Parts & Components
This page gives a quick (though pretty lengthy) overview on the parts and components I recommend for high quality audio applications.
For essentially all application I use 0.6 W, 1 % metal film resistors. I prefer the higher power rating over the more common 0.25 W parts as especially with discrete design there are often uses which require the higher rating and stocking two different types would be cumbersome.
If higher precision (with respect to initial precision, temperature drift as well as long-term stability) and/or a very low voltage coefficient is needed, I use the RC55 series from Welwyn (0.25 W, 0.1 %) or the SMA 0207 from Vishay/Draloric (0.6 W, 0.1 %).
For higher power the T series from IRC offers 1 W to 10 W at still 1 % precision.
There are three common film capacitor types: polyester, polypropylene and polystyrene. As they all have names starting with poly they are often confused—which is a pitty as performance, size and cost can be dramatically different. In the following I will quickly summarise the characteristics of each type in order to hopefully clear things up a bit.
Polyester is the cheapest and most compact type. It comes at usually 10 % or 20 % precision, 10 nF to 10 μF capacity (though values up to 100 μF are available) and with typical voltage ratings from 63 V to 400 V. With respect to dielectric absorption, temperature coefficient and distortion it residues at the lower end of the film capacitor scale (though still way better than electrolytics). Distortion may vary greatly from one type to another and even from sample to sample. Usually distortion will be low without DC bias but is prone to show a substantial increase when a bias voltage is applied. The low price tag make polyester a good choice for power supply bypass applications and other non-critical decoupling applications. The small size is helpful for about any applications where large values are needed.
Polypropylene is usually supplied with 5 % or 10 % precision (with some parts offering even 1 %) and with typical values ranging from 1 nF to 2.2 μF (with some series available up to 10 μF). The voltage rating is high (250 V to 2 kV typically). With about any performance aspect it is superior to polyester. Most important for audio applications is the consistently very low distortion (usually right at the edge of what can be measured), even with substantial DC bias. With these properties it is the choice for any critical application requiring medium capacity values—equalisers, RFI protection filters, DC servos and so on.
Polystyrene is another high performance dielectric. Temperature coefficient, dielectric absorption and distortion is even a tad lower in comparison with polypropylene, though the differences are small. Typical values range from 10 pF to 10 nF with voltage ratings from 63 V to 630 V. Precision is high at 1 % or 2.5 %. Its more frequent use is prohibited by the limited temperature handling—above 85 °C the material weakens. Hence great care must be applied during soldering to not damage the capacitor. Typical applications are similar to the ones mentioned for polyester, with a natural shift towards the lower value range.
Some of the above mentioned film capacitor types are available both metallised and non-metallised. For audio applications the difference will not usually be significant.
Ceramic capacitors are available mostly with three different dielectric: C0G (read C-zero-G), X7R and Z5U. Performance of these dielectrics is very different. As with film capacitors these differences are often neglected so I’ll provide a quick summary for every type.
C0G (often called NP0 as well) is a high quality dielectric. Distortion is almost as low as for polypropylene or polystyrene, while the temperature coefficient is dependably close to zero and hence superior to film capacitors. It is typically available as multilayer capacitor with values from 10 pF to 10 nF and voltage ratings of 50 V and 100 V. Precision is usually 5 %. At extended cost 1 % precision and larger values up to 100 nF are available. Single layer disc types have a lower value range (1.8 pF to 1 nF) at much higher voltage ratings (up to 2 kV). The small size and the excellent electric properties do make C0G a good choice for applications such as compensation capacitors in discrete designs, feedback capacitors in opamp designs and RFI filters.
X7R offers values in the range of 1 nF to 1 μF at voltage ratings of 50 V and 100 V within a very compact package. Precision is usually 10 %. Distortion is in any case high which makes them useless for audio signal path applications. However, the small package provides very low high frequency impedance which is advantageous for power supply decoupling use.
The last dielectric to be mentioned (Z5U) provides even more capacity at low volume—values range from 10 nF to 10 μF, again with voltage ratings of 50 V and 100 V but at lower precision (20 %). The electric properties are mostly pretty poor (again with the exeption of good HF characteristic) and its use cannot be recommended for high-quality audio applications.
Electrolytic capacitors are the only capacitor type which is available with very large values at reasonable size. There are two main types: aluminium and tantalum electrolytic capacitors. Once again they are pretty different with respect to performance, size and application.
Aluminium electrolytic capacitors are available with a wide range of values, starting at 100 nF and ending at about 470 mF. Typical voltage ratings go from 6.3 V to 450 V. Do expect precision to be rather low—20 % is typical. Temperature coefficient is simply horrifying and you can expect a drastic drop in value at low temperatures.
If used as AC coupling capacitors (i.e. as high-pass filter) they do generate appreciable distortion at low frequencies. As a rough guidline I’d expect them to start with serious distortion (above 0.001 % at +20 dBu) at about 10x the cutoff-frequency. Or the other way round: choose the cutoff-frequency no higher than 2 Hz. At the same time this figure will mostly avoid time domain waveform distortion due to low-frequency phase shift. DC bias will encrease distortion even more. Bipolar electrolytic capacitors (both as “true” bipolar capacitor or implemented with two polar ones back-to-back) show greatly reduced distortion and are recommended if space and cost allows their use. Obviously if used e.g. as smoothing or bypass capacitor in power supplies this distortion characteristic is far less problematic.
