If you have the best dual op amps, why not use them?

[CraigS]

I suspect that the issue is more to do with global negative feedback applied well, and the same applied poorly.  A case in point is the excellent AD797 - stunningly low noise, equivalent to a 40 ohm resistor, ridiculously low distortion etc etc.  But it really needs a lot of care to keep the thing stable - decoupling right at the chip (I suspect that is the blue film capacitor right at the device in the photo a few posts back), and correct layout is a must.  Although it is interesting that they use IC sockets - that is usually an admission of "not quite sure" of the design!

 

It is very possible to get any global feedback system wrong - usually layout, decoupling, no series resistor in the output etc etc regardless of opamp.  If they do not actually oscillate at hundreds of kHz to low MHz, they can be on the verge of doing so.

 

So I think global vs local feedback issue is something it is difficult to be definitive about.

 

With a power amp, getting something wrong can cause exciting consequences.  An early version of a Trace Elliot professional bass guitar amp used MOSFETs in the 500W + 250W + 250W amp - and after a week or so the transistors literally blew up at a shrapnel level.  A friend was engaged to sort out the design.  The problem with power MOSFETs is that they can oscillate at MHZ within the device themselves it they don't like the electrical environment they are in - and after a while the smoke comes out.  The solution had nothing to do with global feedback, but in modifying the detailed layout around the output devices and adding the right component network right at the device pins.

[Guest Muon]

Meh...a little is OK if done well

 

Hold on! I'm no designer....

 

user opinion here with me only

[CraigS]

Nothing the matter with subjective opinion!  I've heard high end set ups in the hundreds of k that sound either just like plain hifi, or at worse even like a bag of hammers being rattled.  I hauled my wife around a high end hifi show a while back, and we went into a room where some Mark Cohn that we know well was being played - it sounded like the back of your eyeballs was being scratched.  Mrs S burst out laughing - I had to usher her out:  "But it sounds dreadful!" says she;  "Yes - but those guys in there have worked very hard designing something that sounds precisely that dreadful!"

 

And I've heard low priced stuff that sounds great, not because it measures particularly well, but the important details are right.

[Addicted to music]

I suspect that the issue is more to do with global negative feedback applied well, and the same applied poorly. A case in point is the excellent AD797 - stunningly low noise, equivalent to a 40 ohm resistor, ridiculously low distortion etc etc. But it really needs a lot of care to keep the thing stable - decoupling right at the chip (I suspect that is the blue film capacitor right at the device in the photo a few posts back), and correct layout is a must. Although it is interesting that they use IC sockets - that is usually an admission of "not quite sure" of the design!

It is very possible to get any global feedback system wrong - usually layout, decoupling, no series resistor in the output etc etc regardless of opamp. If they do not actually oscillate at hundreds of kHz to low MHz, they can be on the verge of doing so.

So I think global vs local feedback issue is something it is difficult to be definitive about.

With a power amp, getting something wrong can cause exciting consequences. An early version of a Trace Elliot professional bass guitar amp used MOSFETs in the 500W + 250W + 250W amp - and after a week or so the transistors literally blew up at a shrapnel level. A friend was engaged to sort out the design. The problem with power MOSFETs is that they can oscillate at MHZ within the device themselves it they don't like the electrical environment they are in - and after a while the smoke comes out. The solution had nothing to do with global feedback, but in modifying the detailed layout around the output devices and adding the right component network right at the device pins.

Notice all the IC are on DIP sockets, this has the advantage of easy servicing, very 80/90s layout. The designer has opted subjectively for through the hole components, a contridiction to his philosophy on the his white paper, if this preamp was redesign today, amplifier designers like Dan D'Agastino will have no option but to go with SMD and the layout will be very different. The AD797 has a typical CMRR 120db and a 20v/us, very forgiving on a DIP landscape, the rest of the RC network around it is for RIAA.

On the E14 AUD site the AD797 dip version will not be restocked but it's SMD equivalent is, you can see that SS manufacturers are switching quickly to SMD.

If you proactively control and reduced the possibility of oscillation (self harm) and put in limitation so Vgs is never exceeded, apart from heat management, Lateral Mosfets can be indestructible. I built a DIY 240w @ 8 ohms utilising 2sj176 and 2sk56 Hitachi Mosfets back in 1994 and it's been driving 1.8 ohm ribbons with 2x 10" Eton Drivers in parallel ever since, it's just been place in retirement and replaced by a commercial offering exactly 12 months ago, but I can guarantee if you fire it up it will be usable. That's 21 yrs of use without a service or a cap change!

