Choke input power supply is a great thing! Nowadays I'm refusing to build anything with cap input and currently I'm making a choke input separate PSU for a DAC + tube preamp - 7 separate transformers and 11 chokes in total - it's heavy! 😀
There are some "underwater stones" I would like to share though:
1. Sometimes very high rectifier PIV is needed, especially if the choke and capacitor following the choke are high value and the sum of the Rdc of rectifier + PT + choke are low. When B+ is applied abruptly (no soft start), the capacitor after the choke is not charged and represents short circuit to ground. When the rectifier conducts, high current passed through the choke as it saturates, while the capacitor hasn't charged yet and still represents very low impedance. When the rectifier goes (off) state, the saturated choke suddenly sees a very high impedance circuit and it tries keeping the same current. Guess what can happen?
The choke has to be with a high insulation.
PSUD2 does help.
2. Critical inductance (Lc) vs critical load (Rc) has to be thought carefully. If the criteria are not matched (Lc too low or Rc too high), the following B+ tends to be above the Urms*0,9 choke input rule and tries to reach the Urms*1,41 cap input rule. Make sure you put a minimum critical load - a bleeder resistor across B+ to ground can help.
To prevent wasting power, I've put a string of zener diodes on my amplifier with a heatsink. Their breakdown voltage is chosen so they activate in a case of a fault (power tube failure for ex.)
3. A choke for a choke input duty can be noisy. It swings a very high voltage amplitude and can see large inrush current, so it needs to have very good insulation between windings and good mechanical strength. Sometimes it needs to be poted. It also emits quite more of a magnetic field than a choke after a capacitor, so it has to be far away from sensitive electronics. C core chokes are better in this.
When having spare time, I'm winding my own "higher end" on double bobbin single C core, splitting their bobbins to at least 6 sections for low parasitic capacitance, so that less crap goes into the circuitry and so the choke has a higher self resonant frequency, preferably above the audio band. There's a photo of one of my chokes:
Of course the idea of winding the chokes this way is not mine.
Regards,
Alexander.
There are some "underwater stones" I would like to share though:
1. Sometimes very high rectifier PIV is needed, especially if the choke and capacitor following the choke are high value and the sum of the Rdc of rectifier + PT + choke are low. When B+ is applied abruptly (no soft start), the capacitor after the choke is not charged and represents short circuit to ground. When the rectifier conducts, high current passed through the choke as it saturates, while the capacitor hasn't charged yet and still represents very low impedance. When the rectifier goes (off) state, the saturated choke suddenly sees a very high impedance circuit and it tries keeping the same current. Guess what can happen?
The choke has to be with a high insulation.
PSUD2 does help.
2. Critical inductance (Lc) vs critical load (Rc) has to be thought carefully. If the criteria are not matched (Lc too low or Rc too high), the following B+ tends to be above the Urms*0,9 choke input rule and tries to reach the Urms*1,41 cap input rule. Make sure you put a minimum critical load - a bleeder resistor across B+ to ground can help.
To prevent wasting power, I've put a string of zener diodes on my amplifier with a heatsink. Their breakdown voltage is chosen so they activate in a case of a fault (power tube failure for ex.)
3. A choke for a choke input duty can be noisy. It swings a very high voltage amplitude and can see large inrush current, so it needs to have very good insulation between windings and good mechanical strength. Sometimes it needs to be poted. It also emits quite more of a magnetic field than a choke after a capacitor, so it has to be far away from sensitive electronics. C core chokes are better in this.
When having spare time, I'm winding my own "higher end" on double bobbin single C core, splitting their bobbins to at least 6 sections for low parasitic capacitance, so that less crap goes into the circuitry and so the choke has a higher self resonant frequency, preferably above the audio band. There's a photo of one of my chokes:
Of course the idea of winding the chokes this way is not mine.
Regards,
Alexander.
Attachments
Alexander thanks for sharing those experiences and observations. That choke looks awesome - would love to do that. One day.
The PSU will be in a separate box from the phono stage and I'll be mounting the chokes to maximise distance from each other and at right angles where possible, diagonals where not. I have bleeds in both the psu and the Phono stage, the latter being used to elevate the heaters for v2. Both b+ and b- should see around 34mA.
The PSU will be in a separate box from the phono stage and I'll be mounting the chokes to maximise distance from each other and at right angles where possible, diagonals where not. I have bleeds in both the psu and the Phono stage, the latter being used to elevate the heaters for v2. Both b+ and b- should see around 34mA.
