The Passinwind Open Source Preamp

[bowthing]

I agree that the OPA145 is an engineering breakthrough for bass audio. But there are several really good candidates that are only slightly in the OPA145/2145's shadow.

Using the noise test feature of LTspice I compared a bunch of op amps with 2 different circuits.

Circuit 1 is inverting, no gain, 200K load on the input, 100pF feedback cap and 470 ohm output load resistor. TI calls it Riso and for large capacitive loads it really helps stabilize the op amp from oscillating. TI usually suggests 50-100ohms, but actually for the OPA145 it needs to be higher. I have been testing with 1200p and 2000p cap loads as the worst case. Since I own a Monster cable 20' long that has > 1100p, I am trying out big numbers as a worst case.

The 4 numbers are noise at 20Hz, 1KHz and 10KHz in nV and the last number is total noise taken over 20-20KHz in uV. I am reporting only the best ones, plus two old favorites for comparison:

OPA145 82,80,50, 7.9uV
OPA180 81,81,51, 8uV
OPA202 83,83,52, 8.1uV
LT1012 85,84,53, 8.3uV
LT1097 85,84,53, 8.3uV
LT1112 85,84,53, 8.3uV

LT1351 156,129,81, 12.7uV
LM4250 136,135,110, 16.6uV

Circuit number 2 is non-inverting and this is inherently quieter. Still provides a 200K input impedance, same Cf and Riso.

OPA145 17,12,12, 1.7uV
OPA180 17,13,13, 1.8uV
OPA202 18,14,13, 1.9uV
LT1012 21,17,17, 2.4uV
LT1097 85,84,53, 8.3uV
LT1112 85,84,53, 8.3uV

LT1351 35,17,17, 2.4uV
LM4250 54,53,66, 9.9uV

My takeaways: LT1351 is fine for a low source impedance, lousy for a buffer or an MM 2-band EQ.
OPA991 is a nice product but I left it out because it uses too much battery. Comparable noise to 145, 180.
OPA202 is awesome, uses a little bit more battery than the OPA145 and laughs at capacitive loads.
OPA180 is the also-ran that would be great except that the OPA145 is even a bit better.
LT1012, LT1097, LT1112 must share some architecture to have such matching noise figures, or maybe someone just cut and pasted the same parameters into the 3 Spice models(?) BUT they all use a lot less battery that the OPA products shown, so they would be a good choice for a circuit with a lot of amps.

p.s. Did I emphasize enough that the OPA202 is really great? Hope this helps!

What configuration are you using to test capacitive loads? I can't get the 145 to oscillate.

 

[Bemo9]

Hey Bowthing, Have you used this yet with any of your G&L's? I'm really interested to know how it does with the MFD's. Looking at maybe getting a L2000 down the road. Sorry for the slightly off topic question.

 

[JayGunn]

What configuration are you using to test capacitive loads? I can't get the 145 to oscillate.

My investigation started with this LTspice setup.

Read the datasheets especially from TI carefully and you see that they usually suggest Riso=50 ohms. Try it. Pure oscillation. Here i move it up to 1K and you see borderline stability aka ringing. This schematic is the worst case since Cf is not connected. TI suggests Cf in the 5p to 10p range for many of their op amps. That's for finding the optimum 'zero' to balance out a 'pole' in the very high frequency range of the gain-bandwidth graph. This is stuff that has very good supporting literature, white papers and so on, from TI and Linear Technologies (and its parent Analog.com). Not practical to explain poles and zeroes here. Z-z-z-z-z. But worth reading if this interests you.

But the point is that the Cf=100p slows down the op amp (for a bass preamp who cares?), but without Cf we're toast with this application of the OPA145. Leo used Cf=100p or 120p and Rload=1000 for the LM4250 in the MM 2-band preamp. Those high numbers will make almost any op amp work okay. But that's unnecessarily heavy-handed for some and essential for others. Note that this is only a 600p load. The 23K is roughly a 25K volume control in parallel with the amp's presumed 1megohm input impedance.

Subsequent tests increase Cload to 1200p (likely) and 2000p (unlikely but possible) and determine what are the useful values of Rload and Cf to eliminate oscillating and excessive ringing. You can't settle for a just-good-enough solution since LTspice uses a simulated cap and resistor that work ideally. Your 100p cap in real life could be 95p or 105p so you need some safety margins. I suggest limiting the ringing to about 8%. Around 22% you are on the verge of oscillation.

