IC's get hammered

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Excuse me gentlemen, although I find much about Burson´s to be nonsense there is some truth in the page in post 1 which I also believe can make a difference. The part about transistors being inferior is nonesense and actually the opposite is quite true but resistors and capacitors inside ICs are. Resistors not much of a problem but do you know how caps are formed in ICs ??. They are not caps like we know them in discretes but are reverse biased mosfets (in some cases and this was the best solution some years ago). Looking at their performance and properties they are of similar performance to disk capacitors and not the COG type. Would you use such caps as say compensation ??

Some proper research is in order although modern techniques are much improved.
 
What's particularly funny about the "hammer the ICs" argument is that their very unique "feature" is turned into a bug. The feature being, with modern manufacturing (circa 2000), that the transistor-matching and passive-component matching are within 0.1% on a chip! That is really hard to achieve in discrete circuits (not too hard with the passives, but just TRY to come up with a dozen or two 0.1% matched semiconductors!)

Moreover, when I had my "automatic double-blind" A/B/C switch rig, a whole room full of us (some self-proclaimed golden-ear types) couldn't tell the difference between a bog-standard op-Amp, a super-duper ultra-high end one, and a discrete module. Provided the output trim pots were adjusted so that the gain overall was exactly the same. But this is what digital multimeters are good at, especially when outfitted with a small zero-crossing-point A-law dU meter with good Zener references, and independently battery powered. Easy to adjust to 3½ digits. 0.1 dB accuracy.

Oh - if you were listening for cues like "noise" ... sure. But that's with no signal coming through. Even the IMD on noise+signal is inconsequential when there's a 100 dB difference between the two. I'm quite happy with modern FET input opAmps, even the cheaper ones. The $0.99 wonders (mouser) have 15 nV/√Hz ... and the more expensive ones (such as TI OP1642 at $4/ea) come in at less than 7 nV/√Hz. Amazing devices, really.

Ah well. There are people that think that letting power cords sit in a jug of liquid nitrogen for a few hours (until it boils off, one presumes), then installing them on little risers to avoid capacitive effects of one's carpet, and of course "burning them in" for a few days ... is the cat's meow. My little A/B/C completely-anonmymous (therefore unimpeachable) vote-box got rid of that silver-plated bias!

GoatGuy
 
burson are known for speaking nonsense on this issue and 'their' discrete regs and opamps are not exactly known for their high performance by discrete standards. they are forever posting ridiculous, biased so-called research and technical results on their pages

I purchased a pair of the Burson regulators for the Linear Audio Magazine review -- they were pretty simple devices -- and for some reason (inexplicable in the statistical analysis, or by any measure of PSRR, noise and Zout) they sounded very nice!
 
OTOH, getting the value you actually wanted, within a factor of two, still seems to be a challenge.

Uh, huh ... though not really. Chip passives (resistors) now are within tolerances of 20% wafer to wafer. Doesn't sound terribly impressive, but consider: virtually ALL designs are set up to be insensitive to significant overall variation in Hfe, gm, R ... so long as the RELATIVE values remain closely matched. (This is exactly analogous to high-end tube and discrete amplifier design, where carefully matched, but not absolute-value precise components are used)

The reason why discrete amp makers use high-precision resistors is because they're comparatively cheap now, and because this gets them 'matched' without having to actually 'test-to-match' them. Capacitors are notoriously imprecise, yet, in their slated "duty", their absolute precision is in no way meaningful. Amplifiers are "happy" (and many after-market DIYdesigners "play around") with +20%, +50% caps. ... except for equalization where more precision is necessary ...

But even discrete amp makers "sigh and deal with" the imprecision of their discrete active transistors, and especially tubes. They will match them, and so on ... but REALITY sets in: this is why fine amps have trimpots onboard to allow "bias point" and operating points to be adjusted. All those 0.1% precision resistors are for naught ... when you need to adjust the trimpot 25% to account for a Soviet-manufactured 12AX7 not having as much mu as the GenElec equivalent.

So... the analysis of ICs was fatuous and bogus. Not much more to be said.

GoatGuy
 
I purchased a pair of the Burson regulators for the Linear Audio Magazine review -- they were pretty simple devices -- and for some reason (inexplicable in the statistical analysis, or by any measure of PSRR, noise and Zout) they sounded very nice!
so, are you telling us that the most reasonable way of altering the signal is by adding a power rail modulating device with absolutely unpredictable behavior in a circuit? and we should be surprised that whatever effect it had in your application sounds good, because, well, it altered the signal in some way?
hmm.
 
