Is it normal that prescaler ic HMC434 and pcb under ic is a little hot?

[Terminator3]

I made prescaler board based on HMC434 and other prescaler for divided frequency. At the moment i can't check it's functionality (no input and measurement device available), only power-on test. HMC434 IC and board under IC is a little hot, if somebody have experience, please tell me is that ok?

 

[biff44]

most Hittite parts run very hot. They work at the rated speed, but **** the current down.

are u overvoltaging it? 3 volts

put some rtv on the bottom, maybe get a little heat to the PCB.

They make dividers with a big fat ground slug on the bottom, switch to one of those

 

[Terminator3]

i use 3.3v regulator, is it not ok for this part? from datasheet it is within the range i doubt

 

[biff44]

yes as long as you are under +3.5 volts it should be ok. Are you using the recommended by pass capacitors (1000 pf AND 10 uF) on the supply CLOSE to the chip pin 5? It might be oscillating. a 10 uF cap is an unusual recommendation (normally would me 0.1 uF), so there might be some low frequency stability problems with the chip.

 

[Terminator3]

Here is my PCB:

Also small part "EVAL:" is a picture of evaluation board layout in datasheet, where capacitors not so close. And bigger capacity is closer to the chip..
If i put capacitors very close to chip, as shown in white lines, can it be the problem?

If chip oscillated at low frequency for some time, can it burn out? I turned my board less than a minute, checked 3.3 and 5v. Those HMC434 not the cheapest one, i hope it still works.

 

[Darktrax]

One tip when starting up a circuit that has the potential to self-destruct by low frequency oscillation when you power it up for the first time is to limit the storage energy available to it from the power source. A few tens of ohms, and rely on limited capacitor decoupling in the board. If the voltage catastrophically collapses, you have a short-circuit, circuit error, it's mounted backwards, or maybe the chip did something bad. If nothing bad happens, remove some resistor.

I also like the very fast self-resettable fuse components.
Available from <50mA up to several Amps and many cost less than 50cents.
Almost all suppliers have them (Digikey, Mouser, Farnell, RS, and many others). See one example:
https://uk.mouser.com/Search/Refine.aspx?Keyword=resettable+fuse
If the expensive chip has a problem, some of these can operate in milliseconds, and return to normal after a few seconds if the problem is removed.

Make sure your test connecting leads and unshielded test conditions are not the unwanted oscillation mechanism. A common-mode toroid choke around all the leads where they arrive near the board can save you $$. Keep in mind that the sheer speed pre-scalers work at is likely to make them run hot. The dissipation is proportional to voltage^2, so even a slight drop in voltage will let in run disproportionately cooler.

Don't let it run for more than a second or so until you know what is happening. Put a current meter in the power line. The value had better be near what you expect. Couple some triggered instrumentation to a safe point. You want to gain information about what is going on when it tries to fry itself!

 

[Terminator3]

I bought cheapest available 2GHz frequency counter. Good news is that HMC434 output is 300MHz as expected for 2400MHz input. So i hope it's temperature is normal.

 

[Darktrax]

The (Hittite) datasheet for this device shows output power around -2dBm over 1GHz thru 8GHz for three temperatures..
+25 deg C - It will not stay at this temperature with some cooling arrangement.
- 40 deg C - some places on the planet get this cold.
+75 deg C - that is hot enough to scald and injure! About 50 to 55C would make you want to let go!

Nice to know the little chip can take this, but it is small enough to get hot quickly. 186mW (3V@62mA) on a SOT26 package is not good! I know many designers will let it do that, just because it can, but every function-optimizing fibre of my being tells me that mounting it with a narrow aluminium tab or "finger" pressing down on it, or a maybe small copper tab top shield stuck on with epoxy, or any other imaginative trick you can contrive to move that heat out will improve the circuit conditions enormously.

 

[biff44]

Here is my PCB:
View attachment 94747
Also small part "EVAL:" is a picture of evaluation board layout in datasheet, where capacitors not so close. And bigger capacity is closer to the chip..
If i put capacitors very close to chip, as shown in white lines, can it be the problem?

If chip oscillated at low frequency for some time, can it burn out? I turned my board less than a minute, checked 3.3 and 5v. Those HMC434 not the cheapest one, i hope it still works.

I Do have to ask: is there a continuous ground plane on the back side???

 

[Darktrax]

OK - here is where you can go your own way if you have some good rules in the mind for layout.
There is much information searchable regarding good RF layout. You absolutely have to have a ground plane, and probably on the topside too, except where you have to keep clear of circuit lines. Co-planar waveguide results in thinner lines, but gives you less problem with topside coupling.

At 2GHz thru 8GHz, any structure, even a SOT26 lead, will radiate. The NC and the 0V ground pin are next to each other, so can be used to conduct some heat into the top ground plane, but take care - you might need thermal pad layouts to make it solderable.

Decoupling capacitors should be SMT as close as possible to the power pin, the other side landing on topside ground plane, with generous vias right next to it going through to underside ground. The inductance of vias becomes important. Try and put copper cut from copper tape around the outside to join the groundplanes.

Keep the input lines away from the output, and try to make them transmission line. You can damp unwanted urge to oscillate with low value (5-10ohms) SMD resistors in series, at the expense of a little signal level. Give thought to what it might take to get an approximate match to the outputs. It needs to at least covered with a metal plate shield, or a RF shield salvaged from a dead network or PCMCIA card. Using 1 or 2pF series capacitor as DC blocking also gives reduced ability to oscillate at audio frequencies.

Anything with gain can oscillate, and if it gets going at audio frequencies, it will be banging the rails with all it has. It could well get hot! Use a scope with a approximate 20:1 input contrived of a 1K resistor soldered onto a 50 Ohm coax cable, going to a scope input terminated in 50 Ohms. It gives you a very wideband probe capable of sampling GHz. You won't see the RF, but you will see any imposed audio on top. You need to be sure the signal is clean. Looking at the picture of the evaluation board, I would say it does not really look like a Ghz layout, certainly not the 8Ghz that the prescalar can operate at.
**broken link removed**
http://www.sigcon.com/Pubs/straight/probes.htm

 

[Terminator3]

pcb have ground plane on the back side, connected to top side at four points by thin wires in the corners of pcb. There are no any ground pads on the top side under the chip, i think eval board have no this pad too, because power line goes under the chip.

 

[Darktrax]

The eval board is poor given what it is supposed to do.
Such a layout is not even good for the low frequency end of it's operation.
I am guessing that approx 2Ghz is the highest frequency you are dividing down.
If it (sort of) works, and you want to use it as it is, then at least drill through more places as you can find, and add links.
If it is easy to put a SMD 0805 or 0603 CAP 10nF ~ 100nF at the topside power lead to any ground linked down, that would be good.
Take a copper tab, stuck onto the top of SOT26 down to ground near the power pin - stay away from signal conductors.
Allow the rest of the tab beyond the grounded point to be a heat dissipation tab and shield, landing its other end at the board edge.
This will look ugly, but it will work better. Do not allow the cooling tab to become an antenna!

 

[biff44]

pcb have ground plane on the back side, connected to top side at four points by thin wires in the corners of pcb. There are no any ground pads on the top side under the chip, i think eval board have no this pad too, because power line goes under the chip.

drill holes right next to pins 1, 2, and 4. Solder a short wire from each of those pins to the backside.

You should only be seeing a 14 deg c rise in package temp (70 dec c/w * 3.3V *0.062A). So if it is more, u are simply not getting the heat out of those pins to the backside