John Curl's Blowtorch preamplifier

Status
Not open for further replies.
Jump to Latest
#18,481
john curl said:
From 1968-1973, the B&K 2619 used two 200 meg input resistors in parallel for an equivalent of 100 meg. This is what I requested changed to two 2 gig resistors in parallel for an equivalent of 1gig ohm. We paid extra for this change, but they upgraded all the mike preamps later.
For the Grateful Dead mikes, we had already changed them, ourselves, but they were Nagra based instrumentation mike modules, not B&K 2619 preamps, and they were easier to modify.
Click to expand...

John;
it seems to me I saw 2 mikes vertically aligned on each stand on one picture of the concert. What it was?
 
#18,483
The mikes that the Grateful Dead used in the 'Wall of Sound were 2 B&K 1/2" capsules in modified (to lower noise) Nagra electronic microphone preamp followers, paralleled 'out of phase' to create rejection of the far field, which was the sound system behind them. The polarization voltage was variable up to 200V.
 
#18,485
dimitri said:
Thanks, are they exist in discrete package?
Click to expand...


Of course:

http://www.onsemi.com/pub_link/Collateral/NTY100N10-D.PDF

About 0.15 sq. in. silicon active area. Bias it around 50-60uA for optimum noise results.

Unfortunately stabilizing the bias in the subthreshold regime is difficult and needs adjustments from device to device. It also needs to be accomplished by a very low noise circuit.
 
#18,487
john curl said:
This looks like nonsense to me. Prove your point, or be rebuffed! :geezer:
Click to expand...

Elementary my dear Watson :rofl:

For example http://www.eecs.berkeley.edu/~hu/PUBLICATIONS/Hu_papers/Hu_JNL/HuC_JNL_177.pdf and take a close look at eqs. (38) for strong inversion and (41) for weak inversion aka subthreshold regime.

You have everything in fig. 2 to measure and extract the noise parameters yourself (just replace the HP computer with a PC running LabView).

For a more didactical approach, take a look here: http://books.google.com/books?id=ew...X&oi=book_result&ct=result&resnum=7#PPA319,M1

eq. (11.2.8) shows the 1/f noise in subthreshold regime as inversely dependent to the specific Cox (the gate oxide capacity per unit area) and the device area. The following pages have more insights.

EDIT: I assume you are talking to me 🙂
 
#18,490
dimitri said:
Thank you for a nice link. Professor Hu published 868 works for 37 years, 23.5 papers per year or a paper every two weeks. I'm wondering when he is thinking.
Click to expand...

Well, this is a sad reality. Professors get credits for each and every paper that is written under their supervision. It is not unusual that they wouldn't have a clue on what the work/paper details are, so you'd better not ask prof. Hu about 🙂
 
#18,493
It would be a wonderful thing if we can overcome MOS 1/f noise to a significant degree at audio frequencies. It has always amazed me as to how little seems to be understood about its generation. I know, textbooks may give equations, but reality is something else.
It happens to be a historical fact that in the 1960's, we had a similar problem with jfets, but some professor, somewhere, found one, just one, 2N3819, with very low 1/f noise, so they knew that it could be done. Now, it is relatively easy to have very low 1/f noise jfets, and this has been true for decades.
Now, for MOS fets.
 
#18,494
1audio said:
If you are OK with feedback this is a very interesting option for a transformer input: Lundahl Zero Field Not usable for a mike input since it needs some drive from the source device.
Click to expand...


Basically what this does is to cancel the xformer DC resistance. This concept was patented in around 1982 by Bruce Hofer from AP, and is used in the output transformer circuit of the AP S1 and S2 (don't know about the others).
Presumably, the patent is expired.

Jan Didden
 
#18,500
john curl said:
I wouldn't call it 'elementary' but it is interesting. Thanks!
Click to expand...

John,

It's really simple 🙂 Step for a second away from that math and look from this perspective.

A MOSFET in subthreshold conduction (weak inversion) has the drain current dominated by a diffusion component. This current depends exponentially to Vgs and the slope is ideally log(10)*Vt=2.3*26mV=60mV/decade of current. In reality, it's everywhere from 70 to 100mV/decade of current, and that's one parameter that affects the noise, see below.

Under these circumstances, the MOSFET is essentially acting a a bipolar transistor, but with three major extra features:

- no rbb to generate voltage noise
- beta can be safely considered infinite
- no current noise

So from the three terms in the BJT equivalent voltage noise we have left only one, the term inversely proportional to the collector/drain current. You can safely assume that the MOSFET subthreshold noise expression is

Vn [nV/rtHz] = SQRT(K/Id)

where K is a constant and Id is the drain current (or, alternatively, you can use the subthreshold transconductance). For a BJT the constant is 2kT*Vt (Vt=26mV), for a MOSFET you can define an equivalent Vt based on the subthreshold conduction slope as mentioned above. Note that this equivalent Vt is precisely 26mV for an ideal MOSFET (with 60mV/decade conduction slope).

The only problem is that Id has to be very low, to keep the MOSFET in weak inversion. Here's why power (large area) MOSFETs are convenient: they are in weak inversion for tenths of microamps, while small signal MOSFETs like 2N7000 needs a under 1uA current... Therefore, from this perspective, high power (large area) MOSFETs have less noise than small signal MOSFETs. High power MOSFETs have unfortunately a higher equivalent Vt (and a few other problems), but overall they are (in terms of noise) still better than small signal devices.

I've published these things a long time ago somewhere in an IEEE journal, I can dig for a reference if you want to further pursue this matter.
 
Status
Not open for further replies.
Top Bottom