NE3509 SPICE model?

Have you asked NEC/CEL?

The datasheets aren't bad, for an RF phemt. There are actual DC curves!

We use them as fast switches, and I took a bunch of data - capacitances, Ron at various biases, tempco, extended DC curves - and I'll email you what I have. Some of my data is for NE3508, same thing but twice as big. No Spice model.

They appear to behave like jfets, in that you can ground the source and have the drain swing negative so long as you keep the gate suitably negative.

They also enhance to many times the zero-gate-voltage conductance, numbers like 4 and 2 ohms Rds-on for the respective parts.

Cute parts. I hope NEC keeps making them.

John

I haven't. Was in a rush and have plopped down a Calogic DS-series part because I know it can do the job in this case. But I'd like to be able to use the NE3509 in the future or in case things need to be even faster.

That sounds like a white knuckle ride :-)

Yes, so do I. There are lots of RF parts that can be used in pulse apps but other than companies like PolyFet it can be tough to get SPICE models.

Do you need a Spice model? As switches, these things are pretty simple.

No, just bias the gate a few tenth of a volt positive. Works great.

John

Yes, I do need a model. Reason is that things like Cgd versus channel resistance matter. I was in a hurry so used Calogic parts now, good enough for this application. There is a 2nd source (well, sorta) and there is a SPICE model.

This is a very fast switching app where slopes and ringout behavior matter.

Ok, I'd dare to go to 300mV. But any closer is scary.

The 3509 has a drain capacitance, mostly to the source, that's fairly constant at about 0.35 pF. Compared to the DMOS parts, that's the moral equivalent of zero.

What frequency are you working around?

They won't conduct much gate current below 0.6 or so. I didn't test them past 0.3 because I wanted to drive the gate from ECL, 0.8 volts swing, and I needed about -0.5 or so to turn them off. But +0.5 would be fine, and the drain resistance should go down even more.

PHEMTS seem to have a weird mode where drain current goes way up as you start to forward-bias the gate, almost as if it's starting to behave like a bipolar. I noticed this once but haven't had time to explore the effect.

John

One other thought: a BFT25 makes an interesting RF switch. It's easier to use in shunt mode, with the emitter grounded, of course. They are a fraction of the price of PHEMT or DMOS discretes.

I have data!

John

Don't diss the old DMOS parts, their capacitances are also very low. Ok, not quite 0.35pF and the Rdson is tens of Ohms. But it was good enough in this case and, most importantly, there was a SPICE model.

Roughly between 0.5 and 1GHz.

That would be interesting to explore. Can't you do a contest for students like you did before? They could find out for you. And get an honorable mention in your patent application :-)

Just one problem in my case: I found that even hotshot RF transistors take a noticeable time to get back out of even a slightly saturated condition. Nanoseconds, but still. The other issue is that you can't get them to zip Vce to zero like a FET can with Vds. That is required in most of my cases because of a DC component that must not be disturbed.

It would be cool to have someone who could occasionally do some parts testing, or Spicing, or parts research, or breadboard circuit testing, for interesting ideas that we don't have time for.

But if I did that, I wouldn't get to solder any more at all. I'd need a new PHB hairdoo.

John

Pity. They do have a collector DC offset that is a function of base current, a few 10s of millivolts typically.

Sounds like you're designing a signal blanker, maybe?

John

Do they work better upside down? Inverted transistors can be quicker than the same one right way up, and V_CEsat is very low.

Cheers

Phil Hobbs

I don't about that how ever, I've opened some RF modules that you wouldn't want to operate upside down due to a gel that just sits there with out any apparent support from the housing cover to hold it in place. I don't know how such a module could pass high G test up side down.

Jamie

That would be interesting to try. Capacitance would probably be higher, but some tricky base drive impedance could fix that.

Upside-down transistors can be made to saturate at precisely zero volts, or less than zero if you drive the base hard.

John

This one was more of a fast clamp & DC-restore function. A residual DC-offset would not be so cool. However, at least according to SPICE it'll work with the older devices. We'll see when the board is done, when the rubber meets the road :-)

By "upside down", I mean inverted, i.e. forward bias CB, reverse bias CE. You get a low beta and low breakdown voltage, but much lower feedback and output capacitances and a really low V_CEsat.

Cheers

Phil Hobbs

Upside down, their offset is lower.

Cheers

Phil Hobbs

Oh, I measured the BFT25 reverse beta yesterday. On the ones we have in stock, it's about 4. So I'll be using gaasfets to discharge my ramp and switch the range caps.

The fets are more expensive and need level shifters for their gate drive, minor nuisances.

faster.

matter.

Hmm. 16 did seem a bit high, but a factor of 4 error in BR doesn't give me much confidence in the model.

Sure it wasn't oscillating? Diode-connected BFT25As want to do that when you get somewhere above 100 uA.

Cheers

Phil Hobbs

switch

faster.

matter.

Not sure, but it's still set up (who cleans workbenches before another problem comes up?) so I'll poke around a little more. But the reverse beta increased from about 3 to about 4.5 smoothly as I went from approximate saturation to a couple of volts e-b, so probably it's not oscillation.