Den mandag den 26. september 2016 kl. 22.45.50 UTC+2 skrev Jim Thompson:
Humongous can get you in trouble :P
Den mandag den 26. september 2016 kl. 22.45.50 UTC+2 skrev Jim Thompson:
Humongous can get you in trouble :P
Adding 100mOhm does wonders to the phase margin...
...Jim Thompson
-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 |
There you go.
-- Thanks, - Win
My point would be that a dual transistor is a single low-cost sot-23 package, just like a single transistor. Well-behaved at a low price. I originally suggested an op-amp, a low-cost sot-23, of course, for the same reason. A transistor or an op-amp, nearly the same impact. The dual transistor is a well behaved approach.
-- Thanks, - Win
Are they really dual transistors or two dice in a single package? A few years back, I needed a dual but the only thing I could find that wasn't big bux was a Zetex part, and of course Diodes dropped it when they took it over.
Most of the pats in the AoE-III, Table 8.1b, page 502, are two separate transistors in one package. That's true of both the DMMT3906 and the BCM857DS. However, they're closely- adjacent die parts, with nearly identical characteristics. I'm not sure how much we lost when the monolithic parts disappeared.
For example, the DMMT3906 has 1mV Vos matching and 2% hfe matching. D-K has over 100k in stock at 11 cents/qty 1k and it's an active Diodes, Inc., part.
NXP's BCM857 has 2mV Vos matching and 10% hfe matching. It's low-cost and vibrantly-active.
-- Thanks, - Win
What about the thermal characteristics?
Horrible. Hardly better than two separate packages, apart from the approximate die matching. The late lamented MAT02 and LM394 interdigitated sections to zero out gradients.
For a new laser noise canceller design, I've been looking at the possibility of servoing V_CE to equalize the dissipation in the two halves of the main differential pair. Some time back JL pointed out that Early voltage would make this not work with the transistors I was using. However, SiGe:C bipolars such as the BFP650 have gigantic Early voltages that make this worth a try, I think.
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 hobbs at electrooptical dot net http://electrooptical.net
I forget what the power dissipation was here, but it wasn't a lot.
These sorts of dual transistors make horrible mirrors if the power dissipations aren't matched.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
On Tuesday, September 27, 2016 at 5:38:40 PM UTC-7, John Larkin wrote: ...
jpg
...
There's enough headroom in that design before the PFET conducts to put a re sistor in the emitter of each of the current mirror transistors to improve the balance and reduce temperature effects - you could probably even use se parate transistors without a problem. Or replace one with a diode or just a base resistor to the positive supply.
kevin
Cascode configurations cure all that. ...Jim Thompson
-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 |
Not horrible, just not as good as the grand old interdigitated pair of parts on a die. There's pretty good thermal conduction between the two parts on the die.
It's usually easy to insure both parts have nearly identical thermal dissipation. Get this concomitant benefit automatically by designing for nearly identical Vce voltages, to eliminate Early effect voltage offsets. For example, use a Wilson mirror instead of a simple Widlar mirror for high-voltage outputs.
Or come close in the case of circuit's like Jim's HV LDO, where one BJT has Vbe, the other has Vgs.
-- Thanks, - Win
If you look at the datasheet for the BCV62C SOT23 current mirror, they say that the limit for stability of an undegenerated current mirror is typically 5 mA @ 5V, i.e. the output device runs away at a dissipation of 25 mW.
The tempco of collector current vs differential temperature is about
8%/K, so if the device runs away at 25 mW, an 8% increase must be enough to give a 1 degree differential temperature change.That means that the thermal resistance between devices can't be better than
d(delta T) / dP = 1 K /(0.08 * 0.025 W ) = 500 K/W.
This is considerably worse than theta_JA for an ordinary SOT32, which is usually around 300 K/W.
You can sometimes keep the dissipation the same, but not if you want accurate differential pair performance over a reasonable range of splitting ratios. That's what the MAT0x parts were good at.
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 hobbs at electrooptical dot net http://electrooptical.net
A little bit of emitter ballast R plus a cascode arrangement can make even discrete elements well-behaved. ...Jim Thompson
-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 |
For sufficiently small values of "well-behaved", at least compared with a MAT02 or LM394. Those tricks only work if the two sides are running at similar currents. Maximum-power biasing has the same limitation.
And degenerating a current mirror slows it down dramatically. (The switcheroo Wilson mirror actually contains 3 local feedback loops, which is why this happens.)
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 hobbs at electrooptical dot net http://electrooptical.net
I seem to make mirrors that run nicely at 100's of MHz >:-}
And the Wilson mirror is hardly optimal... besides he's a cretin second only to Solomon ;-) ...Jim Thompson
-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 |
Sure, me too. Now do it with SOT-23s running at 1 uA. ;)
The switcheroo Wilson is a pretty nice circuit. That's the one with 4 transistors, with the diode-connected ones in opposite corners. Cancels out the beta dependence, and the aforementioned three local feedback loops--two diode-connected BJTs and the overall feedback.
Doesn't work for s**t with non-monolithic devices, though.
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 hobbs at electrooptical dot net http://electrooptical.net
... 500 K/W.
Be an optimist, it means it can't be worse than 500K/W !
It also means a 4- transistor variant (cascode current mirror) would work out well, the power hog won't thermally couple excessively to the sensitive pair. Alas, there's no off-the-shelf package of four transistors that really works well for the thermal coupling of the sensitive pair and also decouples the others.
Use a matched pair for the sensitive ones, and something else for the other two. Cascode transistors don't have special requirements.
-- Thanks, - Win
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.