HEMT

latest >invocation of MAX was designed in a number of meetings chaired by my daughter, and >where I only got one vote out of 4 or 5.

And you have nothing to say about either.

John

nts. one of

of its fast

e use

ically

that you

aybe.

EOL or

probably

ough

cases for

Perhaps, but the only reason anybody is still using a matrix printer is becauwse it would cost money to replace it with something better.

e,

hat

me

ut

I haven't coded anything recently, but that doesn't prevent me recognising a defect when I hear about one.

"Retyping all the fields of the existing parts"? That would be a particularly mindless brute force solution. Any moderately sophisticated texst editor could do the translation for you - recognising "space M W space" and replacing it with "space m W space" doesn't require human intervention, though you'd probably want to do a checking session, with the converted data-base reduced to bite-sized chunks and spread over everybody in your organisation who could make sense of the data fields.

Sure, and I wrote a long memo about them in the late 1980's when I was still working at Cambridge Instruments and designing a lot of electronics. The substance of my memo was that production, purchasing and engineering all should be working from a common data-base, but the sort of information that they wanted to drag out of it ended up structured in distinctly different ways. I think I managed to work in a reference to Codd and the third normal form.

It was a waste of time then - because Cambridge Instruments was stuck with an antique IBM stock control system that they'd been using for many years, and the equally antiique hardware on which it ran.

Rewriting the memo today would be just as much of a waste of time - once a company has opted for a particular stock control system, the costs of "retyping all the fields" to move the existing data over to a better system is too intimidating to be contemplated, even for a small and relatively agile company like yours.

-- Bill Sloman, Nijmegen

Of course, they have to think a bit about it, which slows them down and distracts them, and makes them more likely to make mistakes, but you are prepared to throw away the processing time involved.

he latest >invocation of MAX was designed in a number of meetings chaired b= y my daughter, and >where I only got one vote out of 4 or 5.

So long as she doesn't change anythihg you care about.

I've never used an HEMT, much as I'd like to have the opportunity. I did have some fun with 5GHz broad-band bipolar transistors a couple of decades ago, admittedly about the saem time that Farnell took to stocking them as consumer electronics items.

-- Bill Sloman, Nijmegen

latest >invocation of MAX was designed in a number of meetings chaired by my daughter, and >where I only got one vote out of 4 or 5.

Next Saturday we have to move a roughly 500lbs fountain. SWMBO wants it at another location. Of course, per Murphy's law, that new location happens to be 5ft higher and one must negotiate some rocks and tiny free-form steps to get there.

Huh? Tell us, how would you print onto anything that requires a true carbon copy all the way to the signature because a law says so? Such as checks.

I don't want to hear about rocks. Mo decided last weekend that she wanted a rock garden near the street in Truckee. After a little experimenting, she decided she wanted only BIG rocks. It's a slope, so you can't just plunk them down, you have to dig little pits for them. In rocky ground. Then, because she can't visualize things like this very well, we start moving them around...

John

Then one fine day clouds roll in, big old dark ones, and a deluge of rain roars down, the pits wash away, the first rock begins to move on its own ...

My wife might ask me for that sort of job as well. Once. :-)

Here in the desert, any project that involves digging also involves deciding 'What to do with the rocks?' I have a fountain I built, with rocks embedded in a red concrete back. All the rocks came from my yard. I have a rock garden next to the house, with all native rocks. I filled in some of the slope beneath my deck with rocks. I lined another slope with rocks. I have begun covering my small side yard with rocks...

Fortunately, the big ones are only about 12" in diamter. We are in a an ancient flood plain, so the ground is just abount anything that might have washed down from the mountains above!

Charlie

Who knows, maybe one day you'll dig some more, pick up another rock and the whole thing has that bright yellow sheen and is incredibly heavy. Afterwards, you may never have to work again :-)

We have considered that! Some of them do have a kind of 'bright yellow tinge' to them... ;-)

Charlie

Not with my vision problems. I delete emails written in all uppercase or all lowercase without trying to wade through them.

