Looking for: Divide by Two Chip 3GHz+

[Yansi]

Hello!

just looking for a part: prescaler divide by two that would work at 3.26GHz. Matched 50ohm input and output, preferably.

The only thing that seems to be currently available is HMC432. Which fits my needs exactly, however I do not like the price (~10$ a pop)

If you know about any similar ICs, please let me know. Obsolete types may be sufficient, if can still be obtained in small quantities (~10pcs). Ebay/Aliexpress is okay too.

The only other chip I could find was SiGe D602, but I found it nowhere to buy.
The closest one I could find available for lower price is uPB1508, but this is only specified up to 3.0GHz and I would have to "overclock" it a bit. Not sure if it would work reliably and what the sensitivity will be.

Thank you for any tips.

//Sidenote: Am just investigating, if it is easier/better to have a VCO but PLL running at half the frequency, or  having the VCO at the PLL range and then multiply the VCO frequency. I found a decent less expensive solution for the latter (HMC187), but am liking the divider concept much more.

[edavid]

How about MC10EP32 ?

https://www.findchips.com/search/mc10ep32

[Yansi]

I also stumbled upon this one, but that is an ECL flipflop. Not sure about its usability for RF.  From a quick sneak into the datasheet, it would need an enourmous RF drive level to operate (like 13dBm+ at 3.4Ghz and that is not fun). An the input impedance is a mess too. 

Y.

[ch_scr]

I have played a bit with a UPB1507, it would work up to 4.4ghz with the output of my China NWT4000.

[Gerhard_dk4xp]

I also stumbled upon this one, but that is an ECL flipflop. Not sure about its usability for RF.  From a quick sneak into the datasheet, it would need an enourmous RF drive level to operate (like 13dBm+ at 3.4Ghz and that is not fun). An the input impedance is a mess too. 
Y.

No. You need a few 100 mV into high impedance.

[Yansi]

From the 10EP32 datasheet:

PECL 3.3V supply at 25°C:  Vih max = 2480mV, Vil min = 1430mV.  That is a 1050mVpp swing (1050mVpp at 50ohms equals 4+dBm for sure).  Not sure about the input impedance at 3+GHz though, but I would not consider it "high". Even a single puff of capacitance is a substantial load at 3+GHz (1pF at 3GHz is 53ohm. Couldn't find any mention of input capacitance of the gate either)

I have never worked with ECL logic, so not sure really about anything here. Unknown territory for me.

[Gerhard_dk4xp]

That is the MAXIMUM, not the minimum that you may apply.
The inputs are differential amplifiers and when you drive them
too strong, you drive the tail current source out of its
comfort range. That current source is already challenged
by the 3V3 total supply.

If 1V was really required, the chip could not drive its own
inputs. It is save to assume that the specs for the outputs
are the optimum levels, and they offer some room for
attenuation and crosstalk / coupled noise from somewhere else.

Use the 100K family, not the 10K.  100K has temp compensation,
there are even pathological cases where 10k cannot drive
another 10K chip if it is at a very different temperature.

You must drive the input capacitance anyway. Having 50 Ohms
in parallel does not necessarily make that easier, although it
will make transmission lines behave better. The chip itself
does not care. It wants its bias and the AC voltage riding
on it. The effective load is a few pF and 75KOhms IIRC.

regards, Gerhard

[Yansi]

Yes, that is the maximum and that is what you NEED to design for, otherwise it may not work with some parts a little bit off the typical spec and may not work at all at elevated temperatures (needs even more drive level).

Impedance matching on 3+GHz is a thing, if one needs predictable behavior. Without the proper measurement equipment (well, I have some or access to, but only up to 3GHz, with this I am even out of range of the available measurement instruments). Hence why I want to avoid using devices that are not matched 50ohms or those without input S-parameter characteristics. (I can kind of work with S parameters, at least know the basics how to work the matching out).

It seems there is no other option than the HMC432 on the market.  Not happy about that, not at all. This single part will contribute to the final cost significantly. For the price of the HMC432 I could probably buy complete a fractional PLL frequecny synthesizer chip with integrated VCO for this frequency. 

[Gerhard_dk4xp]

No. This is the maximum and you need to design for LESS than that.
It won't work properly with 1.5Vpp.

If you want a 50 Ohm input, then place a 0603 49R9.
You could just as well terminate in 100 Ohm or 150 if your source
can drive that properly. You burn the signal power in the resistor,
the chip just sees the voltage across it; it takes almost no current,
just that what's needed to charge the 2pF, or what the input capacitance
might be. Plus that tiny bit for the 75K bias resistor.

Even the old Motorola ECLIPSE series of ECL was perfectly happy
with 300 mV, and it worked to 4.4 GHz, where my signal generator
maxed out.

The standard ECL output hi/lo difference is 800 mV, and the inputs
are sensitive enough to allow for temperature, noise margin and...
The 3V3 ECL needs even less.

There are lots of ECL and CML parts on the market, and Hittite
was not even a real player in the open market. They used to ask
for a NDA for every tiny data sheet. No wonder they got bought by AD.
The best thing that could happen IMHO.

< https://www.onsemi.com/pub/Collateral/AND8066-D.PDF >
< https://www.onsemi.com/pub/Collateral/AND8020-D.PDF >
< https://www.onsemi.com/pub/Collateral/NB7V52M-D.PDF >
Micrel/Microchip also have a lot.

regards, Gerhard

[Yansi]

50R parallel with 2pF is nowhere near 50ohm impedance at 3 GHz. Just no.

I do not understand why should I design for a "typical" values and ignore the maximum ones. By ignoring the maximum values, I am in fact designing in a problem, that the circuit may fail with several devices. May not be a problem for one off quantity though - but still don't like it.

Due to the 10EP32 devices recently silently disappearing from Aliexpress and while accounting for their full price, I am ending this thread that no divide by two chip now exists other than HMC432.  Sad, but seems true. The full price of the ECL divider is comparable to the Hittite device, but the ECL is not worth the hassle.

Thanks everyone for suggestions.
Y.

[ThomasDK]

The only thing that seems to be currently available is HMC432. Which fits my needs exactly, however I do not like the price (~10$ a pop)

You can buy them directly from ADI for $5.33 in qty 1?

[Yansi]

Sure and then add a $50 or more shipping. No thanks. Will buy from Mouser for $9 + VAT.

[SaabFAN]

ECL-Chips have a curve in the datasheet, showing the area they are comfortable working in. That curve shows the required power in dBm vs. frequency.
That's what you want to take a look at.
The MC12093 I used on my Prescaler-Board get about 0dBm at the lowest input power and run flawlessly (after I fixed the ground-plane issue ).

[technogeeky]

This may be too rudimentary of a suggestion for the original poster, but The Signal Path did a video a while ago where he walked through the input requirements for a divider. Here is the link at pretty much exactly the point he starts discussing the divider's datasheet.

[donmr]

I do not understand why should I design for a "typical" values and ignore the maximum ones.

You don't ignore the maximum input level spec, you design to always exceed the _minimum_  and never exceed the _maximum_ .  The part will work between those two levels.