DF96 - series-connected heaters (whether A/C or DC powered, regulated or not) have a secret "gotcha" that parallel wired heaters do not: nominal current flow differences.
Now tube specs call for (both "conveniently" and on purpose, tho' long forgotten!) very specific milliampere current ratings at the nominal operating voltage. This is so that tubes "of the same current-draw family" can be daisy-chained together, filament-wise. The kind of Competitive-ideal for 1960s and early 1970s tube sets was to daisy-chain them all together to the point where the voltage would equal 117 VAC line voltage. Then another transformer could be eliminated.
But with this comes the gotcha.
For instance, the 12AX7 (in 12 volt mode) requires 150 ma. Therefore it has an operating equivalent resistance of R = E/I = 12/0.15 = 80 ohms. The 12BZ7, pin compatible and sometimes recommended as a high-current substitute, uses 300 ma in its 12 volt configuration. R = E/I = 12/0.30 = 40 ohms.
Now... were you to put that 12BZ7 in series with another 12AX7 (and supposing for simplicity, there are only these 2 tubes in the filament line), then the CCS is going to deliver 150ma, heating the 12AX7 correctly, but not getting close to what the 12BZ7 would want. Clearly... not optimal.
And in the same vein, were you to be using a constant VOLTAGE supply for these series connected tubes (again, admitting it is somewhat silly, as a configuration), then what would happen? The 24 volts (nominal) would be divided by the two equivalent resistances, and the 12BZ7 (40 ohms) would receive 1/3 (8 volts), and the 12AX7 would receive the other 2/3 (16 volts) ... both would be substantially outside their spec range.
This is why - in the end, especially for the TV market, where there was a LOT of impetus to be competitive, price-wise - why many manufacturers eschewed the series-connected tube situation. This, and because as tubes age, that "cathode-to-filament" potential has a tendency to get all squirrely, making for poorer overall reliability of the TV sets. Reputation, of course, doesn't want that... bad for future sales. Remember Dumont?
GoatGuy
Now tube specs call for (both "conveniently" and on purpose, tho' long forgotten!) very specific milliampere current ratings at the nominal operating voltage. This is so that tubes "of the same current-draw family" can be daisy-chained together, filament-wise. The kind of Competitive-ideal for 1960s and early 1970s tube sets was to daisy-chain them all together to the point where the voltage would equal 117 VAC line voltage. Then another transformer could be eliminated.
But with this comes the gotcha.
For instance, the 12AX7 (in 12 volt mode) requires 150 ma. Therefore it has an operating equivalent resistance of R = E/I = 12/0.15 = 80 ohms. The 12BZ7, pin compatible and sometimes recommended as a high-current substitute, uses 300 ma in its 12 volt configuration. R = E/I = 12/0.30 = 40 ohms.
Now... were you to put that 12BZ7 in series with another 12AX7 (and supposing for simplicity, there are only these 2 tubes in the filament line), then the CCS is going to deliver 150ma, heating the 12AX7 correctly, but not getting close to what the 12BZ7 would want. Clearly... not optimal.
And in the same vein, were you to be using a constant VOLTAGE supply for these series connected tubes (again, admitting it is somewhat silly, as a configuration), then what would happen? The 24 volts (nominal) would be divided by the two equivalent resistances, and the 12BZ7 (40 ohms) would receive 1/3 (8 volts), and the 12AX7 would receive the other 2/3 (16 volts) ... both would be substantially outside their spec range.
This is why - in the end, especially for the TV market, where there was a LOT of impetus to be competitive, price-wise - why many manufacturers eschewed the series-connected tube situation. This, and because as tubes age, that "cathode-to-filament" potential has a tendency to get all squirrely, making for poorer overall reliability of the TV sets. Reputation, of course, doesn't want that... bad for future sales. Remember Dumont?
GoatGuy
Your so-called 'gotcha' will only catch incompetent designers, who ignore datasheets. Only a moron would expect a 150mA heater and a 300mA heater to work together in series connection. No secret. Series-connected heaters were routinely used in European TV sets right up to the move to solid-state TVs.
