1
\$\begingroup\$

I have added a .1uF decoupling capacitor between Vcc and GND on a TMS9118 (this is an old video display processor). When I measure the voltage at Vcc with my oscilloscope, I read a pk - pk of 1V. Is that noise level way too high? This is related to a previous question of mine regarding a noisy composite video out signal from this chip: Lack of color displaying composite video

I would also like to note that I'm using a very clean bench supply for power. Pk - Pk on the bench supply is 60mV.

\$\endgroup\$
7
  • \$\begingroup\$ What is the frequency of the 1 Vpp noise? \$\endgroup\$ Jun 23, 2016 at 19:23
  • \$\begingroup\$ @OlinLathrop I can't get a reading of the frequency with my scope. It is a 70 Mhz scope. The clock of the CPU driving the VDP is 4Mhz. \$\endgroup\$
    – Adam T.
    Jun 23, 2016 at 19:48
  • \$\begingroup\$ As a diagnostic, try holding the CPU in reset and see if that affects the power supply noise. Similarly try disabling the 4MHz clock. This would confirm whether the issue is insufficient bypass(reservoir) caps or problem with the supply itself. \$\endgroup\$
    – MarkU
    Jun 23, 2016 at 20:28
  • \$\begingroup\$ Holding reset has no effect - nor either removing power from the CPU board (which includes the clock). I found that removing the 7.xx Mhz crystal from XTAL1 and XTAL2 on the VDP reduced Vpp by ~200mV to 800mV. \$\endgroup\$
    – Adam T.
    Jun 23, 2016 at 21:28
  • \$\begingroup\$ - Make sure that you are measuring correctly. The ground loop on a classic probe also captures noise. 1Vpp is a lot though. - Add a 10uF capacity in addition to the 100nF. \$\endgroup\$
    – le_top
    Jun 23, 2016 at 21:32

1 Answer 1

6
\$\begingroup\$

The only way to know if 1 Vpp is too high is to consult the datasheet. At the very least, the chip must be specified to operate over the resulting supply voltage range. For example, if the supply voltage is nominally 3.3 V, then the chip is seeing 2.8 to 3.8 volts. Is the chip specified to operate over that range?

Even if it is, it may still not be OK. You may not be seeing all the high frequency components, making the total deviation from nominal higher than it appears. Also, the dV/dt can cause problems, even if the absolute voltage is always within spec. dV/dt limitations are often not in the datasheet, so it's best not to push things too far.

That all said, 1 Vpp on a 3.3 or 5 V supply is a lot. Something is probably not working right if you get that much noise. Leave the 100 nF cap there, but add a 1 µF or maybe 10 µF ceramic directly across the power and ground pins of the chip. This is assuming the 100 nF cap is already directly across the pins with as small a loop as reasonably possible.

There is something wrong if there is 60 mVpp at the bench supply, and just a short distance away the same nodes have 1 Vpp on them. This should not be allowed to happen. The connection between the bench supply and the chip should be very low impedance at low frequencies, and the decoupling caps should provide low impedance at high frequencies.

It would be useful to know what the frequency of this 1 Vpp noise is. That would give you hints where it might be coming from, and what can be done about it.

Until you have a good power supply, nothing else matters. This is the first thing you should address. Everything else is a waste of time until the supply is clean.

\$\endgroup\$
2
  • \$\begingroup\$ You should also mention the need to make a good high-frequency connection to the scope. Long probe leads and improper ground connection points can produce the most amazing traces. \$\endgroup\$ Jun 23, 2016 at 19:47
  • \$\begingroup\$ I'm marking this as answered because it answers my question as to whether or not Vpp is too high. Thanks @OlinLathrop! \$\endgroup\$
    – Adam T.
    Jun 23, 2016 at 21:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.