Using MMIC's

[Wendymott]

Hi Peeps... For a long while, I have had a few different flavours of MMIC's, and they can be very useful as block amps when wideband RF gain is required.
However I have come up against a problem, that I hope someone may shed a bit of light upon.
The device is the MAR-06, which has a gain of 18.1 db. it is in a standard config with 6V dc and a 150R current setting resistor and an RF choke.
The problem is.. that I made an amp with two in series, hopefully to give a gain of 36db, I fitted a SMB connector on the output of the first device and one on the second, so I could select 18 or 36 DB gain. It is quite apparent that somewhere I have gone seriously wrong, as... if the second stage is connected via the coupling cap, the gain of the first stage is say 9 dB.
Please advise if I am totally way off and suggest a solution.

[Radio Wrangler]

Beware!

One of the main makers of mmic amplifiers (and schottky diodes) has just obsoleted their entire portfolio.

Mini circuits bought a number of their MMIC amps from them, as well as diode quads for their DBMs

The microwave semiconductor division of HP went with Agilent when the firm split up. Agilent sold them off and their name changed to Avago. Avago recently took over Broadcom in what looked like a sprat eating a shark sort of deal. So now the combined firm (they chose to go by the Broadcom name) have just dumped te entire Avago portfolio.

Is the world going mad?

I know a number of firms who are now frantically planning the redesign of their products, and have sworn to ban anything from broadcom.

Expect a number of minicircuits parts to vanish.

David

[Radio Wrangler]

Oh, and it also looks like a reel of dodgy MAR-6 got retailed by Mainline fiveish years ago. The discrepancy was identified and purchasers refunded, so it looks like mainline did the honest thing. But there may be dodgy parts still out there.

Some MMICs do not take kindly to being cascaded. Oscillation is a possibility.

David

[Argus25]

I think these devices are merely a darlington configuration. I cannot see how these could have a low and predictable output impedance across such a broad frequency range as it would become very dependent on the properties of the choke. So I guess they rely on the output being capacitively coupled into a 50R load resistance/impedance to flatten it. So to cascade them, you would need a 50R load resistor between the two coupling caps to ground. But I'm not sure about this.

I have successfully used larger metal cased "cascadable amplifiers" that do live up to their specs with 50R input and output impedances, they are the Expensive UTO series made by Avantek, Avnet,Teledyne etc.

Motorola made some great modules but the circuitry inside them is not well documented, but its very impressive and perfectly symmetrical in its design, in that the current sinking and current sourcing of the output stage is identical and the input impedance is identical on both halves of the input wave. I'll look for and example. It shows how elaborate they have to be to get a good result.

[Argus25]

I have attached a circuit of a 1 to 250 MHz, 50R in -out Motorola module with 36dB gain and a good power output too, but it shows the extraordinary lengths they went to and the beautiful circuit symmetry to get a good result. You won't find this circuit easily, and you can see the feedback components that helped to flatten the response. It is hard to imagine a pair of darlington configuration devices beating this, but I guess its possible.

[PETERg0rsq]

MAR devices are supposed to be cascadeable, and stable.

What frequency do you want to pass through it?

It will need to be well designed and built to make them work in cascade.

Any feedback between the two stages will cause problems.

How are you constructing it. Will need fabricated pcb, and preferably SMD parts. I did make one cascaded amplifier using "island" construction and SMD parts a few years ago.

The bias line is the most likely route of feedback, so make sure it is well decoupled. Add more rfc's and decoupling capacitors. Pick capacitors that will give a low impedance path to earth at operating frequencies. Screen the two stages if necessary.

I would only use one blocking capacitor between the two stages, tap off the first output via its own capacitor to the sma.

Also the blocking capacitors seem a bit big. Spec sheet suggests 1nF=100MHz, 100pF = 400MHz, 10pF = 1.2GHz

The supply decoupling capacitor is also a bit big, try several smaller capacitors in parallel (e.g. 1u, 100n and a 1n).

Another approach would be to use a MAR6 for the low gain, and an MAR8 (30+dB gain) for the high gain output. This would work better if you need a broad band amplifier.