Alexander - did you sleeve your interconnect and terminal wiring to the choke ? One terminal wire in the photo looks to be very close to the core.
Hello,
The wire is wound and secured, to not go any closer to the core. The distance between the wire and the core is 1mm of air. The other potential wires on the bobbin have 1mm distance of bobbin material.
The wire is wound and secured, to not go any closer to the core. The distance between the wire and the core is 1mm of air. The other potential wires on the bobbin have 1mm distance of bobbin material.
The biggest 'problem' with LCLC… is that depending on that first L, it can seriously sap the output voltage … having gained not much in exchange. If you insist (like I do!) on using a pair of chokes, at the very least put a small C in front of the first L. How small? I like to arrange it so that the first C won't have an RMS ripple more than 5% of the peak working voltage of the secondary. And you have to take into consideration whether you're doing full-wave or half-wave rectification.
For instance, to the first approximation:
F = Hz
I = nominal continuous amperes
VRMS = RMS rating of secondary
T = multiplier 1 for half-wave bridge, 2 for full wave bridge config
Vripple ≈ I/TFC which isn't RMS but is peak-to-trough ripple.
Its not the most elegant or mathematically perfect equation, but it captures the idea: the amount of ripple will be the droop from rectified peak to next peak, which is a function of frequency, type of rectification, mean current flow and capacitance.
So, working it backward, with that 3% in mind:
C = I/(TFVRMS × 1.414 × 0.05)
Now for a real example…
I = 2500 ma. VRMS = 70 V. T = 2 (FWB).
C = 2.5 /( 2 × 60 × 70 × 1.414 × 0.05 )
C = 4,200 μF (use 3,900 μF or 4,700 μF for standard inexpensive values)
Sounds like a lot? Well, the power supply will be delivering in this example two and a half amps continuous. That takes some pretty good sized capacitor to hold and droop only 3% (ripple of 3% of 70 × 1.414 = 2.97 V)
Or, try a classic tube amplifier's power supply, in Europe, 50 Hz mains:
F = 50
T = 2 (full wave)
VRMS = 350
I = 300 ma
C = 0.300 / (2 × 50 × 350 × 1.414 × 0.05)
C = 121 (use 100 μF to 150 μF for standard values)
These recommendations are larger than what one sees (especially for vacuum tube amplifiers), but realistically, if you purchase heavy-duty ultrafast silicon rectifiers, there's no real problem having the larger initial C. That, and your output will be within 2.5% of its zero-current-drain theoretical peak.
GoatGuy
For instance, to the first approximation:
F = Hz
I = nominal continuous amperes
VRMS = RMS rating of secondary
T = multiplier 1 for half-wave bridge, 2 for full wave bridge config
Vripple ≈ I/TFC which isn't RMS but is peak-to-trough ripple.
Its not the most elegant or mathematically perfect equation, but it captures the idea: the amount of ripple will be the droop from rectified peak to next peak, which is a function of frequency, type of rectification, mean current flow and capacitance.
So, working it backward, with that 3% in mind:
C = I/(TFVRMS × 1.414 × 0.05)
Now for a real example…
I = 2500 ma. VRMS = 70 V. T = 2 (FWB).
C = 2.5 /( 2 × 60 × 70 × 1.414 × 0.05 )
C = 4,200 μF (use 3,900 μF or 4,700 μF for standard inexpensive values)
Sounds like a lot? Well, the power supply will be delivering in this example two and a half amps continuous. That takes some pretty good sized capacitor to hold and droop only 3% (ripple of 3% of 70 × 1.414 = 2.97 V)
Or, try a classic tube amplifier's power supply, in Europe, 50 Hz mains:
F = 50
T = 2 (full wave)
VRMS = 350
I = 300 ma
C = 0.300 / (2 × 50 × 350 × 1.414 × 0.05)
C = 121 (use 100 μF to 150 μF for standard values)
These recommendations are larger than what one sees (especially for vacuum tube amplifiers), but realistically, if you purchase heavy-duty ultrafast silicon rectifiers, there's no real problem having the larger initial C. That, and your output will be within 2.5% of its zero-current-drain theoretical peak.
GoatGuy
Last edited:
I can't see how that is a problem for a cleansheet design. The real power available from the transformer is probably a little higher - voltage and current output levels are designed to suit.
It sounds like your preference "I like to arrange it so that the first C won't have an RMS ripple more than 5% of the peak working voltage of the secondary" would make the filter effectively a capacitor input filter, not an inductor input filter. So a very different beast than what the OP was looking for.
