On Thu, 07 Nov 2019 20:40:08 +0100 (CET), "Mat Nieuwenhoven" wrote:
On Thu, 31 Oct 2019 10:08:18 -0400, wrote:
On Wed, 30 Oct 2019 20:33:11 -0700 (PDT), Phil Allison wrote:
wrote:
--------------------
Also, bipolar amps don't use large devices to reduce noise.
** Discrete op-amps use paralleled BJTs at the inputs and fairly high current levels to get voltage noise down low as possible.
Enlighten us -
what magic trick do integrated ones use instead ?
Without getting too wordy, main components of bipolar noise are shot noise, which is a noise in the collector current and goes as the sq rt of emitter
bias current, the thermal resistor noise of all resistance at the base of the signal input bipolar devices, and the IR voltage noise developed at the
base by the shot noise divided by transistor beta times base resistor value. Increasing bias currrent increases shot current noise at the collector
and base, so low noise bipolar amps tend to be biased at lower currents. Base resistance is a function of the area and the length of the current path.
Since the depth of the resistor path in an IC is fixed by the fab process, one way the area and length are optimized is by using long, skinny min
width emitters with base contacts the same length as the long emitters and on both sides on the emitters. Usually the device will have multiple long
emitter stripes so that each long base contact in between the emitter stripes serve both emitters on each side, reducing base resistance by 2. True, a
larger device could continue to reduce base resistance, but other performance parameters must be met so base resistance is minimized by transister
geometry rather than brute force size. Not sure why discreet amps would use high currents for low noise. Shot noise is independent of process and
increases with current. Perhaps they practically eliminated all base resistance so there is no thermal or base shot noise component. They woud still
have the collector shot noise to deal with.Possibly they dealt with it with more transistors and more power eslehwere in the design. Don't know. ICs
don't have the luxery of having the unlimited power and cost available to high quality hifi stuff. Of course, amp designs with high power output
drive must use large transistors and high bias currents, perhaps this requirement reflects through the entire design.
I designed a decent low noise op amp a few years ago, the OPA1662, 3.3nv/rtHz noise densisty. -124db distortion, total noise+distortion 0.00006%,
22MHz GBW, 22V/uS SR, Sig/noise 95db, voltage gain 114db, and ..... 1.5ma power draw total.
Impressive specs. Is this an enhanced version of the OPA1622?
No, this was a new design. It's a two stage amp with a modified Monticelli stage for the output. I designed it about 9 years ago. It's been on the
market quite awhile. I was actually working remotely for Burr Brown just after TI acquired them. All of the support work, product definition, customer
visits, characterization, etc., were done in Tucson by BB (now TI Tucson), but it's manufactured by TI in Dallas. Don't know anything about the 1622.
I stopped doing work for TI in 2014 and haven't kept up. Looking at the dates on the TI web site, the 1622 was announced around 2016, 5 years after
the 1662, so with that and looking at the 1622 block diagram, it might be that the 1622 is a modified 1662, but I have no idea.