I was working at Intersil last year and was checking out what was cool in the portfolio. I came up with ISL28291 - Dual Single Supply Ultra-Low Noise, Low Distortion Rail-to-Rail Output, Op Amp There's also an ISL28290 with lower noise and better bandwidth, but they don't want to put it on the market unless you want millions of them.
I made a phono preamp with the ISL28291's in the msop package. It runs from 3 X AA cells for weeks if I forget and leave it on. The high efficiency LED set at a dim glow uses as much power as the three dual op-amps. I did it as a split passive eq with gain stages of 11. The power switch just toggles the enable pin and the last op amp enable has a cap on it to delay for 5 seconds or so for the filters to settle before the final output is enabled. There's just a tiny tick when it comes on.
1.7nV/Hz^-2
1.8pA/Hz^-2
0.00018% THD
61MHz bandwidth
2.6 mA/amplifier
The ISL28290 is better, but they're not selling it. There was only one wafer made and they don't want to make more unless there's demand for a full run of parts.
I made a phono preamp with the ISL28291's in the msop package. It runs from 3 X AA cells for weeks if I forget and leave it on. The high efficiency LED set at a dim glow uses as much power as the three dual op-amps. I did it as a split passive eq with gain stages of 11. The power switch just toggles the enable pin and the last op amp enable has a cap on it to delay for 5 seconds or so for the filters to settle before the final output is enabled. There's just a tiny tick when it comes on.
1.7nV/Hz^-2
1.8pA/Hz^-2
0.00018% THD
61MHz bandwidth
2.6 mA/amplifier
The ISL28290 is better, but they're not selling it. There was only one wafer made and they don't want to make more unless there's demand for a full run of parts.
Last edited:
I talked about opamps too and they can be paralled in a Faulkner like circuit so where is the deal? What is the new information ?
Trading voltage noise for current noise ? Yes, every decent low noise specialist knows that.
Why not doing this instead of using ONE that has low noise. they do not come better then 0.9nV/qHz. Some have lower 1/F corner but thats it. For me not low noise enough.
When -120dB distortion is not good enough go to compound or use distortion cancelation.
Trading voltage noise for current noise ? Yes, every decent low noise specialist knows that.
Why not doing this instead of using ONE that has low noise. they do not come better then 0.9nV/qHz. Some have lower 1/F corner but thats it. For me not low noise enough.
When -120dB distortion is not good enough go to compound or use distortion cancelation.
Last edited:
Yes, with tubes. I listen privat to tubes.
I managed to make a low noise tube phono pre with selected 6C45Pi but that is another story.
This may be interesting :The Zero-Transistor IC, a New Plateau in IC Design - Application Note - Maxim
I managed to make a low noise tube phono pre with selected 6C45Pi but that is another story.
This may be interesting :The Zero-Transistor IC, a New Plateau in IC Design - Application Note - Maxim
EC8020 is even better but that one costs 500,- $ up when you can get them.
Gm is 60mS on this one.
Gm is 60mS on this one.
Joachim, what Ein do you get with this preamp and the 4R4 Lyra?
Can you link to the post or page in the MPP thread which has the circuit? It will take me some time to go through nearly 1000 pgs 😡
I take it its NOT the circuit on the 1st page
I did not publish the circuit but the measurements. I have to search myself.
I similar idea is the Lepassant circuit that Scott mentioned ( Frensh Picovolt Meter ) but my feedback arangement is different.
You can also look at the Lundahl site and at Walt Jung for some circuits that erode distortion in transformers, albeit at the output.
I similar idea is the Lepassant circuit that Scott mentioned ( Frensh Picovolt Meter ) but my feedback arangement is different.
You can also look at the Lundahl site and at Walt Jung for some circuits that erode distortion in transformers, albeit at the output.
Interesting part - the only caveat seems to be the highish input bias current, max 7uA. Perhaps kgrlee will see that as a pro rather than a con though as it implies no bias current cancelling circuits... 🙂
http://www.diyaudio.com/forums/attachments/analogue-source/265897d1329093376-mpp-diamond-dove-without-colour-fill.pdf
Joachim, this is indeed very impressive. The amp noise is some 10dB below total noise at 100Hz and some 15dB below 50Hz
It's a pity the measurement hides any differences above 1kHz as that's where the most audible differences will be. If we assume the difference is still 10dB at mid & high frequencies (which your test suggests), the noise deterioration is 0.4dB .. ie 1.7dB better than (the other) Gerhad's
What is setting the noise floor in the measurement?
What transformer are you using?
I'm a bit suspicious of the noise of the 20cm balanced shielded cable. 😕
Please, please, pretty please can we have a look at the circuit? 🙂
____________________
The easiest demo is to swap between OP27/37 and 5532/34. They have very similar internal circuits and specs. ... except 5532/4 have their inputs as God intended, without Evil Ib cancelling.
