Famous monologue from Russian movie, "Everything is crap except bees, my son... Well, bees are crap too"
Translation for DIYers: you should not read Blameless books as a recipe to cook perfect amps. It is rather a book that explains weak points of topology, and how to design it optimally from author's point of view, with his set of optimization criteria. Different sets of criteria, different topologies, and different points of view can be valid as well.
Hi,
You tell me. I see them used with 50K or higher value pot's all the time. The NE5532/34 and the Nat Semi LM4562 being the most prominent offenders.
True. Just having a Fet input on chip is not always the answer either.
You need the ones that either cascode the fets - can't think of ready integrated examples, for discretes try goldmund - or use a rush cascode (which is no cascode at all BTW) example is the OPA627/OPA637/OPA827 range from Burr Brown.
On the other hand four pieces of 2SK170 or BF862 matched (match the two top and two bottom devices) and set up DCB1 style for each input of (say) an LT1028 can produce an outstanding result, follow with a nice Mosfet SE Follower (see below).
Of course, if use this we might equally set up one pair of 2SK170 (better a 2SK389 or LSK389 actually) as differential input (I'd even use a resistor tail - 5K1 will do splendidly for the +/-30V rail example) with a pair of PNP transistors (say BC560) as folded cascode with a suitable base voltage.
We add a pair of NPN's as current mirror (say BC550) and a 2SK2700 Mosfet (Crrs =10pF) or something similar with low revers transfer capacitance with a CCS tail (even something as ignoble as a 317 may be gainfully pressed into service here) with around 50mA quiescent current as output.
Use compensation from the VAS out to negative in (not miller please)...
If we can accept that we have to loose some available swing from the two rails (say we run 30V rails and reserve 10V for overhead) we can degenerate the heck out of the current mirror (say 1.4K emitter resistors) for maximum linearity.
And we can have a very large value resistor supplying the folded cascode (say 1K for 3mA Ips current and 7mA VAS current) with (lowest noise per volt) 12V Zenners supplying our PNP Transistor bases (use a 47k resistor to the negative rail here) and still get +26dBU output into 600 Ohm loads.
BTW, I'd estimate DC OLG of this discrete Op-Amp in the 80dB rage (or more, depends a little on our bipolars) with around 35KHz uncompensated OLBW for 80dB DC Gain.
But that would be cheating as we are now discrete and no longer use Op-Amp's and we probably will have minor issues giving 0.0001% THD in an optimal setup on the testbench.
The sound may be quite interesting though, even with a 100K pot and a 10K load (even though the 100K pot would throw away the stupendously low noise of this circuit, heck, even a 10K one would) ...
Ciao T
You tell me. I see them used with 50K or higher value pot's all the time. The NE5532/34 and the Nat Semi LM4562 being the most prominent offenders.
True. Just having a Fet input on chip is not always the answer either.
You need the ones that either cascode the fets - can't think of ready integrated examples, for discretes try goldmund - or use a rush cascode (which is no cascode at all BTW) example is the OPA627/OPA637/OPA827 range from Burr Brown.
On the other hand four pieces of 2SK170 or BF862 matched (match the two top and two bottom devices) and set up DCB1 style for each input of (say) an LT1028 can produce an outstanding result, follow with a nice Mosfet SE Follower (see below).
Of course, if use this we might equally set up one pair of 2SK170 (better a 2SK389 or LSK389 actually) as differential input (I'd even use a resistor tail - 5K1 will do splendidly for the +/-30V rail example) with a pair of PNP transistors (say BC560) as folded cascode with a suitable base voltage.
We add a pair of NPN's as current mirror (say BC550) and a 2SK2700 Mosfet (Crrs =10pF) or something similar with low revers transfer capacitance with a CCS tail (even something as ignoble as a 317 may be gainfully pressed into service here) with around 50mA quiescent current as output.
Use compensation from the VAS out to negative in (not miller please)...
If we can accept that we have to loose some available swing from the two rails (say we run 30V rails and reserve 10V for overhead) we can degenerate the heck out of the current mirror (say 1.4K emitter resistors) for maximum linearity.
And we can have a very large value resistor supplying the folded cascode (say 1K for 3mA Ips current and 7mA VAS current) with (lowest noise per volt) 12V Zenners supplying our PNP Transistor bases (use a 47k resistor to the negative rail here) and still get +26dBU output into 600 Ohm loads.
BTW, I'd estimate DC OLG of this discrete Op-Amp in the 80dB rage (or more, depends a little on our bipolars) with around 35KHz uncompensated OLBW for 80dB DC Gain.
But that would be cheating as we are now discrete and no longer use Op-Amp's and we probably will have minor issues giving 0.0001% THD in an optimal setup on the testbench.
The sound may be quite interesting though, even with a 100K pot and a 10K load (even though the 100K pot would throw away the stupendously low noise of this circuit, heck, even a 10K one would) ...
Ciao T
These still need to have the back gate up against DI to perform. The rest all perfectly good stuff BTW.
John,
Low input Iq means there is low current in the input stage. This tends to impact a range of parameters and non of them in a good way.
For example something like an AD815 has many features that would seem to recommend it for audio use, yet it's input current modulation limits our use, especially with Pots like the Alps conductive plastic ones, which have a quite strong non-linearity in the wiper/track interface (basically terminate them into as high an impedance as possible), as do carbon types.
As far as I can tell cermet pots and multi-turn wire wound potentiometers are free from this kind of problem, which explains my preference for them when all the circuits I used (including discrete ones) featured bipolar inputs...
Ciao T
Low input Iq means there is low current in the input stage. This tends to impact a range of parameters and non of them in a good way.
