0
\$\begingroup\$

I have 4 opamps powered from the LM2664. Two of them (1MHZ 1.6V R-R) are preparing the offset voltage for the pair of the OPA4141 which actually provide the signal for the ADCenter image description here

It works as expected but unforunatelly I see the negative supply voltage "shadow" in the output signal. Yellow - signal on the 3.3uF LM2664 capacitor - red ADC1 output when input is connected to the ground. They are different scale. It is about 30mV PP. (gain 100) enter image description here How to remove this influence.

Remarks - 15pF capacitors are not populated. JP31 is a slider switch with the common pins 9 & 10 (one per side).


EDIT

I have moved the LM.. out of the board. Power is taken from another board. GND is conected via a wire.

But the problem persists: enter image description here The

\$\endgroup\$
  • \$\begingroup\$ Are these IC's being powered by a linear or switching power supply? \$\endgroup\$ – Sparky256 May 17 '18 at 23:03
  • \$\begingroup\$ + linear, - charge pump (inverter). \$\endgroup\$ – P__J__ May 17 '18 at 23:05
  • \$\begingroup\$ Charge pump is not linear. \$\endgroup\$ – Lior Bilia May 17 '18 at 23:14
  • \$\begingroup\$ @LiorBilia plus is linear minus not linear. \$\endgroup\$ – P__J__ May 17 '18 at 23:22
1
\$\begingroup\$

Learn to think about providing "local batteries" at all power pins of your crucial amplifiers.

If your signal-chaing is AC coupled, this is easy.

If DC coupled (yours looks DC coupled), you need to avoid any changes in shared currents. Star-ground-thinking is your friend.

\$\endgroup\$
  • \$\begingroup\$ Thank you. Edited my question. I tried to separate the GND-s but the effect is very limited \$\endgroup\$ – P__J__ May 18 '18 at 16:41
1
\$\begingroup\$

A guess: Current peaks of LM2664 Charge Pump DC/DC converter go through your signal GND wiring and cause noise voltage due the resistance and inductance of GND wiring.

If you look at the datasheet of LM2664, you see that there are three capacitors connected directly to LM2664. No tapping is allowed to other circuitry between the capacitors and LM2664. In addition the wires between your main power supply and LM2664 shouldn't be any part of signal wiring. Most impostant this is for GND.

These requirements are the same as in all device designs which contain analog circuits and logic or pulse circuits. No pulse or logic circuit currents are allowed in analog signal GND wiring.

ADD: Oscilloscope measurements can also generate noise if there's long ground connections for the scope or mains power wiring creates ground loops. You should test does your measurement catch additional noise (have a high frequency signal transformer + DC blockinc capacitor in the input of the scope or feed your circuit with DC from a battery) USB refers you have a computer. How many separate ground connections you have via the computer? One GND connection only is allowed.

\$\endgroup\$
  • \$\begingroup\$ The problem is that i do not know how to separate GNDs. My 3.3uF caps are placed exactly as shown in the DS. \$\endgroup\$ – P__J__ May 17 '18 at 23:15
  • \$\begingroup\$ PS my LM oscillation frequence is 88KHz not 160 as in the DS \$\endgroup\$ – P__J__ May 17 '18 at 23:21
  • 1
    \$\begingroup\$ @PeterJ_01 GND separation is not a must. Only be sure that the wire between LM2664's GND and your main power supply carries no signal between analog parts (including Analog to digital converter). 88kHz pulses are as sharp as 160kHz pulses. \$\endgroup\$ – user287001 May 17 '18 at 23:33
  • 1
    \$\begingroup\$ Analog, digital and power grounds should meet at the power connector only, where you normally have 100 uF electrolytics as power buffers. You may need 3.3 to 100uH inductors at each op-amps power pins to kill current based noise. Plus the usual bypass caps at the power pins to ground. For HF mixed signal boards filtering, power and ground nodes dominates the board layout. \$\endgroup\$ – Sparky256 May 18 '18 at 0:35
  • 1
    \$\begingroup\$ @PeterJ_01 User Sparky256 in his comments added some commonly used mixed signal circuit design details. Analog and pulse signal grounds should meet exactly in one point. Otherwise pulse currents spoil the equipotentiality of the analog GND. If you have ground connections here and there you cannot keep pulse currents out of the analog signal path (half of that is the analog GND) I know several cases where people have a computer in their audio system and then wonder "all is new and premium quality, why there's so much buzz and hum". The system has GNDs connected via USB, audio and power. \$\endgroup\$ – user287001 May 18 '18 at 1:11
0
\$\begingroup\$

Eventually I have found what the problem is. The trimming capacitor was catching the oscillations over the air. As a workaround it was enough to temporary shield the inverter. in the production boards I have moved the inverter more far from the trimmer.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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