The 2N7002 has served me well for many years but at 20MHz general purpose switching the 30nS rise times are just too slow by an order of magnitude. What SMD is readily available with less internal capacity and comes close to the other 2N7002 characteristics? Thanks, Harry
Above those speeds I usually use BJT such as the BFS17A. Those are a bargain. I bet that IRF or Zetex has other and hotter FETs. Whenever it had to be FETs I chose UHF dual-gates such as the BF998 but they won't go to any serious drain voltage.
Thanks Joerg, I would like to use FETs to save gate resistors and current drain when idling. The BF998 looks like a canidate but the 12Vds is limiting and I can give up some speed for >Vds and
I'm getting < 3 ns output edges from 2N7002's at 50 volts drain swing. You do have to drive a low impedance load, and you do have to drive the gate hard. In the over-20 volt range, within its current capability, I've found nothing faster, at least in silicon. My working model of a 2N7002 is an infinitely fast mosfet with source and drain capacitances and some wirebond inductance.
For lower swings, up to 10 volts, GaAs mesfets and phempts are much faster and have far lower capacitances, but they are expensive (a few dollars instead of a few cents) and are depletion-mode parts, so you need negative gate drive.
For higher voltages and blinding speed, GaN is the winner, at $50 per fet to start.
If all you need is 5-6 volt swing, TinyLogic gates and LVDS-TTL receivers can give you 600 ps edges.
It's the same with high voltage FETs where I sometimes have a hard time convincing people to at least try it. Which makes me wonder why FET datasheets are so very conservative WRT switching times.
datasheet shows considerably lower capacitances (the typical c vs v, in the curves), than the Zetex part, especially at low voltages. Normalizing to Rdsonmax, they wind up pretty close.
If you assign a figure of merit of Idmax/Cd or Rdson/Cwhatever, a lot of these small mosfets come out close, with none really outstanding. That's interesting, considering how old the 2N7002 is.
The Fairchild 2N7002 actually gets down to 1.2 ohms typ Ron, at 10 volt gate drive, versus about 1.6 typ for the BSS, so that places it way ahead. I usually design around typicals in key places, because most of my stuff simply wouldn't work if I used max values.
No. Silicon seems to be pretty repeatable in production. Of course there's a risk betting on typical/measured performance, so only a few key performance-pacing parts should be pushed like this. If you design around typicals for all 200 parts on one board, you're guaranteed to be in hot water. On or two, carefully tested and observed, can lead to serious performance advantages.
Just today, we decided to use some -2 volt rated schottky diodes at
-3.3 volts, for a serious performance payoff. They don't seem to leak much below -6, and they have 0.2 pF capacitance, so we'll take that risk.
50pF (max). Oddly, the graph in Fig. 9 shows typical Ciss=40pF.
Annoyed by the discrepancy, I scanned the BSS138 & 2n7002 datasheets from OnSemi, Fairchild, Zetex, and Supertex, and conclude that all appear somewhat inconsistent. It appears likely that actual switching perfomance is better than indicated.
For example, Fairchild shows Q(c) to 10Vgs as 1.7nC for both their BSS138 and 2n7002 alike, despite higher typical capacitances for the BSS138 straight down the line (Ciss, Coss, Crss). Of the two, their BSS138 has a much lower (0.7ohms) typical ON-resistance, lower threshold, and 50% high transconductance.
Of all the parts, Supertex's 2n7002 and Zetex's BSS138 show the lowest Q(c) to Vgs=10v at 1nC and 1.1nC, respectively, while OnSemi specs 3nC(!) for their BSS138--quite a variation.
Ultimately I suppose one has to test 'em to see what the real scoop is and how far they can be pushed.
I suspect that each vendor uses a single wafer and derives a bunch of different part numbers from it, with different data sheets. So it's prudent to stick to one vendor when something works, at least for non-trivial things like ns switching, and buy a reel of the silly things for 3 cents a part.
I missed that. I was quoting from the graphs at Vgs=0V.
Reviewing the various 2n7002 datasheets, it looks like Supertex's part benefits from lower Crss, resulting in Q(c) ~1.1nC to Vgs=10V, versus Fairchild's 1.7nC. From the graphs Crss looks like roughly 2pF vs. 4pF, which is 0.63nC at the tested Vds=25V. It's not free though--Rds(on) increases to 2.2 ohms (Vgs=10V), whereas Fairchild's is about 1 ohm (Fig. 2).
OnSemi and Zetex's 2n7002 datasheets only give worst-case specs; there's not enough information to analyze further.
As for BSS138s, I previously noted Q(c) for the OnSemi part is given as 3nC, but that's at Vds=40V, while the other mfrs spec at 25v. Crudely approximating by taking the Miller capacitance of 3.5pF typ and calculating Qc(40v)-Qc(25v) yields a 1.7nC adjustment. Here's how a few stack up:
(view in Courier or other fixed font) Fairchild OnSemi Zetex Rds(on) 0.7 2.2 1.6 ohms (Vgs=10V) Q(c) 1.7 1.3* 1.1 nC (to Vgs=10V, Vds=25V)
It has just ocurred to me, after all these years duh, that if you drive a 2N7002 from a fast logic gate (or preferably several in parallel), an inductor in series with the gate may well improve drive risetime. Using the old EROT
L = r^2 * c
and assuming, say, 30 ohms and 40 pF, the rough ballpark for L is 40 nH, much more than the baseline wirebond inductance. The Miller dynamics complicates things a tad, so it's best tested by experiment.
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.