Design Article
Selecting resistors for preamp, amplifier and other high-end audio applications
Dr. Michael Belman (Vishay Intertechnology) and Yuval Hernik (Vishay Precision Group)
8/25/2010 3:30 AM EDT
In high-end audio equipment, careful selection of resistors is one of the best ways to avoid or minimize noise and distortion in the signal path. This paper describes the noise generation in resistors manufactured using the various available resistor technologies and quantifies the noise insertion typical for each type.
Noise is an unwanted wide spectrum signal that may be superimposed on any useful signal, including DC. Resistors, like other passive components, are noise sources to various degrees, depending upon resistance value, temperature, applied voltage, and resistor type.
Many experiments have been done to show why some resistors are "noisier" than others. But the only test that audio experts and audiophiles have agreed on is comparing the level of fidelity that results when different resistor technologies are used in actual audio systems.
Noise in resistors
Overall resistor noise has several components. The most relevant for audio applications are thermal noise and current noise.
Thermal noise is notable for being independent of the resistive material. In fact, the thermal noise level is the same for any type of resistor provided the resistances and temperatures are the same.
The voltage power spectral density (PSD) of thermal noise ST [V2/Hz] is uniformly distributed through the entire range of frequencies. It may be presented by the following expression [1, p.76]:
where
R - resistance of a resistor [Ω],
T - resistor temperature [K],
k = 1.3807×10-23 J/K - Boltzmann's constant.
Current noise, on the other hand, has a direct relationship to the type of resistive material. The spectral density of voltage of current noise SE is found experimentally to be directly proportional to the square of DC voltage drop U across the resistor and inversely proportional to the frequency f [2, p.164]:
C is a constant that depends on material of the resistive element and its manufacturing process. The spectral density S of the total noise voltage in the resistor is presented in Fig. 1.
The current noise level in a resistor is commonly expressed in units of µV/V or in decibels (in terms of Noise Index [NI]dB)
where u is root mean square noise voltage over a decade bandwidth, and U is the DC voltage drop across the resistor. Both u and U are measured in volts. The lower the Noise Index, the lower the level of current noise in the resistor.
The Noise Index of resistors manufactured using different technologies is presented in the graph below [2, p.168].
As the graph shows, resistors based on composition resistive materials such as carbon and thick film have the highest level of current noise. Why? Because of the significant non-homogeneity of these resistive element materials. The conduction path in composite materials is formed by the conductive particles touching one another in an isolative matrix. Non-stable contacts in these "touching sites" generate noise when electrical current runs across them.
Thin-film resistors have a considerably more homogeneous structure and consequently are less noisy. Thin films are deposed using evaporating or sputtering of resistive material (for example tantalum nitride TaN, silicon chromium SiCr, and nickel chromium NiCr) onto a ceramic substrate. The thickness of the layer varies typically from 10 to 500 angstroms depending on the resistance value.
The noise in thin films results from the occlusions, surface imperfections, and non-uniform depositions which are more significant when the film is thinner. That is why the thicker the resistive film, the lower the resistance value and thus the lower the noise level.
The lowest noise level is observed in resistors with bulk metal resistive elements: foil and wirewound. Wire is made of metal alloys similar to foil material, but additional noise may come from the junction of the fine wire of the resistive element and the comparatively coarse resistor terminals. In foil resistors, the terminals and the resistive element are parts of the same piece of the foil, so this problem is avoided.
But the major objection to wirewound resistors is their inductance, which results in chopping the signal peaks1, and the significant dependence of resistor impedance on signal frequency. In addition, special attention must be paid to the following effects associated with reactance of wirewound resistors:
- The audio amplifier may oscillate spontaneously at 5 MHz to over 50 MHz affecting audio quality [3, p.22-6].
- Equivalent series inductance (ESL) can cause large phase shifts affecting audio tone [3, p.22-6].
- The wire coil may act as a pick-up of EMI that may surpass the level of usual current noise [2, p.167].
Foil resistors avoid these problems because they are structured using chemical etching of a flat bulk metal foil so that the current in adjacent current carrying paths runs in opposing directions, cancelling parasitic inductance of these paths. Also, path-to-path capacitances are connected in series, which has the effect of minimizing the parasitic capacitance of the resistor. These low-inductance/capacitance resistors are characterized by nonmeasurable peak-to-peak signal distortions.
Footnotes:
1 The effect of peak chopping originates from the entire amplifier circuit that comprises wirewound resistors. ESL in wirewound resistors (sometimes together with parasitics of other components) may result in signal phase shift that is sufficient for amplifier “ringing” on signal transitions (peaks). This high-frequency "ringing" (chopping signal) fills relatively low-spectrum audio signal peaks (chops them). It is illustrated graphically below:
|
|
|
|
|
Robotics Developer
8/25/2010 4:01 PM EDT
Thank you for an informative article! It was nice to have a refresher on the source(s) of noise and brings back memories of my earlier audiophile days.