[its interesting - this idea of modulating the power rails]

As far as EFX boxes go - and especially older tube-guitar-amps go - this is exactly what is done in their power-supply rail, to deliver distortion that is considered musically interesting. Long chains of high-value resistors, modestly small capacitors ... ostensibly to form LP networks blocking stage-to-stage feedthrough, but in reality also drooping and rising in seconds-long RC periods; delivers interesting macro effects all by itself to guitar signals.

Could be a similar effect in output amplifier stages: changes in a nonlinear way, the maximum output slew rate and gain. Kind of like a nonlinear low-cut filter? But more complicated.

Amps as EFX boxes! DF96's most favorite chiding-point topic.

GoatGuy
 
Amps as EFX boxes! DF96's most favorite chiding-point topic.
would you be amazed if I told you that I'm using such amp?
yes, I really am because it sounds better. am I being a hypocrite? no. because it was cheap, looked nice and for some reason makes my system sound engaging.
it has "custom vibration isolation" on the transformer, uses "special treated RCA connectors", custom capacitors and there's some putty on one of the RCAs on the inside. LOL do I care? no. do I think it sounds good because of them? no.
:D
 
I must be Retarded. I Paid for Volume 4. I could have waited for the link to the Important content here on DiyAudio.

Socko, Microtones, Authentic, Height, Awash, Smeared, Weirdness, Textures, Recessed, Yawner, Bloom, Dull, Congested, Veiled, Flabby, Course, Murky...........etc

As an aside, Burson virtually last in all the Measurements but at the top of a list in
"how it sounds".

Quoted as "Subjective comparisons as the final arbiter of Valid judgment", something Stinks. Can we just throw out the Numbers ?
 
Linear Audio - Volume 4 "A Comparative Review of Power Supply Regulator Designs with Listening Tests"
no, I read the link Thorsten L posted at diyhifi.org.
so... you're talking about corroboration, they said it sounds good and so did you.
I'll state again my POV: it is obvious from the measurements that this device alters the output voltage. which equals to the signal modulating itself. so a difference in sound is to be expected.
I think it should be obvious to anyone that the way the device behaves is application-dependent and the simple idea of "best regulator", especially when there's not much actual regulation to talk of is useless.
the fact that you and the LA guys thought it sounded best (best? what is best? most of the times best = different) says nothing.
 
There's a curious thing about the Burson which may have escaped your attention -- pay close attention to the Zout chart.

FWIW, this was a very expert listening panel -- (save for me, I like sitting in the last row so i can go to the boys room.) I didn't score any device. None of the panelists were aware which device was being tested, and the tests were repeated several times. The ordinal rankings didn't change that much and as i've said before the bottom cohort were really stinkers. I was pretty much in awe of how badly some of the regulators could make a cymbal sound.

The line amp tested -40dB PSRR, compare to an OPA637 which is ~-90dB.
 
OK, I just get queasy when I see things like Gain-Bandwidth Product: 8MKz typ, 2MKz min, (max not stated). However that's probably mostly a cap's fault. Also that's for an LM3886, maybe newer stuff is better.

You present an interesting point - marketing versus real-world engineering. At the time of the LM3886's creation, "2 MHz min, 8 MHz typical" was a reflection not on caps, but on the control of gain on the small signal transistors that constituted most of the circuit. IN TURN, the Ft (frequency response) of the transistors depended heavily on the exact mix of bulk-silicon dopants that were part of the wafer more than the implantation process. Nowadays, dopant levels are under much tighter control, and in many cases, even the "bulk" dopants are done by ion-implantation and annealing to get the specs tighter. Thus this, combined with much finer control of dimension and overlap - make for tighter chips.

But that doesn't mean the snake can't bite. So, to keep to the idea that many low-precision applications exist for cheap op-amps, and to "sell the most that they can" is the trick to raise profits... for cheaper op-amps wider marketing ranges are written down.

Companies that make the $4/dual op amps (such as TI with the OPA1642, or the LME49880) ... especially "new" designs like the 49880, are creating chips of amazing sophistication and repeatability. The 49880, for instance, has 7 nV/Hz noise (outstanding), 25 MHz gain-bandwidth product, 0.00003% THD (in other words, "zero"), 15 ohm output impedance, and 15 gigaohm (!!!) input impedance. 0.5 µV 20-20KHz wide spectrum noise. Sufficiently low to act as an excellent microphone preamp, even for moderately low-output dynamic microphones found in high-end recording studios.