For example something like an AD815 has many features that would seem to recommend it for audio use, yet it's input current modulation limits our use, especially with Pots like the Alps conductive plastic ones, which have a quite strong non-linearity in the wiper/track interface (basically terminate them into as high an impedance as possible), as do carbon types.
As far as I can tell cermet pots and multi-turn wire wound potentiometers are free from this kind of problem, which explains my preference for them when all the circuits I used (including discrete ones) featured bipolar inputs...
Ciao T
First set of transistors came today. Now for some matching. I would suspect that is one of the things not done carefully in mid-range consumer amps like I have, er had. Managed to pick up an HCA 1200II last night for a low ball. Every now and again, I get lucky. I missed a 1500 the day before. Sure hope it is "good enough" as steeping up to a Bryston or Krell gets really serious.
I have to make notes on all the above op-amp suggestions. I think T's suggestion comes out like a modern version of the good old LH 0001. Simple 4 transistor buffer. Sure was sweet in it's day (30 years ago). My active preamp, (clone of the Grado head amp) wound up embedded in my speaker measurement jig, so I need to build a new one. If I use a socket, and set up to measure the power amp distortion, I should be able to see differences between various op-amps if what the book describes is correct.
I was surprised at Cordell's measurements of switch and relay contacts. I did hot think it was that bad. It does not say much for the consumer amps assembled with wire-wrap now does it? Anyway, I remember relays are a BIG catch-22 from my old lab days. We had a big DC motor and an emergency power off relay, much like an output protection relay. This was driven by a 90V, 45A amplifier. We had to have very precise measurement against an optical tach at very low current, but it had to withstand opening at full power. Well, for the very low power you need gold multi-point contacts. But gold will blast off at the first arc. We switched to some exotic metal, but I don't remember so what I am asking here is, what? What relay spec is suitable for output protection and still be very high quality? Protection is another of those things totally lacking in my Hafler and Rotel. Come to think about it, in the Creek too.
My rant for today is Anchor porter is now $10 a six out here. Gotta pay to play I guess. Life is way too short for cheap beer.
I have to make notes on all the above op-amp suggestions. I think T's suggestion comes out like a modern version of the good old LH 0001. Simple 4 transistor buffer. Sure was sweet in it's day (30 years ago). My active preamp, (clone of the Grado head amp) wound up embedded in my speaker measurement jig, so I need to build a new one. If I use a socket, and set up to measure the power amp distortion, I should be able to see differences between various op-amps if what the book describes is correct.
I was surprised at Cordell's measurements of switch and relay contacts. I did hot think it was that bad. It does not say much for the consumer amps assembled with wire-wrap now does it? Anyway, I remember relays are a BIG catch-22 from my old lab days. We had a big DC motor and an emergency power off relay, much like an output protection relay. This was driven by a 90V, 45A amplifier. We had to have very precise measurement against an optical tach at very low current, but it had to withstand opening at full power. Well, for the very low power you need gold multi-point contacts. But gold will blast off at the first arc. We switched to some exotic metal, but I don't remember so what I am asking here is, what? What relay spec is suitable for output protection and still be very high quality? Protection is another of those things totally lacking in my Hafler and Rotel. Come to think about it, in the Creek too.
My rant for today is Anchor porter is now $10 a six out here. Gotta pay to play I guess. Life is way too short for cheap beer.
As already said , the 5534 is one of the best op amps for whom
know its best purposes.
75db OLG at 10khz...
OPA134 or OP627 are not as good in this respect....
Invariably , a total disaster , worse than miller compensation
in all possible parameters.
Such a THD figure is trivial with a discrete design provided extreme
care is given to layout and realisation.
See the graph below..
Attachments
I guess that it was too neutral...
Remember that it s a three stages design with at least
three nested feedback loops.
It has more OLG in the audio band than almost any recent op amp,
and even the highly regarded LME49860 is just barely on par
with the ancestor in this matter....
www.ti.com/lit/ds/symlink/pcm1794a.pdf
Remember that it s a three stages design with at least
three nested feedback loops.
It has more OLG in the audio band than almost any recent op amp,
and even the highly regarded LME49860 is just barely on par
with the ancestor in this matter....
www.ti.com/lit/ds/symlink/pcm1794a.pdf
To be fair that was not the intended application. Walt Jung's AD823/AD815 line amp has been well liked by many folks, Walt personally demoed me his prototype and it sounded fine to me but I can't get past how different people's preference in speakers just overwhelms the whole listening comparison experience.
So a Levinson 27 would be dead meat! Guess Acrus would not have gone over very well. What a world we live in. I am waiting for the nostalgia stage; like the old Marantz of the 70's. Spinning that big heavy dial made you think you had something. Or Bakelite in tan or pink. Yea, that's the ticket. Art nuvo amps.
Sfdr
FYI I have a designed and fabricated ~130dB SFDR splayed directional couplers for 40Gb optical transmission The triangle waves are quite precise (hopefully see attached) Bet the curvature in the triangle wave response would be clearly visible
Interesting point .01 % is a spur free dynamic range of only 40 dB; that is well above the noise floor and missing higher order harmonics. Can you actually hear a 40dB 2nd harmonic? I know I found ~.6 % clearly audible from 500 hz to 4khz.
FYI I have a designed and fabricated ~130dB SFDR splayed directional couplers for 40Gb optical transmission The triangle waves are quite precise (hopefully see attached) Bet the curvature in the triangle wave response would be clearly visible
Attachments
Really? I would have thought 80dB. Could you explain to a dummy how you got that number?
- Status
- Not open for further replies.