My Cart
Your Shopping Cart is Empty
How Is "Tube Sound" Produced? The Measurable Physics Behind Valve Warmth
|
0 comments
posted by Vincent Zhang
How harmonic distortion, overload, output transformers and loudspeaker loading can change what you hear.
“Tube sound,” also called valve warmth, can describe a fuller midrange, softer treble or the way a circuit rounds off peaks. Several mechanisms can produce those impressions. The tube stage can change the harmonic spectrum as signal level rises. In a power amplifier, the output transformer and the interaction between amplifier output impedance and speaker impedance can also shape the result.[2][6][8][11]
What “Tube Sound” Describes
It helps to distinguish three applications. A microphone preamp can colour a recording when driven into overload. A guitar amp may be designed to operate there. In a hi-fi system playing within its normal range, the amplifier and loudspeaker pairing can change the frequency response even at low levels.[1][2][6][11]
Russell O. Hamm’s 1973 paper compared recording preamplifiers and surveyed dozens of amplifier circuits. Strong microphone transients in his studio examples could drive preamps into severe overload. He cited conditions approaching 30% total harmonic distortion (THD). After aligning the THD curves at a common distortion reference, he found broadly similar slopes among the circuits tested. The stronger distinction in his samples was how the distortion was distributed across the harmonics.[1][2]
Nonlinear Transfer Curves and Harmonics
A triode’s plate current depends nonlinearly on both grid-to-cathode and plate-to-cathode voltage. Around the bias point, the stage’s input-to-output curve can be approximated by a Taylor series. With a sine-wave input, the quadratic term produces a DC shift and a second harmonic. The cubic term changes the fundamental and adds a third harmonic. How much of each reaches the output depends on the operating point, load, circuit and signal level.[2][16]
Hamm measured the harmonic components as he increased input level to 12 dB above each circuit’s 1% third-harmonic reference. His examples show how different the results can be.[2]
- Two-stage single-ended 12AY7 triode preamp. The second harmonic led and the third followed closely. The fourth began rising about 3–4 dB later on the input-level axis.
- Two-stage single-ended pentode preamp. The third harmonic led. The second began rising about 3 dB later on the input-level axis. The fourth and fifth were also prominent.
- Discrete transistor preamps. The third harmonic led. Several higher harmonics rose beyond a break point and the waveforms clipped nearly symmetrically.
- Hybrid-output op-amp example. The third, fifth and seventh harmonics rose steeply and the even harmonics were strongly suppressed.
What listeners can hear
The second harmonic sits one octave above the fundamental of a single tone. Hamm described it as adding body to an instrument’s timbre. In playback, the audibility of an added harmonic depends on its level, frequency and the rest of the programme. Higher-order nonlinearities can create products farther from the components that mask them. With music, intermodulation products also matter.[2][3]
In Lee and Geddes’s 2003 study, 34 listeners rated 21 audio stimuli. The correlation between the GedLee metric and mean ratings was r = 0.68 for all 21 stimuli. After the authors excluded the three stimuli with the highest metric values, the correlation for the remaining 18 was r = 0.95 (p < 0.001). THD and IMD were not statistically significant predictors in the full set. A THD reading gives the combined level of the harmonics, so the spectrum adds information about what a circuit is doing.[4]
Overload, Headroom and Recovery
Clean headroom is the available signal swing before a circuit reaches a specified distortion or clipping limit. Hamm also discussed a perceptual overload margin. In his informal listening tests, some preamps could be driven farther past a distortion reference before their output sounded objectionable. That observation applies to the sampled circuits and test conditions.[1][2]
- Level changes the spectrum. As the input moves into a more curved part of the transfer characteristic or begins clipping, the harmonic balance can change. Measurements at several levels show how that change develops.[2]
- Transients can expose limited headroom. A close-miked drum or vocal peak can overload a stage if its peak level exceeds the available swing. Gain settings and input attenuation determine when that happens.[1]
- Some circuits have a measurable recovery time. In an AC-coupled stage, grid conduction on a large peak can charge the coupling capacitor and shift the bias temporarily. The capacitor discharges through the grid circuit, so the response depends on preceding signal peaks.[17]
The Output Transformer