There is a seller on ebay selling development boards for cascaded MMIC. I enclose some of his drawings to show what is involved to make them work well! Done like this they should work well in to the low GHz

[Jon_G4MDC]

One of the MAR range was only conditionally stable.
I don't remember which one but I think it might have been MAR6.

[G0HZU_JMR]

I think these devices are merely a darlington configuration. I cannot see how these could have a low and predictable output impedance across such a broad frequency range as it would become very dependent on the properties of the choke.

Because of the internal feedback, it should look like a flat (wideband) 50 ohm resistance when looking back into the device (from an impedance analyser) as long as the input is terminated in a wideband 50R load.

The same applies at the input. If the output is terminated in an external 50R load, the input should look like a (wideband) 50 ohm resistance because of the internal feedback. As long as sensible values are chosen for the bias choke and bias resistor, it should achieve this over a large bandwidth. It's common practice to put the resistor and choke in series and then decouple (at RF) at the cold end of the resistor.

I think that the technology in these devices dates back to the 1980s when they were initially designed/produced by Avantek. I think HP bought Avantek about 25 years ago and the Avantek logo was changed.
Back in those days, the Avantek parts destined for MiniCircuits were shipped in identical packaging but with the logo and text (on the paper backing) changed to Minicircuits. I had a quick look in my stash of ancient MMICs and I've still got a few old (unused) MMIC amps like this in the original Avantek paper+ clear plastic tape packaging from 30 years ago. I've got some ancient MiniCircuits parts in the same packaging.

[G0HZU_JMR]

Originally Posted by Jon_G4MDC

One of the MAR range was only conditionally stable.
I don't remember which one but I think it might have been MAR6.

It's a long time ago but the 'dodgy' old school MMICs that I recall were the MSA0885 (MAR8?) and the MSA1105. These had to be used with caution because they weren't unconditionally stable at all frequencies. The MSA1105 had stability issues at low frequencies and the MSA0885 (and MAR8?) were only conditionally stable over quite a wide range of frequencies.

The MAR6 should be OK but it requires good PCB layout with good grounding of the device in order to keep it stable. Also, the supply decoupling has to be fairly good in order to prevent a feedback path from one device to the other via the supply rail. You also have to be wary of how feeble it is in terms of P1dB. Having two of them in series means over 40dB gain at low frequencies. So you would have to be careful not to overdrive them into compression.

[Wendymott]

Hi Jeremy Peter and Jon. All your comments have been noted...... I think I was being over ambitious to try getting two outputs at different levels... I should use a two stage and attenuate the input level to prevent overloading of Stage 2. I will re-lay the pcb, with ground top and a full ground plane of copper foil on the underside.
The original was just a hand engraved pcb and was a bit rough.. I will add inter stage screening and re evaluate the decoupling.
The original was just to amplify the output from the 2022, to enable me to provide larger signals than the 2022 could provide.
Thanks for your comments.

[G0HZU_JMR]

I think the output compression point of the MAR6 is only a few mW, maybe +4dBm and so it is going to compress below the raw output capability of the Mi 2022. So you would need to choose some other amplifier device if you want wide gain and a high P1dB in order to boost the Mi 2022.

The MAR6 is really just a small signal device optimised for input and output match over a huge bandwidth. It achieves a low noise figure and it is typically biased to run at just 16mA. So it is going to be quite weedy in terms of P1dB and IP3.

[G0HZU_JMR]

Note that it's still worth making the dual MAR6 (40dB gain?) amplifier because as long as you follow the classic design rules for the PCB layout it will make a handy small signal amplifier to have in the workshop. A typical use would be as a precision preamp for your Rigol analyser if you wanted to do any noise figure testing using the Y factor method. It might be worth putting something like a 3dB or 6dB pad in between the two amplifiers to help with the input match though. It would also have to be very well screened in a metal enclosure (or you could make an enclosure using thin sheet/screening metal and a big soldering iron to seam it all together)

It would be great for this because of its low noise figure and also because of the excellent input match when run in a 50R test setup. The preamp needs to have a good input match to minimise uncertainty in the test setup. But you would also need to make or buy a suitable noise source that had a good hot/cold source match. I made my own noise source many years ago and I use a couple of cascaded MMICs as a spectrum analyser preamp. I get really good/consistent results when measuring noise figure from LF through to >1GHz with this setup.

