bypassing fast opamps

[Pars]

I moved this from the build gallery thread cuz it was getting too far off topic. I am playing around with a Rotel 991 CD player and was going to replace the exisiting I/V stage with THS4031 and AD8599 buffers. My plan was to use Aries adapters and socket the opamps, with Pana FM 100uf caps as close to the V+/V- pins, and somewhere in the area of 1-3.3nf across V+/V-.

I had removed the stock opamps, so I put a std. Mill Max socket in one of the I/V spots, and put the Pana FMs on the bottom, going from pins 7-3 (V+) and pins 4-3 (V-). The board has a top ground plane which would be a better spot to run these too than pin 3, but I felt I could get the caps closer to the V pins this way.

I fired it up, and looking on my scope (Tek 465b), I don't see any signs of oscillation. Opamp does not get hot at all. It also doesn't work (most of the time).

When I measured it, the -in was ~0.8V and the +in was 0.85V. The plus input is pin 3, which is grounded. Odd.

I have a test CD with sine waves, etc. on it, so I put that in and put one of the stock opamps in. Nice sine on output. I put the THS4031 back in and it does work briefly, but the output then collapses and does not work after that.

Next plan is to remove the socket and solder the Aries adapter directly into the PCB. Probably put the Pana FMs on the top and use the ground plane.

Any further thoughts/ideas appreciated.

[grenert]

When I put those same chips in my Denon, I paralleled a 0.1uF C0G surface mount chip directly across the FM leads.

[grenert]

Oops, sorry, I used the dual THS4032. Still, parallel smaller bypass might be worth trying. Also, I used a larger rail-to-rail capacitor (2.2 uF polyester).

[Pars]

I didn't have time to get back to this until tonight. I put the Aries adapter right in the PCB with the Pana FMs right on the adapter pins on the board topside and soldered to the ground plane instead of pin 3. Still had the MKP 2.2nf right on the top of the chip from rail to rail. Seems to work fine. No sign of oscillation on the output, sine and square waves looked good and matched the stock chip that I put in the socket on the other channel.

There was some noise (<50mV) that I could see on the PSU pins, so I removed the adapter and rechecked with nothing in. Noise was the same, so it must be the PSU. I guess I am spoiled by sigma22/Gilmore supplies, but I think the noise is inconsequential. I'll do the other channel tomorrow night.

Can I get away with putting the AD8599 buffers on sockets (and adapters)? I haven't really looked at these or the datasheet for them.

Thanks for the suggestions, but I'm not really a believer in bypassing caps. The Pana FMs have good HF characteristics, and bypass caps can lead to other resonances. Plus the way this is implemented, I don't think there is any way I could fit them.

[luvdunhill]

I agree with grenert and I definitely wouldn't discount using a C0G bypass. This is even recommended by various chip manufacturers on datasheets.

[Pars]

Got both channels stable using the THS4031s on adapters directly into the PCB. I wound up putting the AD8599s for the buffer/SE->bal conversion on a pair of BrownDog adapters I had, and I used sockets for these. 100uf on each V pin to ground, 2.2nf MKP right under the socket between V+/V-.

I did some listening and liked what I heard so far. CD player is still torn apart for some additional stuff I am going to do.

The AD8599s seem to be running rather warm, verging on hot; I can put my finger on them but it gets uncomfortable. I didn't notice any oscillation, but I will look again. I may need to pull the sockets and put these right on the PCB as well. I don't think I can fit any COG ceramics bypassing the 100uf Pana FMs.

[Pars]

When I put those same chips in my Denon, I paralleled a 0.1uF C0G surface mount chip directly across the FM leads.

What C0G SMT caps do you (or anyone) use? At 0.1uf, the ones I have seen mostly run $5-$7 each, other than the Murata 81-GRM31C5C1E104JA01 at Mouser, which are reasonable. I had been restricting my searches to 1210 package; these are 1206 package which is why I had missed them before. These are 25V which should be sufficient.

I pulled and socketed the PCM63s yesterday and also pulled the sockets for the AD8599s and will put these right in the board instead. I think I may have a way of doing the SMT caps right on the pins, so will probably try this.

I do have some 0.1uf X7R 1210 package caps, so I may use these. For PSU bypass use, is the C0G versus X7R that big of deal?

[luvdunhill]

I do have some 0.1uf X7R 1210 package caps, so I may use these. For PSU bypass use, is the C0G versus X7R that big of deal?

Perhaps someone else would be more qualified than me on this, but the way I understand it X7R are fine for bypassing, C0G are more appropriate in say compensation or filtering circuits. The latter are "better" though. The difference is the dielectric. X7R have a higher dielectric constant, which results in more capacitance per volume. Thus, X7R are used where the "better" C0G capacitors cannot fit.

Also, search for NP0, perhaps that will yield more results.

[grenert]

What C0G SMT caps do you (or anyone) use? At 0.1uf, the ones I have seen mostly run $5-$7 each, other than the Murata 81-GRM31C5C1E104JA01 at Mouser, which are reasonable.

Those are the same ones I use. Yes, I can't believe how much other caps cost.

With my DAC, I soldered the THS4032 and AD8599 on Aries adapters directly to the board. I never tried using sockets, so I don't know if that would've worked for my setup.

[Smeggy]

oops

Edited November 30, 2009 by Smeggy

[cetoole]

I too use those Muratas.

[Pars]

I tossed together a layout for an adapter board with all caps on it, which would mount from the bottom:

ths4031_adapt.pdf

[luvdunhill]

I tossed together a layout for an adapter board with all caps on it, which would mount from the bottom:

It seems that there are a wider array of SMD polymer caps, perhaps it would be worthwhile to look into these? What's the pad in the lower right corner for?

[Pars]

ground

[Pars]

Put things back together tonight after socketing the PCM63s and replacing the 12 stock 1n4003 rectifier diodes with 11DA10 shottkys. I bypassed the Pana FMs on the Aries adapters with 0.1uf X7R 1210 package SMT caps (soldering those across the cap leads sucked... snapped 3 of them bending leads after they were soldered, leaving me short for the AD8599s). Removed the sockets for the AD8599s and soldered them on the BrownDogs into the board with Pana FM 100uf right on the pins.

Worked right off the bat. Took a look at the PSUs on the scope and remembered that the PCB has the ground plane split into multiple sections, which are tied together with the chassis. Once I put a few of the screws in, the noise spike I had been seeing (~50mV on +/- 15V supplies) was down to 12mV PP. My HP3468 was reading AC of 0.065 mV on both + and -. Listened to a CD (nothing critical) but it sounded good, though I had it up too loud. After the CD was done, I checked the AD8599s and they were cool (had been fairly hot before). Button it up tomorrow and it will be done for a bit. Discrete boards this winter, as well as clock. Will post a pic tomorrow.

[Filburt]

I find the murata np0 work better than x7r for high speed circuits but I guess YMMV. For through hole 100nF I use TDK caps from digikey.

Glad you got it working, apparently. Any particular reason you switched to browndog?

[Pars]

I find the murata np0 work better than x7r for high speed circuits but I guess YMMV. For through hole 100nF I use TDK caps from digikey.

Glad you got it working, apparently. Any particular reason you switched to browndog?

I didn't have any np0 so I used what I had. I only bought 2 of the Aries adapters and used them for the THS4031s... I figured the AD8599s would work OK on BrownDogs which I already had. I may still etch some adapters I laid out which have the caps right on the opamp pins, and a ground plane on board. I don't think I could do these the way Aries does with the pins below the surface though, so the opamp would have to be mounted transversely like the Browndogs.