To clarify, the preamp I built is the 1996 design and I don't agree with Ian's conclusion re the subjective effect of low distortion - it doesn't have "zero" sound quality. That would imply the proverbial straight piece of wire with gain. It screws up the sound and makes it sound compressed and with no image outside of the speaker boxes. I'm keeping it so that I can remind myself on the odd occasion when I need to that Mr Self may be a brilliant engineer/scientist and if optimising simple THD curves and noise were the only considerations necessary to achieving high-fidelity sound his designs would reign supreme, but I'm convinced they aren't. Of course I would like my amp/preamp to have low THD+N but there's more going on that determines the sound quality and I'm sure it's measurable, only not how.
There is some real satisfaction in building such most excellent highly linear preamp. With such high linearity designs the result is usually sound that is unimpressive (lacking the better word). But it is not useless to own preamp with such great specs. It has tone controls and they should be used. Of course, there is a lot of obsolete functions and almost nobody needs preamp with specs equal to specs of precision lab instrument, but most drivers of Aston Martins and BMW Series 7 also do not actually need such cars. If one has time, money and will to build such preamp, then why not?
Let me ask you as an engineer; how that sound quality can be contributed by an extraordinarily low THD+N device if it can not be ascribed
simply to less distortion than your reference? No, I'm not implying random distortion - likely a specific, naturally declining amplitude harmonic
progression, perhaps close to that favoured by Jean Hiraga.
As audiophiles, many of us prefer selected distortion in our amplification that enhances audio detail, image etc. I think if you run a check on
"high-end" amp/preamps you won't find many successful models with fantastic low THD. In fact, "high-end" designers make an intense study
of competitors' distortion "profiles" in order to optimise imaging and tonal qualities in their own.
That's nothing new but modern techniques make it so much easier with just a simulation program, sound card, a file of top designs and decent
listening environment to test prototypes.
Try a simple test; remove your preferred preamp and replace with a fixed pad matched similarly to the amp and source. How's the imaging?
Not so good? Logically, the preamp must add something, even if it is only gain, yet that doesn't address imaging clarity, musicality etc.
in the same way select distortion can - reliably, too.
OK, this wire has loss, not gain but I think you will follow the argument well enough.
The design's distorting all right, just not in a way which is picked up by the usual high level THD+N measurements. Since music normally has a high-ish crest factor, the distortion level that matters is with test tones below -30dB relative to full scale. Also since music is made up of myriad sine waves, its IMD which is of primary importance.
Incidentally owdeo, have you tried the LM6172s yet?
Sorry I don't accept that argument and I'm not one of those that wants my music coloured - I am actually looking for transparency and low distortion. My point is that a low THD+N curve does not seem to guarantee this - there are clearly other measurements that need to be made.
My system sounds far better with a resistive attenuator between the DAC and power amp than with the Self preamp in line. But it sounds better than either with a good discrete preamp inline. There are several issues with using a passive attenuator that could explain this, none of which imply that the good discrete preamp is "adding something" - eg impedance matching, higher distortion at the input stage due to higher series Z, and of course HF rolloff due to input filters and the extra series Z. To me this only proves that the theory is correct - using passive attentuation is both technically and subjectively sub-optimum. I see no reason to assume it is due to added distortion from the preamp.
Thanks, exactly. I don't pretend to have any answers but that theory sounds plausible.
No sorry I haven't. I've been playing with a couple of discrete class A designs to great success - at this point I'm not confident that any opamp is going to sound good enough in a line preamp, but if I do come back to them I will be sure to give them a go.
At the risk of sounding completely subjective in thishas anyone tried using a mix of opamps to "spread the risk" of any particular traits being emphasised.
Hi Mooly,
I'm leaving the Self preamp as is, as it is a very useful reminder to me to be careful of getting bogged down with THD etc as a measure of SQ.
If I was going to try tweaking it I'd probably try using the OPA2604 as this is the only other dual opamp that I think sounds tonally correct in other designs I've tried it in. Ironically in his book Self dismisses it completely due to higher THD in his test configuration - from memory his comments were something like "it's hard to see under what circumstances this opamp would be a good choice". It would be hard to see when using one's ears to evaluate the result is considered pointless!