Similarly, paralleling heaters with different voltage requirements doesn't work either. As this is all obvious, it is unclear to me what point you are making by mentioning it.
Similarly, paralleling heaters with different voltage requirements doesn't work either. As this is all obvious, it is unclear to me what point you are making by mentioning it.
Sigh... I give up. Dude... there are SO many people on this forum that don't have significant clues when it comes to the advice they read. I mean, its pathetic. You obviously not only 'get it', but consider the hapless abrogations of the obvious as bone-headed to such a degree as to not be worth talking about.
My point is - that IF someone thinks they're going to blithely assert that constant current filament sources are good practice, AND there are plenty of caveats that are perhaps too subtle for our good audience, THEN I'm going to remind people why its not so simple.
That's it. The whole point.
Anyway... I'm done. If y'all want to use CCS supplies, by all means do so. Just watch out in the case of filaments in parallel. Constant VOLTAGE supplies are way better because the're not very finicky re: various kinds of substitute tubes (and ones that go open).
GoatGuy
OK. I agree that clueless people do appear on here. They are often those most interested in making tiny, fashionable 'improvements' to a poor design, rather than fixing the design.
I think DC on indirect heaters will give too small benefit.
Many old schematics are available, where even directly heated power tubes are used with AC. (813,...) !
Be 100% sure and verify everything you change in circuit, is not just placebo effect. 🙂
Many old schematics are available, where even directly heated power tubes are used with AC. (813,...) !
Be 100% sure and verify everything you change in circuit, is not just placebo effect. 🙂
Well, it is neither to uncork champagne. 😀
Not if my eyes or the distance, but this here is called ripple.
What is its frequency ?
Ach no ... 'I did not say' this and 'I did not say' that and whatever else .....
GoatGuy,
If at all possible, may we for a moment dispense with the refined disdain and other emotions expressed in your several previous posts?
What I read what you do say, is that a constant current supply to a valve heater (or for that matter any lamp using a tungston/whatever metal heater/filament), forces a higher voltage at the inrush - voltage drop over said heater/filament, I take it.
If Ohms Law is still valid, such higher voltage must indicate a correspondingly higher initial resistance load to said constant current source than somewhat later at steady state, after the inrush is completed. (That is, when, where the load is a lamp filament, it has arrived at operating temperature.)
If correct so far, what source of information on metallurgy indicates that a lamp/valve filament has a higher resistance when cold than when hot? Where can I read about that?
Somewhere in the above logic you must disagree. Please tell us in plain language where, and what the real truth is.
Thereafter you have my permission to indicate what an oaf/moron/dislexic/senile/whatever human being you think I am (even though we have never met) - as far as Moderators will let that pass.
Kindly?
On CCS supplied filaments...
It was pointed out that the situation with filaments is exactly as you cite: they have a lower resistance when cold, and it increases as they warm. This was posted by the first counter-argument.
I was under the mistaken (memory) understanding that tungsten had a somewhat negative resistance thermal coefficient. Another poster showed that with a positive (normal) coefficient, a constant-current source would act as a kind of slow-turn-on device.
I acknowledge both of these points. The subsequent points are what make constant-current under most common (parallel filament) topologies a poor choice. I attempted to go back to the original post, and strike my first two points. However, more than a half-hour had elapsed, and this forum blocks such edits after that.🙁
After reading all this ... I remain convinced that CCS is a terrible idea for parallel-filament designs. It might work well with series-connected filaments, but only should all of them need exactly 150 (or 300, or whatever) milliamps for proper heating. Since tubes' heaters are rated for particular voltages, where the current flow is "nominal" ... it just stands to reason that constant voltage supplies, as banal as a direct transformer winding, or as refined as a CRC filter and a slow-ramp-up constant-voltage supply ... would do the best job.
Now... as to your utility as a fine human being Johan Potgeiter... that is a different question, which I am broadly unqualified to answer. LOL - 😉
GoatGuy
It was pointed out that the situation with filaments is exactly as you cite: they have a lower resistance when cold, and it increases as they warm. This was posted by the first counter-argument.