I gave up on some useful-sounding op amps. In this set of tests I tried out the OPA191 but I wouldn't consider using it. Ditto the OPA180, the LT1637 and some others you're not likely to find interesting. Testing is free, eh? And fun. Hope this helps.

 

[Passinwind]

My investigation started with this LTspice setup.
View attachment 4819125

Read the datasheets especially from TI carefully and you see that they usually suggest Riso=50 ohms. Try it. Pure oscillation. Here i move it up to 1K and you see borderline stability aka ringing. This schematic is the worst case since Cf is not connected. TI suggests Cf in the 5p to 10p range for many of their op amps. That's for finding the optimum 'zero' to balance out a 'pole' in the very high frequency range of the gain-bandwidth graph. This is stuff that has very good supporting literature, white papers and so on, from TI and Linear Technologies (and its parent Analog.com). Not practical to explain poles and zeroes here. Z-z-z-z-z. But worth reading if this interests you.

But the point is that the Cf=100p slows down the op amp (for a bass preamp who cares?), but without Cf we're toast with this application of the OPA145. Leo used Cf=100p or 120p and Rload=1000 for the LM4250 in the MM 2-band preamp. Those high numbers will make almost any op amp work okay. But that's unnecessarily heavy-handed for some and essential for others. Note that this is only a 600p load. The 23K is roughly a 25K volume control in parallel with the amp's presumed 1megohm input impedance.

Subsequent tests increase Cload to 1200p (likely) and 2000p (unlikely but possible) and determine what are the useful values of Rload and Cf to eliminate oscillating and excessive ringing. You can't settle for a just-good-enough solution since LTspice uses a simulated cap and resistor that work ideally. Your 100p cap in real life could be 95p or 105p so you need some safety margins. I suggest limiting the ringing to about 8%. Around 22% you are on the verge of oscillation.

I gave up on some useful-sounding op amps. In this set of tests I tried out the OPA191 but I wouldn't consider using it. Ditto the OPA180, the LT1637 and some others you're not likely to find interesting. Testing is free, eh? And fun. Hope this helps.

I assume you've measured all of this in the real world to confirm that your model is correct? I've found LTspice S/N models to be way off base, FWIW. To the point of not being all that useful actually. But as always, I could easily just be doing it wrong!

 

[JayGunn]

I assume you've measured all of this in the real world

Wish I had a high end analyzer...What I have is an old 50mHz 'scope so I can see some oscillation problems, but I am naively perhaps assuming that LTspice and the models are useful. Actually I do not trust the models on some older parts (e.g. TL061), but I have been assuming that the much more detailed models of the newer TI parts tell the truth. And TI reinforces that by using Spice simulation in some of the TI papers and training videos(https://training.ti.com/ti-precisio...s?context=1139747-1139745-14685-1138805-13952) about the poles they are correcting with appropriate zeroes and vice versa.

I hope in the next few weeks to do some real world testing of half a dozen op amps in a partially SMD version of the MM 2-band preamp. If I do what works in simulation and it works in the real world, then I won't have proven that the simulation is correct, but I'll sleep well. But if I can also induce oscillation when the simulation says it should happen then I'll be able to say with some more confidence that the simulations were worthwhile. That's a lot of work but I really do hope to pursue this to a useful conclusion.

J

p.s. This stability testing is quite different from S/N measurements, I have found some anomalies in S/N simulations too. One example is I recall is comparing noise of the TL061 to that of the TLE2061 which is the "improved" version. I was disappointed to see that the improvement claimed in the datasheet was in Gain-bandwidth (GBW) but the voltage noise density is the same. Yet in simulation of a circuit with low source impedance where the voltage noise is dominant, the TLE2061 is a lot quieter, whereas the specs suggest they should be the same. The model for the TLE2061 is MUCH more detailed and presumably more accurate.

p.p.s. Does anyone besides Charlie (@Passinwind) and me care about this stuff? For any other people who find this electronics stuff entertaining, that TI series of training videos about stability is quite good.

 

[Passinwind]

p.p.s. Does anyone besides Charlie (@Passinwind) and me care about this stuff? For any other people who find this electronics stuff entertaining, that TI series of training videos about stability is quite good.

Jay,

There are often lurkers here who know a whole lot more than I do. I get a sniff of who they are occasionally, you would definitely recognize some big names. By all means carry on, especially with results of the reality checks.