I can see the benefit in having a small level of C as input to an inductor input filter - for the purposes of rectifier induced transient V & I management - but not aiming to reduce the AC input voltage presented to the inductor. Sometimes I've increased that first C to a few uF for the purpose of raising the B+ to suit a particular amplifier - but that was for guitar application, so a very different design rationale to the OP's intent.
Ciao, Tim
Yes, ripple of 5% on the first C means that it is a cap input PSU; for a genuine choke input PSU you must have 100% ripple at that point. One of the points of a choke input PSU is the reduction in output voltage; it is a feature, not a problem! It also makes better use of the transformer VA rating.
The practice of using a few uF for the first cap is surprisingly popular but then you get the worst of both worlds: peaky charging pulses (like C input) and high output impedance at low currents (like L input). Useful if that is what you want, but I suspect that it is often used by people who don't understand how a choke input PSU works.
The practice of using a few uF for the first cap is surprisingly popular but then you get the worst of both worlds: peaky charging pulses (like C input) and high output impedance at low currents (like L input). Useful if that is what you want, but I suspect that it is often used by people who don't understand how a choke input PSU works.
Thanks. Definitely don't want cap led. However interesting to work through the calc and check myself.
Based on the above e formula in my example,
F 50hz
I 34mA
Vrms 424 (check my understanding the secondary is 300-0-300, so Vrms is 600 / 1.414... Not at all sure this is right)
T = 2 (fwr)
Target ripple say 5%
C = 19nF
Given that I want choke led LCLC, as several people have said it's beneficial to have a small cap before the first choke I'll experiment. How would I calculate that?
Based on the above e formula in my example,
F 50hz
I 34mA
Vrms 424 (check my understanding the secondary is 300-0-300, so Vrms is 600 / 1.414... Not at all sure this is right)
T = 2 (fwr)
Target ripple say 5%
C = 19nF
Given that I want choke led LCLC, as several people have said it's beneficial to have a small cap before the first choke I'll experiment. How would I calculate that?
Do the calculation which GoatGuy suggested, but using 100% for the ripple factor instead of 5%. This will estimate the biggest cap you can use and still have something like a choke input supply. Take half this value. Anything between this and zero capacitance should be OK. I am not convinced that any cap here is necessary, for three reasons:
1. there will always be some stray capacitance.
2. provided that the rectifier diodes are reasonable well matched then one will be turning on as another is turning off so there should be no time that the choke is left open circuit so no back emf to worry about.
3. choke input supplies from the 1950s never seem to put a cap here.
1. there will always be some stray capacitance.
2. provided that the rectifier diodes are reasonable well matched then one will be turning on as another is turning off so there should be no time that the choke is left open circuit so no back emf to worry about.
3. choke input supplies from the 1950s never seem to put a cap here.
DF96, RhythMick, TRobbins - You all make valid points, especially that I prefer CLCLC over LCLC, which was the OP's design goal. So, I guess I should apologize, for leading the discussion astray.
TRobbins - Looking inward, I think my real “beef” with LCLC is that it tends toward producing substantial over-voltage when load is close to zero. Noting that nominally, a FWB→LCLC will produce VRMS as DC (less series choke resistance times quiescent current), and slightly less (I*R loss) under double, triple load … it has always bugged me that at the opposite end the arrangement can produce up to 40% higher voltage under light (near zero) current draw. And, even if that isn't "bad" for an amplifier, and is expected within its operating range, then it still represents a significant over-voltage condition which must certs impact dynamic performance a whole bunch.
DF96 - I concur: if the goal is LCLC then one cannot reasonably put a “C” in front and call it LCLC. Contrary to your view “provided diodes … matched … no time choke left open …”, there's alway that dastardly VFORWARD voltage drop of diodes right around 0.7 V to 2.0 V (depending on PIV) to contend with. A little capacitance will solve the back-emf problem … if one's being a purist about it.
Point 3 - 1950s never put cap there - is also well made, tho' I really think it has to be balanced in the context of “caps were expensive, chokes were cheap”. We tend to forget that in practical terms, the 40 μF choke of 1952 cost (according to my DIY audio builder of that era) was like $25 apiece in bulk (today's dollars). And you couldn't get 'em very big in any case. Triple section and double section (single can) was popular. Cost per μF was optimized (I guess the can cost a bunch!)