I can't comment on Feedback because I don't recall ever to have had the opportunity to hear a SS circuit that doesn't even have no feedback applied. Most if not all SS amplifiers I have laid my ears on all have some feedback applied. Opamps is a cheap and easy way to get into audio without the stress on the hip pocket. You have to wonder why most component manufacturers used opamps apart from keeping cost down.

Edited May 11, 2016 by Addicted to music

[Guest Muon]

Considering that the market as a whole is cost driven, maybe It's the main reason.

 

Cost has been a large motivation in every significant change in the audio world.

Edited May 11, 2016 by Muon

[CraigS]

I can see I am going to have to bite the bullet and go surface mount before long.  I have a USB microscope that will press into good use for this.

 

There are adapters to go from SM to DIL that don't cost much.  Either from places like Farnell, or from the usual Chinese suppliers on eBay.  You still have to solder the SM device onto the adapter, but you get to use your pin and hole board without doing a re-layout.

 

The AD797 is better now for stability than it was some years ago, so perhaps they tweaked the design somewhat.  But the Walt Jung superregulators are still prone to instability with the 797 in spite of quite sage hands trying to make it work with that device.

 

The troublesome exploding Trace used BUZ901P/BUZ906P.  Looking through the mod notes again, part of a raft of changes in making it work reliably was changing MOSFETS to 2SK2221/2SJ352 .  Part of the problem with a 500W amp is that the voltage rails sit at +/-90V, so the chosen MOSFETs need to be 180V - 200V devices.  As a professional grade amp, you have to be bulletproof.  It is well summed up by an old mate called Phil Jones, who now runs his own bass guitar electronics/speaker company in the States http://pjbworld.com/ .  In particular the pages on reliability are worth a read http://pjbworld.com/about.html .  

 

In particular "After a PJB amplifier is off assembly line, we test it at full power with continuous pink noise (all frequencies at the same time) for 36 hours non-stop. On top of that, we test it as a full combo or on a loudspeaker load for another 16 hours at 45 degree centigrade (113 degree F) room temperature. Only then is it ready to ship."

Edited May 11, 2016 by CraigS

[Guest Muon]

Resistance is NOT futile

 

I'm sticking to through hole components and point to point in audio

 

Old tech! is the new tech!

 

Ok, I'm going..... no need to call security

[CraigS]

I'm with you there.  But already many semiconductors that were through hole are being discontinued and only available in surface mount.  LM4562 and AD797 being cases in point,

 

Actually looked upon from a technology perspective, more compact layouts ought to be a benefit - less track inductance and resistance for starters.  The last but one big project I was involved with was a space instrument that is going to Mercury, launch next year (I was project manager). Everything in that is surface mount - front end electronics, power supplies and data processing unit.  For performance reasons, reliability, and to reduce size and weight.

[HypnoToad]

DIP 8 adaptors for surface mount op amps are plentiful and it's easy to solder them on, I use the AD797BRZ which only ever came in surface mount, it's simply graded to a tighter tolerance. It's a wonderful chip and the noise is so low it suits all but the lowest output moving coil carts, that is ones under 0.1mv's. But I wouldn't recommend it for MM carts.

 

I know people go on about the near impossible to get 2sk170's and how low the noise is if you parallel them, but we can get too paranoid about noise, I have had people comment that they can hear noise if they turn the volume all the way up and go close to their speakers, I say well try a needle drop at that volume and see if you can still hear it.

 

Cost to performance and ease of design Op Amps have a lot going for them, no wonder they are used a lot, and then there is the it's discreet so it must be better crowd, well there are some great discreet circuits for sure, Nelson Pass does some jaw dropping ones and there are lousy ones too, it's horses for course IMHO, YMMV.

[Guest Muon]

I'm with you there.  But already many semiconductors that were through hole are being discontinued and only available in surface mount.  LM4562 and AD797 being cases in point,

 

Actually looked upon from a technology perspective, more compact layouts ought to be a benefit - less track inductance and resistance for starters.  The last but one big project I was involved with was a space instrument that is going to Mercury, launch next year (I was project manager). Everything in that is surface mount - front end electronics, power supplies and data processing unit.  For performance reasons, reliability, and to reduce size and weight.

Well, size and especially weight are big factors in a project of that kind, as is reliability I admit.

 

I'll not be leaving orbit any time soon so I'll stick to analogue for the time being

[CraigS]

I know people go on about the near impossible to get 2sk170's and how low the noise is if you parallel them,

 

 

Linear Systems has remanufactured the 2SK170 and its complement the 2SJ74.  They call them the LSK170 and LSJ74.  

