I guess it depends what you consider to be enough - is ~40 dB @ 1 MHz enough? At some point, most of your regulators stop working, too.
40dB @ 1MHz translates to a 100MHz opamp? I reckon probably good enough yeah with suitable decoupling.
Yep, I rely on caps at higher freqs.
Yeah, I took that figure from the OPA1611 datasheet as an example. It's a little lower than that, maybe around 37.
ClassAB is de rigueur when opamps are employed. @Bonsai's design is one of the exceptions I believe seeing as he augments the opamps with a discrete OPS.
In some cases I had first the equipment , listen to it and compared with nowadays "high end "....
I don't know of any discrete design that beats an LM4562 or one of its newer derivatives on PSRR (measured please) without resorting to fancy regulators, or cap multipliers (see JC's design for example). As for Samuel Groner's effort - they cost about $80 IIRC and a modern high performance bipolar opamp $3, FET input a bit more.
I would agree you can bring a few things to the party on discrete designs that maybe IC types cannot deliver without some ancillary circuits (e.g. class A buffer, higher output currents) and you might get a bit lower noise in certain very specific applications. The blanket claim that 'discrete is better' does not stand up to scrutiny - like 'tubes sound better' or 'feedback goes round and round the loop'.
LIGO doesn't use discrete opamps. It must be for a good reason.
sorry to be pedantic, but
Class b works best on the same silicon wafer , Integrated chip technology, but there it can surpass class a in performance by large.
I wonder which discrete design can beat opa2227 psrr...I've seen tested lots of SMPS of up to 60kw , 300 000v with all the error , voltage and current control based entirely on opa2227 and we were like 5 guys testing these power sources at the same time with no shielding within 3 meters one to each other, making final discharge tests with shorting out up to 300kv ouputs for 100 times in a raw...When you put your scope in your circuit the harmonic noise level was huge on the IC power supplies as we tested 4 different types of smps with different switching converter and pfc frequencies in each one...
There are though some class b high power amps that can do well with a little bit of class a help.
I wonder which discrete design can beat opa2227 psrr...I've seen tested lots of SMPS of up to 60kw , 300 000v with all the error , voltage and current control based entirely on opa2227 and we were like 5 guys testing these power sources at the same time with no shielding within 3 meters one to each other, making final discharge tests with shorting out up to 300kv ouputs for 100 times in a raw...When you put your scope in your circuit the harmonic noise level was huge on the IC power supplies as we tested 4 different types of smps with different switching converter and pfc frequencies in each one...
There are though some class b high power amps that can do well with a little bit of class a help.
Attachments
Last edited:
DPH has pointed out that I did not measure the performance of the opamp/buffer in class AB which was a good suggestion as I don't have a comparison point unfortunately. If you operate in class A the rail harmonics are low order which, if you are pathological about it, could be argued as a plus. In class AB, there are loads of wideband harmonics on the rails.
I use an AD797 and my QA401 on a good day can get down to about 2ppm - the 797 distortion is well below the noise floor and only comes up when I load it with c. 200 ohms at 0dBV. In most cases it will be driving a >10k load and I presume the OPS will be in class A anyway.
Anyway, the LM4562 class a buffer thingy does sound nice driving a pair of 32 Ohm 'phones but I wont claim its better than anyone else's efforts in that regard.
You hadn't mentioned cost in your original claim that IC beat discrete on PSRR so why bring it in now?
There were two different HDAM circuits.One was diamond buffered, the other was the picture you saw.I think you had the diamond version.
Wait, chris, can't you hear differences of sound quality between various amps and audio circuits, and, sometimes even components ?
If your answer is "No, I can't" you have a problem. Or your system is very bad, or your ears. And if it was so, it would lead at the disparition of all the audio industry.
There is no need for any proof when we can listen. Do-you need a proof of the existence of the moon if you are not blind ?
I dont dismiss integrated circuits, I use them quasi exclusively for line levels.
Used hundreds of them in my analog mixing desks. And it is the way all the records are produced in the world, even the analog way: there is always, somewhere, an integrated circuit in the signal path.
Compare a TL072 with a AD797: yes, they don't sound the same ... and they don't measure the same.
Not saying that TL072 sound bad.
Now, I'm not sure i could hear a difference between a AD797 and a OPA1622 ;-) Just I have to try one of these days.
Anyway, in my mind, it hearts the most elementary logic to say "All integrated circuits sound bad" as well as to say "All integrated circuits sound the same".
Last edited:
The 5534 is a very good Swiss knife OMHO. And can drive lot of loads thanks to its current capability. It measure quite good as well. (some IC, now achieve incredible signal/distortion+noise ratios [Thanks Scott and few others], so why bother with old dinosaures ?).
The only problem is I always funded it sounding a little DULL.
I agree with the poster who mentioned that the 5534 measures better often than for example a 134. Let me add that in real life audio circuits, I measured better performance of this chip than of whole slew of other opamps with better specs, like the 4562. Is it more resistant to lay out imperfections?
To me, that is still a pertinent question, anyone clever enough to come up with an explanation that holds water?
As to dull, dunno. The only way I can make improvements at this level is by taking measurements. My ears don't discriminate between opamps very clearly. Pathological examples excluded of course. Let's say everything better or equally specced than the 5532/4
Last edited:
Not necessarily so on DI or other modern processes, there are amps with Ccomp split 1/2 to each rail.
This is not to slight you in the least because you do/did some fine things, but I want to paraphrase Bruno: good engineering starts by defining specs. Also trying to define the way to achieve those is asking for trouble.
I don't see how bringing the cost in changes any claim - its just a fact of life.
(BTW are you still based in Hangzhou?)
I seem to recall this was just a vanilla VFA front end - LTP, Isource, single transistor+diode mirror driving IIRC straight into the VAS. I was quite surprised because they had made a big deal about HDAM at the time.
I used to get a c. $1000 'allowance' every 2 years to spend in the Philips (who owned Marantz at the time before selling it off to DM Holdings). I even got a cassette tape deck - 1998 - what was I thinking?
- Status
- Not open for further replies.