[G6Tanuki]

When I've done stuff with these MMICs I've generally found that gain's worryingly easy and cheap to come by but stability when handling 'unknown' sources and sinks isn't always so much fun.

3dB pads between stages help. As does a 3dB pad on the output - so the MMIC never has to 'see' a nastily-reactive load at LF or UHF.

Just make sure the resistors you use in your pad are truly non-inductive: cheap carbon-film types are often made using a spiral track machined around the 'barrel' of the resistor. Several turns of track on a 3mm-diameter former - that's a VHF/UHF inductor!

[Wendymott]

Hi Jeremy and G6T.... Thanks for the additional Info. The schematic from Peter shows an interstage pad.....and I note the comments about overload Jeremy. As there is nothing on TV.. I will lay out a pcb tonight.
I have a nice range of SMD 0603 resistors now so I can make 3dB pads or any other that comes to mind. Watch this space.

[Biggles]

On a few of my experiments into the magical world of MMICs I have found that I had to reduce the supply voltage by quite a bit to reduce instability. I think this probably just reduced the gain but watching the output noise dramatically fall to an acceptable level on a spectrum analyser was proof enough for me. My PCB design was probably at fault but I got the impression that although the gain figures/simplicity certainly look attractive on paper, you have to be very careful with layout, particularly when cascading. 40dB is quite a bit of gain...but you already knew that.

[Argus25]

Originally Posted by G0HZU_JMR

If the output is terminated in an external 50R load, the input should look like a (wideband) 50 ohm resistance because of the internal feedback. As long as sensible values are chosen for the bias choke and bias resistor, it should achieve this over a large bandwidth.

So that means it does need the 50R termination resistance I suggested between the two cascaded devices ? (which was not on the posted diagram) or will terminating just the output device be satisfactory because that should establish a 50R load at its input to correctly load the input MAR device ?

[G0HZU_JMR]

No, I don't think it should have the 50R between the devices. I think this 50R shunt resistor would cause the output impedance to go up because of the damping impact the shunt resistor has on the feedback. So looking back into the MMIC collector it would then look more like 80 ohms typical rather than the desired 50R. It would also mess up the input impedance to the first stage. I'd expect this to go up too. This all assumes the input and output terminations of the 'test gear' are a wideband 50R of course.

or will terminating just the output device be satisfactory because that should establish a 50R load at its input to correctly load the input MAR device ?

Yes, I think this is correct

[G0HZU_JMR]

Originally Posted by Biggles

On a few of my experiments into the magical world of MMICs I have found that I had to reduce the supply voltage by quite a bit to reduce instability. I think this probably just reduced the gain but watching the output noise dramatically fall to an acceptable level on a spectrum analyser was proof enough for me. My PCB design was probably at fault but I got the impression that although the gain figures/simplicity certainly look attractive on paper, you have to be very careful with layout, particularly when cascading. 40dB is quite a bit of gain...but you already knew that.

Even with just a single stage MMIC that is unconditionally stable up into the UHF region, it's generally a good idea to avoid having an equivalent of a short circuit at the input and/or output at the same time 'in the many GHz region'. Otherwise it might be possible for devices like this to oscillate. But usually, there would be sufficient ESR in a class 2 ceramic blocking cap (small SMD 10nF?) to negate any negative resistance at the input of the device up at these frequencies. But all this still relies on having lots of ground vias dotted around the device.

...start to compromise on the grounding and maybe (somehow) allow the equivalent of a decent short up at several GHz to appear at the terminals of the device and it could oscillate up at maybe 6GHz or so.

[Jon_G4MDC]

So it was. MAR8.
Thanks for the reminder.

[Wendymott]

Hi peeps... On the comments so far.. I have made a pcb based on the attached schematic..the pcb is shown as single sided, but will have a full copper shield on the top side............ lets see where it goes.... I am not interested in use above 200 Mhz.....