This including, the listening comments, are perfectly possible. Looks like only Bonsai bothered to follow up the Kingston thread I linked to.
With due respect, it is unlikely that more than 2 people on this thread can layout, decouple & ground stuff to a sufficiently high standard to realise the full performance of 5534 let alone da new uber OPAs .. and one of them is Guru Wurcer 🙂
Be VERY suspicious if you are measuring much more distortion than you expect especially if it seems to come & go with level. Often this is the only visible sign of Stability/Oscillation. Difficult to find as connecting ANYTHING, including 10x scope probes, changes the stability.
_______________
In Jurassic times, when MM preamps were of serious commercial interest to me, I did a full analysis of MM noise, and found you'll be lucky to get 1dB better noise compared to 5534 with anything else . You have to take into account the substantial Inductance by looking at noise in 1/3 8ve bands, applying noise weighting to taste, and then RMS sum. There's a NS or AD application note that details all this.
5534 is almost tailored for MM use. 😀
____________________
Grados have some of the lowest inductance of MMs but are otherwise flawed. The ADC XLM mk3 has probably the best balance of performance including noise .. better than the earlier XLMs.
And yes. ALL MM makers rely on substantial electrical peaking for flat response. This has been the case since the V15 type II. (Before, they were doing it but didn't realise it.) IIRC, Shure did an AES paper describing the type II including this.
A supa quality MM preamp MUST have some easy facility to adjust i/p capacitance. 47k is what the MM makers design for but the required capacitance varies. I like the QUAD method where you can get at the input bits to solder the odd NPO/COG ceramic. Switches are EVIL in this sensitive spot.
Last edited:
i know a little about low noise design
my last pro gig was designing strain gage instrumentation - 120-700 Ohms for our transducers, typical Av 4000 - at the end of 50' cables - in medical and industrial applications - looking with 16 bit ADC - spent many hours plotting the noise, histograms, fft - lots of "onion peeling" to get within a few dB of Johnson
I passed up the AD797 in favor of the LT1028 on a close reading/extrapolation from the datasheets - it appeared the LT has a slight edge in I_n 1/f corner for our resistance range
but out customers never demanded the best possible DC performance, all dynamic applications - I never got to use AC bridge drive, synchronous demodulation - a real career disappointment
my last pro gig was designing strain gage instrumentation - 120-700 Ohms for our transducers, typical Av 4000 - at the end of 50' cables - in medical and industrial applications - looking with 16 bit ADC - spent many hours plotting the noise, histograms, fft - lots of "onion peeling" to get within a few dB of Johnson
I passed up the AD797 in favor of the LT1028 on a close reading/extrapolation from the datasheets - it appeared the LT has a slight edge in I_n 1/f corner for our resistance range
but out customers never demanded the best possible DC performance, all dynamic applications - I never got to use AC bridge drive, synchronous demodulation - a real career disappointment
to answer the original post ..
- Pay LOADSA attention to earthing, decoupling & stability. You need to understand Kingston's thread.
- OPAx134 is the least fussy of da uber OPAs if you haven't got Guru Wurcer's facility with the above. Good for HiZ generally and when you dun wun Ib. See what came best for Kingston.
- Choose horses for courses .. the correct OPA for the source. This is NOT trivial.
- There's good reason why 5532/4 is a favourite.
- DON'T use LM4562, LMExxxxx for LF filters & single supply.
@ kgrlee
Yus i dud reed it 🙂
I like 2 x rail to ground & rail to rail.
Bingo !
Double bingo !!
Well i've been implimenting such things for years. I've also mentioned on numerous occasions in various threads on here that, you can't beat having Individual ground/common returns to the star point, from EACH component.
Ideally you then have an audio star point for the Actual audio components ground/common returns = Clean. And also a power supply star point for OPA's + & - pins, & also for the OPA's bypass capacitors ground/common returns. You only link the 2 stars @ the power supply. The mains earth "can" connect to this point, & chassis.
Yus i dud reed it 🙂
I like 2 x rail to ground & rail to rail.
Bingo !
Double bingo !!
Well i've been implimenting such things for years. I've also mentioned on numerous occasions in various threads on here that, you can't beat having Individual ground/common returns to the star point, from EACH component.
Ideally you then have an audio star point for the Actual audio components ground/common returns = Clean. And also a power supply star point for OPA's + & - pins, & also for the OPA's bypass capacitors ground/common returns. You only link the 2 stars @ the power supply. The mains earth "can" connect to this point, & chassis.
OPA2134 or LF3553
the 353 is old but it is way better sounding then the TL072. it is difficult to hear the difference between the two at low gains, however the OPA2134 is about 1000 times lower distortion.
the 353 is old but it is way better sounding then the TL072. it is difficult to hear the difference between the two at low gains, however the OPA2134 is about 1000 times lower distortion.
For low voltage noise purpose, once I put 4 C1844 (0.5nV/√Hz) in parallel, which makes 0.5/ √4 nV/√Hz, at the input stage.
For low current noise i.e. for good I-V conversion of photodiode signal, FETs with Teflon pot.
I would do still the same thing this year.
I will never use OPamps at the input stages for extremely low noise purposes.
For low distortion, I found RC4558 the best fit in my circuit
A Low Distortion Amplifier Based on a "Tandem" Correction Scheme Powered by a Modifie
because it doesn't get oscillation easily.
For low current noise i.e. for good I-V conversion of photodiode signal, FETs with Teflon pot.
I would do still the same thing this year.
I will never use OPamps at the input stages for extremely low noise purposes.