RhythMick - if you don't want a C in front … you don't want a C in front. If that's for some sort of perceived sonic purity - hey! Go for it, mate! If you don't mind the VRMS under-load droop (or the opposite, loadless rise toward Vpeak), then again … go for it! Its not like there aren't hundreds of examples of LCLC working just fine, especially from a “clean sheet” perspective.
Thanks all for the discussion. As Uncle said once, “if capacitors were free, we'd never use chokes. If chokes were free, never use capacitors. Then there are those dâmned resistors (chuckle…)”
GoatGuy
TRobbins - Looking inward, I think my real “beef” with LCLC is that it tends toward producing substantial over-voltage when load is close to zero. Noting that nominally, a FWB→LCLC will produce VRMS as DC (less series choke resistance times quiescent current), and slightly less (I*R loss) under double, triple load … it has always bugged me that at the opposite end the arrangement can produce up to 40% higher voltage under light (near zero) current draw. And, even if that isn't "bad" for an amplifier, and is expected within its operating range, then it still represents a significant over-voltage condition which must certs impact dynamic performance a whole bunch.
DF96 - I concur: if the goal is LCLC then one cannot reasonably put a “C” in front and call it LCLC. Contrary to your view “provided diodes … matched … no time choke left open …”, there's alway that dastardly VFORWARD voltage drop of diodes right around 0.7 V to 2.0 V (depending on PIV) to contend with. A little capacitance will solve the back-emf problem … if one's being a purist about it.
Point 3 - 1950s never put cap there - is also well made, tho' I really think it has to be balanced in the context of “caps were expensive, chokes were cheap”. We tend to forget that in practical terms, the 40 μF choke of 1952 cost (according to my DIY audio builder of that era) was like $25 apiece in bulk (today's dollars). And you couldn't get 'em very big in any case. Triple section and double section (single can) was popular. Cost per μF was optimized (I guess the can cost a bunch!)
RhythMick - if you don't want a C in front … you don't want a C in front. If that's for some sort of perceived sonic purity - hey! Go for it, mate! If you don't mind the VRMS under-load droop (or the opposite, loadless rise toward Vpeak), then again … go for it! Its not like there aren't hundreds of examples of LCLC working just fine, especially from a “clean sheet” perspective.
Thanks all for the discussion. As Uncle said once, “if capacitors were free, we'd never use chokes. If chokes were free, never use capacitors. Then there are those dâmned resistors (chuckle…)”
GoatGuy
Think about what a choke will do when it finds its source of current being cutoff as the AC waveform approaches zero. It will try to force current to continue by dropping its input terminal voltage below zero! Hence there is always a diode switched on. This is a feature of choke input PSUs which the books never seem to mention, and which I took years to realise. Spice should model this OK; I'm not sure about PSUD2.GoatGuy said:
A small cheap film cap would cure the problem, if the problem existed. Much cheaper than a choke, even back then.
Let's be clear about this. There are two good reasons why someone might use a choke input PSU:
1. he wants 0.9xRMS instead of 1.4xRMS as the DC output
2. he wants to make best use of the transformer VA rating
The downside is that he then needs to ensure that he meets the minimum current spec, in order to get a constant output voltage (apart from resistive drops).
I suppose there could be a third reason for a guitar amp:
3. he wants to violate the minimum current spec, because he wants a very saggy PSU and this is what choke input does with too little current drawn.
The forced commutation between diodes (and PT windings) is somewhat abrupt. Softening the dI/dt through the diodes, and dV/dt across the diodes by the use of low value cap is the benefit in my book. PSUD2 does a fair job of showing that, but just idealises the inductor.
The added capacitor at the front of an LC filter needs a suitable VAC rating at twice mains frequency - for B+ supplies that was probably not within the ratings of 1950's caps except maybe for oil/film caps.
I certainly prefer the better PF obtained with a choke input filter - from the perspective of reducing higher frequency main harmonics and the ubiquitous flat-topping caused by capacitor input filtering - but even then PF is not 1.
Certainly the issue of startup voltage is a problem for valve amps, before B+ loading kicks in - a point that can be glossed over when choosing B+ operating level and capacitor voltage ratings.
The added capacitor at the front of an LC filter needs a suitable VAC rating at twice mains frequency - for B+ supplies that was probably not within the ratings of 1950's caps except maybe for oil/film caps.
I certainly prefer the better PF obtained with a choke input filter - from the perspective of reducing higher frequency main harmonics and the ubiquitous flat-topping caused by capacitor input filtering - but even then PF is not 1.
Certainly the issue of startup voltage is a problem for valve amps, before B+ loading kicks in - a point that can be glossed over when choosing B+ operating level and capacitor voltage ratings.