 

The LSK170 is easily available in all gm grades, but the LSJ74 at this point you can only buy in 100 off from the manufacturer or via diyAudio (tried to post the link, but for some reason it fails to add to the post - google it if interested)   Linear systems have also remanufactured the n-channel low noise dual FETs 2SK389 (0.9nV) and 489 (1.8nV), again renamed LSK - again freely available.  They have not yet started manufacture of the complement 2SJ109 yet.

Edited May 12, 2016 by CraigS

[HypnoToad]

Linear Systems has remanufactured the 2SK170 and its complement the 2SJ74.  They call them the LSK170 and LSJ74.  

 

The LSK170 is easily available in all gm grades, but the LSJ74 at this point you can only buy in 100 off from the manufacturer or via diyAudio (tried to post the link, but for some reason it fails to add to the post - google it if interested)   Linear systems have also remanufactured the n-channel low noise dual FETs 2SK389 (0.9nV) and 489 (1.8nV), again renamed LSK - again freely available.  They have not yet started manufacture of the complement 2SJ109 yet.

 

You can also get the BF862 which some prefer.

 

I have built some of Nelson Pass designs using the original 2sk170's and they sound dry and sterile to my ears, YMMV. As I said I think people go overboard worrying about the last few decimal places of noise.

[Guest Muon]

2sk170 dry and sterile?

 

Never found them like this, but then I have never used them in a Pass design, maybe the dry and sterile aspect was a result of the choice of other parts used including passives.

 

Edit: typo

Edited May 12, 2016 by Muon

[CraigS]

You can also get the BF862 which some prefer.

 

 

Yup.  A radio frequency surface mount n-channel FET.  The hint is that RF transistors tend to be low noise, because the noise generating base spreading resistance is a bandwidth-limiting parameter.  Kind of the same thing with biplolar switching transistors with very low saturation voltage - base spreading resistance needs to be low to ensure this.  So low noise transistors can end up that way by an accident of their primary application.

 

The problem in using RF transistors is keeping the things from breaking into self oscillation - but if you can do that the BF862 is a fine device at least as good as a 2SK170 for noise.

[HypnoToad]

The problem in using RF transistors is keeping the things from breaking into self oscillation - but if you can do that the BF862 is a fine device at least as good as a 2SK170 for noise.

 

It's recommended to use a 100ohm stopper resistor on the gate. You can get SOT-23 to DIP-3 adapters already made on Ebay if you want to use them on existing PCB's and the BF862's are very cheap.

 

Anyway getting back to Felix's original question, there are some very good op amps around and if people took the time to optimize them instead of just throwing them in as a replacement for what they have they would be surprised at the results IMHO, but I'm just a guy who messes with electronics.

[catman]

G'day all, just elucidating on a point that I made earlier op amp noise can be dependent on passive component values around the feedback loop.  For example the OPA2134  is slightly 'hissy' in the DIY ESP P06 but mostly noiseless in the Bruce Heran op amp phono stage.  In fact in that circuit the OPA2134 is quieter than the LM4562!  Regards, Felix.

[Guest Muon]

DIP 8 adaptors for surface mount op amps are plentiful and it's easy to solder them on, I use the AD797BRZ which only ever came in surface mount, it's simply graded to a tighter tolerance. It's a wonderful chip and the noise is so low it suits all but the lowest output moving coil carts, that is ones under 0.1mv's. But I wouldn't recommend it for MM carts.

 

I know people go on about the near impossible to get 2sk170's and how low the noise is if you parallel them, but we can get too paranoid about noise, I have had people comment that they can hear noise if they turn the volume all the way up and go close to their speakers, I say well try a needle drop at that volume and see if you can still hear it.

 

Cost to performance and ease of design Op Amps have a lot going for them, no wonder they are used a lot, and then there is the it's discreet so it must be better crowd, well there are some great discreet circuits for sure, Nelson Pass does some jaw dropping ones and there are lousy ones too, it's horses for course IMHO, YMMV.

Some people worry about noise that they will never hear in normal operation.

 

Strange if you ask me.

[CraigS]

I guess the question is this - if the music has a dynamic range of 75dB with an excellent pressing, you need to have an amp that is quieter than that.  The analogy is that CD and digital sources have a dynamic range of around 100 - 110dB, and I don't think you would consider buying a high end CD player or streamer that had noise greater than that.

[Tranquility Bass]

G'day all, I'm talking about LM4562 dual op amps which are by all accounts about the best of the present crop for audio use, at least judging by the data sheets!  So after doing a few wiring repairs on my existing DIY ESP P06 phono stage, I whipped out the existing dual op amps and put in two LM4562's from my existing stocks. 

