Discrete Opamp Open Design

Disabled Account
Joined 2012
-- once corrected, the 2nd and 3rd harmonic were both at -89dB. Meaning, very close to identical distortion with badly mismatched JFETs ...

But, and a mighty big one, the offset is at the mercy of the BF862s. If perfectly matched, LTspice predicts an excellent DC null up to 75degC; above that temp the offset goes a-wandering; on the other hand, if using a severely mismatched pair, with gain of 100, the offset varies between -2.4V and 2.4V in the range 7deg to 47deg (default is 27deg)!

One thing, this has been an excellent learning exercise for me - thanks to all ... :)

Frank

So, you learned that matching the input diff pair isnt all that important to thd but for the lowest/best DC parameters matching is important. For this topology and using high neg fb. Thats great. Your sim shows you should be able to get away without using a dc servo on low gain apps. Just as Scott said it should :) -RNM
 
Last edited:
Disabled Account
Joined 2004
> The BF862 in my experience can not go beyond 4 paralleled devices with even good DIY assembly.

Why is that ? Oscillations ?
How about using a 0402 gate stopper next to the gate pin ??
I know there would be a noise penalty.

SMD inductors are not very linear ??


Patrick

I was speaking of true paralleling where you try to preserve the bandwidth, they are prone to oscillate at multi-Mz frequencies (probably parasitic Collpits mode). An MRF91 would oscillate at 2Ghz with just an inch of wire on the emitter and collector.

There is no gate current to speak of at audio frequencies, and certainly no DC current so I don't see the problem.
 
Disabled Account
Joined 2004
None at all. Just my demo of what can be done if you have a clear idea of what's important for good sound and ignore all Golden Pinnae considerations. :D

It's a specialist application, now only of interest to vinyl fanatics with low output moving coil cartridges. Probably still the lowest noise such device in the known universe. It's taken more than 30 yrs for someone to come close. What Mr. synaesthesia tells me is that certain devices that I thought were Unobtainium may actually be available. :)

.06nV preamps exist, even more exotic and specialized. Transformers are your friend.
 
Last edited:
We might be getting carried away with these thermal issues. A simple potting in the right stuff might be enough.

Diamond filled epoxy? :D


My direct experience is that potting sounds horrible for complex dynamic signals, like audio, where one requirement for expression/condition is never the same as the next. this, for micro signals and for macro signals. (microphone and speaker)

Clocking circuits, which require repeatability in environment, or tracking circuits of almost any kind (depends on the requirements of the tracking, obviously), they can and do benefit from potting. They can be micro adjusted back to a norm, after the given potting has settled to final cure. (months) If time is a concern, then a low level oven cure for a day or two can take the cure to +90% cure rate and then the micro norming can take place. Low level oven is critical, due to lattice structure linkage deviation under high speed cure rates. Deviations should be quite minor after that. This can get product out the door sooner, to cut the cycle shorter, for all the right reasons. Makes for expensive gear, but also good gear. Secondary concern is epoxy adhesion to substrate, in situ. Breakage or shift, on the job. Consider micro etching or even micro scarring (blast, sanding, whatever), for better adhesion.
 
Last edited:
Did some quick sims of the split/bootstrapped compensation (jcx). I found with an even split (2-200pF in series) with the center bootstrapped was an improvement, however the bootstrap had to be derived at the output of the emitter followers just before the output pair or the thirds were worse. The distortion numbers start getting silly.

Indeed the settling time is very compromised by this.

Hi Ken, I'm honored. There is very little signal current in any resistors but the feedback network.

Sorry Scott. Did I insult via default silence? I'm not aware of your 'full' background, but I'm aware of 'some' of jneutron's.
 
My direct experience is that potting sounds horrible for complex dynamic signals, like audio, where one requirement for expression/condition is never the same as the next. this, for micro signals and for macro signals. (microphone and speaker)

Clocking circuits, which require repeatability in environment, or tracking circuits of almost any kind (depends on the requirements of the tracking, obviously), they can and do benefit from potting.


The clocking and potting is tricky. Transient mass loading (mechanical) can cause deviation. Besides the given load response of the potting under optimal conditions. The nightmare of repeated brute force and ignorance as a testing/design regimen becomes the projected norm of that doorway of invention.

Edit: There is a simple way past the potting issue for tracking and clocking. But that is the part that is relatively unknown. Otherwise it would have been published as product by now. Perhaps it is, as I don't get out much. (read the trade rags/papers, and so on)
 
Last edited:
well I guess youre screwed then… are you removing all the cases on every device, resistor, cap, dac with some sort of hardcore acetone, so you are left only with lead-frame?

One should do this, at least once, so they can get a good handle on what the cases do the sonic presentation. This, of course, in a circuit that does not require the given cases for thermal reasons. To try one against the other, in a controlled environment. This, properly handled... constitutes a fairly decent single cause analysis.

I should go away again, for a while, before I'm seen as saying something stupid. To leave while the getting is good. :p. Comments on nano diamonds at a few dollars per pound is bad enough.
 
