JE-990 Help

[Manifold]

Hello everyone,

I'm new here and was wondering if someone might be able to offer me a little assistance on a JE-990 Circuit that I've been working on. I decided to try and learn more about discrete opamps so I thought the 990 was a good circuit to try and build. Anyway, I built one on perf board and to my surprise it worked perfectly the first time. So, I decided to draw up a PCB for another one. I did and now the PCB version is giving me all kinds of trouble.

The problem I'm having is that when I set the circuit up in a voltage follower configuration with all of the components in I get an output that is about 1/10th the input and is offset by -15V. When I take out CR1 & CR2 the output is equal to the input but is still offset by about -2V. On my prototype that I made, when I set it up in a Voltage follower configuration, it's at unity gain and there is no offset. So I was wondering if anyone had ever run into anything like this before. As far as I can tell there are no leads touching each other and no traces touching either and everything seems to be connected properly. I've A/B'ed it with the perf board one I did and it looks the exact same to me. Any toughts or help would be greatly appreciated. Thanks.

Brian

 

[PRR]

Since you have a working model, and a model with a gross problem, set them side by side, no-signal (grounded input) and go around with a DC voltmeter. I'm sure you will find some DC voltage that is way off.

However I'm not sure the 990 is fully happy at unity gain. Wire it for gain of 2.

The bit about CR1 CR2 should be a clue but nothing leaps to mind yet. Do look closely at the Q10 R4 Q3 R5 area, because if things get "better" without the input diodes then something may be very wrong up there.

 

[pstamler]

Sounds like a fried active device, possibly (half of?) the input pair. Try substituting another unit and see if the problem goes away.

Peace,
Paul

 

[Manifold]

Thank you both for the suggestions. I will try both of those and see what I come up with.

Brian

 

[Manifold]

It worked. Thank you both. It was both a fried transistor and it was one of the ones PRR mentioned. So thanks again. The advice is greatly appreciated.

 

[Samuel Groner]

However I'm not sure the 990 is fully happy at unity gain. Wire it for gain of 2.

Is there a simple way to make it more stable at unity gain?

What about skipping L1/L2, making R1 and R2 bigger (and perhaps modding R9 and C1)? Would make it noisier but who cares @ unity gain?

Samuel

 

[clintrubber]

W.r.t. unity gain & stability, the '990 has then less than 40 degrees of phase margin (and that might be quoted without additional cap-loading, don't know). Why not keep the opamp itself as is and apply 'measures' externally ? Like configuring for gains of 6dB or more(+additional attenuation if you need the unity gain) or add a cap across the feedback-resistor.
More about stability etc in the Jensen-990/AES-paper. You have it ? If not, you can send a request to the nice people there and a few weeks later you have it in your mailbox.

Please keep us informed about your PCB-design & plans - as I understood it people experimenting with it had problems there and it sounds like you have a good layout.

BTW, there are also layout-info/suggestions in the mentioned paper. I guess they don't mind sharing a tiny bit (if you don't know it already): 'node common to Q2,3, Q5-base, CR4,R9 is Hi-Z, so keep area as small as possible' & 'Q8,9 draw much current so connect to supply terminals directly' - both will be no surprise though).

Bye,

Peter

 

[Manifold]

Thanks for all of the suggestions. I actually don't need this opamp to run at unity gain though. I was just testing to see if it worked properly by setting it up as a voltage follower, because it was quick and easy. I didn't realize though that that opamp was not very stable at unity gain, which I find interesting. I did get mine working though. After PRR mentioned that it might not be stable at unity gain, I configured it differently. I set it up similar to an API 312 and it seems to work well.

Also about the PCB. I will definitely keep everyone up to date on the lay out, although I must warn everyone that I'm not trying to fit the PCB into a 2520 footprint. The board I did is about 2" x 2". I decided that I wanted a 312 clone and didn't need to have it fit into 2520 footprint so why not make the board a little bigger and give myself a little more room to work with and also that allowed me to make the traces a little bigger. I was trying for a while to do that, but kept running into problems, just as everyone else seems to.