A conventional tube power stage uses an output transformer to reflect a low-impedance speaker load as a higher load at the tube plates. The impedance ratio is the square of the turns ratio. The primary and secondary are electrically isolated, so the speaker is separated from the plate supply.[8]
Primary inductance and the effective circuit resistance set the low-frequency response together. At the high end, leakage inductance, winding capacitance, and source and load impedances affect bandwidth and can produce roll-off or peaking. Windings and core size are selected for the intended load, signal level and bandwidth.[8][10]
Single-ended bias and magnetic asymmetry
In a transformer-coupled single-ended stage, the standing plate current flows through the primary and magnetizes the core. An air gap helps the core accommodate the DC magnetization, at the cost of lower primary inductance for the same winding and core. DC bias can make the magnetic response asymmetric and introduce even-order distortion. The transformer’s contribution combines with any same-order distortion from the tube stage according to relative phase. Its size depends on the core, frequency, signal level and driving impedance.[8][9][10]
For a hi-fi output transformer, strong core saturation is an overload condition that can raise low-frequency distortion. In a balanced push-pull stage, opposing quiescent currents largely cancel their DC flux. A well-balanced pair can also cancel much of the even-order distortion generated in that pair. Mismatch and distortion from other stages can leave even harmonics at the output.[8][10]
Output Impedance and Loudspeaker Loading
In Bussey and Haigler’s guitar-amplifier comparison, the solid-state power amp had an output impedance below 0.1 Ω and the tube amp was above 5 Ω. These are measurements of two designs. Damping factor is the nominal load impedance divided by the magnitude of the amplifier’s output impedance at the stated frequency. The value depends on the amplifier, tap and test conditions.[6][11]
A loudspeaker’s impedance varies in both magnitude and phase. Together with amplifier output impedance it forms a voltage divider. In the linear equivalent circuit, the speaker-terminal voltage is the source voltage multiplied by Zspeaker / (Zout + Zspeaker), with both impedances evaluated at the same frequency. A substantially resistive source impedance tends to make voltage relatively higher around impedance peaks and lower around dips. Bass resonance, crossover features and treble impedance can all affect the shape. The difference may be audible depending on the pairing and the size and width of the response change.[5][6][7][11]
With low source impedance, the speaker’s impedance curve has less effect on the amplifier’s terminal voltage. Speakers still differ in acoustic response. For a fixed amplifier and speaker operating in their linear range, the voltage-response change can be approximated with EQ matched to that pairing.[6][11]
Feedback and Noise
Global negative feedback
In a stable voltage-feedback loop that senses the output, feedback can reduce distortion and output impedance where loop gain is sufficient. When the transformer is inside the loop, its phase shift affects the feedback and compensation the design can use. The treble response still depends on transformer bandwidth, load and the rest of the circuit.[6][8][19]
Noise depends on the circuit
Space charge can reduce plate-current shot noise relative to full shot noise. Flicker noise, grid-current noise, and in multi-grid tubes partition noise can also contribute. Which contribution dominates depends on the tube, bias, source impedance and measurement bandwidth. One published table reports about 0.6–1.4 µV RMS input-referred for selected small-signal triode samples over 30 Hz–8 kHz. A comparison with a FET circuit requires comparable source and measurement conditions.[14][15]
Can FETs or DSP Reproduce It?
A single-ended JFET stage can generate second-harmonic distortion from the approximately quadratic part of its transfer curve. Its bias, degeneration and load affect the result. Reproducing a particular tube circuit may also require matching its frequency response, clipping, recovery and the load it was designed to drive.[17][20]
DSP can model both the static transfer curve and time-dependent effects such as bias recovery. Universal Audio’s Century Tube Channel Strip provides tube-style saturation, EQ and optical compression in an original software design. Its 175B and 176 plug-ins emulate specific studio compressors. Universal Audio reports that its plug-ins were indistinguishable from those compressors in its own blind listening tests.[12][13][17]
