What is the latest of this OP Amp by Scott Wurcer. I find it hard to see the final word and a schematic.
If I understand Scott Wurcer he is playing at a league of his own. And this tend to make Op Amp not very basic or simple. But neverless it must be good. Because the man has been doing this to earn his daily bread.
Inform me please of a schematic showing the complete Op Amp so far.
Thank you 🙂
If I understand Scott Wurcer he is playing at a league of his own. And this tend to make Op Amp not very basic or simple. But neverless it must be good. Because the man has been doing this to earn his daily bread.
Inform me please of a schematic showing the complete Op Amp so far.
Thank you 🙂
Hi lineup, I posted a .pdf a couple of weeks ago with my current schematic and description of operation. I forgot the post number.
To me it's not that complicated and I think there are benefits to having the comp cap not referenced to either supply.
Another contender is a FET based JE990 with modern components, this makes sense since many of the ones out there are based on that and it would be nice to put a complete tested version out into the creative commons.
To me it's not that complicated and I think there are benefits to having the comp cap not referenced to either supply.
Another contender is a FET based JE990 with modern components, this makes sense since many of the ones out there are based on that and it would be nice to put a complete tested version out into the creative commons.
Last edited:
I agree with both of the above. Quite a reasonable complement of parts for the purpose and the performance.
I suspect Guru Wurcer will Muntz loadsa stuff once he gets his hands dirty with the input stage of SW-OPA.
If I may make a teensy contribution .. You only need a single diode D1 on the i/p to avoid 'output reversal' on severe overload.
I'm using Cordell's 2n4401/3 models cos they give the worst results both for SW-OPA and my hacked FET990. Also, the real things will give 3.162Vp into 15R without bursting into flames .. at least for a short while 😱
This newbie is really impressed with LTspice on this project. Only suspicious of the i/p stage. Even with fas42's improved BF862, there's stuff I don't believe but I have relatively little experience with FET i/p stages. Maybe try some much simpler circuits with the same i/p stage.
I also don't believe the Vbe on fas42, Cordell, Wurcer & Staccatiss' c2240/a970, BC546/556, n4401/3 BJT models compared to 1n4148/1n914
I'd like to replace a diode connected BJT with 1n4148 in my current mirror. If my (and Scott's) experience is kosher, this shouldn't be a problem as 1n4148 will just show a little more series R. But the models show more volts on the diode connected BJTs. Extra leakage & noise dun matter here.
If I may make a teensy contribution .. You only need a single diode D1 on the i/p to avoid 'output reversal' on severe overload.
I'm using Cordell's 2n4401/3 models cos they give the worst results both for SW-OPA and my hacked FET990. Also, the real things will give 3.162Vp into 15R without bursting into flames .. at least for a short while 😱
This newbie is really impressed with LTspice on this project. Only suspicious of the i/p stage. Even with fas42's improved BF862, there's stuff I don't believe but I have relatively little experience with FET i/p stages. Maybe try some much simpler circuits with the same i/p stage.
I also don't believe the Vbe on fas42, Cordell, Wurcer & Staccatiss' c2240/a970, BC546/556, n4401/3 BJT models compared to 1n4148/1n914
I'd like to replace a diode connected BJT with 1n4148 in my current mirror. If my (and Scott's) experience is kosher, this shouldn't be a problem as 1n4148 will just show a little more series R. But the models show more volts on the diode connected BJTs. Extra leakage & noise dun matter here.
Attachments
Hi Lineup,
Mr. Wurcer's latest is at #1568, with a slight mod to prevent latch up at #1776. The circuit discription is at #1675. In case you are wondering for a fully complementary frontend, that frontend is at #454, (though the symbols for P channel JFETs were wrong.)
Besides, he also published a FET input version of the JE990 discrete OPAMP at #1510
Regards,
Edward Tam
Du.uuh! Didn't see that. Guru Wurcer's mod does everything my pathetic contribution does .. & more 😱
Gonna delete my post. 😡
DAMN! Can't delete old posts! Forever shamed!
Thanks for keeping track, this weekend is a little busy but I hope to try the basic input stage by next weekend.
Ok. I'm back in the land of OZ (USA). Got pcb? Got prototype ? It's been 5 weeks, gentle men. I've been all over the place and did a lot in 5 weeks. Whats the progress on the opa in that time? Thx-RNM
This is a hobby project, we spend what time we can spare on it. Hopefully it's not like Gibbs' boat in the basement.
I guess guilt doesnt work. How about - have to finish it becuase you are not retired yet? ["Retirement - never having to finish anything because everyday is Saturday"]
Would rude work? I'll try that next. 🙂
Would rude work? I'll try that next. 🙂
Nope, rude will just entice a response in like kind. I would hope the process has almost as much value as the outcome. There have been lots of recommendations for good discrete devices and I hope folks are picking up some nuggets from the design process.
Always welcome my friend, we should do an SY home invasion sometime. My last visit was ruined by Sandy.
I am enjoying the info and sim and all, also. FYI [Just pushing - gently - to keep it going at more than a snail's pace.]
Signed - impatient for more.
Signed - impatient for more.
Last edited:
JFET 990
I'm posting what is very near a final [*] design for a JFET 990. Such beasts have appeared in the past so what's different?
