1. Home
  2. Electronics Design
  3. Clocks and baluns

Clocks and baluns

Question for the RF-knowledgeable: I intend to use AD9510 for clock distribution, as it has fery low additive phase noise. My problem is, according to the datasheet performance is degraded if input exceeds 2 V peak-to-peak, and the low jitter TCXO I am considering has the typical either 3.3 V or 5 V output. Someone suggested I use a 2:1 or 3:1 balun, but I'm concerned about bandwidth, since the clock is square wave output (because the AD9510 datasheet says it needs high slew rate on the inputs for optimal performance). I'm also not sure how to handle transmission line effects. With a PCB track impedance is easy to control, and I normally use series termination resistors at the source on high speed lines to get a decent waveform, but I'm not sure what inserting a balun will do, how to choose the balun, and where to locate it for best performance (at the clock or at the distribution IC's input). The TCXO I'm using has about -110 dB phase noise at 10 Hz offset from the fundamental of 24 MHz, and I'd like to not get more than 5 dB increase from the overall distribution circuit--transformer+AD9510. What's the best approach?

Reply to
Prune
Loading thread data ...

Why not just use a (symmetrical) resistive divider?

--
Regards, Joerg

http://www.analogconsultants.com/
Reply to
Joerg

Hi, thanks for the fast reply. I was concerned about input capacitance of the IC making a lowpass with added resistance and limiting slew rate, though it looks like it's only 2 pF... The clock is specified for HCMOS output, 15 pF and > concerned about bandwidth, since the clock is square wave output

Reply to
Prune

Please post underneath the previous text, makes it easier for others to follow (and respond).

It doesn't really matter but if you place it far away you must calculate the resulting input impedance and size the traces accordingly. IOW so they have the correct characteristic impedance. It is generally best to place RF stuff close together in order to minimize EMI.

--
Regards, Joerg

http://www.analogconsultants.com/
Reply to
Joerg

Good grief, a 60 page datasheet for a clock buffer!

What I'd do is

long run? osc--R1---------------------------+---C1-----CLK1 | | +--CLK1b | | R2 C2 | | | | | | gnd gnd

where the impedance of [long run] = R2. Tweak R1 for desired signal swing into the adi chip.

John

Reply to
John Larkin

John Larkin wrote in news: snipped-for-privacy@4ax.com:

Well, it's a few more things than a buffer; it has PLL and dividers and is programmable. I do however just need a very low additive jitter distribution. From the clock I need to drive four AD1955 ICs, several SN74LVC1G79 single flip-flop ICs, and an Ethernet transformer (for sending the clock to a DSP board). Initially, I was thinking of using a tree of SN74LVC1G34 single buffer ICs, which I can match by hand for propagation delay. I don't know if that would really give worse jitter than the AD9510 IC. Certainly avoids the hassle of programming and soldering a complex IC.

In regards to your schematic, what if it's not a long run? I can easily place the TCXO right besides the ditribution IC. I'm also not sure how values for C1 and C2 are computed, and if these should be the same size.

The AD9510 has four outputs configurable to CMOS, but two of those have delays with a non-zero minimum, so I can't send one CMOS output to each of the four AD1955. Is there a way to connect LVDS outputs to CMOS input without degrading performance? The minimum digital input for logic high on the AD1955 is 2 V.

Reply to
Prune

A short run is even better. Then it's just a small voltage divider. make R1+R2 as low as the tcxo can reasonably drive, 200 ohms total maybe.

Nearly any cap values will do. 10 nF, things like that maybe. The ADI thing furnishes its own DC bias, so really just wants some AC-coupled drive.

John

Reply to
John Larkin

John Larkin wrote in news: snipped-for-privacy@4ax.com:

The clock specifies HCMOS output and load capacitance but not load resistance. Can I be sure that 200 ohms total R will be OK? Or should I have an AC termination on the clock line, with a cap between R2 and ground?

Reply to
Prune

That wouldn't hurt, and would eliminate the average DC load on the XO. If you don't mind the parts, that's even better. But sonce 200 ohms isn't a heavy load, it would work either way.

HC parts can usually drive pretty hard. But we have seen some cheap Mouser-type XOs that had fairly weak drive but screaming (as in 600 ps) edges, and produce mediocre swings when seriously terminated. That wouldn't be a problem for you, since you have swing to spare.

John

Reply to
John Larkin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.