Sign in to Reply
Guru of Grounding
8/26/2010 2:12 PM EDT
Many audio circuit designers seem to ignore the fact that, with regard to noise, all resistors are the same EXCEPT when DC flows through them - which, of course, causes the "excess noise". Most applications of resistors in audio circuits see no DC, yet many audiophiles believe that an expensive resistor will result in lower noise. Of course, low excess noise resistors are useful in applications where there DC does flow - the plate load resistors in vacuum-tube circuits, for example. This article seems to underplay that fact.
Sign in to Reply
milliganp
8/27/2010 7:32 AM EDT
The final graph is very dramatic, and a salutary reminder to do one's calculations. However to dissipate 0.3w in a 1K resistor would require 24v RMS, which is typical of a power amp rather than a low level circuit. In a low level circuit with output of 1v RMS and R2 =2k2 the stabilised resistance change would be 3ppm for the bulk resistor and 0.5ppm for thin film.
Sign in to Reply
kdboyce
9/7/2010 2:51 AM EDT
It is great to build very high quality audio amplifiers! But...that quality can be ruined by the speaker attached to the amp. So while this article is a welcome reminder about the importance of component selection, don't over-design the amp and under-design the speaker. And make sure the supply noise is not a contributor.
Oh, and the overall gain that is required will dictate what the noise performance of the amp stages should be.
Many applications simply do not require the ultra-low noise performance. The high performance is needed in the professional audio, recording, and measurement industries, and also for the audiophiles among us.
Sign in to Reply
Joe Geller
10/12/2010 12:36 PM EDT
This article was informative in describing noise considerations beyond thermal and current noise, such as temperature-induced nonlinearity.
However, the bold print lead in is “Experiments have sought to show why some resistors are ‘noisier’ than others”, suggesting perhaps the actual measurements of thermal and excess noise are naïve or useless in their incomplete nature. But, the authors then devote a page to thermal resistor noise calculations (J.B. Johnson, “Thermal Agitation of Electricity in Conductors”, Physical Review, vol. 32, July, 1928, page 97), and restate that carbon resistors are noisier than thin film resistors, and that metal foil resistors are best, seemingly well known information, even known to many non PhD “experimenters”.
The description and caution that one should consider temperature-induced nonlinearity in power applications (and don’t forget inductance, especially if you see spontaneous RF oscillations, a good point) is all helpful information. Probably many designers have not taken temperature-induced nonlinearity into proper account, where there is enough power dissipation for temperature-induced nonlinearity to be a factor. (As an earlier post points out, temperature-induced nonlinearity is likely to be insignificant in many small signal applications.)
On the other hand, since our 2007 JCan experiment article (on measuring thermal and excess noise) began with the line “You have probably heard someone say that some resistors are “noisier” than others”, the lead in title to this article, in my opinion, felt like a bit of a snobbish remark towards those of us “experimenters” who go to the effort to actually measure parameters. It is unclear why that negative approach was needed for an otherwise informative aggregation article.
Sign in to Reply
rpell2
10/15/2010 12:28 PM EDT
A reader has emailed the following comment:
On Page 36 of the Sept 27 EE Times story about "Selecting resistors for high-end audio applications", there is the following statement in the lower part of the left-hand column:
"The major objections to wirewound resistors are their inductance, which results in chopping of the signal peaks, and ----------."
That is the biggest pile of $%*$& that I have read in many years. If an inductor is a linear device, and all wirewound resistors that I have seen in 40+ years of electrical engineering have linear inductance, because they are not wound on iron cores, there is no way that the inductor will chop signal peaks, but leave the remaining portion of the signal alone/untouched. I fully agree that wirewound resistors can very easily be inductive as well as resistive, but there is no way they can "chop" signal peaks.
H. R. Hofmann,
Past President of the IEEE EMC Society,
President of Hofmann EMC Engineering
Sign in to Reply
rpell2
10/17/2010 12:41 PM EDT
The authors appreciate the feedback and have added a clarification to the article (see footnotes).
Sign in to Reply
jcdrisc
4/21/2011 9:46 PM EDT
Mr Hofmann is right. They certainly ARE linear inductive.
A w.w. resistor load from a MOSFET source follower caused oscillations. Tank. Capiche ?
I would add as well as Vishay's very good parts the British Company Welwyn is their standard for many resistors of all kinds.
Sign in to Reply
Nic Cohen
7/27/2012 4:19 AM EDT
What a superb, article and what incredible amplifiers to boot, not literally of course!
Nic
www.kdweb.co.uk
Sign in to Reply
Michael Dunn
3/2/2013 8:54 PM EST
When was this published...April 1?
Sign in to Reply
neevkidman
4/30/2013 5:37 PM EDT
Resistors and amplifiers are the best tools for any audio player. An audio player is not working means it has any problem in one of these two.
Neev
http://www.supremeconsultant.com/buy-dissertation-uk.html
Sign in to Reply