Are the datasheets of Hittite believable, Such as that HMC547LC3 ?

[saulbit]

Recently I have been using the hmc547LC3 from Hittite to make a spdt circuit. The PCB is RO4350, 0.254mm, almost the same as the eval board in the datasheet. However, when I test the circuit, the conversion loss is about 10dB when frequency is about 20GHz and the isolation is about 10dB. What is wrong with the circuit. The soldering is a big question and may be the problem. However, have someone used the qualities in the datasheet 100%? The PLL can't and the attenuators can't, either. Any advice is welcome. Moreover, Hittite is an excellent company, I just can't make the puzzles out.

- - - Updated - - -

Here is the datasheet of HMC547LC3.

 

[jiripolivka]

I have used a number of MMICs from Hittite and all performed as specified or better. I would recommend to buy the evaluation board if you have not much experience with using MMICs at mm waves. You can learn from the manufacturer on how to properly install the MMIC and what components are used around it. A bad SMT capacitor can spoil it all.

 

[kspalla]

this could be either improper mounting and/or not having sufficient GND vias under the component.

 

[biff44]

Something like a switch....I would bet that the data sheet is pretty close to the true performance. But realize that the performance of something like this is highly dependent on the pcb layout. If you have too few vias, poor underchip solder attach, lines close together, data lines coupling your rf input, etc etc, you will get poorer performance--especially switch isolation! Analyze a FET switch for isolation, and then add 0.5 nH of ground via inductance and analyze again, and see how many dB of isolation you will lose.

On the other hand, things like amplifier chips in packages....I always assume I will get a few dB less gain or output power than a data sheet says (independent of which manufacturer you choose to use!)

Show us a picture of your board layout. Maybe we will see something.
Rich

 

[ashiir]

I think the datasheet is reliable even if it only shows the best performance (which means that it maybe hard to get that performance; only proper condition can get that!)
There are so many things to make its performance more worse than that in the datasheet such as PCB layout, how well solder attach and so on.
What I always do is not compare it numerically but I just check the output that it agrees with the datasheet or not. If it agrees with the datasheet, it will be ok.

 

[tony_lth]

Hittite is a reliable company you can count on.
If your performance is not good, all the foresaid factors need to be considered.

 

[Element_115]

Did you get the Eval board ? If you did and it works, then you could replace the part
from the Eval board with the part on your board. You will see if the poor performance
follows the part or the PCB.

Chances are that you have a bad layout or the RF Connector is not soldered well.

Good luck.

 

[FvM]

However, when I test the circuit, the conversion loss is about 10dB when frequency is about 20GHz and the isolation is about 10dB. What is wrong with the circuit.

Reviewing your previous posts, unexpected results of microwave designs seem to be a constant somehow...

 

[RF_Jim]

Recently I have been using the hmc547LC3 from Hittite to make a spdt circuit. The PCB is RO4350, 0.254mm, almost the same as the eval board in the datasheet. However, when I test the circuit, the conversion loss is about 10dB when frequency is about 20GHz and the isolation is about 10dB. What is wrong with the circuit. The soldering is a big question and may be the problem. However, have someone used the qualities in the datasheet 100%? The PLL can't and the attenuators can't, either. Any advice is welcome. Moreover, Hittite is an excellent company, I just can't make the puzzles out.

....

It is important you read the data sheet carefully and note every detail and specification. Often things are spec'd because processing artifacts remain a little 'active' (provide leakage paths) on the substrate of the IC.

Such things as E-beam bleed-off 'paths' (resistances) are sometimes still present on the wafer (even though a 'cut' with an etching process in the area of the E-beam bleed-off path has been performed) and the IC manufacturer will allude to this with a 'leakage' current value specified somewhere ... it is very important that design include a DC choke or perhaps a high enough value resistance to provide a 'path' for this leakage current yet not affect RF.

Not only have I seen the E-beam bleed-off 'paths' detailed on the 'level' plots (while involved with IC design reviews at TI's GaAs facility before purchase by TriQuint) but later on experienced a problem in productin owing to a customer design that did not allow a path for the leakage current and the *zero* value of attenuation was always anomalous (the leakage current caused a bias on gates in the IC and was inserting anomalous amounts of attenuation).

Also, the ICs have been through at least one stage of test, if not two today: (1) On-wafer DC tests (usually before backside grind and via-etching) and (2) on-wafer RF tests. We used Keithley instrumentation and KLA probers (probe stations) to perform the DC tests at the time at TI and were just ramping up to do production-scale on-wafer RF-probe testing.

Jim