In a separate 1981 double-blind study, Bussey and Haigler compared a Fender Twin Reverb with a specific solid-state system. They matched gain, frequency response and clipping level as applicable. In the power-amp tests, 1 of 12 listeners reliably detected a difference without clipping and 2 of 6 did so with clipping. In the complete-system comparison, none of 9 reliably detected a difference. The listener who had consistently identified the tube amp’s buzz was unavailable for that comparison. The authors linked the buzz to power-supply ripple intermodulation in the tested amp. These findings apply to that guitar rig and those conditions.[11]
With a specific loudspeaker, a model can include the measured response of that amplifier–speaker pair. Driving a different passive speaker through a low-impedance power amp changes the physical interaction, so reproducing the response may require different modelling or EQ.[6][11][17]
What the Evidence Supports
| Common claim | What the evidence supports |
|---|---|
| Tubes clip more gently. | In Hamm’s samples, the aligned THD curves had broadly similar slopes. The audible overload margin varied across the tested designs.[1][2] |
| Tube distortion is mainly even-order. | Hamm’s single-ended triode example had strong second and third harmonics. His pentode example had a dominant third harmonic. Balance can reduce even-order products in a push-pull pair.[2][8] |
| Second harmonic always sounds warm. | Audibility and perceived character depend on the amount of distortion, frequency and programme material.[2][3][4] |
| Output transformers can affect tone. | Finite bandwidth and magnetic distortion can change the output. Core design, drive level and the connected circuit set the magnitude.[8][10] |
| High output impedance always boosts the bass. | Relative changes in speaker-terminal voltage depend on the impedance curve, including its bass, crossover and treble features.[6][11] |
| Tubes have a lower noise floor than FETs. | Noise comparisons require specified devices, source impedances, operating points and bandwidth.[14][15] |
| Digital processing cannot reproduce tube behaviour. | Models can reproduce selected behaviours. A comparison depends on the target circuit, its load and the test conditions.[11][13][17] |
Practical Takeaways
- Match the speaker to the rated tap. Start with the speaker’s nominal impedance and the amplifier manufacturer’s instructions. Compare other taps only if the manual permits it.[18]
- Check the transformer at the intended load and level. For a single-ended design, include the rated DC current. Check low-frequency bandwidth and distortion under those conditions.[8][10]
- Look at the spectrum across levels. Compare individual harmonics at more than one output level and frequency, especially if the amplifier will be driven close to clipping.[2][4]
- Measure the speaker-terminal response. A resistor provides a useful reference. The intended speaker or an impedance-simulating load reveals the effect of source impedance.[6][11]
- Match levels when comparing gear. If the goal is to isolate distortion or overload behaviour, check the frequency response and clipping level under the test load too.[11]
Output Transformer Examples
Compare a single-ended and a push-pull model using IWISTAO’s listed specifications. Match the primary impedance and speaker taps to your circuit.
IWISTAO · WVTR-OT5W
Single-Ended Transformer
Rated power5 W each
Primary impedance5 kΩ
Speaker taps4 Ω / 8 Ω
Primary DC current50 mA
12 H primary inductance. Sold as a pair.
Shop WVTR-OT5WIWISTAO · WVTRPP-20W
Push-Pull Transformer
Rated power20 W
Primary impedance6 kΩ
Speaker taps4 Ω / 8 Ω
CoreZ11 EI steel
Choose with or without an ultra-linear tap.
Shop WVTRPP-20WFrequently Asked Questions
Why can a single-ended amp have a strong second harmonic?
An asymmetric transfer curve can produce even-order harmonics. In a balanced push-pull pair, much of that pair’s even-order distortion can cancel. Bias, load and signal level still matter.[2][8][16]
Do tube amps always clip more gently?
Clipping behaviour depends on the circuit and which stage overloads first. Hamm observed a wider range of subjectively tolerable overload in some of his preamps. The result applies to the conditions he tested.[1][2]
Why does the same tube amp sound different with different speakers?
Both speakers have their own acoustic response. If the amplifier has substantial output impedance, differences in the speakers’ impedance curves also change the voltage delivered at different frequencies.[6][11]
Shop Output Transformers
Choose a model for your output stage and check the winding options on its product page. Specifications are from IWISTAO.
IWISTAO · WVTR-OT5W
Single-Ended Transformer
Rated power5 W each
Primary impedance5 kΩ
Speaker taps4 Ω / 8 Ω
Primary DC current50 mA
12 H primary inductance. Sold as a pair.
Shop WVTR-OT5WIWISTAO · WVTRPP-20W
Push-Pull Transformer
Rated power20 W
Primary impedance6 kΩ
Speaker taps4 Ω / 8 Ω
CoreZ11 EI steel
Choose with or without an ultra-linear tap.
