-
This is v2 of my stereo preamp board. There a 3 op amp stages and I'm seeing a 481 kHz oscillation on one channel, at TP3, on the output of the 2nd op amp. It's a gain stage and the other channel is working as expected so I'm wondering if it might have something to do with how I routed the signal path for the misbehaving channel.
On the board layout you can see that the signal leaves TP3 on it's way to the tone stage (R16) but there's a via across a pair of power rails, +15VDC and - 15VDC. There's also a via on the other channel that crosses just the - 15VDC rail but no oscillations on that channel so maybe not. And you can see some sort of modulating influence on the oscillations as well. Note that the time base is different for the scope shots... to show the signal more clearly.
-
If it's really something self oscillating, placing a finger near the origin will change the frequency.
-
Op amps really like to oscillate if they have reactive loads.
A simple check/fix is to connect a lowish value resistor in series with the output pin.
The oscillation should cease. -
Op amps really like to oscillate if they have reactive loads.
Edited
A simple check/fix is to connect a lowish value resistor in series with the output pin.
The oscillation should cease.
Thanks xavier60 for the tip. Putting my finger on the op amp caused a small change in frequency so I it seemed like that could be the cause. I cut the trace on the output going to the tone stage (I made sure to do it in a place that is easy to repair) and added a 10k resistor but it didn't change the oscillations. I also tried it with a 47 ohm resistor and still no changes.
Taking a closer look it now seems like R13 may be the culprit. It's so close to R11 that I can't actually place my finger on R13 only but putting my finger there causes big changes in the oscillation.
Testing the resistance of both in circuit I show 47k for R13. But for R11 I show 440k in one direction and 350k with the leads reversed. I know it's not conclusive when you're testing in circuit but looking at the schematic I would think that if the op amp is impacting the value of R11 than I would expect it to also impact the value of R13. -
Op amps really like to oscillate if they have reactive loads.
A simple check/fix is to connect a lowish value resistor in series with the output pin.
The oscillation should cease.
By " in series" I assume you mean after TP3 in my schematic, to the right? -
Screaming aside.
The TL072 has a very high input impedance, and you have a polarised cap on the non-inverting pin ONLY. This is likely to slowly float-about or rise and pin a rail - you really should have a small load (perhaps 1M) on that to 0V to keep it centred on 0V.
If you create an HPF (high-pass filter) with the 33uF and say a 1M, you'll have unnesaarily low cut-off, so you could make the cap much smaller. Also the load is very small, do you really need a TL072 buffer between the output of the digipot and the gain TL072?
TL072 has a built in output resistor, and I've not found them to be squealers (unlike some of the low-Z high speed opamps) - perhaps you want to drop a few pF across the feedback resistor to lower the gain at HF.
-
Quote
author=hummusdude link=topic=283479.msg3583494#msg3583494
Testing the resistance of both in circuit I show 47k for R13. But for R11 I show 440k in one direction and 350k with the leads reversed. I know it's not conclusive when you're testing in circuit but looking at the schematic I would think that if the op amp is impacting the value of R11 than I would expect it to also impact the value of R13.
Something isn't right there. Assuming you are testing with the circuit unpowered, an in-circuit test of a resistor can never measure more than spec, only the same or less. So if you are correctly measuring high resistance across R11, then the resistor itself is likely open--which would obviously be a likely cause of your issue. -
U4A pin 3 has no DC path! I can't believe it works at all like that. (Same on other channel.)
Jon -
Ok...now for my next noob mistake...I posted this last night when it was late and I was tired and I neglected to make sure I was including the most recent and up to date version of the schematic. Very, very sorry for the confusion. I'm attaching here for clarification rather than to edit the beginning of the post. Not sure if this is the best etiquette but seems better than changing the attachments after several people have already commented.
Basically, I added two 10k resistors, R51 and R52, at the recommendation of floobydust, to get rid of the 2.5VDC offset at the non-inverting inputs to the gain stage. R52 is actually a THT resistor that is jumped over the board...it's not connected by a trace to ground as the pcb indicates (but it is grounded in the same place). I did that because it was the easiest way to show the added resistor without reworking the rest of the traces. I hope to get this board working but if I have to do a redesign this will be updated properly.