Hi all,

I just find out this post looking for a way to drive a pHEMT with a ECL logic. I don't know much about pHEMTs but, can I use them in a DC Coupled circuit like I would do with a MOSFET?

I would like to connect the output of an ECL gate (like an EP89, for example, as John suggested) directly to the gate of a pHEMT and then an Rd load resistor to the drain and take an amplified (and inverted) version of my logic signal on the drain terminal (a simple common-source configuration).

Anything wrong with this configuration? Do they work without complex matching networks and bias tees? It would be great for my application since I need a DC coupled signal path and a component like a pHEMT with fast switching times.

Thanks a lot,

Michele

--------------------------------------- Posted through

That works. If you ground the source of the fet (which you should) you can set the Vcc of the ECL part to get the fet gate swing right.

Consider using an MC10EL89 or an EP89 as the gate driver, '89 outputs connected directly to the phemt gate with a zero-length trace. They swing almost 2 volts. Vcc around +1 or so would enhance the phemt gate a bit, which makes them turn on harder and faster. The '89s have differential inputs, so it's not hard to translate whatever logic levels that you have into the '89 inputs.

Outputs will fall fast when the phemt turns on; risetime will depend on the pullup resistor and capacitances.

How much drain current do you need? There are phemts that get up into amps.

The scope trace in this pic

ftp://jjlarkin.lmi.net/DSC01371.JPG

is a 1 GHz, 5 volt p-p square wave, from a breadboard. It's a SOT-89 phemt driven from an EL89. The drain is pulled up by a current source and clamped with schottky diodes to set the high level. That became this:

John

Thanks a lot! That is what I was looking for.

Time for me to look for the right phemt and do some experimenting.... some of the parts you listed are very promising.

Thanks again for the tips.

Michele

--------------------------------------- Posted through

Breadboard like this!

ftp://jjlarkin.lmi.net/BB_fast.JPG

John

I am impressed that such a prototype board could actually output a 1 GHz, 5 V square wave. Very good design, indeed!

I was looking at the board and I could recognize the ECL gate with the non-standard power supplies, the next IC must be the pHEMT and I recognize the clamping diode connected to the V clamp voltage.

From what I think is the drain of the transistor, I see a long red wire that connects to V+ through a resistor: is that the drain pull-up resistor? It's incredible how everything works, I was expecting such long wire to create a lot of inductance and slow down the transitions plus it might add ringing and oscillations.

In my class on RF electronics they always thought me to keep traces as short as possible and to minimize parasitics but apparently it is not always necessary :-)

Michele

some

--------------------------------------- Posted through

That matters when you need the trace to not act as a stub, but you can use that to your advantage as well. Short stubs can be modeled as capacitors or inductors, and if they have resistors, you only have to match the impedance

Satellite (microwave) hardware is a wonder of circuitry. Stubs, links, bends, all sorts of crazy things are handy at a single frequency. They typically use 1/4 wave lines as bias tees, so there's a bypass cap on one end and the MMIC or what have you at the other end. Up around 20GHz, these structures are only a cm long or so, so you can jam a lot of filters and amps into a little box.

Tim

You only need to do that on the actual RF/sensitive traces! (Assuming they are not in fact striplines anyway...). So the long wires and part leads you see on the photo will all be on "low frequency" signals. On JLs breadboard you can see that the signals that matter are indeed as short as possible, with coax brought right up to the device pins for example.

Jim Williams' AN47 is a great primer in high speed breadboarding techniques (although it predates surface mount parts). See page 26 and the appendices.

There's a small (0805 maybe?) ferrite bead sticking up under the end of the V+ wire; it's not visible in the picture. The wire itself is an inductor of sorts. For seriously good pulses, the ideal drain drive is a fast active current source with some inductance in its output. Ferrite beads are OK, more complex things are better, like the Piconics conical inductors.

If it's a matched impedance, like the 50 ohm output line, it doesn't need to be super short. And a long, very skinny trace makes a high-Z line, which can be useful for injecting bias into a lower-Z structure.

Fast time-domain stuff is different from RF, because you can't resonate out parasitics. DC-to-4 GHz is harder than 4 GHz.

John