I was under the mistaken (memory) understanding that tungsten had a somewhat negative resistance thermal coefficient. Another poster showed that with a positive (normal) coefficient, a constant-current source would act as a kind of slow-turn-on device.
I acknowledge both of these points. The subsequent points are what make constant-current under most common (parallel filament) topologies a poor choice. I attempted to go back to the original post, and strike my first two points. However, more than a half-hour had elapsed, and this forum blocks such edits after that.🙁
After reading all this ... I remain convinced that CCS is a terrible idea for parallel-filament designs. It might work well with series-connected filaments, but only should all of them need exactly 150 (or 300, or whatever) milliamps for proper heating. Since tubes' heaters are rated for particular voltages, where the current flow is "nominal" ... it just stands to reason that constant voltage supplies, as banal as a direct transformer winding, or as refined as a CRC filter and a slow-ramp-up constant-voltage supply ... would do the best job.
Now... as to your utility as a fine human being Johan Potgeiter... that is a different question, which I am broadly unqualified to answer. LOL - 😉
GoatGuy
Last edited:
Meanwhile back at the LM317..... what's wrong with an LM317 as a regulator for a filament? Are we that picky? The LM317 has got to put out a fairly decently clean 6.3v with that kind of constant load. (?)
Its going to put out very nice 6.3 ... if you calc the resistor-network wisely. Using a 5K1 resistor from output pin to adjustment pin, and a 25K adjustable trimpot ... you can quickly dial in whatever voltage amuses you. Then if you like, turn the whole thing off, and measure the pot setting. Replace with a resistor for long-term stability.
You can also get the slow-turn-on by simply putting a capacitor in parallel with the trimpot of a value that depends more on the adjust-pin leakage current than on the value of the trimpot! I just did a napkin calc, and got a 1 sec RC time constant for a 47uF 6.3V cap.
GoatGuy
You can also get the slow-turn-on by simply putting a capacitor in parallel with the trimpot of a value that depends more on the adjust-pin leakage current than on the value of the trimpot! I just did a napkin calc, and got a 1 sec RC time constant for a 47uF 6.3V cap.
GoatGuy
GoatGuy,
Ah! Thank you kind Sir!
I have found a quite unexpected variation in heater current (up to 25%) for the same type tube, even of the same make! Being easier to supply dc at higher voltage lower current (thus all heaters in series) I nevertheless had to abandon the idea (excepting the 12A*** types), using 6,3V instead (5,5V across some heaters, almost 7V across others with constant current!). Actually one would suppose that the meaningful parameter is the cathode temperature, the closest to which I came by feeding constant power into a number of heaters.
Result, neither the same voltage nor the same current for that to occur! Blimey; I would have thought there is tighter control over heater wire guage for the same tubes. Thus one settles for 6.3V; possibly the 'equal power' concept is academic. Voltage spread is at least defined in the specs, not so current.
Thus constant voltage .... as said above. I use regulators mainly to get rid of ripple spikes after the necessary large value of storage capacitor. Smaller capacitor, high enough input voltage to regulator to always keep spikes above output, clean dc. (I did in my early years find some interference from the h.f. content of a rippled/spiked dc in a sensitive input stage.)
But I believe I am rather off topic regarding regulator/no regulator. Interesting to read other views.
Ah! Thank you kind Sir!
I have found a quite unexpected variation in heater current (up to 25%) for the same type tube, even of the same make! Being easier to supply dc at higher voltage lower current (thus all heaters in series) I nevertheless had to abandon the idea (excepting the 12A*** types), using 6,3V instead (5,5V across some heaters, almost 7V across others with constant current!). Actually one would suppose that the meaningful parameter is the cathode temperature, the closest to which I came by feeding constant power into a number of heaters.
Result, neither the same voltage nor the same current for that to occur! Blimey; I would have thought there is tighter control over heater wire guage for the same tubes. Thus one settles for 6.3V; possibly the 'equal power' concept is academic. Voltage spread is at least defined in the specs, not so current.