For low distortion, I found RC4558 the best fit in my circuit
A Low Distortion Amplifier Based on a "Tandem" Correction Scheme Powered by a Modifie
because it doesn't get oscillation easily.
Last edited:
Off the grid right click on the Y axis. This should bring up a pop up with 'Format Axis". Click on format axis. Click on the scale tab. Look for 'Value X axis crosses at' Set the value.
Works on exel 2002. But you know how micro loves to change things.
Cheers,
Davada, thanks for that. Its my work PC - its quite a recent version that seems to work differently. I tried it on my old Excel version (c. 2000) and it works - but not on the latest version. You have to go to the design tab, which comes up when you right click off the grid. BTW, trying to protect cells is also a pain in Excel. I want to lock everything except the cells where data is input. Massive rigmarole. No wonder people prefer the other options from those other guys down in the Bay area. Ditto Word. OK for writing letters or fiction. Add some graphics, plots, diagrams and its all fall down. A real PITA. Who the #$%$ designs this stuff?
End of rant.
End of rant.
Last edited:
"5534 is almost tailored for MM use."
Yes - its the low current noise of 0.4 pA/rt Hz. This makes a big difference 10 kHz to 20 kHz. Most bip opamps are > 1 pA/rt Hz and that ends up swamping the noise voltage with high L sources.
I think a Linear Systems JFET with 1nV/rt Hz and negligible IIRC input current noise will pip most opamps - provided the second stage is carefully designed - e.g. no carelessly configured current mirror load.
I am working on the phono equalizer for my new preamp - 10 BF862's in parallel with a similar input structure to Denis Colin/Syn08. I think that will do it noise wise. MC a different story or course - seems parallel bipolar is the way to go, or stacked AD797/LME49990.
Joachim still has not shared his 20 parallel method circuit yielding 0.22 nV/rt Hz
😉
Yes - its the low current noise of 0.4 pA/rt Hz. This makes a big difference 10 kHz to 20 kHz. Most bip opamps are > 1 pA/rt Hz and that ends up swamping the noise voltage with high L sources.
I think a Linear Systems JFET with 1nV/rt Hz and negligible IIRC input current noise will pip most opamps - provided the second stage is carefully designed - e.g. no carelessly configured current mirror load.
I am working on the phono equalizer for my new preamp - 10 BF862's in parallel with a similar input structure to Denis Colin/Syn08. I think that will do it noise wise. MC a different story or course - seems parallel bipolar is the way to go, or stacked AD797/LME49990.
Joachim still has not shared his 20 parallel method circuit yielding 0.22 nV/rt Hz
😉
Is that measured or sim?
I thought I had the lowest noise MC headamp in the known universe at <0.28nV/rtHz measured. It uses a giarnormous total of 2 devices. 🙂
Yes. That might pip a NOS Mullard 5534 (made by Southampton virgins) with a MM cartridge for RIAA but not by more than a dB regardless of the weighting you use (or don't use).
From the king himself. "Oh no, everything in windows is completely configurable."
I think we have to deal with formats in the tool used to generate the graph. Apps like word get a bit mapped copy. That was MS's way of dealing with importing.
We have bloatware like adobe for that.
Maybe his number is even lower. I think his claim to fame on this one is he's using opamps (is it 20 pV rt Hz? Where are you Jochaim?).
Were you using the Rohm low rbb devices? The MJE13005 also make good low resource amplifiers I heard because of their low Rbb but the hFE is abysmal. And nowadays most of them are made in crappy fabs and assembled in sheds . . .
Did you work at Southhampton? Facility there is long gone I am afraid.
Last edited:
Endless paralleling of devices is not particularly sophisticated design. Take 100 1nV op-amps in parallel and you have .1nV, at some level so what. The Lepaisant pre-amp does .06nV with transformers. Philips also presented an IEEE paper on a .25nV pre-amp using an optimized low rbb npn pair.
Last edited:
It's the user interface that needs a thorough upgrade and modernizing. Example: rotate the page . . . You have to go into a menu tab and set page breaks.
Why?
Right click,select 'rotate page' and the job should be done. Ditto moving pictures around and a few dozen other things.
But, I'm digressing here and going off topic. Better to move this to the lounge.
Last edited:
Yes, I would agree with that - I think Jochaim is doing it in his measurement amplifier. Can get expensive as well
Got a link to where he claims 0.2nV/rtHz? I was amused this Millenium to find my <0.28nV/rtHz was still the quietest in the known universe after more than 30yrs. But I'm always willing to be proven wrong. 🙂
I used Hitachi 2sa108x & 2sc245x circa 1980. A number of people have built my circuit with the Rohm devices. Alas most of these supa lo rbb devices are now Unobtainium. You can still get some npn ones but the pnp ones only exist in certain secret hoards.
I've tested a lot of medium power devices (and so did Baxandall) but I no longer recommend them. They have to be selected for 1/f noise and the yield is poor.
I was only a satisfied customer. The Mullard Southampton factory made both my favourite LN OPA at that time and also my favourite LN FET.
When I returned to LN electronics circa 1990, Philips 5532/4s came from foreign climes where the virgins weren't and were no longer consistently LN
Doesn't have to be sophisticated, if it works & gives good performance 😉
Yes 😀🙂 It's surprising how little we've come, in some respects, in 30 or so years !
Even if it were possible to design/produce a device/s that produced no noise, "unobtanuim" we would still be subject to the thermal etc noise/s of the other components in the circuit/s ! So realistically, i don't think we will see much improvement in that department !
It's in Yahoo MicBuilders as MCamp.pdf in my ricardo Files directory. You have to join.
- Status
- Not open for further replies.