Incidently I heard from an ex-Burr Brown applications engineer that Neve chose this opamp for use in some of their mixing consoles due to its SQ, so perhaps I'm not the only one.
The OPA2604 has a degenerated JFET input stage - I reckon this is more linear where it really matters (low-level and also at HF) than the undegenerated bipolar stage in the NE5532.
Incidentally, if you can get hold of them (they're obsolete), HA-5222 sounds very good to me as a line stage provided the inputs and power are protected from RF. I put this down to the dielectric isolation.
Incidentally, if you can get hold of them (they're obsolete), HA-5222 sounds very good to me as a line stage provided the inputs and power are protected from RF. I put this down to the dielectric isolation.
Last edited:
That translates simply to 'mismatch' and the amplifier input filtering is unchanged. I believe a transformer or a fixed pad of a few passives can adequately emulate the match, albeit with a fixed loss. Still, this may suit a digital source with only a few 10s of ohms output impedance, since the days of 47k line impedance in domestic audio and needing preamplifiers optimised for it, have long gone. So I agree that passive attenuation with 10-100k pots connected in the traditional manner will likely sound sub-optimal. I find it so, anyway.
Otherwise, a fixed pad at realistic impedances of ~200R might be closer to ideal. To establish or compare the audible effects of preamplifiers, or any devices, I think there should be a better reference device than just a preferred type. If it may only be an inexpensive 10 min. lash-up, why not at least consider it?
For that matter, digital level control is now incorporated in multi-ported DAC or DSP assemblies which connect directly to the amplifier without any traditional preamp. I think that removing the focus of criticism lets sense prevail rather than concerns about undetectable IMD effects or certain opamps sounding better. If only the essential analog output devices of the DAC are present, just as with CD, Digital Media players etc. then preamplifiers and the necessary measurements to demonstrate the superiority of particular types become largely irrelevant. i.e. We can and should first know how no preamp sounds.
By implication you're claiming IMD effects (I'm not claiming they're undetectable, only you're implying this) and some opamps sounding better are both nonsensical?
I also like the OPA2604, even though it has measurable THD shortcomings, especially in non-inverting configurations. Nevertheless, its distortion is quite low. I tend to prefer JFET op amps anyway. How an op amp is used can make a difference as well. Asking the usually class-B output stage of an op amp to drive relatively low impedances is not a good practice in my mind. This can be an example of where the myopic pursuit of extremely (and unnecessarily) low noise can come at the expense of good sound quality.
Cheers,
Bob
Mr. Cordell,
It is nice to see honest disagreement between designers here in this forum. I say this as I reread your book on amplifier design and I also have Doug Self's books on filter design on my shelf. Thank you for the information about this particular opa and why you can overlook a slight negative about the THD.
It is nice to see honest disagreement between designers here in this forum. I say this as I reread your book on amplifier design and I also have Doug Self's books on filter design on my shelf. Thank you for the information about this particular opa and why you can overlook a slight negative about the THD.
Not really; In the first instance, I'd expand by suggesting that unmeasurable distortion will be undetectable or indistinguishable from that which can be measured. I could be wrong, but unless the detector only looks at fixed harmonics, the result of significant IMD should also be evident.
The OPA2604 was nominated as sounding different but I'm not deriding either issue. I'm actually suggesting the whole caboodle of component performance is sidelined, whilst the issue of what is being compared is clarified. I know what Self's 1996 pre is but I only have seen a schematic for one of his discrete line amplifiers posted here for the OPs comparison. Consider that in a full preamplifier like Self's, there might be dozens of such buffer/amplifiers, leading to possibly similar unsatisfactory results. My imagination, however is not accurate so I have suggested the alternative of a far simpler comparison, in the absence of detail.
This was no criticism of your contributions, which I find variously interesting, unconventional and illuminating. Nor am I criticising others. However, I don't think the A-B comparison of 2 apparently very different preamplifiers shows what the base or reference level of performance potential is or should be here.
