Hi, all,
There are a lot of great small, fast NPNs out there, such as the BFG25AW, but for PNPs, you're more or less stuck with the BFT92 or BFT93. Or am I wrong about that?
The '92 and '93 have rising beta in the low milliamps, nothing specified at lower currents, and they're a bit on the chunky side--output capacitance of 0.6 pF or more.
Anybody know of nice fast PNPs with f_T > 2 GHz, output capacitance
IIRC PNP's inherently have higher fT. Your problem is likely saturation and charge storage?
What are you trying to do? ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
specified
This one has constant HFE down to 100uA and lower but the Cob is 0.5p
NE97833 maybe, but not remarkably better than BFT92.
I wish there were p-channel PHEMTS.
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
Cascodes in fast differential noise cancellers. I have to use them closed loop with a BF862 to get rid of their voltage noise at low photocurrent, so low CB capacitance is a big plus, and beta linearity limits the cancellation performance.
Noise cancellers are easy to get right at 1 MHz and 0.5 mA, but get progressively more challenging when you go faster at lower current. I'm trying to get shot noise limited performance and >55 dB cancellation from 50 uA to about 5 mA with a >10 MHz bandwidth.
It has a $2 ARM in it to handle the tweaks, and two completely separate front ends that get switched in and out with relays.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Am 17.07.2012 21:59, schrieb Phil Hobbs:
Intersil HFA3096 family perhaps, the naked chips seem to have an advantage. :-)
regards, Gerhard
specified
That's a nice part, but a good 10x too much money, unfortunately, and its future has to be pretty precarious at that price. The Intersil HFA3128 gets me 5 decent PNPs for $4-ish, but still nothing in the 0.25 pF range.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Me too. We'd need another few entries in the periodic table, though, because hole mobility in GaAs is pathetic--lower than in silicon.
For the main diff pair in the low current front end, I may use SiGe:C discrete devices and servo V_BE of one of the transistors to keep the dissipation constant. As you pointed out some time back, that doesn't work in normal RF BJTs due to Early effect, but something like a BFP640 has VAF ~ 1 kV--pure magic. Thermal isolation would require some thought.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
NE97733? 8.5GHz fT (typ) 6 (min). Remarkably more $$.
Am 17.07.2012 23:12, schrieb Phil Hobbs:
If you compare the capacitance for the npn / pnp in the HFAxxx you get CCB 300 fF / 600 fF for the pnp and 200 fF / 500 fF for the npn for die vs. packaged.
So, the package alone has 300 fF.
regards, Gerhard
Thanks, that's interesting. The BFG25AW is typically 0.2 pF packaged, so I gather that Isil did something dumb there.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
specified
At a buck thirtyfive? Too much money? Shazam! You must have started my kind of designs now :-)
[...]-- Regards, Joerg http://www.analogconsultants.com/
specified
<~;)
It's for licensing, to try to get a little recurring revenue so that I can branch out a bit from just selling my time.
A buck thirty-five, forsooth? Thou varlet? ;) I'm expecting to pay not much more than that for the STM32F100R CPU that runs the calibration and tweaks.
Total BOM including case and board should come in just below $200 at
100-piece prices, but we'll see. That's assuming $25 for three swoopy photodiodes, $20 for the board, and $25 for the box. Lots of engineering has to go in to make it that low--calibrating all the Ree' values, all the betas, and so on. The basic topology does most of the work, but it needs both AC and DC tweaks to get to the corners of the performance envelope. I think I can do almost all the tweaking using TL084s, 2N7002s, and a $3 dual 12-bit serial DAC. (The STM32F100 has a couple of 12-bit DACs built in.)Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
specified
<~Ah ... Hobbs & Son, Optical Works :-)
$25 for the box? That must be one fancy box. Same with the board, $20 would buy a big fat multi-layer in China. Unless the quantities are very small, of course.
Problem is that for RF PNP there really ain't no market to write home about. If you need only 10MHz BW or a tad more why a gigeehoitz transistor?
-- Regards, Joerg http://www.analogconsultants.com/
specified
<~I padded it a bit so that I could have some wiggle room on other stuff, in case I need a fancier calibration DAC or something like that. I'm planning to use either the 12-bit serial one or (ideally) a few PWMs with delta-sigma extensions a la Tim W.
The box needs to be aluminum with a light-tight gasket and a sliding shutter with three holes, so I can do the AC calibration.
I need the beta error and phase shift to be small (like 0.3% / 0.2 degree) at 10 MHz with as low a collector current as possible. With no tweaks, that needs an f_T of about 3 GHz. With tweaks, I can probably get that down to 300 MHz or maybe 100 MHz, but no lower. The f_T goes like I_C in the low-current limit, so if it's 2 GHz at 1 mA, I hit the wall around 150 uA. A smaller transistor would help a lot.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
specified
<~As Joerg pointed out, it's really only $1.35 in real life--I was seeing $4ish. That nice linear beta at low currents is pretty attractive. Now if the Ree' is reasonable, and I can figure out some method to return the base current to the collector, at least for small errors.
Something that turns a linear error into a quadratic one would be just the ticket. Darlingtons theoretically do that, of course, but the current-starved driver stage is so slow that the benefit is lost.
Then I could run well at f_T = (10 MHz)*sqrt(300) ~170 MHz, even without tweaks.
Do you have any tricks up your sleeve for that one, Joerg?
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
the BFT92 or
milliamps, nothing specified
side--output
capacitance &lt;~
Dunno about Joerg but one way to eliminate base current error is to use a JFET buffer- so it's a JFET-PNP darlington.
Am 18.07.2012 04:18, schrieb Phil Hobbs:
and gain falls with 12 dB / octave since it is the product of two transistors. That makes them feedback-unfriendly.
IIRC, Hollister showed a structure with a diode over BE of the second transistor in his wideband amplifier book. Cost some gain at dc but made it well-behaved at RF.
regards, Gerhard
...
maybe:
the fT-doubler circuit patented by Carl Battjes "US patent 3,633,120: Amplifier Circuit." September 16, 1973.
Thanks, Jure Z.
Thanks, that's a good one. It replaces each of the input devices with a diff pair, with the other input grounded and the other collector connected to the opposite side. That effectively doubles the amount of output current you control with a given base current, at the price of needing twice the input voltage swing, i.e. doubling the input impedance. That's good news for voltage amps.
In this case, what I have to work with is a photocurrent, which I need to preserve very carefully in order to get high-accuracy noise cancellation, so I don't see how to use the f_T doubler there.
The JFET Darlington is a good idea too, but I'd have to use a P-channel in order for it to behave like a PNP, and P-JFETs are scarce and horribly noisy. (Of course any JFET except for a BF862 is horribly noisy.) It would also have the same problem as the vanilla Darlington, namely very low first-stage transconductance due to low drain current.
I'll probably have an offset current tweak anyway, so maybe what I'll do is to use a BFT92 Darlington, and bias the first stage with a sub-Poissonian current source of a couple hundred microamps, i.e. a 200k resistor to the 40V photodiode bias supply, and then carefully take that current out again with another resistor to the opposite supply. With 40 volts across the resistor, the flatband noise power will be 800 times below the shot noise of the bias current, rising a bit with frequency as beta declines. That'll be a few dB noisier than the plain BFT92, but there should be room in the noise budget. If not, I can wrap a BF862 common-source amp around it to keep the emitter still.
I can probably use the same trick on the current mirrors that generate copies of the three photocurrents for tweaks and ancillary measurements.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.