Thanks again to everyone for their idea's and help.

 

[clintrubber]

Also about the PCB. I will definitely keep everyone up to date on the lay out, although I must warn everyone that I'm not trying to fit the PCB into a 2520 footprint.

Note that this is not meant to clone peoples layouts or other infringements on ip, but you're aware of this thread ? Again, I'm not suggesting this to copy stuff, but since pics of general orientation are there, I don't think it's unethical to bring that thread to your attention. Again, please let me know if it's not OK.

Oh yes, the thread :wink:

http://www.groupdiy.com/index.php?topic=5439

 

[Manifold]

No, it's quite alright that you mentioned that thread. I actually did see that thread earlier. I might try and get the PCB to fit the 2520 footprint later, but right now I wanted something that would work and not something I would have to keep trouble shooting in order for it to work. I've been working on this for a while and I started off trying to fit it into the 2520 footprint, but like I said kept running into problems, so I decided to make it larger for now and then maybe later go back and try a smaller one. Thanks for the ideas though.

 

[omeiko]

There is actually one working je-990 in 2.5.2.0- footprint floating around. I've build one with bd139/bd140 and it seems to be stable. I skipped the inductors because I couldn't find them easily. I don't know how it performs against the original one but it works and thats enough for me.

Here it is:

http://www.groupdiy.com/index.php?topic=1401&highlight=je990

-Okko

 

[PRR]

> Is there a simple way to make it more stable at unity gain?

I would not even attempt it.

The 990 comes from a tradition where a very few amplifiers do a lot of work each. Amplifiers are expensive (or used to be, before the 709 came down in price). You used amplifiers to AMPLIFY. Unity-gain... why would you want that?

The 990 started as a project to use new silicon with a very low-impedance (low-flaws) input transformer. In an era and place where hot condensers and loud acts were rare. So it would only be used with significant voltage gain.

Adjustable compensation is a royal pain. With fixed compensation, you can't get best results over a wide range of gain. If compensated for very low gains, the high-gain performance is poor, and if compensated for high gain it is unstable at low gain. Deane did a great job of compensating for a wide range of gain. But it gets awful marginal at unity gain.... that was not considered very likely.

You can build a good "unity gain" amp with a 2:1 resistor divider and a gain of 2 amplifier. This is often a happier plan than just strapping the output to the input. It runs at noise-gain of 2, so even if (as is usually true) it is kinda marginally stable at unity gain, it will be stable. You usually need an input resistor to ground for DC reference, and often a series resistor to protect against overdrive blow-up. The 2:1 voltage divider is low-impedance than an open input, but in audio it is usually the capacitive impedance that limits us, usually to under 50K. A pair of 22K input dividers gives 44K input impedance. It present 11K impedance to the pin, but if you keep that node tight that is not a problem.

> setting it up as a voltage follower, because it was quick and easy.

Seems reasonable. But only since the 741 came out. Back in the Old Days, you'd never assume an op-amp could be wired as a straight follower. The B-B tube opamps didn't even suggest you use the non-inverting pin: everything was inverting. The "+" pin was often hard-wired at the socket to a pot and mercury battery to trim offset. Note that an inverting opamp will rarely be run at noise-gain less than 2.

While it appears from Clint's remark that a gennie 990 is stable at unity gain unloaded, I don't know if a fresh-built reproduction on a new layout will be stable. And if it isn't, it may whine at 50MHz, too high to see on the scopes we usually have on an audio bench. If the oscillation is asymmetrical, all you may see is an odd DC offset and screwy output, not the MHz oscillation. That's why I suggested working at "some" gain.

Also: stuffing some resistors around it makes it harder for your signal generator to just blast-through some shorted junction, and produce "output" that isn't really amplification. When Manifold said the output was 1/10th of the input, I suspected something like that.

 

[Manifold]

Thank PRR for the info. It's always great to have a little more knowledge than you started off with. Thanks to everyone else as well who gave suggestions for everything.