Yes, we don't care more than about the transistors non linearities inside ICs. IC is a black box to be considered as a whole. The only resistors we have to care of are outside the IC: the feedback bridge ones. And they are under you choice.
There are IC resistors in the DACs, and good DACs measure VERY low distortions levels.
You never answered my query earlier regarding which type you used.
What is the internal? Are they bog standard electro's or are they switchmode types with low ESR/ESL?
jn
I have about 10 types coming in, all sorts, but all radial aluminum electrolytics 1000 uF at 35V+
Which resistors? N diffused, P diffused, N well, P well, polysilicon, thin metal film? Or you mean the CMOS implementation for linearised floating resistors, with a value depending only on the β of a matched set of nMOS transistors?
Please be specific.
We don't use those things in any critical place. Any gain network would use thin film resistors and the linearity pushes a few ppm or so (if that), and yes I said that days ago and jcx and I both produced examples.
Last edited:
Just spent an hour listening to the MBL system with Tape Project copies of master tapes played back LOUD on their mid price system with AD 797 in the preamp in Montreal . In 40 years. I haven't heard better even with my Blowtorch clone and Martin Logan CLX with JL Audio fathom 113 subs even with HD Tracks down loads. I'd beg for copies of those tapes of master copies ( Led Zep, Beatles, Jackson Brown, )Hugh Maekela,) wonderful!. EXTREMELY, HIGHLY RECOMMENDED if you ever get a chance to hear it. NYC show with 7.1 and 4 JL Audio Gothams and Sony K4 projector coming up . - We been ripped off for 40 years of compressed consumer garbage
Last edited:
Pavel it is not like you to be imprecise. A standard 1% metal film resistor at 1/4 power has about .00001% distortion. Excluding carbon composition distortion at 1/4 power ranges from .001% to .000001% from my sampling. Now that is for 1000 ohm values. It turns out that higher values when used at 1/4 power have higher distortion. So a typical design number for a decent resistor might be .0001%.
Now I am sure many here would claim to be able to hear that!
But as a practical issue given the four stages of a typical sound reproduction system (Playback, preamp, power amplifier and loudspeaker.) One must consider the cascade. So with .002% reproduce, .001% preamplifier, .05% amplifier at typical operating level and 1% loudspeaker distortion the system would have distortion components of -300 db. at the 16th order distortion. Allowing for musical energy being at a maximum around 125-150 hz. and rolling off by about 18 db. to the sixth harmonic, allowing for an increase in hearing sensitivity (inverse Fletcher Munson) of another 12 db. and finally allowing for our hearing able to pick out a signal down by 30 db, means that the resistor distortion for quality resistors would only be 300-(18+12+30)=240 db. down.
Now if you do the same weighting for really lousy resistors and more than just second order harmonic, you might get close to where it is an issue. Of course the solution is don't use lousy resistors.
Which raises the issue of do the resistors used in monolithic circuits behave badly enough to reach a problem level. Since I haven't measured such resistors and what their actual contribution to total distortion would be when placed in the actual circuit topology, the answer is I don't know.
Now I do know a number of techniques for improving the quality of IC based topologies and have played with which commercial parts do perform better.
As an aside to the mention of National Semiconductors audio specific LM family. The LM4562 has a real issue with RFI. Now if I built a discrete amplifier that had that issue the cost of fixing it would be small and the correction could be applied to all future production. So what is the cost of fixing an IC in production? (Yes you can use external fixes, but that is not the issue.)
Scott we cross posted. I am not concerned with the feedback network as I presume that is external. The issue is with the other resistors which are not as critical.
Last edited:
There are many types and many places where the resistor quality does not matter much. No time to go over all of them, just pointing out that there are precision resistors available. Maybe I should send some to Ed for testing.
Where they are not critical, or they produce no (or little) distortion, because they do not deal with variable voltages following signals, or their non linearity can cancel the active devices ones, or they are part of the open loop, so their (low) distortions are canceled, like the transistors non-linearities, hundred times higher, by the feedback.
Why did-we care about feedback network quality ? Because it is the only place where distortion of the network will be not canceled, but added.
- Status
- Not open for further replies.