Bonsai, please post an audio circuit you have tried where NJM4562 gives worse results than LM4562. What's the lowest distortion level you can sensibly measure?
Are you prepared to do a little experiment for us? No accusations etc. Just fact finding.
NJM4562's topology is indeed ancient and is in fact the same as evil 4558. Almost all the NJM stuff is developed for audio. In many ways, their philosophy resembles that of another ancient OPA, the venerable 5532. So no pA bias currents.
But pA bias is rarely of importance in audio.
My apologies. Cos kunt reed rite en kont, I tot 20V/us = zillion 😡
Have you tried overloading LM4562 configured as a simple Sallen & Key HP filter at about 50-100Hz? I'm sorry if I wasn't specific enough in stating where I thought LM4562 was contraindicated. 😡
Bonsai, have you done any comparative tests on RFI immunity with different OPAs? Even the very lenient EU tests will show big differences with unbalanced domestic stuff.
I do have some small experience with 'bip input m/c head amps' including probably the quietest one in the known universe. 🙂 To take advantage of this of course requires RFI to be reduced to a sufficiently low level.
Perhaps I misunderstood. Are you recommending some other OPA for better RFI immunity? I'm assuming, one takes all sensible passive precautions appropriate to a high quality product.
PS Some nice amps on your website.
Last edited:
I got some good measured results on the X-Altra Mini line stage - you can see that in the PDF on my website. There's also a simple buffer stage with measurements as well.
The bias current story is important if you want to direct couple. The 5532/34 quotes circa 200nA and that means it has to use capacitive coupling for all but the lowest source impedances. I am feeding the LM4562 directly from a 10k log pot on my line stage. A 5532 in the same socket is noisy when the plot is adjusted, as would be expected.
WRT noise, yes in absolute terms a JFET front end is better, but when all is said and done once housed up, with the application of standard passive filtering and good interconnect cables, I think in most cases it's a non issue. Besides, if RFID is that bad that you have to plumb for a JFET input stage, you probably have other more serious issues to worry about.
Anyway, only my 2 pennies worth. Happy 'audioing'
🙂
The bias current story is important if you want to direct couple. The 5532/34 quotes circa 200nA and that means it has to use capacitive coupling for all but the lowest source impedances. I am feeding the LM4562 directly from a 10k log pot on my line stage. A 5532 in the same socket is noisy when the plot is adjusted, as would be expected.
WRT noise, yes in absolute terms a JFET front end is better, but when all is said and done once housed up, with the application of standard passive filtering and good interconnect cables, I think in most cases it's a non issue. Besides, if RFID is that bad that you have to plumb for a JFET input stage, you probably have other more serious issues to worry about.
Anyway, only my 2 pennies worth. Happy 'audioing'
🙂
Andrew, I see your X-Altra Mini One V2 is indeed the ideal vehicle and you have the requisite equipment to investigate this.
Would you be prepared to replace the LM4562s in one channel with NJM4562 and redo your careful measurements?
I quite understand if you find this an imposition and refuse as the OPAs on test really need to be soldered in place. DIP sockets are often evil when looking at differences of this order.
I accept that the DC offsets with NJM4562 may make this inappropriate for a final design (cos scratchy pots mainly) but it would be nice to know if THD, noise etc are up to LM4562.
Would you be prepared to replace the LM4562s in one channel with NJM4562 and redo your careful measurements?
I quite understand if you find this an imposition and refuse as the OPAs on test really need to be soldered in place. DIP sockets are often evil when looking at differences of this order.
I accept that the DC offsets with NJM4562 may make this inappropriate for a final design (cos scratchy pots mainly) but it would be nice to know if THD, noise etc are up to LM4562.
Last edited:
Sorry, I do not have access to the AP anymore which is a great pity.
I have no doubt the NJM would show higher distortion, and so would the 5532.
I have plugged in some 5532's and even a TL072 to have a listen. They don't sound quite as good, but I suspect there is some optimizing to do to get the best sound from them. That said, I personally do not believe the LM4562 sounds better because it's distortion is lower. It does just sound very good to my ears.
That's a shame. I was hoping for confirmation of NwAvGuy: Op Amp Measurements You'll note his gain staging is very similar to your X-Altra Mini One.
BTW, I'm enjoying your website. Your clear explanation of design choices is refreshing after the mumbo jumbo of certain pseudo prophets. 🙂
The requirement for strict RFI compliance was from Broadcast organisations. Certain applications, eg microphones, need even more paranoid RFI bomb-proofing than these. IM very HO, this is one of the factors that separate great microphones from the toys.
Television lighting was probably the most hostile RFI environments when I was pretending to do electronic stuff in my previous life. Today, mobile phones are probably the most evil. There are very expensive active 'monitor' speakers which announce to all & sundry that you have just received an SMS message. 😱
It is called 'lack of input protection' in the IC. We used to have to protect IC's with Zeners and diodes. I guess we forgot over time, as most modern IC's do not suffer this problem, in my experience.
I've damaged bipolar input pairs by taking the node on the far side of the substrate potential. Some of the chips I tested had no additional components diffused in to prevent forward biasing the isolation junction. At the extreme, forcing the input pair E-B junctions into avalanch always did some nasty permanent degradation to the device.
Some of the devices I tested would put the input bonding pad directly on a designed in junction. Kinda like putting a shottky in, but higher vf. Course, that made the pad a tad more sensitive to T/S wirebonding, cracks there would kill the device.
jn
What??? 20Vp-p is +/-10V at +/-15V power supplies!! For gain +1 !!!
And it was only a test circuit, not any "design". I wanted to push it, to see different topologies capabilities.
To me, the part does not meet specs. It is unable to withstand +/-10V 200kHz square at the input for gain +1. And it was RC filtered. Please read more carefully. Once again the test. It was tested by sine OR by square. Not both together.
Last edited:
Pavel,
I may miss something, also can't see the value of the pot in your cct, but if you put 10V pk into the B opamp it will see diff inputs of several volts, no?
jan
I may miss something, also can't see the value of the pot in your cct, but if you put 10V pk into the B opamp it will see diff inputs of several volts, no?
jan
It was the 1st opamp that has always gone. The input to 2nd opamp (with gain 2x) was divided 2x. The 2nd opamp served only to monitor output of 1st opamp and to drive RC filter (R1,C1) to get DVM DC reading.
One can see the comment "overall gain = +1" in the image of the test circuit.
One can see the comment "overall gain = +1" in the image of the test circuit.
Last edited:
PMA..
During the square wave test the device started increasing it's DC offset voltage, 5.3 volts DC at 200Khz.
I'd be worried about toasting the input pair at that drive level. The DUT is clearly no longer able to keep the negative input node close to the positive input. So I'd suspect sending one of the E/B junctions into avalanche.
jn
With this I completely agree, JN. But I wanted to know how the LM4562 behaves under severe conditions, as it was quite new part and it was very very highly regarded. I had still preferred OPA2134 and tried to find what's different.
PMA, sometimes just listening to different op amps seems to get more results. '-) That's what I did with the JC-3 phono preamp.
jcx
The lumped element model you have used assumes a prop velocity of infinity, I do not see any reference to length or prop speed. Does your software have the ability to include delays between nodes? For example, can you toss 2 nSec bidirectional delays in with the resistor elements?
I would suspect that some high speed PC board software modelling packages would allow the use of long striplines.
My point being, I can design two cables with different lengths, but yet a ten element lumped model would be identical between the two.
Richard,
Do you have a link to that 90 degree crossing angle inductance cancelling? I'l really like to read it. Your description of their design doesn't seem capable of inductive cancellation.
jn
Last edited:
- Status
- Not open for further replies.