[SI-LIST] Re: Decoupling of Oscillator
- From: pwelling@xxxxxxxxxxxxxx
- To: bmgman@xxxxxxxxxx, Anand.Kuriakose@xxxxxxx, si-list@xxxxxxxxxxxxx
- Date: Wed, 13 Nov 2002 08:34:58 -0700
Anand,
From an experience at a previous company, I'm reminded of the first PC
EGA/VGA Video controller cards that were produced by many vendors. As some
may recall they used to have multiple oscillators on them for the different
video pixel clocks and horizontal/vertical refresh rates. The initial ones
(and I evaluated many cards at the time) had serious FCC compliance problems
which many vendors resorted to output filters to make compliance which
affected video quality. A large majority of the vendors that didn't have
problems had used Pi filters on their oscillators. This prevented the
oscillator contamination of the VCC planes and prevented coupling into their
RAMDAC output device. Many of the cards started to put filters on the RAMDAC
power pins and that soon became a requirement from the RAMDAC vendors.
The EMI profile was obvious when you looked at it, you could easily identify
each oscillator (all of them were always enabled - and the one they wanted
was selected by a mux in the video controller chip). The 3rd, 5th, and 7th
(some up to 19th) harmonics were always visible. These harmonics plus the
selected clock usually caused problems for compliance as they were impressed
on the video cable signals or cable shields.
The Pi filters helped with the power supply pin noise but, not the
oscillator harmonics that were injected into the grounds. The harmonics on
the ground planes would shoot through the RAMDAC too. That was when they
started to use ground moats for the RAMDAC to provide isolation. This too
became a standard in RAMDAC datasheets. Some provided Pi filters for the
ground but that didn't work well logically because of the return path
inductance and Signal Integrity (not even coined back then) induced
problems.
A real boon to the video controller market was the PLL programmable video
clock generators (like ICS) that had the ability to programmed with
multiplexed clock divide ratios to give them only the clock(s) they needed.
This only had 2 frequencies to worry about, the 14.318 MHz reference, and
the Video output clock frequency from the PLL. This reduced emissions
dramatically as all of the separate free running video clock oscillators
frequencies (some had up to 6 oscillators) had been reduced to 2 frequencies
plus the oscillator for the video logic. Sometimes the video logic (serial
shift register clock) caused problems too.
My point is that Pi filters are good for oscillators because they keep the
emissions off of the planes which might couple to I/O cabling which causes
EMC failures. Since oscillators draw relatively little current, a high
impedance bead can be used to provide high isolation. Remember that as
current is pulled through surface mount beads, the impedance (isolation) is
reduced nearly logarithmically. You are making a new supply for the
oscillator. Use high frequency ceramic and low frequency Tantalum
decoupling. The Tantalum will filter the lower frequencies that make it
through the bead causing jitter that you may not be able to track out of
your PLL. A caution here, be careful using PLL based clock oscillators (like
JITOs) feeding a PLL circuit. Their jitter may be multiplied up in the PLL
circuitry ie... don't cascade PLLs. The series resistor will reduce edge
energy (besides providing termination) to the destination device which helps
quiet the destination device noise too. Make sure that you provide a good
return path back to the oscillator.
The Pi filter not only prevents contamination of the planes, it provides a
clean "living area" for the oscillator enabling high performance.
Just my $0.03 (cost of living).
Philip Ross Wellington
Mgr. Signal Integrity & EMI
L-3 Communications CSW
-----Original Message-----
From: Mike Brown [mailto:bmgman@xxxxxxxxxx]
Sent: Wednesday, November 13, 2002 6:54 AM
To: Kuriakose, Anand; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Decoupling of Oscillator
Anand,
It is relatively common for the oscillator to be treated as just another
load, and omit everything in the filter except the HF capacitor.
I think the main purpose of the dedicated pi filter for the oscillator
is protection of the oscillator, because the other loads are relatively
immune to LF/MF variation in their supply voltages. That doesn't say
that the filter doesn't provide other functions as well.
The pi network mentioned will certainly keep the oscillator current
spikes out of the supply. Is that its main purpose? Every load driven
by the oscillator is also drawing current at the oscillator frequency,
or 2x the oscillator frequency. The other loads don't use a dedicated
overt pi filter, but rely on HF decoupling. Why not the oscillator?