Guru Wurcer's recommendations on FET matching etc for SW-OPA are applicable.
Q2 is shown at the +ve rail cos I suspect it could be part of a dual diode with D3.
D1 is to prevent one manifestation of latching on overload.
This is actually a 2 stage amp. The i/p & current mirror are the Transconductance Stage.
C4 turns the VAS & Class B o/p into the single integrator which sets the OL bandwidth etc. This is Cherry's solution.
The input to the integrator is very sensitive and anything connected to it must be clean. Q16's Cob is unfortunately not clean as it's modulated by Vcb. The Q15 cascode takes it out of the picture.
In the original 990, the emitter follower feeding the VAS does something similar. Unfortunately, that topology requires a diode across the Miller capacitor to avoid latching which re-introduces the non-linearity.
Keeping the bases of Q10/15 at Hi Z (not connecting anything there)
But the Cherry capacitor C4 is referenced to the +ve rail which reduces +ve rail PSR. The 'cure' is injecting an identical compensating current via C5 into the current mirror. I've used this dodge in a Jurassic low feedback PA and so does Guru Wurcer in the AD797 datasheet.
Using TI-an1516.pdf methods
CMRR 126dB to 20kHz
-ve PSR 125dB to 20kHz
+ve PSR 129dB falling to 93dB@20kHz
C4/5 should be 15p for 10x & greater gain
47p for 3x
100p for 1x
Need to do more work on stability with reactive loads but preliminary sims show similar behaviour to AD797. R18=10R on the output seems to work well but further optimisation and better recommendations are possible.
Q9 allows Iq to be varied over a wide range and should be strapped & superglued to Q10/15.
Re-arrangement of the current sources is to avoid latching. C6 is there for the same reason rather than to improve PSR
Do I believe this SPICE sh*t?
As apart from the 15R results, the distortion products are mostly around -100dB, it is likely that this and SW-OPA will be as critical as AD797 and OPA627 for layout & decoupling. In many situations, layout & decoupling is the MOST important factor in noise & distortion.
But most of the improvements have been in the direction I expect from previous experience so yes, I believe it may have similar distortion to SW-OPA. Not so sure about CMR & PSR though.
More to come.
[*] in as much as anything which hasn't been built can be called final 😉
I'm posting what is very near a final [*] design for a JFET 990. Such beasts have appeared in the past so what's different?
- According to LTspice, it should have distortion into 600R up to 20kHz at SW-OPA levels. ie much better than JE-990
- one less amplifying stage so should (??) be more stable.
- Needs less Iq than SW-OPA for low distortion into 15R
- +/- 60V/us slew
Guru Wurcer's recommendations on FET matching etc for SW-OPA are applicable.
Q2 is shown at the +ve rail cos I suspect it could be part of a dual diode with D3.
D1 is to prevent one manifestation of latching on overload.
This is actually a 2 stage amp. The i/p & current mirror are the Transconductance Stage.
C4 turns the VAS & Class B o/p into the single integrator which sets the OL bandwidth etc. This is Cherry's solution.
The input to the integrator is very sensitive and anything connected to it must be clean. Q16's Cob is unfortunately not clean as it's modulated by Vcb. The Q15 cascode takes it out of the picture.
In the original 990, the emitter follower feeding the VAS does something similar. Unfortunately, that topology requires a diode across the Miller capacitor to avoid latching which re-introduces the non-linearity.
Keeping the bases of Q10/15 at Hi Z (not connecting anything there)
- avoids the evil Class B distortions enumerated by Self with his Lo Z drive and critical Iq settings
- maximises slew as nothing slows it down at that point
- reduces xover distortion to the minimum even with small Iq. eg This circuit has less distortion into 15R with Iq 7mA than SW-OPA with 10mA
But the Cherry capacitor C4 is referenced to the +ve rail which reduces +ve rail PSR. The 'cure' is injecting an identical compensating current via C5 into the current mirror. I've used this dodge in a Jurassic low feedback PA and so does Guru Wurcer in the AD797 datasheet.
Using TI-an1516.pdf methods
CMRR 126dB to 20kHz
-ve PSR 125dB to 20kHz
+ve PSR 129dB falling to 93dB@20kHz
C4/5 should be 15p for 10x & greater gain
47p for 3x
100p for 1x
Need to do more work on stability with reactive loads but preliminary sims show similar behaviour to AD797. R18=10R on the output seems to work well but further optimisation and better recommendations are possible.
Q9 allows Iq to be varied over a wide range and should be strapped & superglued to Q10/15.
Re-arrangement of the current sources is to avoid latching. C6 is there for the same reason rather than to improve PSR
Do I believe this SPICE sh*t?
As apart from the 15R results, the distortion products are mostly around -100dB, it is likely that this and SW-OPA will be as critical as AD797 and OPA627 for layout & decoupling. In many situations, layout & decoupling is the MOST important factor in noise & distortion.
But most of the improvements have been in the direction I expect from previous experience so yes, I believe it may have similar distortion to SW-OPA. Not so sure about CMR & PSR though.
More to come.
[*] in as much as anything which hasn't been built can be called final 😉
Attachments
Last edited: