From datasheet:
"The noise and distortion performance of the OPA165x is exceptional in applications with high source
impedances, which makes these devices an excellent choice in preamplifier circuits for moving magnet phono
cartridges."
Devil is in the details - the 2.9nV figure for the 1656 is at 10kHz, not 1kHz...
At 1kHz voltage noise of OPA1656 is 4.3nV (typical)
At 1kHz voltage noise of OPA1642 is 5.1nV (typical)
But the 1/f knee of the OPA1656 is much higher around 2kHz, compared to 100Hz or so for the OPA1642, which makes the 1642 a clear winner at lower frequencies. And none of these figures is guaranteed by testing on manufacture it seems - there is no maximum noise spec.
Both have low current noise being FET opamps so only the voltage noise performance is going to discriminate them.
At 1kHz voltage noise of OPA1656 is 4.3nV (typical)
At 1kHz voltage noise of OPA1642 is 5.1nV (typical)
But the 1/f knee of the OPA1656 is much higher around 2kHz, compared to 100Hz or so for the OPA1642, which makes the 1642 a clear winner at lower frequencies. And none of these figures is guaranteed by testing on manufacture it seems - there is no maximum noise spec.
Both have low current noise being FET opamps so only the voltage noise performance is going to discriminate them.
When you take into account RIAA and psychoacoustic weighting, the OPA1656 outperforms the OPA1642, see post #15.
You should aviod to have X7R in the signal path (only for decoupling) if you want extremely low distortion. X7R has also high dielectric absorption which has an influence in analog/digital conversion for instance. I would never come with the idea to use X7R in critical analog circuits.
Also X7R are pizzo-electric, so microphonic, as well as THD generators. Avoid, particularly in high gain environments like a phono stage.
If you really want low noise AND a J-FET input, either go to x2 ADA4625-1, or put the matched pair of JFE2140 from TI in front. Yes, finally there are matched J-FETs of very low noise back in the industry after Toshiba got cold feet... Not as quiet as a pair of PNP ZXTP2027F, but still: < 1nV/^Hz @ 20Hz J-FETS!
If you really want low noise AND a J-FET input, either go to x2 ADA4625-1, or put the matched pair of JFE2140 from TI in front. Yes, finally there are matched J-FETs of very low noise back in the industry after Toshiba got cold feet... Not as quiet as a pair of PNP ZXTP2027F, but still: < 1nV/^Hz @ 20Hz J-FETS!
If you want as low a noise level as possible in between records, first replace the 47 kohm resistor with a circuit producing a 47 kohm resistance with less than thermal noise (electrically cold resistance), then worry about the input stage. With a source inductance of 500 mH or so, the termination resistor contributes more to the RIAA- and A-weighted noise than the OPA1656.
If you only care about the noise with a record playing, then don't bother, record surface noise will dominate anyway.
If you only care about the noise with a record playing, then don't bother, record surface noise will dominate anyway.
To get an idea of the distortion level take a look at fig. 1 in this article Bateman Cap Sounds. And compare it with fig. 2.
I have 10u XR7s at the input of my TPA3255 design which measures 0.0006 THD at 1kHz. I guess it depends how and where you use them. Same level as with 10u Panasonic polars:
https://www.diyaudio.com/community/...pffb-and-single-ps.352069/page-7#post-6800776
Last edited:
I will try to measure soon the distortion levels of the amp with 100u XR7s at the output, see if I can see a difference between them and a short. For now, I can tell you that this sounds fantastic.
Large values on the input don't see signal voltages as they are much lower in impedance than the load (47k resistor in may cases). 10uF at 1kHz is 16 ohms, so its effect is diluted by 3000-fold or so at 1kHz if into a 47k load - the measured 0.0006% is thus only revealing that the 10uF caps are 1.8% THD or better. And with low signal levels distortion is going to be less anyway.
It might be more revealing to see if the THD rises towards the bass end - this is typical of coupling-capacitor distortion effects as the voltage across the cap rises with falling frequency as its impedance rises.
For the RIAA equalization network the caps will see signal voltages and there's won't be this dilution of distortion so the cap distortion is much more critical.
I did a THD measurement across the audible band and there was no raise in distortion towards lower frequencies. The worst THD was around 8kHz.
Another way to visualise how you can get THD: When there is no voltage across the capacitor, there is no THD, as there isn't in the passband of any filter (except all pass). As as you approach the filter response knee, you get a phase shift, which puts a varying voltage across the capacitor. Then this voltage is not linear with capacitance, creating the THD. It is worse the larger percentage of the rated voltage of the capacitor.
So if you ensure the bandwidth is far away from the filter knee OR the voltage is tiny compared to the capacitor voltage rating, the THD may not be measurable in a noisy circuit. So your 10uF into 47k the HPF is 0.34Hz. This is nearly low enough to be THD free, depending on the capacitor voltage rating. But as the siganl is so small at low frequencies, it will be lost in noise. The 100uF on the output will depend on the next product's input impedance, which is likely >10kohm, so again fine, unless you use 6.3V capacitors.
The 100uF in the feedback loop is problematic. It's breakpoint at 12Hz means the phase shift at 20Hz and above is not insignificant. So do not use X7R here for best results. Then again I would not chose Nichicon Muse non polar either, but that is just an subjective choice. Everything else here is objective.
For the filter caps: Definately NP0. You can get 27n 1% from Kemet in 0805, 8n2 and 7n5 1% in 0603 and up.
However, again X7R are microphonic. So the input caps are going to see 60dB gain at LF. So you are making an active mic with an RIAA response. So don't use X7R.
So if you ensure the bandwidth is far away from the filter knee OR the voltage is tiny compared to the capacitor voltage rating, the THD may not be measurable in a noisy circuit. So your 10uF into 47k the HPF is 0.34Hz. This is nearly low enough to be THD free, depending on the capacitor voltage rating. But as the siganl is so small at low frequencies, it will be lost in noise. The 100uF on the output will depend on the next product's input impedance, which is likely >10kohm, so again fine, unless you use 6.3V capacitors.
The 100uF in the feedback loop is problematic. It's breakpoint at 12Hz means the phase shift at 20Hz and above is not insignificant. So do not use X7R here for best results. Then again I would not chose Nichicon Muse non polar either, but that is just an subjective choice. Everything else here is objective.
For the filter caps: Definately NP0. You can get 27n 1% from Kemet in 0805, 8n2 and 7n5 1% in 0603 and up.
However, again X7R are microphonic. So the input caps are going to see 60dB gain at LF. So you are making an active mic with an RIAA response. So don't use X7R.
Thanks for the detailed suggestions. Output caps are probably 50V and there should be no DC to polarize them anyway, so I think it should be good. I could probably even short them since there is already a dc blocking cap on every input down the road.