Also +/- 15 volt output swing.

Now that's one hell of a chip, for $2.02, qty 25, at Mouser, "retail". Dual too.

GoatGuy
 
Excuse me gentlemen, although I find much about Burson´s to be nonsense there is some truth in the page in post 1 which I also believe can make a difference. The part about transistors being inferior is nonesense and actually the opposite is quite true but resistors and capacitors inside ICs are. Resistors not much of a problem but do you know how caps are formed in ICs ??. They are not caps like we know them in discretes but are reverse biased mosfets (in some cases and this was the best solution some years ago). Looking at their performance and properties they are of similar performance to disk capacitors and not the COG type. Would you use such caps as say compensation ??

Some proper research is in order although modern techniques are much improved.

the part about the capacitors is nonsense as well and it seems you are well behind the current state, fets are still used for less precise decoupling applications up to a point, as with fractal caps which apparently (I only just heard the term today) are better at minimizing the coupling to the substrate, which left unchecked can cause an effective short between metal layers and thus unreliable yield. there are also MiMCAPs and PiP caps, which are both thin film processes, with MiM still being the highest linearity afaik. I dont have access to the design rules for the highest technology processes, as these require NDA, so there may be even better technologies.

how do I know this? because i've been doing some cursory research into MPW (multi project wafer) short run IC prototyping runs. places like MOSIS act as brokers/intermediaries between you and Foundries such as OnSemi, IBM, Global Foundries, offering up to 65nm processes where you share in project runs with other people for as little as 2500-3000USD for 50 dies in a wafflepack.

its a very long way off for me, but I find the possibility very interesting and intend to work towards it as a long term goal. its not a quick, or risk-free process as a noob, the MPW runs generally take a few months as they have to fill up the run with other people, as well as not being high priority. somewhat like the Batch PCB of IC manufacture. its one shot though, if there is a mistake, youve blown your money.

if you need more than this in an IC I would be surprised, the recent emergence of thin film caps as discrete parts (for example Vishay RFCS passivated die package caps), shows you just how linear they can be and that is VERY linear, in some respects more linear than film/foil caps. the resistors and active devices as mentioned are quite excellent.

I purchased a pair of the Burson regulators for the Linear Audio Magazine review -- they were pretty simple devices -- and for some reason (inexplicable in the statistical analysis, or by any measure of PSRR, noise and Zout) they sounded very nice!

yeah well I also had several of them a couple years ago, I sold all but 1, which sits in the back of one of my drawers unused, I didnt like them and rejected them long before your article. I thought they were pretty ordinary
 
yeah well I also had several of them a couple years ago, I sold all but 1, which sits in the back of one of my drawers unused, I didnt like them and rejected them long before your article. I thought they were pretty ordinary

I didn't expect anything remarkable from them, but they never got knocked out of the first cohort. I created the schematic from the PCB -- you can do the same -- nothing special there.

It's just a supposition, but the Zout is pretty flat, bumps up in the most sensitive band for our hearing, then trends down. It's not as if it were designed by a bunch of MIT engineers for an F-22, it just works in a serendipitous fashion quite pleasing to the ear.

I never achieved the results WJ obtained with the Sulzer regulator. This I will attribute to the change in semi's over the intervening decades.

The other really surprising event was how nice the off-the-shelf Linear Tech devices sounded.
 
sounds like its an example of what we are talking about, a sort of effects module that sounds good in some circumstances due to its specific and probably unintentional 2H non-linearity. I would have to construct a special amp to try it again, because most of my gear is very high PSRR, so a comparison with that would be a bit pointless. so that comparison is unlikely to ever happen.

yes the LT regs are very good for what they are, I use a bunch of lt1963A and lt1764A as preregulators in my dac, not final stage, but very good for first stage regulation. they are followed by some flea-like buffered references with common mode filtering on the input and differential filtering on the output.

I havent used the 3015 though, despite looking into them at one stage, so its interesting that you guys found them good. I guess I just find the flowery audiophile lingo used to describe the regs in the article a bit offputting. Describing sound is difficult at the best of times, but I dunno, the typical cliches are to my mind not terribly useful.

I was looking at the 3015, but i'm more interested in the new range of parts from Ti the +/- 1A TPS7A4700/TPS7A3301 and their lower current MSOP powerpad counterparts TPS7A3901/TPS7A3001
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.