This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LMK04832-SP: voltage controlled oscillator connection to lmk OSCin

Part Number: LMK04832-SP
Other Parts Discussed in Thread: LMK04832EVM-CVAL, ADC12DJ3200EVMCVAL, LMK04832, LMK04828

Hello, how are you?

I´m working with LMK04832-SP and I need to connect a voltage controlled oscillator

 to OSCin pins and I use a dual loop configuration.

I have a VCXO ACMOS, voltage controlled oscillator, with a single ended output. I don't have an Outout_P and an Output_N. Like the LMK development kit.

We can see the VCXO and LMK connection, in the Kit LMK04832EVM-CVAL, in the next picture. The clock signal is connected to pin N, LMK 44 pin, not to pin P, LMK 43 pin, and has an network with the  components R58 C72 R59 R60 C71.

Also we can see the same network, with the  components R58 C72 R59 R60 C71, in the ADC12DJ3200EVMCVAL kit.

I have some questions about that:

 

1_ Why is the oscillator single ended signal connected to pin N not to pin P, like we can see in Figure 2 and Figure 3? 

 

2_ What is the function of the network between the voltage controlled oscillator output and the LMK input?

 

3_ Can I connect the oscillator single ended output to OSCinP without a network Like the Figure 1, as it says in the data sheet in figure 1 ?

 

Thanks you

Regards

Facundo

INVAP S.E.

  • Hello Facundo,

    1. The VCXO vendor originally released an LVPECL version of the VCXO that used differential inputs. Eventually they released an HCMOS version that had better phase noise and we switched to using that device. The pinout for the HCMOS version populates pins 1, 3, 4, and 6 of the original pinout of the LVPECL version, with the output coming out on original pin 4 (connected to OSCin_N). In summary, it was for convenience with the existing pinout. There is no requirement for single-ended signals to use one or the other pin, either one is acceptable.
    2. This network is intended to allow compatibility with many different termination requirements, such as CML (the pull-ups) or LVPECL (the pull-downs) - it also allows for some level shifting configurations with LVPECL at different supply voltages if required on another oscillator we ended up not using. The 100/50 resistor divider in the signal path reduces the HCMOS amplitude to 1/3rd of the full amplitude to help comply with the single-ended input swing requirements on OSCin; conceptually this could be 50/50 or 50/100 and it would still be compliant. I recall that when we did this on LMK0482x, it was intended to reduce peak current stresses on the oscillator - but we didn't ask whether we needed to do that until much later (and it turns out it was unnecessary). On LMK04832, despite knowing we were not at risk of stressing the oscillator, we retained the structure because we saw that the higher order harmonics of the larger swing HCMOS signal were creating spurs at the phase detector. Reducing the amplitude and the slew rate of the signal helped to reduce these spurs. The two AC-coupling capacitors are to prevent DC loading on the oscillator and to prevent DC loading on the input stage. In theory the second AC-coupling capacitor is not required, since the device also internally AC-couples the input signal, but the DC average voltage at the resistor divider output will quickly drift to 0V with just the first AC-coupling capacitor, and the signal swing will begin to clip on the ESD structure. The second cap keeps the DC value at the pins close to the internal bias level. Remaining components (mostly 0Ωs) are just to provide the same set of options for external signal sources as on-board ones.
    3. Yes, you can connect single-ended output to OSCinP with the network like the referenced figure - provided the oscillator can accept high-impedance termination and meets the signal swing requirements. Otherwise you'll want a 50Ω load AC-coupled to the signal source and to the OSCin pin. I suppose you could also use two 100Ω resistors in series from VCC to GND and merely AC-couple the input signal, letting the DC bias sit at VCC/2 at the OSCin pin - but of course this adds supply current and creates a path for ground-referred supply noise to couple on the input, so I don't recommend it. In theory you'd also want to AC-couple the unused OSCin pin to ground through equivalent loading (50Ω) to minimize capacitively coupled reflections from one pin to the other, but in practice we've found this unnecessary.

    There's a slightly better depiction of acceptable input circuitry in section 10.3 of the LMK04828 datasheet - it's a different device, but the commercial version of LMK04832 is p2p with LMK04828, and they share a lot of similar circuitry (including the input stages) so the advice in LMK04828 10.3 is applicable to LMK04832(-SP) as well. At some point these diagrams will make their way into the LMK04832(-SP) datasheet.

    Regards,

    Derek Payne

  • Hi Derek

    Thank you Derek, your answer was complete and very understanding.

    Regards,

    Facundo Cosentino