Thus constant voltage .... as said above. I use regulators mainly to get rid of ripple spikes after the necessary large value of storage capacitor. Smaller capacitor, high enough input voltage to regulator to always keep spikes above output, clean dc. (I did in my early years find some interference from the h.f. content of a rippled/spiked dc in a sensitive input stage.)
But I believe I am rather off topic regarding regulator/no regulator. Interesting to read other views.
Johan, thank you for the information. Indirectly, you confirmed something which I encountered, thought was a "fluke", but you too have experienced: that for a given heater voltage, the current draw of exactly the same 'model' of tube varies by ±25%!!!
One would nominally hope that the manufacturers would have tighter specs, but from what I hear, even seemingly insignificant impurities in the tungsten alloy that is used to draw the wire, and a few percent variation in the die opening (as it wears...) cause this variation. Tube making machines however don't have adaptive algorithms that feed exactly the right amount of wire ... they do it just by formula.
That said, your notion of running the filaments at constant power is (again, unsurprisingly) the thing that's working the best. Bravo! Unfortunately, in the realm of small analog electronic devices, there is no easy "multiplier" to transform voltage and current into "power" ... which then could be linearly regulated.
[This is not exactly true ... there are a large number of simple FET-as-variable-resistor multiplier circuits such as http://www.kingjoe.net/image004.gif out there...
[With such a circuit, one could pretty trivially reconfigure an LM317 to become a constant-power device ... utilizing a small-value (4.7 ohm) resistor in series with the output as a current-sense signal, and of course the output itself as the other source for the multiplication to get 'relative power' (uncalibrated exactly, not necessary to do so)]
Indeed, if one were to invest in making these circuits on a small PC board, I imagine with quad op-amps you could stuff at least 2 of them, with a few "driver transistors" to do the regulation. In fact, I'm going to think about how to minimally solve this one.
Good luck...
GoatGuy
One would nominally hope that the manufacturers would have tighter specs, but from what I hear, even seemingly insignificant impurities in the tungsten alloy that is used to draw the wire, and a few percent variation in the die opening (as it wears...) cause this variation. Tube making machines however don't have adaptive algorithms that feed exactly the right amount of wire ... they do it just by formula.
That said, your notion of running the filaments at constant power is (again, unsurprisingly) the thing that's working the best. Bravo! Unfortunately, in the realm of small analog electronic devices, there is no easy "multiplier" to transform voltage and current into "power" ... which then could be linearly regulated.
[This is not exactly true ... there are a large number of simple FET-as-variable-resistor multiplier circuits such as http://www.kingjoe.net/image004.gif out there...
[With such a circuit, one could pretty trivially reconfigure an LM317 to become a constant-power device ... utilizing a small-value (4.7 ohm) resistor in series with the output as a current-sense signal, and of course the output itself as the other source for the multiplication to get 'relative power' (uncalibrated exactly, not necessary to do so)]
Indeed, if one were to invest in making these circuits on a small PC board, I imagine with quad op-amps you could stuff at least 2 of them, with a few "driver transistors" to do the regulation. In fact, I'm going to think about how to minimally solve this one.
Good luck...
GoatGuy
Do you know what is the ripple I am getting with the unregulated power supply?
Does it really matter? Think hard!
What you r doing is like witch hunting!
May be you want to enlighten me on the following:
What do I do if the ripple is
2mVpp 1MHz
5mVpp 200kHz
10mVppp 400Hz
0.04Vpp 1kHz
0.1Vpp 50Hz
or all of the above couple together. SO what is the implications?
I do not know if you are measuring the voltage or trying to measure the scope.
Last edited:
No, I was just trying to help, like many, many others.
You're the one who asked for help, even having all the answers.
If my experience in DC heaters, do not like, or do not serve you, I'm sorry.
I withdraw from this thread.
DC regulation on heaters. Its not a bad concept. but the things that have to be taken in account.
