LM6172 oscillating

[Saurav]

I replaced the OPA2134s in my active line level crossover with LM6172s. The power supply is regulated +/- 12V, I have caps from the opamp power pins to ground (don't remember the value right now, they're small). For some bizarre reason (probably because I read the multimeter wrong), I had 16Kohm series resistors at the outputs of the opamps (and I wondered why the dynamics seemed softer ever since I added the XO). Anyway, with the 16K resistors, the LM6172s worked fine. I opened it up to replace the series resistors with some 50 ohm Vishay/Dales that I bought for that purpose. That made the opamps starts squealing, and the scope shows it's oscillating. It also seems to be radiating a lot of noise, because moving my hand around changes things, and using an alligator clip lead to short the input RCA just made things worse, probably because it turned into an antenna - the scope showed oscillations at the input pins too. I tried adding bigger caps to the power uspply pins, but no luck. The circuit is on a breadboard right now, it's not on a PCB. The opamps are directly on the breadboard, no sockets. The decoupling caps have their leads cut down so they're flush with the board, not standing up on their legs.

Any ideas? Do I just need more series resistance? More/better/different decoupling? I'd bought a bunch of opamps to try, and I have some AD826s in there now, and they don't squeal (there does seem to be a bit more of low-level buzz though). I'd rather use the AD826 with 50 ohm resistors than the LM6172 if it's going to need a series resistor in the thousands of ohms. What's a maximum for the amount of series resistance I can try? I liked the sound of the LM6172 in the circuit, but I really don't think using it with 16K resistors is what I want to do.

Another thing I noticed is that these make an audible thump when I change volume levels on my transformer volume control (probably teh collapsing field, my switch is break-before-make). The OPA2134 didn't do that... or maybe the 16K resistors swallowed it.

Thanks in advance for any help.

Saurav

[macboy]

Try adding a small capacitor in parallel with the feedback resistor. This will add a pole at 1/(2*pi*R*C) where R is the feedback resistor of course. Try making the pole around 10 * 20 kHz, and if it stabilizes things, work up from there if you want to push it even higher out of the audio band.

You could also try adding a small cap between the two input pins of the opamp, which should help to remove RF. You can also add a cap from the signal input to ground (bypassing Rin, your input resistance) to help with RF. Finally, 50 ohms may not be sufficient to decouple the opamp from a very higher capacitive cable attached to its output. Try 100 or higher.

The thump may indicate a DC offset at the output. Once you stablize things, check for that.

[Saurav]

Wow, lots of good tips. Thanks.

The thump may indicate a DC offset at the output. Once you stablize things, check for that.

I'd checked for a DC offset with the OPA2134s, forgot about it with the new opamps. I'll look into it.

Finally, 50 ohms may not be sufficient to decouple the opamp from a very higher capacitive cable attached to its output. Try 100 or higher.

It's about 2-3 feet of video/digital coax cable. How bad can that be in terms of capacitance? The datasheet for the LM6172 said to start with 50 ohms, so I bought resistor values at 50 ohms or lower. I didn't expect I'd need to go higher. OK, I'll try some in the hundreds of ohms.

I'll try out your other suggestions too. Thanks once again.

[greyhorse]

Check out Jan Meier's reply on Headwize.com for hints:

http://headwize2.powerpill.org/ubb/s...2&srch=lm6171;

As well as his implementation of the LM6171 in one of his projects:

http://headwize2.powerpill.org/proje...=meier_prj.htm

[Saurav]

Excellent, thanks a lot. Now I have lots of ideas to try.

After looking at the schematic, I have some questions. Here's the schematic I'm working from (the high-pass section is what I'm dealing with here, I left the OPA2134s in the low pass section since I'm not too concerned with that):



Differences - mine are single opamp stage and therefore 2nd order. I got lazy and ran out of space and opamps; I should try 4th order some day. Anyway... the output and -ve input are tied together, and the feedback resistor goes to the +ve input. This worked fine with the OPA2134s, but doesn't work with the LM6172s. I'm not sure I can just add the feedback network + resistor ground on the -ve input that I see in Jan Meier's schematic? That doesn't sound right.

So... I can add the input RF cap, change the input series resistor, and move the feedback tap to after the output series resistor. I'd appreciate it if someone could tell me what exactly to do with the inverting input.

Thanks,
Saurav

[SY]

OK, I'll be a nice guy and not wonder why you're asking for trouble by using 100MHz opamps...

There's a bunch of stuff on this topic in the National Semi app notes. Or see the chapter on opamps in Bob Pease's "Troubleshooting Analog Circuits." He shows a nifty method of tailoring noise gain while keeping unity signal gain by attaching a series RC (or just a resistor) between the inverting and noninverting input.

There's also a lot of good stuff at the National Semi site on decoupling the output to better stabilize with capacitive loads. And get Pease's book.

You'll be very happy when you go to 4th order and use a Butterworth-squared (Linkwitz-Reilly). Makes crossover integration much easier.

[Saurav]

OK, I'll be a nice guy and not wonder why you're asking for trouble by using 100MHz opamps...

Just trying to be adventurous. This got lots of comments as being a good sounding opamp that was hard to tame, both of which sounded like desirable traits to me And that's why I got the others, in case I was unable to tame it.

Thanks for the references. I did look at the schematics/appnotes in the datasheet for this opamp. I'll go study their website more thoroughly.

[SY]

National's app notes have made many a fine evening's read. That's where I learned the basics of thermal design, noise theory, and feedback in solid state circuits. I'd give you some links for your specific app, but doing the search and multiple window thing with a 28.8 modem is a bear. Where's Phred when you need him?

Oh, one more thing: breadboards. BAD for stability, especially in the megahertz range,

[Saurav]

I figured there was a reason why all commercial products weren't built on breadboards But, how much better will a generic holes-in-a-grid PCB be? I think I'd need a well-planned out dedicated PCB to really make a difference?

I found one app note about building high frequency filters, and found some schematics that look exactly like what I'm trying to do. The biggest difference I see is that mine connects the output to the -ve input, while these have an Rf between output and -ve, and an Rg between -ve and ground. If I understand this right, these two would set the overall gain in the regular way, while the Rs and Cs on the +ve input would make the whole thing a filter. If I have that right, my questions are:

* Even though the datasheet says this opamp is unity gain stable, would it be happier if I gave it a modest gain factor of 2 or 3?

* I guess I should use resistors in the few kohms to build this voltage divider/gain setter network? How critical is the quality of the resistors here? Let's say, when compared to the series resistors in the input and output? Are they "as much in the signal path" as the series Rs?

* Given that I use a passive linestage and a 2A3 SET amplifier, would it be sacrilege to use these preamps to add a little gain to my system and improve the overall gain structure for the old LPs with low recorded levels?

Thanks for the help.

Saurav

[SY]

It may well take a properly designed, groundplaned circuit board to tame that opamp. A perfboard won't have as much stray coupling as a breadboard, but it may still not be good enough.

You can retain your topology by some noise gain tricks. What you'll do is put a resistor in series with the input lead to the noninverting input, then run an RC series combo from the output (tied to the inverting input) back to the junction of the series resistor and the noninverting input. The values I recall, which made sense for a 6dB noise gain, were 10K for the input resistor and something like 5K for the noise gain shaping resistor. Its series cap doesn't have to be too large, maybe .01uF or so. If your layout allows, this can stabilize things. You might also think about an LR parallel network in series with the output (like a power amp).

I'll see if I can dig up the app note number or a link to the RC trick. I first read about it in a Pease paper maybe 20 years ago, so I'm sure there's an official App Note.