Didn'r realize that you're still messing around with the MM2B preamps, I moved along from that many years ago. The original design didn't have a series output resistor at all and that was my very first mod. I've never built one with the LM4250 and probably never will.

 

[Kinkh]

p.p.s. Does anyone besides Charlie (@Passinwind) and me care about this stuff? For any other people who find this electronics stuff entertaining, that TI series of training videos about stability is quite good.

Not that I'm a particularly "important" or active member on TB, but I do enjoy lurking around and reading stuff like this.
I may have temporarily stopped working with electronics, but they're still somewhat part of my job, and every new bit of info is useful...
So just keep up the good work at informing us lurkers!

 

[bowthing]

I'm interested! Love to dig into this stuff when I have time. I have built preamps with the 145 and no Riso (mostly non inverting) and I haven't seen (or heard) any oscillations, except when killing the power. I've tested with large caps and long cables, so I just assumed I was good. I guess I don't know how to induce oscillations! Thanks for all the info JayGunn! A lot to look at.

 

[bowthing]

The biggest problem I found with the 145 was the common mode voltage range. With a battery at 7V and biased to 3.5V, any positive signal is out of range (it is V+ -3.5V), so it starts distorting badly. I got around it by biasing much lower, and it works well. Kind of odd that they say it has a 4.5V single supply operating range.

 

[Passinwind]

Trying a different enclosure format and adding a new feature while I’m at it:

 

[Bemo9]

Oh that looks really good Charlie!

 

[12bass]

Will there be a WTFHPF?

 

[Passinwind]

Oh that looks really good Charlie!

I did one in a color scheme like yours too.

 

[Passinwind]

First WTHPF prototype is done and ready to make some noise tomorrow:

Everything looked as expected in bench testing, and a new board revision for the HPF section should be showing up here tomorrow.

 

[Bemo9]

Have to say Charlie I'm a little disappointed that a Washingtonian like yourself chose Oregon Duck colors....

 

[Passinwind]

Have to say Charlie I'm a little disappointed that a Washingtonian like yourself chose Oregon Duck colors....

The box goes purple in the right lighting, it could almost be a Husky theme!

 

[Passinwind]

First WTHPF prototype is done and ready to make some noise tomorrow:


View attachment 4835337


Everything looked as expected in bench testing, and a new board revision for the HPF section should be showing up here tomorrow.

Got in a nice long test session yesterday with passive and active fretless fours and an active fiver. I used a pretty aggressive voicing spec for this build that I may want to back off just a little for the open source version, but the HPF really adds a lot of versatility and I would highly recommend using it for pedal builds. The Gorva C65 is just a little too small for builds with a standard stompbox footswitch, but 1590BB, !590BB2, or the Gorva S90 should all be fine.

These are the new HPF boards, which at 2" x 1.08" should fit in a 1590A box OK I think. DigiKey's board fabricator sent me a couple of extras so I will probably try that right quick.

 

[Passinwind]

^ New HPF board is in final testing, so I'm hoping to finally get all the new board shares at OSHpark up in the next couple of days. At this point I do not really recommend building the original version, the new one is much much better and parts supply issues will bring too many compromises with the old one. If you're not yet comfortable with surface mount soldering, it might be time to get to work on that!

But there's a good chance that Pedal PCB will go with a new through hole version, and I will totally leave that up to the PPCB owner and the members of his pedal building forum to decide on that. I'll hand him the initial build spec this week and we'll see how things develop from there. I'll add as much or as little input as desired, but in any case I'm really looking forward to seeing a new design fork with someone else driving the bus.

 

[Passinwind]

I've started working on a new section in the WIKI: Second Version (2022) | TalkBass.com

Board share links are now active, as is the Mouser section of the BOM. It might be a good idea to wait a day or three before actually ordering anything though...just sayin'.

 

[Passinwind]

It might be a good idea to wait a day or three before actually ordering anything though...just sayin'.

Heh, that escalated quickly! A few potentially good opamps for pedal builds have just come back into stock at Mouser. I changed the BOM for the moment to call two I've worked with quite a bit in my amp builds: OPA1641 and OPA1642. I'll be ordering a few of those and a couple of sets of other ones this morning and doing new test builds by early next week. None of these chips are terribly expensive, so I'm keeping my fingers crossed that they work out.