Matched RF amplifier sections...to modularize or not to modulize 2

[wd4dui]

Hey all you smarter than me RF people! I am assisting in the developement of broadband circuitry and keep insisting on building discrete RF amplifier/ attenuator modules that can be attached to a main circuit board. We are having some problems with our current configuration consisting of the amp/attenuators built onto the main PCB. Because of minute differences of the components, caps, resistors, amps, solder amounts and other devices, it is very difficult to obtain a similar response across the octaves of desired bandwidth. There are four RF strips on each main board that are combined and compared. I keep harping on discrete rf sections so the responses of each can be swept and matched and then installed onto a main PCB. It is extremly important that the response of each of the four amplifier sections be as close as possible within 0.5dB. It does not matter how lumpy and bumpy the response is just as long as all four amp sections are the same. Just because an amplifier section is copy and pasted to the PCB in the design phase does not mean all four will be of the same response. Them Minicircuits amplifiers are good but not all exactly the same....the same part number of course.
Sooooo, does modularizing the RF sections make sense to anyone or am I barking up the wrong tree? I need the response to be better than 0.5dB difference across 3Ghz of bandwidth.

Thanks & 73,

Rob
WD4DUI

 

[nbucska]

it is not a trivial task to match the gains of two amp-s
within .5 db for this BW.

Can you get around it somehow ? ( e.g. valibration )

<nbucska@pcperipherals DOT com> subj: eng-tips
read FAQ240-1032

 

[Higgler]

My former company had to laser trim resistors on a Crystal Video Receiver to match gains over the 6-18 ghz range within 0.5 dB. Lots of measurements into a computer, then the software picked the closest sets of data. Just matching two units is alot tougher than building a dozen and finding a matched pair.

Don't forget to test over temp.

You may want to try adjusting phasing from one amp to another to see the effect. That can change the shape of the curves.

Good luck,
kch

 

[pstuckey]

You could use a voltage varible amp or voltage variable attenuator in each path. Measure each path over freq (and temp) and create a control table to match the differnet paths.

Digital control is also possible, many companies have digital attenuators with 0.5 dB step sizes (not quite good enough for your app), but Hittite just released a part with 0.25 dB step which may work for you. HMC539LP3 is a 3mm x 3 mm, dc-4 GHz, 5 bit, 0.25 step. I just ordered one to evaluate for an interference cancellation system i'm working on (hasn't come in yet). Also you don't mention phase match requirements, but be aware that there will be a phase difference for different attenuations.

Peter

 

[Comcokid]

I would go with pstuckey recommendation to use a voltage or programmable attenuator on each output. For a detector, use a log amp detector from Analog Devices (their AD83xx family). These have about 1db accuracy over a wide input range, but repeatability at a specific dBm level is much better. Linear Technology is also making similar detector devices.

 

[groundhog1]

I agree that .5dB variation over Octaves of bandwidth is a very hard task...

Whenever I see these type of extreme specs, my first inclination is to go back upstairs.

Go and revisit the system design. What caused this requirement? Is it based on accurate assumptions? Do you really need .5 dB? Maybe you really need +/-2dB.

That may be the easiest thing to do.
Just go and correct the original requirements.

groundhog1

 

[biff44]

Of course, you do realize that mini circuits amps are pretty much the bottom of the barrel for consitency. Don't know if they still do, but for the longest time they simply sold agilent reject lots.

It is not clear what your frequency range is. "3 ghz bandwidth" does not tell us much.

 

[wd4dui]

Maybe a bit moreinfo would be helpful. Freq range is 500 to 3000 Mc. I started looking at the the PCB with all the components removed and signal path jumped. The loss and flatness is not too bad. Had about 4Db from In to Out after normalizing my test cables. I am now installing one part at a time and measuring the loss and ripple, removing the part and installing another part in the path and measuring.....It's going to a L-O-N-G day!!! I am also looking at some different resistors as there are some in the RF path used as combiners and splitters.

Measured the return loss of the first amp (ERA-3) and was not very good, anywhere from -15 to about -5 Db, now know the mini circuits amps are not great. At least this version of the board does not oscillate....that's another story.

Thanks for all the advice!!!

73
Rob
WD4DUI

 

[biff44]

If it was me, I would go with something like a Hitite HMC311LP3

http://www.hittite.com/index.cfm?bo...=leftnav&subID=61,59,60,128,134,142&subHead=2

It has a nice grounded bottom package, so the package parasitics will be repeatable. Also, it has operation to 6 GHz, so it will be more repeatable up to 3 GHz. I would print on the bias structure, using only a chip resistor, to make the bias line as repeatable as possible. And I would be sure that all the amps I bought came from the same production lot number.

 

[wd4dui]

Hey Biff,

Thanks for the info on the Hitite amps. We have already built the boards using the ERA-3 amps. As I stated before, I think the PCB has problems as the resulting response of the RF path is ugly. I jumped an unpopulated board and swept the thing. UGLY!!! Either the board house did not build to our specs or our specs (trace widths, ground plane distance, wrong board material (FR4) and other determinants) are wrong. I have to figure out how to "flatten" the responmse of our present boards if that is even possible. This is reason for my orginal post. If all the stages were modularized and connected via coax and appropriate connectors, then we could sweep each stage seperately and make corrections to our layout. It would be less costly and certianly easier to determine where the impedance bumps are occuring.

I had a nice full head of hair before I started this; and now? Getting thinner by the hour.

Please....more advice.

73
Rob "flatter is better?"
WD4DUI