my big surprise I found it to be very dependant on source impedance
Here are the results:
0R source: -118dBu EIN (no mistake here, it's really slightly noisier when shorted than loaded 200R)
200R: -118.5
470R: -114
1k: -112
Open circuit (actually 22k): -97
That is contrary to all other experiments I had conducted on FET's.
Is it a particularity of ultra low noise FET's to have a noise current that seems to be of the same order of magnitude as bipolars?
I don't have any info on the topology, but the manufacturer assured me they indeed use FET's. The gain is very load-sensitive so I suspect it may be a basic LTP with the outputs directly on the drains.

Yes, JFETs with low voltage noise typically show higher current noise. That's because of their higher gate leakage and drain-gate capacitance (which causes some feedtrough to the gate). But there need's to be another reason, these effects won't be significant at 1 kOhm source Z. How does the spectrum of the noise look like? I presume they somehow messed up the bias of the transistors (that runs on single-ended supplies, right?) which causes higher input current noise.

Samuel

I would look at the in circuit Vgs to make sure the gate diode is well reverse biased.

It does seem logical that a larger area diode would have more leakage, but I agree with Sam should be insignificant at modest low impedances. Another possibility is bad parts from a device process contamination? Perhaps even some PCB contamination? (just a guess).

JR

I can't tell you how the spectrum looks (I don't have the AP at home), but I can tell that it sounds more like pink than white, with a definite trace of 1/f.
The unit runs on phantom, so yes it's single-supply. Each leg draws 3mA.

Unfortunately, the unit is sealed. Would have to destroy it to see what's in.

That's strange.  Surely indicative of being biased(or being used) incorrectly, if indeed it uses a FET as the name implies.

While Jfets which have higher voltage noise will have lower current noise and vice versa, generally Jfets will still have much lower (input) current noise than (input) voltage noise in ratio, however a lot of them are still quieter than some BJTs in both areas these days.

Also, for such a low source impedance from a ribbon motor, I wouldn't think you would want to use a JFET fronted amp.  I would think the voltage noise would be killing you.  Perhaps the designer was thinking only of loading? 

abbey road d enfer said:

but I can tell that it sounds more like pink than white, with a definite trace of 1/f.

Click to expand...

May be technological 1/f noise corner is somewhere up.
It can be shifted to several kHz in some transistor unit.
It means defective transistor (bad passivation,. SiO2,...)
Swap the transistor and listen.

May be also Miller fedtrough, what is your Collector (drain) circuit???
When you have OP amp a use small condenser form output to minus input
you have also better performance for open input.


BTW gain current of the JFET (which is source of shott noise) is
(Dang Luong Mo:Excess Noise in a junction gate FET, Proc. IEEE, Letter. 1166 (July 1970))
Igx=A*sqrt(Vd)*Id*exp(B/sqrt(Vd))
On the picture can be seen, that Vd for optimal Ig would be as low as possible - here some 3V.
Try to use cascode in your amp. Bipolar upper transistor is enough.

Svart said:

Also, for such a low source impedance from a ribbon motor, I wouldn't think you would want to use a JFET fronted amp.  I would think the voltage noise would be killing you.  Perhaps the designer was thinking only of loading? 

Click to expand...

That's what I thought too, but look here http://www.groupdiy.com/index.php?topic=37547.0

xmvlk said:

Swap the transistor and listen.

what is your Collector (drain) circuit???

Try to use cascode in your amp. Bipolar upper transistor is enough.

Click to expand...

Not my design... and as I said before, the unit is sealed, I can neither see nor change anything

I had understood, or at least assumed, that these were aimed to give some boost to older 50R and 200R ribbon mics rather than hi-Z ribbons. I also had a hunch at the time that a product like this could have been inspired by this thread (although I don't have time to reread it all now)

http://www.groupdiy.com/index.php?topic=5743.40

And this is of interest.

http://headfonz.rutgers.edu/BR-pre.gif

Stewart

zebra50 said:

I had understood, or at least assumed, that these were aimed to give some boost to older 50R and 200R ribbon mics rather than hi-Z ribbons.

Click to expand...

They put emphasis on the rather high input impedance of the thing (22k). And they have a related product that is really Hi-Z
http://www.tritonaudio.com/index.php?option=com_content&task=category&sectionid=4&id=14&Itemid=32

And this is of interest.

http://headfonz.rutgers.edu/BR-pre.gif

Click to expand...

My original idea was to use a THAT 340, that would do 4.6 nv/sqrtHz at 20dB gain. My design has already been confirmed by them; all I need is actually do it. But since I'm a lazy old fart, I felt I just could buy something, but it doesn't live up to specifications.

abbey road d enfer said:

They put emphasis on the rather high input impedance of the thing (22k).

Click to expand...

So they do - I hadn't seen that!
Fedhead

They also say "When mated with a high-end dynamic microphone like a Sennheiser MD441, FetHead really shines"
The MD441 has 200R output, according to the datasheet.
MD441 datasheet

So is it for lo-Z and hi-Z mics?

OT; that's a new variation on the "inline barrel preamp" I'd not yet seen.  Maybe I'm missing something, but the noise specs on these devices seem always much poorer than a proper preamp.  I would think they'd counter some of the benefits they propose, though the input Z may have merit with some ribbons. 

Interesting discussion. 

Usually, the main problem is having an input noise at least as good (hopefully better) than a standard preamp. With at least 40dB gain, any decent pre has an EIN of -125dBu. For a number of practical reasons, a barrel pre has to be restrained to ca. 20dB. With a bipolar design at 20dB gain, it is very difficult to achieve a better figure than -122.
I must admit I have been lured; my reasoning was: using ULN FET's with 1nV/sqrtHz would produce -130 intrinsic noise, and I was naive enough to think that the noise current would be negligible. It turns out it is not the case. In fact I reckon a bipolar design can produce at least equal or better noise performance than this FetHead thing.

The Fethead seems to measure better than the noise quoted for the other variations.  I think one of the other variations quotes -110 dBu with 200 ohm source.  I have a hard time seeing how a 20 dB barrel preamp with that noise figure can improve the performance of a low output mic.  My gut says these things may be most useful for driving very long lines in noisy environments, and in that case show most obvious signal improvement. 

The BigAmp DI looks interesting.