I am at a loss to explain why this method returns MUCH worse results than the one I have always used, which follows the formula for noise of
ent=√(Vn squared)+(In×Rs)squared
where ent*= Total noise voltage; Vn*= Voltage noise*(nV/Hz); In*= Current noise*(pA/Hz); Rs*= Source resistance (Ω)
and then computing signal-to-noise ratio with
=-20*LOG10((signal level in mV/(((SQRT(POWER(Vn,2)+POWER(In*Rs,2))*141.35))*1/1000000))))
The 141.35 is the square root of the signal bandwidth of 19,980 (20 Hz to 20K Hz)
I'd say Bonsai's spreadsheet is b0rked. For one thing, the values that should be in the S/N (1k) column have ended up in S/N (100R), at least those are the ones consistent with the e_n,tot (1k) column. The other S/N columns look dodgy to me - the LME49990 with the highest current noise of the bunch mysteriously turns in one of the best results @10k. Not to mention that S/N should come out positive like that.
I can't tell what exactly is going on as I can't see what's in the cells, them being protected and all.
At 10 kOhms, my own calculator gives -107 dBV worth of output noise for the LME49990 (20 kHz BW), -114 dBV for the NE5534A if entered with the correct 3.5 nV/√(Hz), and -113 dBV for the OPA627. (Rf = 10 kOhm, Rg = 100000 kOhm, others = default.)
At 100 ohms, LME49990 = -132 dBV, NE5534A = -126 dBV, OPA627 = -119 dBV.
LM4562/LME49710 Vnoise in spreadsheet needs to be corrected as well (2.7, not 4.7).
I can't tell what exactly is going on as I can't see what's in the cells, them being protected and all.
At 10 kOhms, my own calculator gives -107 dBV worth of output noise for the LME49990 (20 kHz BW), -114 dBV for the NE5534A if entered with the correct 3.5 nV/√(Hz), and -113 dBV for the OPA627. (Rf = 10 kOhm, Rg = 100000 kOhm, others = default.)
At 100 ohms, LME49990 = -132 dBV, NE5534A = -126 dBV, OPA627 = -119 dBV.
LM4562/LME49710 Vnoise in spreadsheet needs to be corrected as well (2.7, not 4.7).
Last edited:
In general if you are worried about noise you have sufficient gain to push the signal up in level before the next stage can add more noise, so compensation isn't needed and the extra open-loop gain this affords helps control distortion. If you aren't needing to boost a signal above the noise and have low gain you'd switch from the NE5534A to the NE5532 dual opamp and get the benefit of the dual package.
You've omitted the Johnson noise of the source, √(4.Rs.kT) which depends on the resistive portion of the source impedance. The current noise has to be multiplied by the absolute value of the source impedance, not the resistive part, note, which matters for complex source impedances like MM cartridges.
ent = √(Vn^2 + (In.|Zs|)^2 + 4.Rs.kT)
Yes - Mark is correct - the spread sheet takes noise voltage and noise current and considers the generator resistance in calculating the S/N.
You need to enter the noise voltage and noise current figures for the opamp you are looking at into the table.
Once you consider noise current and source impedance the S/N of some very low noise bipolar input OPA's degrades very quickly - the 5534 being one of the exceptions.
For most of the 4562 bipolar types, best to use them in applications where Rsource is not more than say 5k Ohms. After that the FET types look better.
You need to enter the noise voltage and noise current figures for the opamp you are looking at into the table.
Once you consider noise current and source impedance the S/N of some very low noise bipolar input OPA's degrades very quickly - the 5534 being one of the exceptions.
For most of the 4562 bipolar types, best to use them in applications where Rsource is not more than say 5k Ohms. After that the FET types look better.
Last edited:
Do any of you use ne5534/32 in unity gain line stages or just impedance adapters at line levels instead of lm4562 or similar ? Well I dont use it and i tried them both ...I never use lm4562 for high gain stages, i go with the low noise old Japanese parts anytime in high gain stages because high gain stages automatically involve the use of some sort of transducers which have limited slew rates by their nature .
Line level unity gain circuits are the circuits where the difference is audible and striking and just do it guys, you'll never use ne55xx in a unity gain stage against lm 4562 in your life!
Line level unity gain circuits are the circuits where the difference is audible and striking and just do it guys, you'll never use ne55xx in a unity gain stage against lm 4562 in your life!