Another aspect to remember is their relatively short life—given proper design, aluminium electrolytic capacitors are the first components that fail. It is wise to run them cool (i.e. placement on the PCB is important), to use parts rated for 105 °C and to avoid small sized capacitors as they dry out faster than larger packages. Usually I try to avoid packages with less than 8 mm diameter. In addition to this it is beneficial to not use the capacitor too close to its rated voltage—about 50 % overrating is a sensible rule of thumb.
All mentioned limitations apply to tantalum electrolytic capacitors as well. Distortion tends to be even higher though and they have a tendency to catch fire when they fail. On the other hand, they show much lower leakage currents and improved high-frequency impedance. For audio applications there is usually little reason to use tantalum capacitors instead of aluminium types. An exeption is e.g. timing circuits for dynamic range processors where low leakage is required for long release times.
Inductors And Ferrite Beads
Transformers still offer some advantages over transformerless circuits. For input transformers, their excellent RFI immunity is—at least in my opinion—the most important parameter for todays world. Good to excellent low-frequency CMRR, high common-mode input impedance and very high common-mode input range are important advantages as well, though at least the first two (if not all) of these aspects are achievable by careful transformerless design as well. Output transformers allow easy interfacing with both balanced and unbalanced inputs due to their at least at low frequencies fully floating nature (again, current transformerless implementations do well emulate this property, though their common-mode range is usually limited by the power supply voltage).
On the other hand, transformers show several deficiencies with respect to audio performance. Low-frequency distortion and high-frequency response variations with different source impedances are the top two on my list. In addition to this, the DC resistances of both primary and secondary winding add thermal noise to the audio signal.
If best technical specifications are asked for, there is no way around Jensen Transformers. This company offers a wide range of audio transforms, all with good to excellent performance. Cinemag Inc. offers a comparable but lower cost range of transformers at a similar performance level—with the exeption that for most parts they forgot to present recommendations for zobel networks to reduce amplitude response peaking and square wave ringing.
My standard small signal diode is the 1N914B from Fairchild Semiconductor (equivalent to 1N4448). It offers tighter specifications over the more common 1N4148. If very low leakage is needed the FDH300 from the same manufacturer is helpful. For medium power rectifier and protection applications I use the 1N4004 standard rectifier diode. For special applications asking for the 1N4004 current and/or voltage rating but low leakage current the 1N4007GP might be chosen.
As zener reference (or low power protection) diode I use the BZX79 series from Phillips. In contrast to many other zener diodes they are available with 2 % tolerance. Depending on the application at hand voltages between 3.0 V and 6.2 V are best suited as reference diode. At 5 mA zener current the 3.0 V diode offers a temperature coefficient (–2.1 mV/K) which is closely matched to that of a bipolar transistor Vbe. This is very helpful for making current regulators with very low temperature coefficient. On the other hand, the impedance is rather high with about 80 Ω, making supply rejection and load regulation relatively bad. A 6.2 V zener is much better with this respect at about 6 Ω impedance. However, the temperature coefficient is now around +2.3 mV/K. The 5.1 V zener residues somewhere between these two extremes. The temperature coefficient is close to zero with –0.8 mV/K and the impedance 40 Ω.
For protection and voltage regulator applications asking for a bit more power I usually use the 1N53xxB series from ON Semiconductor.
The “standard red” LED has interesting properties as biasing element. Its temperature coefficient is close to that of a bipolar transistor Vbe and the impedance at medium forward currents (say 5 mA) is relatively low (typically < 10 Ω) as is the forward voltage drop of about 1.6 V. This makes them a good choice for temperature compensated current regulator applications where the drop-out voltage needs to be lower than what’s possible with a 3.0 V zener diode. I usually use the HLMP-6000 from Avago Technologies. Forward voltage and impedance measurements of this part can be found in the Miscellaneous section.
My standard bipolar transistors are the BC550C/BC560C and the 2N4401/2N4403 pair. The BC550C/BC560C parts offer very good hFE and slightly higher brakdown voltages in contrast to the 2N4401/2N4403. The later complementary pair however excels at voltage noise, which makes it perfect for input stages optimised for low source impedances (e.g. transformerless microphone preamplifiers). I prefer to source these parts from ON Semiconductor, but I’ve successfully used Fairchild Semiconductor parts as well.
If higher breakdown voltages are needed, I often use the complementary MPS8099/MPS8599 pair from ON Semiconductor. Alternatively 2SC2240BL and 2SA970BL from Toshiba provide even higher breakdown voltages along with very low noise and high hFE. In addition to this, Toshiba parts are known for their tight process control resulting in good consistency of transistor parameters (e.g. Vbe).
The NPN transistor MPSA18 from ON Semiconductor has very high hFE which is perfect for applications such as voltage amplifier stages in discrete opamps.