The other change is that the op amps are not TL072, but OP1656. I didn't change the schematic because KiCad doesn't have the same style op amp symbol in it's library and I was too lazy to create my own. The pin outs are the same so seemed like an harmless fudge. Obviously, that's not the case and it led to my cardinal sin of posting an inaccurate and outdated schematic. This will be #348 on the list of mistakes I won't ever make again!
Assuming I haven't thoroughly alienated everyone with my oversight, here's the rest of my post with questions.
bdunham7...I agree it's most odd to see different readings in opposite directions. I think it's most likely human error even though I've tried several different angles to get on the pads and I keep getting the same result. In either case, the measured value does seem to point to an open in the resistor so that's easy enough to replace.
Regarding the placing of R52... the first time I added it I used a 1206 smd because I had it and it fit nicely between pin 3 of U4 and the ground pad on R11. Having fixed the DC offset problem I now focused on the oscillation and I assumed I had created some sort of inductive loop situation by grounding R52 to the ground pad on R11. So I reworked R52 using a THT 10k resistor so I could easily ground it elsewhere. I picked a power rail ground as I had for R51 thinking that it would not interfere with the signal since it hadn't had that effect on the other channel. My question here is, is there any reason why I shouldn't put R52 with an smd and ground it to the ground pad on R11? That would make the addition of the 10k resistor a lot more tidy on the board.
fcb...QuoteThe TL072 has a very high input impedance, and you have a polarised cap on the non-inverting pin ONLY. This is likely to slowly float-about or rise and pin a rail - you really should have a small load (perhaps 1M) on that to 0V to keep it centred on 0V.
If you create an HPF (high-pass filter) with the 33uF and say a 1M, you'll have unnesaarily low cut-off, so you could make the cap much smaller. Also the load is very small, do you really need a TL072 buffer between the output of the digipot and the gain TL072?
I know the size of the cap, and that it's not a film cap, is less common. I used this set up because I was following an app note from Maxim that has these values in their sample circuit for how to create an offset for use with Dpots that won't tolerate negative voltages. Someone else commented on the buffer, U3, as perhaps being superfluous, and I had the same thought I but I put it there because it was in the app note, and because another person on this blog recommended I keep it. I'm sure the fine folks at Maxim assumed that anyone using their circuit would know enough to account for the issue you mention. In my case, I didn't and so here I am with a Frankenstein circuit I'm hoping to make work. To that end, there is currently a 10k resistor to ground, R52, on the channel that is acting up. It was put there to address the DC offset problem and is working well on the other channel. Is this what you are referring to? fc would be about 482 mH. For sure that's very low but I thought the filtering effect here was an unavoidable side effect of using a decoupling cap and not necessarily something that was needed to filter out unwanted frequencies. Is there a downside to having such a low cut off?
-
U4A pin 3 has no DC path! I can't believe it works at all like that. (Same on other channel.)
Jon
Are you referring to R51 and R52 in the updated schematic? They were added because it wasn't working. I bread boarded the DC offset circuit with the buffer and it worked perfectly but I didn't know about the need for such a DC ground path when connecting to an op amp. -
bdunham7...I agree it's most odd to see different readings in opposite directions. I think it's most likely human error even though I've tried several different angles to get on the pads and I keep getting the same result. In either case, the measured value does seem to point to an open in the resistor so that's easy enough to replace.
No, it's common to see different readings in different directions with in-circuit testing because there are semiconductor and leaky electrolytics involved. So if you saw 10K in one direction and a lower value in another, I wouldn't be too concerned. It is, however, not possible to have a higher value than the resistor in any direction in a fully de-energized circuit. I'm not sure why a 10K resistor would go open at random. Is it the correct part? -
That makes a lot of sense.
The part is correct. It may be that it is out of spec (less likely) or that I didn't get it soldered properly and hence the weirdly high readings (more likely I think). This is my first smd project so still on the steep part of the curve for working with smt. I plan to pick it up again in a few hours and replace it altogether.
Thanks for the feedback! -
Both gain stages are now working very well... I get a very clean sine wave with the correct amount of gain at TP3 and TP4. I replaced R11 with a new resistor even though it tested good out of circuit and now the circuit works perfectly...up to the gain stage. Strangely, the replacement R11 tests at 10k in circuit so I can only guess there was some problem with the first solder job.
On to the tone stage!