Continue to Site

Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

  • Congratulations waross on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!

1.8-2.3GHz 2 stage amp pcb design

Status
Not open for further replies.

2GHz

Electrical
Jul 21, 2008
5
GB
Hi there,
I have been assembling 2 stage amps using minicircuits VCOs (ROS 1700 and 2500) rf output to stage 1 sirenza SXA389BZ or SXB4089Z on .031"FR4 via 2oz microstrip, and stage 2 in the same way to SPB 2026Zs.
I've been following the sirenza datasheets very closely to get tracklengths correct, but get HUGE losses between stage 1 and 2... Been using exact components for impedance matching, and talking to wes paxman at sirenza (rfmw) to get things right. NO LUCK - 2nd set of pcbs wasted...
Instead of observing track lengths, could I try just making everything as close together as possible in an attempt to minimise losses, despite the obvious increases in destructive interference due to non proven striplengths?
I can't afford a network analyser (startup company trying to get a part working prototype) but I've got an ok spectrum analyser. I also have ok single stage amps working...
 
Replies continue below

Recommended for you

there's not much loss in a short length of FR4 microstrip.
FR4 changes alot with temperature by the way.


Huge losses never occur in normal circuit board layouts. You'd get more help with a photo of your circuit and links to the components you're using.

good luck,

kch
 
here's a picture of the board in varying states of assembly, the relevant point is where you see the green wire attached to an SMA connector - which is my crude attempt to "sample" the rf power at the junction of the first and second stages of the amp. you can see the ros 2500 vco

leading to the sirenza sxa389bz
which leads into the spb2026z

does that help?
 
here's a closeup - again - the green wire is my attempt to locate the missing rf! you can see it's connected to the input side of the 2nd amp. I know rf doesn't like going round corners, but I have got signals this way in the past - my board losses are so large (30dB or more) that this technique may help. I have the same identical problem with 2 attempts at this(pcbs and components duplicated) - so I don't think it's component failure...
wes paxman in the past has suggested that around 2GHz it is very easy to get big losses on microstrip to coax connections, so maybe my pcb layout is to blame - though as you'll see from the datasheet links - I'm very close to the sirenza spec...
 
 http://files.engineering.com/getfile.aspx?folder=b00f81e3-219b-4ca0-ad77-89a2f12a8b38&file=IMG_0116.jpg
The soldering is atrocious. Those are not striplines, they are lumps of solder.

And it looks there might be a short circuit at the top right of the righthand black chip.


 
thanks for the compliment!

well my soldering technique is to reflow the non sensitive components - you will see they are not bad at all - no excess solder. The short circuit you point out is intentional - if you see the datasheet I hope you'll agree!

I spoke to the engineers at sirenza - the chip manufacturers - and they said you could use extra solder on the ground pins in order to get the very best heatsinking via thermal vias to the backplate - so that is intentional.

I agree the sma connector connected to the green wire is a mess, but it is not part of the unit, and was all I had to hand...
 
Is this a multilayer?

I designed a 802.15.4 on double-sided .031 FR4 that worked very well. Afterwards, I did a multilayer version that was amplified on both the TX and RX - most of the layout was similar to the 2-side version. Yet, the amplified multi-layer version, which should have had 20 dB additional link gain proved to have 15 dB LESS!. After beating around with the design on the bench for a few days, and not finding what was wrong, I cut one of the boards in two. The PCB house had made the inner layers with the wrong thickness making the 50 Ohm traces about 15 Ohms! Even though my impedance traces were about 3cm in total length, the impedance mismatch was more than enough kill any gain in the design.

Solder lumps at this frequency for the layout you have shouldn't hurt you more than a couple dB. Make sure your impedance traces are correct.
 
thanks for the input - i think you may have a point, because my losses are about 30-40dB, which must be lost between stage 1 and stage 2 - only 0.75" - but with 2 caps in series with the transmission line, and a couple of resistors/inductors/caps leading off sideways to ground - which are nevertheless as per the sirenza specs.

I know they are for impedance matching, but could the sideways resistors/caps/inductors be siphoning rf power away as they would in a dc circuit?

My focus has been on playing with the passives, but maybe it's the pcb and stripline.

My thinking is that maybe I should bring the 2 amps as close together as possible, leaving enough room for only one dc blocking cap in series, and perhaps a couple of impedance matching elements sideways.

any thoughts?
 
Well, yes that construction is pretty dismal! But that is how engineers learn, so do not take it to heart. General comments:
1) the "green wire" goes to where? It is not that clear from the data sheet, but I suspect that the input to the SPB chip is not DC blocked (i.e. there is a dc bias voltage at the input, and if you load it down with something like a 50 ohm power meter, the 2nd stage gain will vanish!)

Even if this wasn't the problem, you would have to remove the connection to the 2nd stage before you could get anywhere close to a meaningful power measurement out of the "green wire".

2) What is with all the giant size components? There should be nothing bigger than an 0603 chip anywhere near the RF path. You are probably having component frequency resonances.

3) What is with the giant topside ground plane? Do you use a blowtorch to solder down those tiny chips to it? You have made a non-solderable design! Anywhere a tiny chip has to go back to ground, use a small (maybe 30x30 mil pad) with one or two via holes to the ground plane on the back.

You might have blown out the parts with excess heat.

4) BTW, is there a ground plane on the back?

5) The output sma connector looks to be a low frequency type, with a horrendous attachment to the ground plane. Need to respin and use a better connector.

5) I suspect that the bias lines are sucking out a lot of power. Try different types/values of inductors and see if your power gets better.

6) these components are static sensitive. Have you blown them out by a) not using a static strap at every stage of handling/testing, b) soldered them down with an untested soldering iron with 50V AC at the tip, c) sat on them by accident?
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Top