For lowest voltage noise (almost) nothing beats the 2SC3329GR and 2SC1316GR from Toshiba—it’s quoted as 0.6 nV/√Hz in the datasheet, which is a very good figure. Unfortunately the higher hFE rating (BL suffix) seems to be almost impossible to source.
MJE170/MJE180 or the equivalents with higher voltage rating (MJE171/MJE181 and MJE172/MJE182) from ON Semiconductor are my favourite medium power devices. Alternatively I use BD135-16/BD136-16 (BD137-16/BD138-16 and BD139-16/BD140-16 respectively) from Fairchild Semiconductor in cases where the hFE specification is important.
For small power applications the FPN560A/FPN660A complementary pair from Fairchild Semiconductor offers good hFE at pretty high speed. Alternatively MPSW55/MPSW56 from ON Semiconductor or Fairchild Semiconductor provide higher breakdown ratings.
Junction Field Effect Transistors
There are only few JFETs available which offer sufficiently low noise for critical audio applications. The most common part is the 2SK170 N-channel transistor from Toshiba with typically 0.95 nV/√Hz voltage noise. It is ranked in three different Idss classes (GR, BL and V suffix). Fortunately enough, Linear Systems now offers a second source with their LSK170—A, B and C suffix are the corresponding Idss ranks.
Toshiba used to offer a similar dual version as well, the 2SK389 (same Idss ranking as 2SK170). In contrast to the single devices it provides higher breakdown voltages but slightly higher noise. Much to the chagrin of the audio world this part is now obsolete. Linear Systems offers a functional replacement with the LSK389, though with a different footprint and lower breakdown voltages.
InterFET has a range of very low noise N-channel JFET parts. Amongst them the IF3601 and the dual version IF3602 look very interesting. The manufacturer claims a typical voltage noise of 0.3 nV/√Hz at 100 Hz which is by far the lowest figure I’ve seen for a single transistor.
My standard audio opamps are the NE5532 and NE5534, usually in the graded A suffix version and sourced from Texas Instruments. They provide low voltage noise and distortion, sufficient (though not spectacular) slew rate, good CMRR and PSRR as well as a low price tag and wide availabilty. In addition to this, they happily drive relatively low loads and run on supply rails up to 44 V. Due to the uncompensated input bias currents they are limited to applications with medium or low feedback network impedances though. Pots or faders may not like the bias currents as well.
For applications requiring very low bias currents JFET input stage opamps are often used. They do have the additional advantage of higher slew rates compared with similar BJT amplifiers (as a consequence of the lower FET transconductance) and reduced sensitivity to RF demodulation. At the same time, common mode distortion and input capacity modulation with high source impedances will usually be much more of a problem in contrast to bipolar amplifiers, asking for even more care (and knowledge) from the designer. Personally I’d recommend the OPA2134 and OPA134 from Texas Instruments—if better DC offset specifications are needed, the OPA2132 and OPA132 appear to be simply graded versions of the amplifiers mentioned before.
TL071 and TL072 are FET amplifiers (usually from Texas Instruments again) often seen in medium and low price audio gear. I found these amplifiers very useful for auxiliary duties such as LED drivers or full-wave rectifiers. However, their poor noise and distortion performance makes them pretty useless for serious audio signal path applications.
If we step towards the high end of integrated circuit amplifiers there’s no way around the AD797 from Analog Devices. Properly used, it is hard to beat distortion wise (what could be a better recommendation than the fact that Audio Precision uses this IC for their 2700 series state-of-the-art distortion analysers?). DC precision, slew rate, voltage noise as well as CMRR and PSRR are very good too. In medium and high impedance circuits the high current noise will quickly become the dominating source of noise though. In addition to this, stability requires special attention—the applications section of the datasheet will be very helpful in this respect.
The OPA627 is another high performance opamp from Texas Instruments, this time with FET input. Impressively high slew rate (55 V/μs typical) and overall satisfying specifiations make this a very good (and equally expensive) all-round audio opamp. A decompensated version (OPA637) with even higher slew rate is available.
Pretty new aspirants for audio designs are the LM4562 and LME49710 from National Semiconductors. Both with respect to performance and price they are situated somewhere between NE5532/NE5534 and AD797.
There is a surprising number of manufacturers out there that offer discrete operational amplifiers. Amongst them, John Hardy Co. offers the 990C (and the slightly simpler versions 990 and 990A), which is a very good bipolar audio amplifier. Forssell Technologies sells several JFET-based models such as the JFET-992 and JFET-993, both having good reputation.
The designs described on these pages are not commercial in any way. This means that I do not sell PCBs, kits, or anything like that. It also means that any for-profit use of the information on these pages is strictly prohibited—you may not sell PCBs, kits (neither partial, nor complete) or finished (or unfinished) units of the designs on this site, without the express written consent of Samuel Groner/SG-Acoustics.
Notice that all information, schematics, layouts etc. are supplied as is, and that I can in no way be held responsible for its accurateness, functionality or even safety. Samuel Groner/SG-Acoustics shall not be responsible and disclaims all liability for any loss, liability, damage (whether direct or consequential) or expense of any nature whatsoever, which may be suffered as a result of, or which may be attributable, directly or indirectly, to the use of or reliance upon any information, links or service provided through this website.