0
\$\begingroup\$

I have made a very simple inverting opamp amplifier to convert an output of 0-3.3V from a DAC to somewhere between -10 to +10V. enter image description here

My reference (REF_1.65) is 3.3V/2 made simply by a 10k/10k divider of the 3.3V supply (buffered with 100nF cap). Both are not shown here. So when going with DAC_REF from 0 to 3.3V, this voltage is going from positive to negative as expected. However the output has more ripple than expected. The output V_REF_X is used as a reference Voltage for an instrumentation amp. The point is, to roughly set the ReferenceVoltage (it does not have to be very exact) but stable, as ripple is directly added to the actual (sensor-)signal.

Sidenote: To get a better grasp on the ripple, I made an AC-amplifier by using an opamp with amplification of 18k/1k noninverting (I had some free pins and soldered a series cap to in+ and two resistors for the amplification. The scopeshots below are all made with signals amplified by this helper-amplifier, taken directly from the output). Furthermore the Volts/div are wrong by a factor 10 in the pictures, I used 1/1 probe, but had 1/10 set on the scope. So 50mV is actually 5mV.

Lets get to the Scope shots: First, TP9 (the output) has a very rippely output: enter image description here

In contrast, when measuring my supply (same 18x Amplifier used) it looks quite ok: enter image description here

The output of the most right opamp has really much noise (pin 14). Note the 200mV instead of 50mV scale: enter image description here

Although the inputs (DAC_REF_X, REF_1.65V) look ok as well: enter image description here

My layout looks like this:

enter image description here

How come this simple circuit produces so much ripple. Note: I am looking at the ripple (the low frequency change of voltage), not the noise, which is ok.

(The noise is some kind of 200kHz Signal, probably from my buck converters. If you think I should investigate this, I will)

Update 1: Ok, in-between-result: the 100pF made the noise on the first opamp go away, so this is progress:) Secondly, something in my workplace seems to induce the 200kHz "noise" into my circuit (or rather the probe), and as I thoroughly twisted my supply cables, it got better. Third: I added 100pf on +15 and -15 to gnd, but the problem remains. However, after the excess noise on opamp one was gone due to the 100pF||18k, I saw the ripple (as shown in Pic 1) clearly on the output of opamp one (that would be picture 3 looking like pic 1 now). And lastly: I have three channels, and had the strange phenomenon that all three channels stopped having the ripple alltogether (although I made changes to only one channel). But it came back after a few power cycles.

Update 2: I have a capacitance multiplier for my +- Supply of the opamp. I experimented around with adding 100pF, 100nF to see if this is of any help but turns out, my supply is not the problem, however the input signal is. I just did not see it, as the opamp (right one) amplifies it so much that I can see it only after having it amplified. Furthermore, before adding 100pF||18k, I only had noise on this output, so I could not see that my signal had ripple already at the first stage. This means: 1.65V and DAC-out are the problem, which I did not suspect at first as I have a dedicated 3.3V supply with the NCP161, which should be very quiet and ripple free. See here: enter image description here

The input 5V are from a SMPS but I hoped the NCP would get rid of all trobules on the input 5V.

The problem now is, that I cannot seem to get the 3.3V quiet although I added 100uF. My next guess would be some starpoint-ground problems, but I gave this extra thought when designing my board, but maybe I made a mistake here still. Thanks for everyone helping me!

\$\endgroup\$
9
  • 3
    \$\begingroup\$ C24 (1uf) risks instability in that unity-gain buffer...could be the source of 200kHz. \$\endgroup\$
    – glen_geek
    May 29 at 1:53
  • 2
    \$\begingroup\$ Normally, for relatively simple circuit diagrams, the main signal flow path is from left to right. It doesn't affect the analysis, it just makes the diagram easier to read. \$\endgroup\$
    – Chu
    May 29 at 7:39
  • \$\begingroup\$ @glen_geek I removed the output caps alltogether, it didn't change the the behaviour at all. Same noise on output of the right opamp, same ripple on the output of the left one. \$\endgroup\$
    – jjstcool
    May 29 at 19:34
  • \$\begingroup\$ @Chu Yes, you're right. In my case, there was a reason to do so, but in general I aggree. \$\endgroup\$
    – jjstcool
    May 29 at 19:36
  • \$\begingroup\$ @jjstcool For supply decoupling you would use much more than 100pF. Usually 100nF near th opamp. And in addition put an 100uF electrolytic cap on each supply to GND if you don't have already, but this can be placed wherever you've got space for them.for scoping it helps to zoom in with the time a lot to see the shape of the ripple-> this helps to narrow down on the cause. \$\endgroup\$
    – tobalt
    May 30 at 5:03
3
\$\begingroup\$

You can't drive a cap directly in this way.

Add a small resistor of e.g. 33 Ohm in series after U7C

If you don't have space/footprint to do this, then replace the cap with a cap+resistor series combination.

Also you mentioned, that already the first opamp is noisy. I see you decoupled +15V to -15V.. this is not much use if your output is GND referenced. Each supply should be decoupled to GND.

It can be though that the cap makes the left opamp oscillate and this affects the right opamp as you don't have decoupling caps.

\$\endgroup\$
5
  • \$\begingroup\$ Hey, thanks. I removed the cap altogether, but the behaviour did not change a bit. Any other ideas? \$\endgroup\$
    – jjstcool
    May 29 at 19:32
  • \$\begingroup\$ Do you have any other capacitor load on the output ? And you should add decoupling as I wrote in the answer. Can you scope the supplies ? \$\endgroup\$
    – tobalt
    May 29 at 19:36
  • \$\begingroup\$ Picture 2 is the supply. I removed everything, so the output is basically in free air. I will add some decoupling with respect to gnd on the V+ and V- later tonight and update it here. But as I have a quiet supply, I don't have much hope... \$\endgroup\$
    – jjstcool
    May 29 at 19:39
  • \$\begingroup\$ Hmm 🤔 another suspicion: R12 looks a little big. The TL74 is not the fastest amp but try adding a little cap in parallel to R12, 10-100pF \$\endgroup\$
    – tobalt
    May 29 at 19:45
  • \$\begingroup\$ Hey, I put an update in my original post (too long for the comments). Thanks for bearing with me! \$\endgroup\$
    – jjstcool
    May 29 at 22:27

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.