MMIC tolerances

For the classic InGaAs darlington mmic, you force bias current into the output pin, so it doesn't have any choice in the matter. I've found the gains (Mini-Circuits, Sirenza, W-J) to be very consistant, nothing like 2:1 span.

You *can* also poke current into the input pin to shift the bias point. Sometimes that's handy.

Any self-respecting mmic is unconditionally stable.

John

Interesting, do you have an test data showing how much improvement biasing the input pin makes? It seems like this method would only be valid if the input was not matched to 50 ohms with the previous stage, no?

Do you remember how much difference you saw in gain?

In my case more than 0.5dB would throw this application off the rocker. Then I'd have to use an opamp and in my case I'd have to use it in non-TIA configuration. THS4021 or something like that in standard non-inverting fashion with the photodiode into a resistor at the non-inverting input. Reason is that I was just informed that the PD version can change and I don't want things to become unstable because the new one has a different capacitance.

Yes, I agree. But the very low NF varieties seem to all have that danger zone up in the inductive part of the Smith chart. Actually, so do many "roll your own" low noise amps.

As long as the part is still in its linear range, pushing the bias won't change the impedances much. For example, if the expected signal swing range is small, you could pull up the input a bit, pulling down the output, and run the part at lower voltage and power. I use these parts in time-domain apps, and if I know that the signal is predominantly in one direction (say, pulses from a microchannel plate or a photodiode) I can bias the output in the opposite direction and get both lower power dissipation and more peak signal swing.

Most of the ERA-type mmics are simple inside: just a darlington with a feedback resistor, easy to model at DC.

John

If you need extreme gain accuracy, I wouldn't use any mmic. They wouldn't be temperature stable to a fraction of a dB... they're just darlingtons with crappy feedback resistors. So an opamp, either as a tia or just using a load resistor followed by a positive-gain stage would be a lot more stable.

Have you looked at the THS4302 series? Vicious little beasts. I was thinking of dumping a photodiode into a grounded resistor-inductor series pair, and amping that with a THS4303 maybe. Tweak the L to cancel some of the pd capacitance and extend the bandwidth a bit. I'd consider a t-coil, if I understood them better, which I don't.

After looking at some other MMICs I have just deleted that part of the schematic. Even the good stuff from Mini Circuits is slightly above tolerance spec for this app. At least it's not like in the olden days when I'd have lots of eraser turds on the floor now :-)

So it's the load resistor opamp combo. Somehow that doesn't look high-tech and cutting edge in terms of cost but, oh well, it'll work.

I don't need to go that high, just to 100MHz. Also, I'd like to get away without more regulators and the THS4021 can live nicely with +/-12V rails.

A peaker coil? That would be cool but this amp is a true hotrod. 12GHz GBW, wow. Possibly the only way to create a reliable inductance here would be to use one embedded in the layout. Just make sure its magnetic field doesn't see any part of the opamp feedback.

Probably this is the frequency range where you might benefit from a discrete design with one of those 45GHz RF transistors. Problem is that many of these are from EU manufacturers and I have experienced procurement nightmares there. Great products but very poor marketing.

T-coil? I've only heard that in connection with hearing aid coupler circuits in churches and public auditoriums.

I wish people would look up existing terminology before they make up names. A real, classical t-coil has nothing to do with deaf people.

If you couple the plate of a tube to the grid of the next one, the plate load resistor and Cp+Cg have a time constant tau, and the bandwidth is 1/2.2tau. Adding an inductor in series with the plate load resistor is "shunt peaking" and improves things roughly 40%. Adding another between plate and grid is "series peaking" and helps more. An ideal t-coil improves bandwidth 2.8 times the basic circuit.

John

Yeah, nowadays marketeers create such terminology. Some of the major hearing aid companies use the word T-coil. But the term "deaf people" isn't correct either since it's only useful for folks with some remaining hearing ;-)

I have never really understood why magnetic loops are needed in the first place. All the industry would have had to agree upon is a common method of close range LF transmission.

Interesting. I've seen coils in line with the plate but when they were used to muffle a load capacitance they were often called peaker coil. But most of the time it was there to mute unwanted oscillation way above the operating frequency (coil wound around a resistor).

Thomas Lee's got a nice write-up of the different ways to do Inductive Peaking including the Bridged T-Coil in: The Design of CMOS Radio-Frequency Integrated Circuits, Second Edition (Hardcover)

I also liked his first chapters on the early history of Radio Circuits as well but a lot of people don't like what they call his "disdain of rigor".

Robert

Amazon now want my email address before allowing to preview. What are they thinking? Or, are they thinking?

Yeah, not scientific enough probably. I like those books. Have to look next time at Borders. Although I am not sure I'll go there anymore because ours has shrunk the EE section to almost zero :-(

Other good books that disdain rigor:

AoE

Phil Hobbs' "Building Electro-Optical Systems". Full of quirky quotations, cartoons, lore. Also dense with good stuff.

Thomas Lee's "Planar Microwave Engineering." But I don't know why he calls it "Planar", since it covers most everything.

All three are worth reading cover to cover.

John