You'll find 5532/4 noise always slightly better than OP27/37 (even though today, without 5532/4s hand carved by Mullard's Southampton virgins, the spec is slightly worse). The difference can be up to 3dB.
The only time when you'd see equal noise is in fully balanced topologies where today, you'd sensibly use a device optimised for this.
I recently found that Guru Wurcer's AD797 has evil Ib cancelling even though the datasheet suggests otherwise. 😱 This may be why Andrew (Bonsai) finds it better between 700R -2k5 when the datasheet & Jensen App. Notes assume Ropt is 450R.
Anyone know of 1nV/rtHz audio OPAs WITHOUT Evil Ib cancellation?
Joachim, this is indeed very impressive. The amp noise is some 10dB below total noise at 100Hz and some 15dB below 50Hz
It's a pity the measurement hides any differences above 1kHz as that's where the most audible differences will be. If we assume the difference is still 10dB at mid & high frequencies (which your test suggests), the noise deterioration is 0.4dB .. ie 1.7dB better than (the other) Gerhad's
What is setting the noise floor in the measurement?
What transformer are you using?
I'm a bit suspicious of the noise of the 20cm balanced shielded cable. 😕
Please, please, pretty please can we have a look at the circuit? 🙂
____________________
This is a real problem.
The easiest demo is to swap between OP27/37 and 5532/34. They have very similar internal circuits and specs. ... except 5532/4 have their inputs as God intended, without Evil Ib cancelling.
You'll find 5532/4 noise always slightly better than OP27/37 (even though today, without 5532/4s hand carved by Mullard's Southampton virgins, the spec is slightly worse). The difference can be up to 3dB.
The only time when you'd see equal noise is in fully balanced topologies where today, you'd sensibly use a device optimised for this.
I recently found that Guru Wurcer's AD797 has evil Ib cancelling even though the datasheet suggests otherwise. 😱 This may be why Andrew (Bonsai) finds it better between 700R -2k5 when the datasheet & Jensen App. Notes assume Ropt is 450R.
Anyone know of 1nV/rtHz audio OPAs WITHOUT Evil Ib cancellation?
it would also be great to discuss new techniques and ideas regarding distortion cancellation methods.
THx-RNMarsh
THx-RNMarsh
Last edited:
That cable picked up noise. You see in the deeper regions it is more or less a short.
I really had to make a special short plug to measure the noise without cartridge and wire.
With a stage like that and when you really crank the system you can hear diffent noise from diffent cables, cartridges and connectors.
In the treble the noise plots converge because the action of the shunt RIAA comes into play.
It really has zero gain in the treble and not R1/R2 plus 1 like a series feedback RIAA.
It also does not have the 14dB ( at 20kHz ) dynamic range disadvantage of a passive RIAA.
I really had to make a special short plug to measure the noise without cartridge and wire.
With a stage like that and when you really crank the system you can hear diffent noise from diffent cables, cartridges and connectors.
In the treble the noise plots converge because the action of the shunt RIAA comes into play.
It really has zero gain in the treble and not R1/R2 plus 1 like a series feedback RIAA.
It also does not have the 14dB ( at 20kHz ) dynamic range disadvantage of a passive RIAA.
Michal Fremer made a video on HighEnd2014 where i open the box of the proto and give a run over. I do not know when and if he will publish it. The exact details are my secret though, sorry.
The transformer is made to my specs. One problem in a diffential circuit is symmetry so i have stringend specs about the wire resistance athough there is some lokal feedback to erode the wire resistance diffence or at least reduce it.
I trim the RIAA with teflon trimmers.
The transformer is made to my specs. One problem in a diffential circuit is symmetry so i have stringend specs about the wire resistance athough there is some lokal feedback to erode the wire resistance diffence or at least reduce it.
I trim the RIAA with teflon trimmers.
"Anyone know of 1nV/rtHz audio OPAs WITHOUT Evil Ib cancellation?"
DS mentions somewhere that is the input source impedances are unbalanced, its an issue although I do not recall the exact detail. Scott does not see this as a problem - and the noise performance of th AD797 is very good.
I really think the input bias currents need to be on the low side,
DS mentions somewhere that is the input source impedances are unbalanced, its an issue although I do not recall the exact detail. Scott does not see this as a problem - and the noise performance of th AD797 is very good.
I really think the input bias currents need to be on the low side,
What is the Transformer step up Ratio for your 4R4 Lyra Atlas?
______________
AD797 is indeed a Wonder of the Silicon World.
When I first saw the spec, I thought, "how clever of these AD gurus to get such super Beta devices in its input that you didn't need Ib cancelling!".
Alas, datasheet speak with forked tongue 😡
- Status
- Not open for further replies.