 

Yeah it sounds pretty good and the hiss is noticeably less so I guess that I'll leave them in there.  I mean with 140 db open loop voltage gain, high slew rate (20 v/us) and ultra low noise and distortion (.00003% THD), they've got to be better, right?  Regards, Felix.      

 

what were the existing opamps and how much did you pay for the 4562's ?

 

cheers

Edited May 18, 2016 by Tranquility Bass

[catman]

G'day mate, I had used most of the well known 'high performance' dual op amps, but mainly OPA2134's.  The LM4562's are a few dollars more than the others, but still 'affordable' but I hear rumours that they will soon be unavailable in through hole pin configurations, so buy a few now while you still can!  Regards, Felix.     

[CraigS]

G'day all, just elucidating on a point that I made earlier op amp noise can be dependent on passive component values around the feedback loop.  For example the OPA2134  is slightly 'hissy' in the DIY ESP P06 but mostly noiseless in the Bruce Heran op amp phono stage.  In fact in that circuit the OPA2134 is quieter than the LM4562!  Regards, Felix.

 

Could that be the wrong way round?  

 

The ESP phono has a lower arm resistor of 4.7k, so the noise is dominated by the current noise of the LM4562 ( = root (2.7E-9^2 + (3.1E-12 x 4700)^2) = 14nV/rootHz, which is more than the purely voltage noise of the OPA2134 of 8nV/rootHz. So the OPA2134 should be quieter in the ESP by about 5dB.

 

In the Bruce Heran single stage, the lower arm resistor is 330 ohms (actually with such a low value, the cartridge resistance adds too, but leaving that aside) so the current noise is a minor contribution - following the calculation above the total noise for the LM4562 is 3nV/rootHz, which less than the 8nV/rootHz for the OPA2134.  So the LM4562 should be quieter by 8.5dB

 

Oddly enough, from a noise perspective, the ESP-specified NE5532 produces 6nV/rootHz in his design - which is less than either the LM4562 or OPA2134 in that design.  In the Heran design it is noisier than the LM4562 but quieter than the OPA2134.

 

In summary, from a noise perspective (so not commenting on other audio considerations) is looks like: use the NE5532 in the ESP and the LM4562 in the Bruce Heran.  The Heran (with LM4562) is a lower noise design than the ESP (with NE5532) by about 6dB based on the three chips above.

[catman]

G'day mate, interesting analysis but doesn't work out that way, believe me I've tried it.  I think that I've mentioned this before that my oft used Shure M97xE with its high coil inductance/ resistance seems to contribute adversely to the P06 hiss/noise figure. 

 

I've proven this by substituting other lower inductance moving magnet cartridges and noted lower hiss levels.  Would some sort of input buffer stage be an advantage here?  Regards, Felix.

 

Edit:  With reference to the ESP P06 which op amp stage are you referring to, the first (bass eq) stage or the second 'gain' stage, or both?   

Edited May 18, 2016 by catman

[CraigS]

I can see I'm going to have to knock up these designs and so some measurements when I find a spare day.

 

Might try spice simulation first....

[CraigS]

OK, Spice analysis (TI TINA) for the Heran single op-amp design, including the M97xE R = 1550ohms and L=650mH from the spec sheet.  SNR's relative to 5mV at 1kHz

 

LM4562 SNR

100Hz 88.5dB

1kHz 86.4dB

10kHz 84.8dB

 

OPA2134 SNR

100Hz 83.5dB

1kHz 81.29dB

10kHz 80.3dB

 

So in the Heran design the LM is quieter than the OPA by 5dB, 5.1dB and 4.9dB at the three frequencies.  That is a bit less than my quick calculation above (8.5dB) because that did not take account of the cartridge

[CraigS]

Spice simulation results for the ESP, same cartridge.  All referred to 5mV at 1kHz

 

LM4562

100Hz 74.9dB

1kHz 73.6dB

10kHz 71.3dB

 

OPA2134

100Hz 78.1dB

1kHz 76.3dB

10kHz 73.2dB

 

NE5532

100Hz 78.9dB

1kHz 77.1dB

10kHz 73.9dB

 

In the ESP design, the OPA2134 is quieter than the LM4562 by 3.2dB, 2.7dB and 1.9dB.  The NE5532 as predicted is the quietest, by 0.8dB, 0.8dB and 0.7dB as compared to the OPA2134.  Given that the NE5532 is as cheap as chips....