Last edited:
diyAudio Member RIP
Joined 2005
One should do this, at least once, so they can get a good handle on what the cases do the sonic presentation. This, of course, in a circuit that does not require the given cases for thermal reasons. To try one against the other, in a controlled environment. This, properly handled... constitutes a fairly decent single cause analysis.

I should go away again, for a while, before I'm seen as saying something stupid. To leave while the getting is good. :p. Comments on nano diamonds at a few dollars per pound is bad enough.
"Call me irresponsible..."

Science: Quantum oscillator responds to pressure | News | ECN Magazine
 
diyAudio Member RIP
Joined 2005
Both seem to be available as SMD duals also - MCH5908 and another. Vds is low at 15V max, but Idss goes up to 30 mA and Yfs above 30 mS. No idea about matching.
Yea! More duals. The voltage limitation is not serious for most of what we do in here, as cascoding is usually necessary anyway to reduce dissipation and voltage-variable capacitance effects.

The real puzzle for me is what is going wrong on the noise performance, based on the reported measurements. At those transconductances it should be significantly better.

I hadn't realized that On Semi had taken over some (all?) of the Sanyo discrete semis. I hope they keep making the fast low-C medium power bipolars, despite the almost negligible market for CRT deflection drivers.

EDIT: Once again they tie the channels together :( And the data are just reproduced from the single device, with the absurd "noise figure" plot that is practically independent of source resistance. And they only emphasize the savings in space, yet with the common channel connection would seem to be addressing conventional diff amps. Hmmmm.... If they pluck from adjacent positions on the wafer then we have something competitive or better than the 2SK2145 dual, especially if the noise gets a bit better. I wonder if those reported measurements might have had some contamination from resistor "current" noise, if they were made from the amplifiers themselves? If the 0402 parts were thick film this is possible.
 
Last edited:
Member
Joined 2002
Paid Member
This guy seems to get away with 8 in parallel.

Hand solder 0402 parts? - That's hard! Brad Plunkett made me do it a few times:eek:

If you look a little further in his article you will see that he uses ferrite inductors to stop oscillations of the BF862. I have 8 Toshibas used on a phono stage but you have to be careful with all the parts and layout. The BF862 is made to use at MHz.
 
Disabled Account
Joined 2004
Yea! More duals. If they pluck from adjacent positions on the wafer then we have something competitive or better than the 2SK2145 dual, especially if the noise gets a bit better. I wonder if those reported measurements might have had some contamination from resistor "current" noise, if they were made from the amplifiers themselves? If the 0402 parts were thick film this is possible.

I expect cluelessness on the part of the data taker. Is there some kind of resurgence in AM radio coming? :D:eek: How do we get them to decide that it needs matched pairs of complimentary JFET's?
 
diyAudio Member RIP
Joined 2005
If you look a little further in his article you will see that he uses ferrite inductors to stop oscillations of the BF862. I have 8 Toshibas used on a phono stage but you have to be careful with all the parts and layout. The BF862 is made to use at MHz.

Ah that is Popa's website! I knew it looked familiar. It was his design that I recalled as using 1uH L in the gate leads. It's well worth study, although I and Vogel have some issues with a few things.

I echo his comments about surface mount. It's easier than it looks, particularly if you are not quite settled on specific values of parts --- the changeout is much less destructive than trying to desolder leaded components in plated-through holes. But yes the better parts are usually very expensive.
 
diyAudio Member RIP
Joined 2005
I expect cluelessness on the part of the data taker. Is there some kind of resurgence in AM radio coming? :D:eek: How do we get them to decide that it needs matched pairs of complimentary JFET's?
If only more of us could read Kanji, I suspect a number of mysteries would resolve.

And yeah that AM thing! Is the EM environment in Europe somehow a lot cleaner in that band than it is here? I thought AM was all but dead, except for news stations. There are a couple of music stations around here, including a classical one, with the latter's fidelity just execrable. :grumpy:

As far as inducing a market for matched pairs, I guess cars need to get a lot longer so we could attempt differential diversity reception :D
 
Disabled Account
Joined 2004
It's a specialist application, now only of interest to vinyl fanatics with low output moving coil cartridges. Probably still the lowest noise such device in the known universe. It's taken more than 30 yrs for someone to come close. What Mr. synaesthesia tells me is that certain devices that I thought were Unobtainium may actually be available. :)

Out of curiosity which version the 2SXXXX or the BXXX, one runs at 500ua and the other at 1ma, I don't see 3dB noise figure from 5 Ohms at those currents?
I have seen some folks use power transistors with 1 Ohm rbb to get there even then at more current.
 
diyAudio Member RIP
Joined 2005
Just checked AM is alive and well in China that and how poorly FM propagates explains everything. Not to mention shortwave, I listen to Radio Beijing at home.:)

Aha! Yes, coupled with the relatively low amount of ~1MHz EMI in the more rural regions. Very interesting! By the way, the BF862s I'm testing are made in China, not sure where.