Shop WVTRPP-20WFind More
References
- Hamm, R. O. “Tubes Versus Transistors: Is There an Audible Difference?” Journal of the Audio Engineering Society 21(4), 267–273, 1973. Reprint, Part 1. https://butleraudio.com/tubesvstrans1.html
- Hamm, R. O. Same paper. Full journal copy with Figures 2–14 and the discussion of overload. https://milbert.com/Files/articles/TvsT/tstxt.pdf
- Geddes, E. R., and Lee, L. W. “Auditory Perception of Nonlinear Distortion – Theory.” AES 115th Convention, 2003. https://www.gedlee.com/Papers/Distortion_AES_I.pdf
- Lee, L. W., and Geddes, E. R. “Auditory Perception of Nonlinear Distortion.” AES 115th Convention, paper 5891, 2003. See Tables 5 and 6. https://gedlee.azurewebsites.net/Papers/Distortion_AES_II.pdf
- SoundStage! Getting Technical, May 2004. Paul Barton interview on amplifier and loudspeaker interaction. https://www.soundstagenetwork.com/gettingtechnical/gettingtechnical200405.htm
- Atkinson, J. “Questions of Impedance Interaction.” Stereophile, 1993. Measurements with real loudspeakers. https://stereophile.com/content/questions-impedance-interaction
- Toole, F. E. “Loudspeaker Measurements and Their Relationship to Listener Preferences: Part 2.” Journal of the Audio Engineering Society 34(5), 323–348, 1986. https://aes.org/publications/elibrary/elibrary-page/?id=5270
- Whitlock, B. “Audio Transformers.” Handbook for Sound Engineers, 3rd ed. Sections 1.2, 1.3 and 2.1.5. https://www.jensen-transformers.com/wp-content/uploads/2014/08/Audio-Transformers-Chapter.pdf
- IWISTAO. “Single-Ended Output Transformers: Core Size, DC Bias, and the Art of the Air Gap.” https://iwistao.com/blogs/iwistao/single-ended-output-transformers-core-size-dc-bias-and-the-art-of-the-air-gap
- Lundahl Transformers. “FAQ.” Single-ended air gaps, inductance and output transformer load. https://www.lundahltransformers.com/faq/
- Bussey, W. S., and Haigler, R. M. “Tubes Versus Transistors in Electric Guitar Amplifiers.” IEEE ICASSP, pp. 800–803, 1981. https://leachlegacy.ece.gatech.edu/papers/TubeVsTrans.pdf
- Universal Audio. “Century Tube Channel Strip: The Complete Guide.” https://www.uaudio.com/blogs/ua/century-tube-channel-strip-the-complete-guide-uad-explore-free
- Universal Audio. “UA 175B & 176 Tube Compressor Collection.” Product description and manufacturer blind-test claim. https://www.uaudio.com/products/ua-tube-compressor-collection
- HomeTheaterHiFi.com. “Dual Triode Vacuum Tubes 12AX7 and 12AU7.” See the reported sample noise results over 30 Hz–8 kHz. https://hometheaterhifi.com/technical/technical-reviews/dual-triode-vacuum-tubes-12ax7-and-12au7-why-they-are-so-popular-for-use-in-audio-amplifiers/
- Pearson, G. L. “Fluctuation Noise in Vacuum Tubes.” Bell System Technical Journal 13(4), 634–653, 1934. https://www.nokia.com/bell-labs/publications-and-media/publications/fluctuation-noise-in-vacuum-tubes/
- Santagata, F., Sarti, A., and Tubaro, S. “Non-Linear Digital Implementation of a Parametric Analog Tube Ground Cathode Amplifier.” DAFx, 2007. https://www.dafx.de/paper-archive/2007/Papers/p169.pdf
- Pakarinen, J., and Karjalainen, M. “Enhanced Wave Digital Triode Model for Real-Time Tube Amplifier Emulation.” IEEE Transactions on Audio, Speech, and Language Processing 18(4), 738–746, 2010. https://research.spa.aalto.fi/publications/papers/IEEETASLP-wdftube/
- McIntosh. MC275 Stereo Power Amplifier Owner’s Manual. Speaker connection and load matching. https://www.mcintoshlabs.com/-/media/Files/mcintoshlabs/DocumentMaster/us/Legacy/MC275v1-VINTAGE-E-OWNERS.pdf?revision=b68599ac-a1d9-4b57-b2b0-9e5f1969a423
- Analog Devices. Application Note AN-356. Effect of negative feedback on linearity and output impedance. https://www.analog.com/media/en/technical-documentation/application-notes/an-356.pdf
- Analog Devices. “RFI Rectification Concepts.” MT-096. Square-law JFET example. https://www.analog.com/media/en/training-seminars/tutorials/MT-096.pdf
Related Posts
0 comments