74HCT9046 VCO phase noise?

[Andy F Z]

Hello:

I wonder what is the lowest phase noise that can be obtained from the
74HCT9046's VCO, running at f_0=16MHz. For instance, the relative VCO
signal power at f_0+-100Hz, +-500Hz, +-1000Hz.

Thank you for you help.

Andy.

 

[Tom Bruhns]

Hello:

I wonder what is the lowest phase noise that can be obtained from the
74HCT9046's VCO, running at f_0=16MHz. For instance, the relative VCO
signal power at f_0+-100Hz, +-500Hz, +-1000Hz.

Thank you for you help.

Andy.

Are you asking about the raw VCO performance, or the VCO in a locked
loop?

Cheers,
Tom

 

[Andy F Z]

Are you asking about the raw VCO performance, or the VCO in a locked
loop?

The raw performance. But if the data for a particular, well tuned locked
loop is available - that would be useful as well.

Thanks,

Andy.

 

[MooseFET]

The raw performance. But if the data for a particular, well tuned locked
loop is available - that would be useful as well.

IIRC: If you consider the noise to be a voltage source at the control
input, you can say that there is a 100nV/sqrt(Hz) wide band noise and
a 500Hz 1/F corner. The values will vary quite a lot from maker to
maker.

Assuming the control voltage location for the noise source, suggests
that the noise is in terms of frequency noise and not phase noise. At
some highish frequency, actual phase noise from within the VCO will
start to show at the output.

Monitoring the output with a counter that samples at a 10Hz rate, I
have seen variations under 10PPM in the frequency.

When you close the loop, the phase margin of the system matters a lot
to the amount of noise you see. If you use the self bias in the the
phase detector's input amplifier, you take a fairly major noise hit.
You are better to use an external comparitor and feed the device with
a sharp square wave. Doing this I have seen frequency noise levels
under 1 part in a billion per sqrt(Hz). My measurement was done at
about 2MHz.

 

[Andy F Z]

MooseFET wrote:
....

IIRC: If you consider the noise to be a voltage source at the control
input, you can say that there is a 100nV/sqrt(Hz) wide band noise and
a 500Hz 1/F corner. The values will vary quite a lot from maker to
maker.

What point in the input control circuit do you mean? The ratio
dVCO_Frequency/dVCO_IN of the 74HCT9046A depends on the R1C1 product.
I have not come across other makers of this chip than NXP/Philips.

Assuming the control voltage location for the noise source, suggests
that the noise is in terms of frequency noise and not phase noise. At
some highish frequency, actual phase noise from within the VCO will
start to show at the output.

In my prototype, the output signal density is essentially spread in a
few kiloHertz area around the center frequency, even with a very very
overdamped control loop, I tend to conclude it's the noise within the VCO.

Monitoring the output with a counter that samples at a 10Hz rate, I
have seen variations under 10PPM in the frequency.

When you close the loop, the phase margin of the system matters a lot
to the amount of noise you see. If you use the self bias in the the
phase detector's input amplifier, you take a fairly major noise hit.
You are better to use an external comparitor and feed the device with
a sharp square wave. Doing this I have seen frequency noise levels
under 1 part in a billion per sqrt(Hz). My measurement was done at
about 2MHz.

Thanks you for the information!

 

[MooseFET]

MooseFET wrote:

...


What point in the input control circuit do you mean? The ratio
dVCO_Frequency/dVCO_IN of the 74HCT9046A depends on the R1C1 product.
I have not come across other makers of this chip than NXP/Philips.

It is a variation on the 4046, 74HC4046. The exact part may not be
made by many people but lots of things very like it are.

The signal on pin 9 is the control input to the VCO. This is where I
referenced my noise measurement.

In my prototype, the output signal density is essentially spread in a
few kiloHertz area around the center frequency, even with a very very
overdamped control loop, I tend to conclude it's the noise within the VCO.

I find this result a little unexpected. What was the gain crossover
frequency of you loop? Near the carrier, the VCO's output noise
should not be what effects the resulting noise. The VCO's noise is
within a servo loop. If you model the VCO's output as a voltage
instead of a phase and think about your average op-amp circuit, I
think you will see what I mean.

 

[Andy F Z]

...
It is a variation on the 4046, 74HC4046. The exact part may not be
made by many people but lots of things very like it are.

Oh, yes, I realise that the VCO part must be the same in the 74HCT9046.

I find this result a little unexpected. What was the gain crossover
frequency of you loop? Near the carrier, the VCO's output noise
should not be what effects the resulting noise. The VCO's noise is
within a servo loop. If you model the VCO's output as a voltage
instead of a phase and think about your average op-amp circuit, I
think you will see what I mean.

In fact, I have a big N divider in the loop, the reference frequency is
about 1 kHz, so the servo loop must not be able to reduce the noise in a
few kHz vicinity of the carrier. I have now taken a look at the VCO's
signal with the loop disconnected, its near carrier spectrum appears to
be pretty the same as with the loop, and too bad for my purpose...

Thank you.

 

[Winfield]

Oh, yes, I realise that the VCO part must be the same in the 74HCT9046.


In fact, I have a big N divider in the loop, the reference frequency is
about 1 kHz, so the servo loop must not be able to reduce the noise in
a few kHz vicinity of the carrier. I have now taken a look at the VCO's
signal with the loop disconnected, its near carrier spectrum appears to
be pretty the same as with the loop, and too bad for my purpose...

Thank you.

You're using a well-bypassed SMD part? Bypassed with both ceramic
and low-esr electrolytics? The VCOs in the '4046 PLLs are excessively
sensitive to high-frequency supply ringing (e.g., from two ceramic
caps
resonating with the supply-trace inductance, which I've been able to
observe by careful scope observations), and the '9046 VCOs are no
doubt much the same.

 

[MooseFET]

Oh, yes, I realise that the VCO part must be the same in the 74HCT9046.


In fact, I have a big N divider in the loop, the reference frequency is
about 1 kHz, so the servo loop must not be able to reduce the noise in a
few kHz vicinity of the carrier. I have now taken a look at the VCO's
signal with the loop disconnected, its near carrier spectrum appears to
be pretty the same as with the loop, and too bad for my purpose...

I still find it strange. Have you looked at the supply voltage with
the scope? Maybe there is ripple on it.

 

[Andy F Z]

...

You're using a well-bypassed SMD part? Bypassed with both ceramic
and low-esr electrolytics?

Bypassed by ceramic caps, and with an inductance in the power rail... I
will try an electrolytics. In the meantime, I have tried a 1u film, but
it has had no effect.

The VCOs in the '4046 PLLs are excessively
sensitive to high-frequency supply ringing (e.g., from two ceramic
caps
resonating with the supply-trace inductance, which I've been able to
observe by careful scope observations), and the '9046 VCOs are no
doubt much the same.

I will try this chip in an optimized testbed, the current layout does
not have a solid ground plane, for instance.

Thank you.

 

[Andy F Z]

I still find it strange. Have you looked at the supply voltage with
the scope? Maybe there is ripple on it.

No, no obvious ripple on the power rail. But it looks like this VCO is
extremely sensitive to the supply voltage/rail (at least when pushed to
the limits: at 15-16 MHz). I will try it in an optimized, "quiet"
testbed, maybe with a local voltage regulator.