Good point. Any idea what the cause might be ?
They are not showing the same figures for the OPA4140 anymore.
Patrick
They are not showing the same figures for the OPA4140 anymore.
Patrick
All parameters on the spec list are unsigned quantities i.e. taken to go either way. The Ib comp should be obvious from the collector current required to get the low noise (~500uA per side absolute min. but like LT1028 ~900uA in reality to account for reasonable rbb). Full LT1028 schematic is published (1986) ISSCC.
Back gate, classic DI has real "glass" isolating everything and modern SOI can have less good dielectrics, etc.
Scott,
I have been looking for a while and could not find any low-noise FET-input opamps from AD for +/-12V rails or above.
Any suggestions of what I might have missed, or any reason why there is no competitive product to say OPA627 ?
Patrick
I have been looking for a while and could not find any low-noise FET-input opamps from AD for +/-12V rails or above.
Any suggestions of what I might have missed, or any reason why there is no competitive product to say OPA627 ?
Patrick
AD4627/37, still working on reducing 1/f a little. 743/45 should still be in the catalog?
a question is, is it "bad" compared to any other low noise fet op amp solution - the AD743/745 has large input C, a few other fet input op amp datasheets warn of the input fet C Vcm modulation too - OPA134 has a paragraph but no plot
since no one else plots it the same we really don't know - many low distortion audio op amp datasheet plots are done with the super low Z "distortion magnifier" noise gain circuit that hides the effect
the "Difet" OPA627/637 is unusually good in this regard - Groner includes a test of this input impedance nonlinearity in his work SG-Acoustics · Samuel Groner · IC OpAmps
Walt had an article showing how composite amps can largely overcome the effect http://www.analog.com/static/imported-files/application_notes/742022599AN232.pdf
while Dimitri went even further: http://www.epanorama.net/sff/Misc/O...g to reduce distortion in op-amp circuits.pdf
depending on the bootstrap technique - you can get negative input Z over a frequency range if the bootstrapping has to be rolled off for to stability (if bootstrappinng both rails or the one with the Ccomp you can bootstrap away its effect too at unity gain)
Last edited:
Yes, much more expensive that 25$, but perhaps in any moment you can think in use discrete opamps:
Discrete Op Amps | Sparkos Labs.comSparkos Labs.com
http://sparkoslabs.com/wp-content/uploads/2013/12/SS3601_SS3602.pdf
Yes, yes, this here is a fine specimen. 🙂 2.7nV / rtHz @ 20KHZ
Its an old picture though. These devices now come with soldermask.
Noise in op amps is dominated by the input differential pair, and any RE degeneration resistors associated with it. What I have found is that as a general rule BJTs have lower noise than FET types. (And better matching, linearity, etc. The only drawback is that they have higher input bias current)
Beware of super low noise ( < 2nV / rtHz) devices. Noise performance like this is usually only obtainable by not using any degeneration resistors in the input differential pair. No degeneration resistors mean worse linearity. It also means a lot more loop gain that must be dealt with from a compensation standpoint. I guess the point is that you can trade off linearity for low noise if you are willing to make that sacrifice.
Another common practice is to shunt RE resistors with inductors. This isn't done in op amps because they can not put inductors inside of ICs - its more of a power amp thing. The inductive shunting effectively removes the RE resistors at audio frequencies, which gives lower noise, and then it "puts them back in" at higher frequencies, usually above the audio bandwidth. This is how the excess gain that was obtained by essentially not using RE resistors (or rather, inductively shunting them) is burned off to maintain stability. It also creates a pole in the loop response. This technique also has the affect of sacrificing linearity for noise, as the inductive shunting essentially removes RE degeneration within the audio band.
not so fast there - the higher global loop gain at low frequency with the L shunting the degeneration makes up for the "reduced linearity" by reducing the input error signal - by the same amount as using the gain as input degeneration - so no audio frequency linearity is sacrificed in the end
the general principle is the "best", "most linearizing gain" is global loop gain - dividing the same gain into local loops is at best as wash in negative feedback amplifiers - and sometimes the global loop higher gain will be much better
the cost is the higher order loop gain roll off to meet the same small signal stability and possible complication clipping recovery in nonlinear operation
Last edited:
Any suggestions for a package limitation?
sot23-5
20-20 hindsight says I should have gone to at least an soic-8
sot23-5
20-20 hindsight says I should have gone to at least an soic-8
This is only partly true, I think. ( correct me if Im wrong ) ITs partly true that the increased loop gain would help counteract some of the loss in linearity, but I dont believe that its a 1: 1 thing. Input stage distortion tends to rise (have a slope of ) 18dB / octave, and the distortion reducing affects of NFB tend to fall at a rate of 6dB / octave - or whatever the slope caused from compensation is. 6dB slope implying a single pole compensation scheme.
I think at really low audio frequencies, like in the hundreds of HZ or less, you are correct that the increased loop gain counteracts much of the increased distortion from the omission of RE resistors. However at higher audio frequencies, this distortion is worsening faster than NFB can counteract it. Especially since loop gain is falling by this point.
You mentioned something else - that compensation would have to be changed to maintain the small signal bandwidth and stability margins. Doesnt this mean that the extra gain picked up by the omission of RE resistors must be burned off by compensation - and therefore wont exist to counteract the loss in linearity ? Especially if the small signal bandwidth is kept the same ?
Ive looked at this in SPICE quite a bit. And Ive found that reducing REs results in more THD, and whatever amount of loop gain that was picked up by doing so was not sufficient to fully counteract the loss of linearity. Certainly at higher audio frequencies.
Might be the increase in distortion with lower or no RE's is because the input stage linearity is much less without the RE's. Of course, the loop gain is high, so the delta input signal is small, but if its approaching the linearity limits you will get increased distortion.