Today I was discussing with my daughter one of the problems in analysis with attempting to solve an engineering problem. Working with chemical compounds it has been shown to me that in specific applications a masking compound was added to hide a specific chemical compound. Trying to use a process such as interferometry to identify the unknown compound was next to impossible due to the masking affect of the added compound. I see a very clear correlation here with the electronic testing for noise generation. If there is a masking noise created by a secondary process the underlying noise you are looking for can not be separated out of the masking noise generator. It just becomes hidden below the noise floor setting the upper limit. There must be many situation in electrical noise testing and also distortion measurements that you just can't get to the point where the separation of noise components is possible. Perhaps this is one of the main reasons that THD and other distortions are so hard to correlate with sound quality, the hidden components hidden by a higher value component?
If your distortion products are buried in the noise then you really have to wonder if they are of any consequence. Doug Self's 1996 Precision pre-amp represents the sound signature of a NE5532 OPAMP, and the way that it was implemented or put to use. It is really that simple. Anything else is just a bunch of nonsense.
As it is, this thread is supposed to be about Doug's new pre-amp, not the '96 version, although both are somewhat similar. This new pre-amp was an attempt to show the audience that with some different design techniques, one could reduce the noise to a smaller level. The new pre-amp sound signature would be different than the '96 version as he has resorted to using some newer National OPAMPs in the signal path, reason why as he pointed out in his Elektor articles.
As it is, this thread is supposed to be about Doug's new pre-amp, not the '96 version, although both are somewhat similar. This new pre-amp was an attempt to show the audience that with some different design techniques, one could reduce the noise to a smaller level. The new pre-amp sound signature would be different than the '96 version as he has resorted to using some newer National OPAMPs in the signal path, reason why as he pointed out in his Elektor articles.
rsavas,
I agree if a noise is -120db then I don't know if my dog could hear that! Just a different way of looking at the ways in which distortion measurements don't always tell the whole story. I have Doug's books, most anyway and I do appreciate how he thinks in a rational way. I haven't heard either of his preamps so can't make any comments on the sound quality. There are so many subjective way that people listen to music and also so many combinations of components that even then it would be hard to compare the results one person get verses another. Loudspeakers are going to be the greatest variable in the entire equation.
I agree if a noise is -120db then I don't know if my dog could hear that! Just a different way of looking at the ways in which distortion measurements don't always tell the whole story. I have Doug's books, most anyway and I do appreciate how he thinks in a rational way. I haven't heard either of his preamps so can't make any comments on the sound quality. There are so many subjective way that people listen to music and also so many combinations of components that even then it would be hard to compare the results one person get verses another. Loudspeakers are going to be the greatest variable in the entire equation.
I'm kinda glad this thread gets revived every now and then. I enjoy the knowledge and opinions of those way more versed than I. I can say I did build the new pre, and am extremely happy with the sound. No, not happy, giddy! I never thot a pre could ever sound so smooth and effortless. I spent buckets of cash on it, and it's been the best project by far that my hobby money could buy.
In fact, it will probably always be the crown jewel of my audio gear collection, as I just can't afford the high-price of the name brand badges for esoteric equipment. I don't fault those that can either. I'm running the pre into a simple little class D amp I built, with inexpensive DefTech bookshelf speakers. The sound is simple marvelous!
Also, thanks to Ian and rsavas for making me aware of one of DS's commercial works, the Cambridge Audio Azur 840A Integrated amp. Even though it is dated it looks tempting enough to want to try.
Rick
In fact, it will probably always be the crown jewel of my audio gear collection, as I just can't afford the high-price of the name brand badges for esoteric equipment. I don't fault those that can either. I'm running the pre into a simple little class D amp I built, with inexpensive DefTech bookshelf speakers. The sound is simple marvelous!
Also, thanks to Ian and rsavas for making me aware of one of DS's commercial works, the Cambridge Audio Azur 840A Integrated amp. Even though it is dated
it looks tempting enough to want to try.Rick