oscillator noise - (sustaining engineering) 1

[UtrasoundGuy]

hello all.
we have been using an oscillator with tristate (enable) function in our design for some odd 10 years. The company purchased enough to sit on the inventory up to about now as it was cheaper to do this. The newest batches that came in (slighly different version with better tolerance MXO45T-3C-5M0000. - vs older MXO45T-2C-5M0000) are creating distrotions in our images in which the oscillator is creating a pulse sequence for. I can trace this back to the oscillator not starting properly or having some type of noise on it, both of which can occur.

Ive attached a part of the schematic if anyone is willing to take a look at it. More can be added as needed.

http://files.engineering.com/getfile.aspx?folder=d23d5d8a-4216-452a-a382-4e8b7fc1a5cf&file=shot.bmp

If also attached a some images of what the noise looks like. In this PDF, the green represents the supply voltage to the oscillator, the blue is the OE (output enable) signal, and pink is the output of the oscillator. There should be many pink output pulses for each blue since the 5MHz is faster than the much slower output enable signal. There is approximatly 250 output enables a second.

http://files.engineering.com/getfil...-f825-4f25-958b-d92504dc8c21&file=TGCtips.pdf

What Im confused about is why there are negative spikes for all of the oscillators except the 2C. How is this possible? This was measured many times over, all Im replacing is the oscillator, probes fixed in place. Im not sure its just RF induced noise.

I figured it was some type of impedance mismatch or filtering issue. Ive tried increasing filter storage caps across all the power supplies on the PCB to 47uF (from 0.1uf) right across the terminals of the osciallor with no luck. Ive also tried shunting the voltage to the OE and output with 1k and 1k//0.1uf without any luck either. I dont get it. Any help is appreciated.
Jim

 

[VEBill]

According to the CTS data sheet (PDF here), the only difference between the '2C' and '3C' is the tolerance of the frequency. So that in itself probably isn't the issue.

The data sheet has a footnote right next to the OE spec, linking to AN 014-0002-0. This described two different OE approaches and that change might (?) be exactly what the problem is.

You'd better establish contact with an Applications Engineer at CTS, and hopefully they can help you to sort out a solution.

Good luck.

 

[Comcokid]

With high-frequency bypassing sometimes more capacitance is worse. It all comes down to capacitor type, lead-inductance, and trace-inductance. If you calculate the inductance needed for a self-resonance circuit and say 10uF and 47uf, you will find that above 1 MHz you need an impossibly small lead inductance to make the capacitor worthless. I don't know if your design is SMD or through-hole, but try this. On chips where noise can affect your design, bypass with 1000pf or 10,000pf ceramic surface-mount capacitor with any lead extension to the pins made up of 1/8" wide flat pieces of copper tape or foil (low inductance connections). I some times use thin (0.010 to 0.032") single-sided PCB material cut with sissors.

The other question is how you are probing your PCB to see the signals. To get a really-clean scope waveform, the flying ground lead is a no-no. It is also inductive and will add false ringing to what you see. You need a spring-type ground, or again copper tape to the ground kept short as possible.

 

[VEBill]

To ensure that this tidbit isn't overlooked, I'll post this explicitly here.

The above referenced AN states for PN MXO45T...
'OSCILLATOR STOP' type: OE time = 10 milliseconds (!!!)
...or...
'OSCILLATOR RUN' type: OE time = 100 nanoseconds

Both types are "approved" by CTS. For your "convenience" (LOL) it appears that they fail to distinguish the two types in the PN.

If the OE is being toggled at 250 Hz, then the 'OSCILLATOR STOP' type of component would be perpetually in start-up time (250/s x 10 ms = 2500ms/s exceeds 100%).

Bypassing probably isn't an issue if the board / bypass design worked with the previous generation. Unless CTS changed something else too.

The AN isn't dated, but it's probably within the past decade. It might be interesting to confirm.

 

[UtrasoundGuy]

thanks Bill and ComcoKid

VEBill, very good find. I think this is a big part of the issue. Now that I think about it, although the design is working and has been for so long (10+ years), my evaluation of it is that it has never really been working right. This might be part of why the chip is so difficult to replace.

The 5MHz output is turned on each time the OE is turned on (around each 3ms) but what ive noticed is that the output between each output enable sequence never goes back down to 0. It like there is an output capacitance that is discharging which is probably related to this 10ms delay time. That is part of why I was trying to put in the shunt resistors to lower the RC time constant but ideally this is internal to the chip im guessing. Ive picked out a few new chips will try.thanks

 

[UtrasoundGuy]

as for the flying leads and ground inducatance to the probe, I know that they can show up in the signals. I had previolsy tried cutting the leads very short on oscillators to try to reduce this and even put in some thicker copper to the ground plane. This could still be an issue but I think there is a bigger issue at play which I think bill hit.

 

[iop995]

Where did you placed scope probe for 5V signal?
There are capacitors (one of 47-100microF and one or two ceramic 10nF) placed across MX power pins?
If you reduce frequency of OE signal to specified limit (100Hz) all are ok?

 

[UtrasoundGuy]

its starting to look like this is just RF noise.
However, why it is larger amplitude for some chips than others is unknow

but thanks for the help everyone

 

[UtrasoundGuy]

But we eventually used 5 different types of programmable oscillators and were able to get all to work in our existing 10 year design. I think we just got a bad batch of oscillators from the vendor. The noise turned out to be somewhat unrelated to the issue.. It turned out if too long of a time was given for the OE, the output would hangup for a few cycles with this one chip that we had used for 10 years. All the programmables worked. We lost a month of work due to this issue as we had consoles we had to return and fix.