Another place to look is at your input capacitance and source resistance.
You may want to look at your comp if you see this happening - it will get worse at HF.
Another place to look is at your input capacitance and source resistance.
You may want to look at your comp if you see this happening - it will get worse at HF.
a complication is that most amps distortion is dominated by the output stage and it's varying current gain rather than input tanh nonlinearity - so you won't see any difference in most amp sims with more realistic output stages
with perfect Spice dependent source output the bjt diff pair tanh function can be seen in sim, the effect of input degen bypassing at audio with the inductor is clear - the distortion from the bjt tanh nonlinearity is much reduced at audio frequency by shorting the degen, applying the gain to the global loop - the 1st nonlinear term in tanh is x^3 so the % distortion goes down by ~ gain^2
with perfect Spice dependent source output the bjt diff pair tanh function can be seen in sim, the effect of input degen bypassing at audio with the inductor is clear - the distortion from the bjt tanh nonlinearity is much reduced at audio frequency by shorting the degen, applying the gain to the global loop - the 1st nonlinear term in tanh is x^3 so the % distortion goes down by ~ gain^2
Last edited:
A lot of good comments here so far!
I was designing my own op amp based phono stage some time ago, and I thought the LME49990 would be a good part for this purpose; low noise and low distortion according to the data sheet. I had PC boards made and built one. First I had to spend time getting the oscillations to stop and when I did I tried playing some records. Sound quality was excellent to me, but it sure was noisy when not playing a record! I didn't have to have my ear close to the speakers, either. I tried a Grado cartridge I had on hand and the noise level decreased significantly over using the AT440MLa. I got a crash course in the difference between voltage noise and current noise from this experience. So I went looking through the various TI offerings, and aside from the ridiculously expensive OPA627 I found the OPA1641 should work well from a noise standpoint. Using this part did in fact get the noise much lower, but the sound seemed harsh, for lack of a better description. I think it has something to do with the non-linear Cin of this type of device as described earlier in this thread. I am currently using LME49710s in this circuit, and they seem to perform well, although somewhat noisier than NE5534 or OPA1641; those two are about equal in terms of noise in this circuit.
The next thing I have planned is to have the third iteration of the board layout made and build several phono stages that will be identical except for the op amp used, and perform measurements and listening tests on them. I have OPA1641, NE5534, LME49710, and LME49990 on hand for this.
I was designing my own op amp based phono stage some time ago, and I thought the LME49990 would be a good part for this purpose; low noise and low distortion according to the data sheet. I had PC boards made and built one. First I had to spend time getting the oscillations to stop and when I did I tried playing some records. Sound quality was excellent to me, but it sure was noisy when not playing a record! I didn't have to have my ear close to the speakers, either. I tried a Grado cartridge I had on hand and the noise level decreased significantly over using the AT440MLa. I got a crash course in the difference between voltage noise and current noise from this experience. So I went looking through the various TI offerings, and aside from the ridiculously expensive OPA627 I found the OPA1641 should work well from a noise standpoint. Using this part did in fact get the noise much lower, but the sound seemed harsh, for lack of a better description. I think it has something to do with the non-linear Cin of this type of device as described earlier in this thread. I am currently using LME49710s in this circuit, and they seem to perform well, although somewhat noisier than NE5534 or OPA1641; those two are about equal in terms of noise in this circuit.
The next thing I have planned is to have the third iteration of the board layout made and build several phono stages that will be identical except for the op amp used, and perform measurements and listening tests on them. I have OPA1641, NE5534, LME49710, and LME49990 on hand for this.
I really don't believe such commentary - doing the numbers we see the OPA164x datasheet distortion graph is rising to 1 ppm @ 1 kHz from below that, reaching 10 ppm at 10 kHz, 3 Vrms Vcm for 600 Ohm series R
some MM carts do have Rseries in that range, most not even 5x more
the parasitic C modulation is a smooth function of Vcm, the distortion is expected to be low order if you don't get within a few Vt of the supplies
no MM cart has 3 Vrms output, < 100 mV if the stylus is actually maintaining contact with the groove - so the op amp input C modulation caused distortion will be in the single digit ppm even stacking all factors
the predicted level of input Vcm distortion from the OPA1641 isn't likely to ever rise to an order of magnitude (or 2) below the record groove noise floor with even few kOhm Rseries MM phono cartridges
you may hear a difference - it may persist in a blinded test - but the stated cause and magnitude of the effect doesn't jibe
Last edited:
you could add in another factor of 2 or 3 if you believe the electrical Q is that high
the loading C is supposed to be chosen to give flat frequency response after all - are phono cart manufacturers compensating even 10 dB of mechanical system roll off?
in any event we could recalculate with the 47 kOhms - always going to be there for MM preamp input
my numbers argument has a good order of magnitude to spare before the groove noise floor and we start arguing if phono playback of high amplitude, high frequency cuts ever manages say -80 dB 2nd,3rd harmonic distortion from stylus, cantilever, suspension tracing/angle geometric effects
while it seems silly for such low Vcm - we could still bootstrap away the Vcm nonlinearity with composite op amp circuits we both have explored
the loading C is supposed to be chosen to give flat frequency response after all - are phono cart manufacturers compensating even 10 dB of mechanical system roll off?
in any event we could recalculate with the 47 kOhms - always going to be there for MM preamp input
my numbers argument has a good order of magnitude to spare before the groove noise floor and we start arguing if phono playback of high amplitude, high frequency cuts ever manages say -80 dB 2nd,3rd harmonic distortion from stylus, cantilever, suspension tracing/angle geometric effects
while it seems silly for such low Vcm - we could still bootstrap away the Vcm nonlinearity with composite op amp circuits we both have explored
Last edited:
- Status
- Not open for further replies.