Its current spikes aren't that much different than the other loads.
I think the question here is "Who's being protected from what?",
followed closely by "What's the most cost-effective way to provide the
protection?"
Regards,
Mike
Kuriakose, Anand wrote:
>Mike,
>
>You said "that filter network is not to keep the oscillator
>spectrum out of the power supply, but to keep power supply noise from
>modulating the oscillator."
>
>I an addition to isolating the power rail noise from the VCC of the OSC,
the
>current spikes generated on the VCC rail due to the OSC switching is being
>decoupled by the decoupling scheme mentioned below.
>
>That is to say that the filter isolates noise originating from either sides
>of the filter, which is precisely what one should do to make all devices
>connected to the common power rail operate without causing too much of
>noise.
>
>Please correct me if i am wrong.
>
>Regards,
>Anand.
>
>
>
>-----Original Message-----
>From: Mike Brown [mailto:bmgman@xxxxxxxxxx]
>Sent: Tuesday, November 12, 2002 7:42 PM
>To: zhang_kun@xxxxxxxxxx
>Cc: si-list
>Subject: [SI-LIST] Re: Decoupling of Oscillator
>
>
>
>Zhangkun,
>
>You are right - that filter network is not to keep the oscillator
>spectrum out of the power supply, but to keep power supply noise from
>modulating the oscillator. Jitter will be introduced into the system
>timing if this modulation occurs. Some jitter will unavoidably occur
>but the filter, including the tantalum cap, will minimize the amplitude.
> The noise frequency will be determined by the timing of the loops in
>the software, which change the power loading periodically. Any load
>variation above the regulator cutoff frequency is a possible noise source.
>
>I prefer to isolate the power to the oscillator. Others don't, and they
>get away with it if their system is not jitter sensitive. I once built
>a system with two oscillators and no isolation and found the resulting
>jitter due to the asynchronous noise to be intolerable. Isolation
>solved the problem.
>
>Regards,
>Mike
>
>Zhangkun wrote:
>
>
>
>>Hi all
>>
>>I meet one question about decoupling of oscillator. In our design, the
>>
>>
>power supply of OSC is always isolated by one PI filter. In the side near
>OSC, there are always one tantalum capacitor of 10uF and several ceramic
>capacitors of 0.1uF or 0.01uF. As we know the resonance frequency of
>tantalum capacitor is about 3MHz. If the OSC is 50MHz, the spectrum will be
>speaded at 0, 50MHz, 100MHz, 150MHz, etc. There will be no power in the
>frequency range between 0 and 25MHz. Therefore, I think I could remove the
>tantalum capacitor. Is there something wrong?
>
>
>>I think it will have nothing to do with the affection from OSC to outside
>>
>>
>circuits. I am worrying about the affection from outside circuits to OSC.
If
>there is some noise of 2MHz and my OSC is of 50MHz, there will be
modulation
>between noise of 2MHz and clock signal of 50MHz. The output of OSC and the
>clock signal will be affected by the noise of 2MHz. The bead will not
>isolate the noise of 2MHz.
>
>
>>I want to know I could remove the 10uF tantalum capacitor or not. Why?
>>
>>By the way, is there some people who do not use bead to isolate the power
>>
>>
>supply of oscillator?
>
>
>>Best Regards
>>Zhangkun
>>2002.11.12
>>
>>
>>------------------------------------------------------------------
>>To unsubscribe from si-list:
>>si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>>
>>or to administer your membership from a web page, go to:
>>//www.freelists.org/webpage/si-list
>>
>>For help:
>>si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>>
>>List archives are viewable at:
>> //www.freelists.org/archives/si-list
>>or at our remote archives:
>> http://groups.yahoo.com/group/si-list/messages
>>Old (prior to June 6, 2001) list archives are viewable at:
>> http://www.qsl.net/wb6tpu
>>
>>
>>
>>
>>
>>
>>
>
>
>
>------------------------------------------------------------------
>To unsubscribe from si-list:
>si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>
>or to administer your membership from a web page, go to:
>//www.freelists.org/webpage/si-list
>
>For help:
>si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>
>List archives are viewable at:
> //www.freelists.org/archives/si-list
>or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
>Old (prior to June 6, 2001) list archives are viewable at:
> http://www.qsl.net/wb6tpu
>
>
>
>
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
