I need to use a divider-by-3 device in my product.
It requires to output a signal with 50% D.C.
Input is a 25-100MHz SE signal.
Output is a divided-by-3 differential signal.
Can you confirm the following:
The /3 of the HMC794 does produce a 50% output. However, it looks like your 25M-100M input frequency is below the HMC794's min input frequency
R Counter outputs are typically not 50%.
You can also look at the LTC6954. This device has a /3, excellent ESD protection and the web price is lower than the HMC794 (as of 12/10/19).
Thank you for your reply.
The input to the divider-by-3 is a square wave, thus the HMC794 should be able to handle it up to very low frequencies(hmc794 datasheet, p. 8: "For input frequency lower than 200 MHz, square wave input signal is recommended").
The LTC6954 is bigger in size. It will be very hard for me to replace the current 4.9 x 3mm with a 7 x 4mm device.Is there a version of it with lower count of dividers?
In my design a division of 96 is required for my input signal (800MHz-2.7GHz sine wave).So, I am using a by-8-by-4-by-3 chain of dividers (in that order).How is the ESD protection of the HMC434? Should it be enhanced with an external protection? A better divider-by-8 device?
Thank you !
The square wave (higher slew rate than a sine wave) should be ok for the HMC794 at those frequencies.
The HMC434 ESD can be assumed to be Class 0. From my experience most ESD events happen during assembly before the part is installed on the board. You can add external protection on your board, but not sure this solves the larger issue (i.e. assembly line ESD rating...).
The LTC6953 maybe another option? It supports up to 4.5GHz and can /96. It is a larger package, but maybe it can replace your other 3 dividers and be an overall smaller solution. If you need access to the /8, or /4 outputs also you can synchronize other outputs with these divider settings. Not trying to change your design, just letting you know about other possible options.
Eventually, I will need only a divider-by-8.
My design has an input SMA, which connects to the by-8 device, with its output connecting to internal board circuitry.The HMC434 is a HBM Class 0 ESD device.
I am concerned about a user connecting a RF cable and discharging into the device, causing an ESD event.
Your concern about ESD events during assembly is also noted. Thank you for raising this.
What is the best way to handle this?
Use some ESD protection devices? That might affect the SI of my signal (2.7GHz max).
Another, more robust, divider-by-8 device?
For assembly you will need to work with your assembly site to make sure they are setup for class 0 ESD. I'm honestly, not sure what all this entails (other than ionizers...), but I know some customers have done this. Once installed, I would think we could come up with some sort of way to have some on board protection, that shouldn't degrade performance. I'll ask around.
We also have the HMC988 and HMC862A which have better ESD and /8 capability. Simpler devices than the LTC6953 above. Both are differential in/differential out, but work just fine single ended in/single ended out if the unused pins are terminated properly. Something like this would be my preferred route, if the price wasn't a variable...
The HMC862A is more expensive than the HMC434E ($17 versus $5).
However, it has significantly higher ESD protection.
As discussed above, I would prefer the cheaper HMC434E with external, on-board, ESD protection.I am waiting for inputs from you about suggested ways to achieve that.
Thank you Chris !
For 2.7GHz input, something like a 18nH-22nH Toko inductor (shunt to GND) would work. The goal is to have a low resistance for the lower frequencies (for ESD) and higher resistance to GND for the 2.7G signal. You also want a decent amount of current carrying capability for ESD event.
Have a great holiday break! If you have anymore questions try to get them in before tomorrow...
I was not accurate. The frequency input is in the range of 800MHz-2.7GHz.The maximum input frequency is 2.7GHz (I specified the maximum, so when using any kind of TVS or ESD diode, the capacitance will need to be taken into consideration).
The plot thickens. Probably stating the obvious, but the wider & lower bandwidths make this a tougher solution. I can't say I have a previously used solution for this BW. Maybe buffering the input to the HMC434 is another route, but this adds costs also.