A swinging choke is a good answer to that particular issue; it may sound fancy and complicated, but it can be as simple as 2 or 4 E's magnetically shorting the gap for a E-I core.
Regarding the small input C, a possibility I think would be worth exploring is to treat it as a snubber, with some series R: it will resist the voltage increase in a large measure, and help damp unwanted resonant oscillations. To be tested obviously, that's just an idea thrown in the air...
I want whatever sounds right, whatever makes feet tap when I have friends around, whatever makes my heart sing in those rare quiet listening moments. So far the best I've heard is choke input, but one of the advantages of DIY is the ability to listen to advice and experiment, and that I plan to do.
Unfortunately Percy the devil seems to have pissed on my soldering station - dammit.
Unfortunately Percy the devil seems to have pissed on my soldering station - dammit.
I can see that the issue of snubbing each PT secondary half-winding (assuming full-wave CT style) needs to be kept in mind. The winding current at diode commutation is not zero, as in a capacitor input filter, and so leakage inductance energy is far more prominent.
Similarly, any discontinuity at the point of diode commutation will cause an additional choke voltage disturbance, let alone the inherent large step in dV/dt. However, I doubt there is any substantial period of time, when the choke is not part of the PT secondary circuit, for the choke to resonate through that small C and back via the LC filter capacitance. Most largish filter inductors would have a SRF in the low to mid kHz - except of course for Alexander's custom choke (did you measure L and SRF Alexander?).
Abrupt in whose terms? Much worse than a cap input PSU?trobbins said:
I think we can agree that the choke does not suffer any discontinuity in current (which is what most people seem to fear), but the power transformer secondaries will so it is they which may require some help.
An exception: when the DC current draw is too low then the choke will be cutoff for part of the time. In fact, when you do the maths, the condition for critical current is that the choke minimum current during the AC half-cycle reaches zero. However, as the choke has reached zero current while connected there is still no back emf to worry about.
For an LC filter with continuous conduction operating condition, the conducting ss diode will have something like the average DC load current passing at the time of commutation. The dI/dt during commutation depends on the dV/dt of the mains frequency waveform at zero crossing (the highest dV/dt portion of a sine) raising the other ss diode voltage to on-state-level, and choke current commutating over.
A capacitor input filter is likely to reduce ss diode current to zero at a more sedate pace (the mains frequency dV/dt transitioning the diode voltage from on to off is lower), and the current is already ramped down to zero.
LTSpice would be needed to quantify a comparison, and component value choices would need to be made that allowed a fair comparison. PSUD2 doesn't allow a good comparison, but does indicate that choke input filter will cause harder switching diode conditions. That is the main reason why I reckon some awareness is needed for choke input filtering with ss diodes.
Choke current will want to remain constant, but choke shunt capacitance will have some influence due to choke dV/dt having its highest level at mains zero crossing (diode commutation), and so there will be a transient flow of current at that time. Not a topic I've seen in reference books or tech papers - probably as the effective choke capacitance is likely only a few hundred pF, and valve diodes would have smoothed out the diode commutation portion of time and dV/dt, and who ever did choke input filtering once ss diodes were common place!
A capacitor input filter is likely to reduce ss diode current to zero at a more sedate pace (the mains frequency dV/dt transitioning the diode voltage from on to off is lower), and the current is already ramped down to zero.
LTSpice would be needed to quantify a comparison, and component value choices would need to be made that allowed a fair comparison. PSUD2 doesn't allow a good comparison, but does indicate that choke input filter will cause harder switching diode conditions. That is the main reason why I reckon some awareness is needed for choke input filtering with ss diodes.
Choke current will want to remain constant, but choke shunt capacitance will have some influence due to choke dV/dt having its highest level at mains zero crossing (diode commutation), and so there will be a transient flow of current at that time. Not a topic I've seen in reference books or tech papers - probably as the effective choke capacitance is likely only a few hundred pF, and valve diodes would have smoothed out the diode commutation portion of time and dV/dt, and who ever did choke input filtering once ss diodes were common place!
Last edited:
For UF4007 : Avg I is 1.0A. Surge is 30A, C is 17pF
For UF5408 : 3A, 150A, 36pF
MUR8100E : 16A, 100A, 30pF
So I go back to my original question. Bearing in mind I'm a relative newbie anyway but for semiconductors I've no clue - what are the criteria to use to select a diode (or bridge rectifier) for use in this design ?
- Status
- Not open for further replies.