Funny how the original question about LM318 thd turns into a discussion about noise & other designs.
Everyone suffering from ADHD these days? 🙂
I did go looking for thd graphs, I found on the LT318 they show noise curves but no thd numbers either. LT318 is still in production, but not the graded "A" version.
Everyone suffering from ADHD these days? 🙂
I did go looking for thd graphs, I found on the LT318 they show noise curves but no thd numbers either. LT318 is still in production, but not the graded "A" version.
I always found the 318 to make a great oscillator unless treated with TLC and good layout. Modern chips seem easier to use.
The 318 was used IIRC in one of the earlier big Musical Fidelity Amps with the mosfet outputs.
Nice and fast for the day and external comp meant you could get up to all sorts of tricks. Modern OPA’s run rings around it though. Spoilt for choice nowadays on just about any parameter you care to choose.
About the only oldy that can still hold its head high is the 5534!
Nice and fast for the day and external comp meant you could get up to all sorts of tricks. Modern OPA’s run rings around it though. Spoilt for choice nowadays on just about any parameter you care to choose.
About the only oldy that can still hold its head high is the 5534!
I'm surprised LM833 fared so well!
I've used NE5532, LME49710/20/LM4562 in the same circuit without changes, and audibly find no difference! Testing to these levels and understanding the implications in such depth is currently beyond my ability🙂
In fact, I have some LT1115 (If I recall) that are reputedly quieter, but I've never had reason to try then, yet.
I've used NE5532, LME49710/20/LM4562 in the same circuit without changes, and audibly find no difference! Testing to these levels and understanding the implications in such depth is currently beyond my ability🙂
In fact, I have some LT1115 (If I recall) that are reputedly quieter, but I've never had reason to try then, yet.
What, no difference, there must be something wrong with your ears & brain? 🙂 Need to go to the audiologist and tell them that you can not tell the difference between -100 and -120dB of thd 🙂 Say other's somehow can hear differences, where I can't 🙂
I've never seen distortion measured in such a fashion---ppm---??? Could I then infer that the 318 is 38.5 db worse than the 4562? (2.1 ppm/0.025 ppm). Lm4562 is -130 db THD (0.00003%); so the LM318 is -91.5 db THD??
Not the input stage, definitely not. Would like to repeat that I have measured consistently <=0,00015% distortion in a composit amp structure, with the LM318 as input stage. The output stage of it could be less performant.
The text doesn't say what the measurement frequency was, so I'm almost inclined to think it's DC. This would give parts with high DC open-loop gain an advantage, i.e. typical precision opamp topologies (including NE5532 and the like, but also OP07). If you were using an early RF opamp like the LM318, maximum DC OLG probably wasn't one of your primary concerns. It's not much different for audio.
I'd stick with tests à la Samuel Groner.
I'd stick with tests à la Samuel Groner.
You are absolutely correct! Thanks for spotting my error. So I just need to add the 4.RskT. Should that not be the SQUARE ROOT of 4 * K * T * Rs?
BGW used LM318 in all PA amplifiers Model 100 / 250 / 320 / 500 / 620/ 750
Hum and Noise level > 106 db in this amps
I have hear this amplifiers >20 years live in different Club Installation,
powerful strong clean bass responce and clarity of sound in mid and highs
have beat a lot of competitor products.
The LM318 in BGW PA Amplifier input stage had excellent sound in live performance, is rugged
and reliable in overload conditions, you can believe it or not.
Last edited:
Well, I remember that, in the early-to-mid 70's, all the rage was about IC SPEED. So, the LM318, with its outstanding 70v/µSec, was the hot ticket, and was used in the ATR-100 tape recorder audio electronics. The NE5534 had not come out by then; otherwise they would have surely used that.
It is measured with a 80Hz triangle wave, thus purely linearity.
Of course, linearity impacts directly the THD, but they are not completely interchangeable numerically
What can i expect is ound quality
if exchange in the original circuit LM318 with NE5534a ?
Could it be audible in live performance ?
Later design using NE5534a,
the schematic show to exchange LM318 with NE5534a,
I have to remove 1 x Overcompensation resistor and capacitor in series 20pf + 4,7K Pin6 to Pin1
and change 2 resistor value from 1 K to 4,7 K
Last edited: