My Cart
Your Shopping Cart is Empty
Winding Your Own Filter Choke for Tube Amplifiers-A More Rigorous Engineering Guide
|
0 comments
posted by Vincent Zhang
Published by IWISTAO
From first principles to a finished component — how to design, wind, assemble, and verify a high-quality choke for tube-amplifier power supplies and related high-voltage circuits.
Table of Contents
- Why Use a Choke?
- How a Choke Works
- Types of Chokes in Tube Amplifiers
- Design Targets: Current, Inductance, DCR, and Insulation
- Core Choices: EI, Toroidal, and C-Core
- The Air Gap and DC Bias
- Inductance, Turns, and Wire Sizing
- Winding Procedure
- Assembly, Gapping, and Varnishing
- How to Measure a Finished Choke Properly
- Three Worked Examples
- Troubleshooting
- Practical Design Tables
- Conclusion
- References
1. Why Use a Choke?
After rectification, the B+ rail of a tube amplifier is not pure DC. With full-wave rectification on 50 Hz mains, the dominant ripple component appears at 100 Hz. If this ripple is insufficiently filtered, it can modulate the amplifier stages and produce audible hum.
A simple capacitor-input supply can reduce ripple, but it also has trade-offs:
- High peak charging current into the first capacitor
- Poorer regulation under changing load
- Higher stress on rectifiers and transformer windings
- Greater HF switching noise and EMI from pulse charging currents
A choke adds series inductance, which resists rapid current variation. In a choke-input or CLC (π) filter, that inductance reduces ripple current, lowers charging-current peaks, and often improves regulation.
A More Accurate Way to Describe Ripple Reduction
It is tempting to treat an LC section as an ideal second-order low-pass filter and write attenuation directly as a function of cutoff frequency. That is useful as a high-frequency approximation, but it is not exact for real tube-amplifier power supplies.
For an ideal undamped LC network, the resonant frequency is:
f0 = 1 / (2π√(LC))
For example, with:
- L = 10 H
- C = 47 µF
then:
f0 = 1 / (2π√(10 × 47 × 10-6)) ≈ 7.3 Hz
Since 100 Hz is far above 7.3 Hz, the LC section will strongly attenuate ripple. In the idealized high-frequency region, a second-order response falls at approximately 40 dB/decade. That explains why a 10 H / 47 µF section can provide dramatically more ripple reduction than a capacitor alone.
However, in a real amplifier supply, the actual attenuation depends on:
- load resistance,
- choke DCR,
- capacitor ESR,
- rectifier source impedance,
- and damping/Q of the network.
So the key engineering point is this: a properly chosen choke-capacitor section can reduce ripple much more effectively than a capacitor alone, but the exact dB figure depends on the complete circuit.
2. How a Choke Works
A choke is an inductor, and its basic law is:
V = L(dI/dt)
This means the voltage across the winding is proportional to how quickly current is trying to change. In a power supply, ripple current is an alternating component superimposed on a DC load current. The choke impedes that alternating component while allowing the DC component to pass.
Energy storage in the magnetic field is given by:
E = (1/2)LI2
That stored energy helps smooth current flow between rectifier charging intervals.
Why a Magnetic Core Is Needed
At 100 Hz, an air-core inductor with several henries of inductance would be physically enormous. A ferromagnetic core raises the inductance enormously for a given number of turns because magnetic permeability is much higher than that of air.
But the advantage comes with a limit: magnetic saturation.
If flux density B is driven too high, permeability collapses and the effective inductance drops sharply. In a choke carrying DC current, saturation can be caused not only by ripple current, but by the DC component itself. That is why a practical power-supply choke almost always needs a gapped magnetic circuit.
3. Types of Chokes in Tube Amplifiers
The original classification is useful, but it is worth stating more carefully.
3.1 Power-Supply Filter Choke
This is the most common type. It sits in the B+ line and carries substantial DC current. Typical ranges are:
- 5–30 H
- 50–400 mA
- DCR chosen according to allowable voltage drop and power loss
This type almost always requires an air gap.
3.2 Screen-Supply Choke
Used to decouple a screen grid rail from the main B+ line. Current is lower, often only tens of milliamps. Because DC current is lower, the core can be smaller and the design can favor higher inductance.
3.3 Anode (Plate) Choke
A plate choke is used as a high-impedance load in certain stages, especially in preamps and some specialty output or driver circuits. Its design priorities are usually:
- high inductance,
- low copper loss,
- low distributed capacitance,
- adequate insulation,
- and correct behavior under DC bias.
A plate choke is not automatically a no-gap device. If it carries significant unidirectional DC, then gap design must still be considered. Whether the core is interleaved, butt-stacked, or deliberately gapped depends on the actual operating current and the allowable AC swing.
3.4 Cathode Choke / Cathode Bypass Choke
Less common, but used in some designs to provide AC impedance in the cathode circuit without relying entirely on an electrolytic bypass capacitor. Current may be low, but linearity and impedance at audio frequencies become more important.
4. Design Targets: Current, Inductance, DCR, and Insulation
Before choosing a core or winding a turn, define four design targets.
4.1 DC Current Rating
The first target is the maximum continuous DC current the choke must carry without unacceptable loss of inductance or overheating.
A sensible DIY rule is to design for about 120% of expected quiescent current.
Typical ranges:
- preamp-only supply: 5–30 mA
- screen supply: 10–50 mA
- small SE output stage supply: 50–150 mA
- medium PP amplifier supply: 150–300 mA
- larger PP amplifier supply: 300–500 mA
4.2 Inductance
Higher inductance generally means better ripple suppression and smoother current, but only within practical limits of core size, copper loss, cost, and DC bias handling.
A commonly cited relationship is:
L = 1 / (4π2f2C)
That is better understood as a resonance-based reference value, obtained by rearranging the LC resonance equation. It tells you where the LC resonant frequency would sit, but it does not by itself determine the minimum useful choke value for a real CLC power supply.
For example, with C = 47 µF:
L = 1 / (4π2 × 1002 × 47 × 10-6) ≈ 54 mH
This is mathematically correct, but 54 mH is not a practical substitute for a 5–10 H choke in a typical B+ filter. It only shows the inductance corresponding to a 100 Hz LC resonance point.
For real tube power supplies, practical starting points are more like:
- 5–10 H for many B+ filter chokes
- 10–20 H or more when current is lower and stronger smoothing is desired
- values chosen together with capacitor size, allowable sag, rectifier limits, and DCR
4.3 DC Resistance (DCR)
Copper resistance causes voltage drop and heat:
Vdrop = IDC × DCR
Ploss = IDC2 × DCR
These equations are exact and should always be checked.
A reasonable design aim is often to keep the choke’s DC voltage drop within roughly 5–10% of B+, though the acceptable value depends on the amplifier’s target operating point.
Likewise, copper-loss power should stay within a level that the winding and enclosure can dissipate safely. In many DIY cases, keeping copper loss to a few watts or less is a good starting point.
4.4 Insulation and Voltage Rating
The choke must survive both steady-state and startup conditions. During warm-up, B+ can exceed the normal operating voltage before the tubes begin drawing current.
A good conservative rule is to rate winding insulation to at least:
1.5× expected peak DC stress
Proper interlayer insulation is especially important in supplies above a few hundred volts.
5. Core Choices: EI, Toroidal, and C-Core
5.1 EI Laminated Core
For DIY work, EI laminations remain the easiest and most forgiving choice.
Advantages:
- easy to source,
- easy to gap,
- mechanically robust,
- tolerant of experimental adjustment,
- straightforward to rewind or modify
Disadvantages:
- more leakage flux than toroids,
- can buzz if not clamped well,
- larger and heavier for a given performance target
For most first-time builders, EI is the best starting point.
5.2 Toroidal Core
Toroids have low external stray field and can be very efficient magnetically. But when the design must tolerate DC current, introducing a controlled and repeatable air gap is difficult.
So the correct engineering statement is not “toroids are unsuitable,” but rather:
Toroids are more difficult to use for DC-biased choke service, especially in DIY construction.
They can still work in certain low-current or specialized roles, but they are usually not the easiest choice for a first choke build.
5.3 C-Core
C-cores can provide excellent magnetic performance and low leakage flux, but they are less convenient for many DIY builds and often cost more.
5.4 Core Materials
Specific permeability figures vary widely with alloy, rolling direction, frequency, gap, and flux density. They should be read as typical ranges, not fixed constants.
For most low-frequency tube-amp choke work, silicon-steel laminations remain the standard practical choice.
6. The Air Gap and DC Bias
This is the heart of choke design.
A DC current through N turns produces magnetizing force:
H = NI / le
Without a gap, a high-permeability core can saturate at surprisingly modest DC current. Adding an air gap lowers effective permeability and makes the inductance more stable under DC bias.
Effective Permeability
A useful approximation for a gapped magnetic circuit is:
μe = μr / (1 + μrlg/le)
From it:
lg = le((1/μe) - (1/μr)) ≈ le / μe
when μr ≫ μe.
This approximation is valid and practical for first-pass design.
Practical Meaning
A larger gap:
- reduces effective permeability,
- lowers inductance for a given turns count,
- increases saturation tolerance under DC bias,
- often makes the design more predictable
So the usual trade-off is:
- more gap = more DC tolerance, less inductance
- less gap = more inductance, less DC tolerance
This is why power chokes almost always end up as a compromise among core size, turns, gap, copper resistance, and current rating.
7. Inductance, Turns, and Wire Sizing
7.1 Inductance Formula
For a gapped magnetic circuit, a useful engineering approximation is:
L = (μ0μeN2Ae) / le
Rearranging for turns:
N = √(L le / (μ0μeAe))
These formulas are appropriate for initial design work, provided all dimensions are kept in consistent SI units.
7.2 Window Fill
Real bobbins do not allow 100% packing efficiency. A conservative fill factor around 0.35–0.45 is often more realistic for hand winding, especially if interlayer insulation is used.
7.3 Wire Gauge
Wire sizing must satisfy three constraints at once:
- current density and temperature rise,
- window fill,
- acceptable DCR.
For a DIY choke, slightly heavier wire is often beneficial if space allows, because it lowers copper loss and keeps temperature down.
But heavier wire also reduces turns capacity. So wire size must be chosen together with target inductance and core window area.
8. Winding Procedure
A practical choke winding process looks like this:
-
Define the target
Decide current, inductance, allowable DCR, and insulation class. -
Choose the core and bobbin
EI laminations are easiest for first builds. -
Estimate turns and gap
Use the inductance and effective-permeability formulas for a first-pass design. -
Check window fill
Confirm that turns, insulation, and wire gauge fit the bobbin realistically. -
Wind in neat layers
Keep tension consistent, but do not over-tension the enamel wire. -
Use interlayer insulation
Particularly important in high-voltage designs. -
Bring out secure terminations
Mechanical reliability matters as much as electrical performance. -
Assemble the core with an initial gap
Shim the gap consistently and clamp the laminations firmly. -
Measure and adjust
Final gap often needs empirical refinement.
9. Assembly, Gapping, and Varnishing
After winding, the mechanical build matters.
- Clamp laminations tightly to reduce acoustic buzz
- Keep the air gap symmetrical and repeatable
- Use suitable varnish or impregnation when possible
- Avoid sharp edges that may damage insulation
- Mount the choke away from power transformers and rotate magnetic axes where practical
10. How to Measure a Finished Choke Properly
This is one of the most important corrections.
A small-signal LCR meter reading at 1 kHz may be useful for a quick comparison, but it is not enough to validate a power-supply choke intended for 100/120 Hz service under DC bias.
A proper evaluation should include:
- DC resistance
- insulation integrity
- temperature rise
- inductance under conditions close to real use
- behavior under DC bias
So instead of saying “measurement without DC is meaningless,” the more precise statement is:
Small-signal, zero-bias inductance measurement can be informative, but it does not fully predict real performance in a DC-biased 100/120 Hz power-supply choke.
Recommended Checks
DCR
Measure with a good ohmmeter or four-wire method if possible.
Inductance
Prefer measurement at low frequency, ideally near the intended operating range, and compare zero-bias and biased results if possible.
Bias Behavior
If the choke is intended for 100–300 mA DC service, then verifying inductance under representative current is much more meaningful than relying on an unloaded bench reading.
Temperature Rise
Run the choke at rated DC current and check that it stabilizes within a safe temperature.
Mechanical Noise
A quiet electrical design can still be a poor finished component if the core buzzes audibly.
11. Three Worked Examples
The following examples should be treated as design illustrations, not universal recipes.
Example A: 300B Single-Ended B+ Choke
Target:
- 120 mA
- 10 H nominal
Illustrative solution:
- EI-86 class core
- about 1600 turns
- AWG 30 class wire
- total air gap around 0.34 mm
- DCR around 86 Ω
Check:
Vdrop = 0.12 × 86 ≈ 10.3 V
Ploss = 0.122 × 86 ≈ 1.24 W
This is a sensible result for a 380 V class supply.
Example B: EL34 Push-Pull Supply Choke
Target:
- 250 mA
- 7 H nominal
Illustrative solution:
- EI-114 class core
- about 1200 turns
- AWG 26 class wire
- total air gap around 0.52 mm
- DCR around 28 Ω
Check:
Vdrop = 0.25 × 28 = 7.0 V
Ploss = 0.252 × 28 = 1.75 W
Example C: Screen Supply Choke
Target:
- 30 mA
- 20 H nominal
Illustrative solution:
- EI-66 class core
- about 3200 turns
- AWG 36 class wire
- total gap around 0.09 mm
- DCR around 380 Ω
Check:
Vdrop = 0.03 × 380 = 11.4 V
Ploss = 0.032 × 380 ≈ 0.34 W
This is acceptable in many screen-supply or decoupling roles, where current is low and a modest voltage drop is not critical.
12. Troubleshooting
Choke Runs Hot
- DCR too high for the current
- Current exceeds design value
- Insufficient ventilation
Inductance Lower Than Expected
- Gap too small and core saturating under DC
- Turns count lower than intended
- Measurement method unsuitable
Ripple Rejection Worse Than Expected
- Insufficient downstream capacitance
- Incorrect load assumptions
- Actual inductance under bias much lower than nominal
- Excessive rectifier/source impedance interactions
Mechanical Hum or Buzz
- Loose laminations
- Inadequate clamping
- Magnetostriction
- Poor mounting practice
Very Low Resistance or Shorted Turns
- Enamel damage during winding
- Layer insulation failure
- Winding compression or abrasion
13. Practical Design Tables
The following quick-reference tables are practical estimates rather than hard limits.
Approximate Core-Class Starting Points
| EI Class | Typical Current Range | Typical Inductance Range | Typical Use |
|---|---|---|---|
| EI-57 | up to ~50 mA | up to ~20–30 H | preamp, light screen supply |
| EI-66 | up to ~100 mA | up to ~10–20 H | low-current B+, screen or plate choke |
| EI-86 | up to ~200 mA | up to ~5–15 H | medium-power B+ choke |
| EI-96 | up to ~300 mA | up to ~5–12 H | higher-current B+ choke |
| EI-114 | up to ~500 mA | up to ~3–10 H | large PP amplifier supply |
These values depend strongly on gap, turns, copper fill, allowable DCR, and target temperature rise.
Example Application Guide
| Application | Current | Typical Choke Target |
|---|---|---|
| 2A3 / 300B SE B+ | 60–120 mA | 5–10 H |
| EL34 SE B+ | 90–150 mA | 5–10 H |
| EL34 PP pair B+ | 150–250 mA | 5–7 H |
| KT88 PP pair B+ | 180–300 mA | 5–7 H |
| Screen supply | 10–50 mA | 10–20 H or more |
Again, these are starting points, not absolute design laws.
14. Conclusion
Designing and winding a filter choke for a tube amplifier is one of the most instructive magnetic-design exercises in audio. It forces the builder to confront the real interaction of:
- inductance,
- DC bias,
- saturation,
- copper loss,
- insulation,
- and mechanical construction.
The most important refinements are simply these:
- do not treat ideal LC attenuation formulas as exact real-world predictions,
- do not confuse LC resonance-based calculations with a true minimum practical choke value,
- do not assume a plate choke is automatically ungapped,
- and do not rely solely on small-signal inductance measurements taken without DC bias.
Get those points right, and the rest of the process becomes much more dependable.
A well-designed DIY choke can absolutely be an excellent component — provided it is designed for the actual current, actual frequency range, actual insulation stress, and actual thermal limits of the amplifier it will serve.
Find More
The Role of Choke Coils in Tube Amplifiers
15. References
- Aiken, R. — Chokes Explained, Aiken Amplification Technical Information.
URL: http://www.aikenamps.com/index.php/chokes-explained - Turner, R. — Audio Filter Chokes: Design and Construction, Turner Audio (2017).
URL: https://turneraudio.com.au/audiofilterchokes.html - Gamma Electronics — The Design and Construction of Low Frequency Chokes (2024).
URL: https://www.gammaelectronics.xyz/coil-design_5.html - Fitzsimmons, F. — Building Audio Frequency Choke Coils, Radio Craft, October 1932. Archived at RF Café.
URL: https://www.rfcafe.com/references/radio-craft/building-af-choke-coils-october-1932-radio-craft.htm - Erickson, R.W. & Maksimović, D. — DC Inductor Design Using Gapped Cores, in Fundamentals of Power Electronics, 3rd ed., Springer (2020).
URL: https://coefs.charlotte.edu/mnoras/files/2013/03/Transformer-and-Inductor-Design-Handbook_Chapter_8.pdf - Snelling, E.C. — Soft Ferrites: Properties and Applications, 2nd ed. Butterworth-Heinemann (1988).
- Crowhurst, N.H. — Basic Audio, Volume 1, John F. Rider (1959).
- Lundahl Transformers — Choke Product Catalogue.
URL: https://www.lundahltransformers.com/chokes/ - Hammond Manufacturing — Choke Application Notes & Product Data.
URL: https://www.hammfg.com/electronics/transformers/choke - Radio Designer’s Handbook, 4th ed., F. Langford-Smith (ed.), Iliffe & Sons (1953).
- diyAudio Forum — Choke Design Guide thread (2023).
URL: https://www.diyaudio.com/community/threads/choke-design-guide.405905/ - GroupDIY Audio Forum — Air Gap: Graphical Solution for Optimising Inductors.
URL: https://groupdiy.com/threads/air-gap-graphical-solution-for-optimizing.52744/ - Colorado State University — Lecture 33: Inductor Design, ECE 562 Power Electronics.
URL: https://www.engr.colostate.edu/ECE562/98lectures/l33.pdf - IWISTAO Blog — The Role of Choke Coils in Tube Amplifiers (2025).
URL: https://www.iwistaoblog.com/2025/03/the-role-of-choke-coils-in-tube.html - Munro, D. — PSUD2: Power Supply Unit Designer v2.
URL: https://duncanamps.com/psud2/
0 comments