1. its better to build it on a separate transformer than the plate supply. the reason behind this is the heater winding does tax the transformer and putting a regulator tax it even more. and in worse case senario caused poor plate voltage regulation.
2. due to the voltage fold-over effect. the regulator needs current headroom as much as 50%. without this headroom, the semiconductor operating temp would be high enough that the component would not survive without a heat sink.
3. RF snubbing is needed. As there is a lot of RF noise that generated by these types of regulators and most of the time that noise has a tendency to effect the gain of high mu tubes.
1. its better to build it on a separate transformer than the plate supply. the reason behind this is the heater winding does tax the transformer and putting a regulator tax it even more. and in worse case senario caused poor plate voltage regulation.
2. due to the voltage fold-over effect. the regulator needs current headroom as much as 50%. without this headroom, the semiconductor operating temp would be high enough that the component would not survive without a heat sink.
3. RF snubbing is needed. As there is a lot of RF noise that generated by these types of regulators and most of the time that noise has a tendency to effect the gain of high mu tubes.
DavesNotHere
In our DIY world where we can slap on gobs of capacitance and our linear regulators are almost always going into a static load are our linear regulators generating RF noise?
I'm just a hobbyist and not an engineer so I am really trying to understand a mechanism for this if it is a problem.
In our DIY world where we can slap on gobs of capacitance and our linear regulators are almost always going into a static load are our linear regulators generating RF noise?
I'm just a hobbyist and not an engineer so I am really trying to understand a mechanism for this if it is a problem.
DavesNotHere...
[1] NO: a properly specified transformer with 15% to 20% capacity headroom is in no way "taxed" by either nominal heater current, nor inrush heater current. We're not talking about cheap Chinese (Korean, whatever) guitar amps for kiddies, but "real" amps, audiophile grade. Conscience driven design. Puhleeze... don't need FUD here.
[2] INRUSH current (not voltage fold-over!) causes higher-than-nominal current to flow when tube filaments are cold. That can, and should be accounted for. A simple gate-to-source shorted MOSFET (depletion-mode) can act as a $0.75 current limiter. Cheap enough that each tube can have its own!, yet all still driven by a constant-voltage supply.
[3] Pure FUD (fear, uncertainty, doubt for the newbies). There is nothing at all about semiconductor voltage regulators, with the usual "follow the manufacturer's casebook design" recommendations for both pre-regulator and post-regulator small-value electrolytic capacitors ... that would lead to such a conclusion. EVEN if used "naked" (without capacitors), the high-resolution oscilloscope readings show only a few dozen millivolts of noise at frequencies above 100 kHz. Hence... why even a small ceramic disk capacitor is enough to eliminate. Don't oversell the RFI thing, dude. Its disingenuous.
Folks... its simple: if you want to use a regulated DC power source for your heaters ... by all means it is harmless enough to do, and will make your circuit happy. Remember though: the inrush current on a 12AX7 which nominally likes 150 milliamps when warmed up, may be as high as 500 milliamps (and some higher!) ... this is significant enough that when quite a few are placed in parallel, that one really does need to make sure that the regulator(s) are stout enough, and well-heat-sinked enough to handle the additional load.
Use of depletion mode MOSFETS-as-current-regulators is an excellent way to keep the inrush current contained, as well as having the benefits (if any) of voltage regulation. Finally though ... be aware that since tube filaments all heat up at different rates (due to the mass of both the filament windings and the surrounding cathode cylinder), that if you're going to use current-limiting, it should be "per tube". It can be by "class of tube" if you're being that scrimping in cost... but at something like 50 cents apiece... depletion-mode MOSFETS are amongst the cheapest add-ons you can buy.
GoatGuy
[1] NO: a properly specified transformer with 15% to 20% capacity headroom is in no way "taxed" by either nominal heater current, nor inrush heater current. We're not talking about cheap Chinese (Korean, whatever) guitar amps for kiddies, but "real" amps, audiophile grade. Conscience driven design. Puhleeze... don't need FUD here.
[2] INRUSH current (not voltage fold-over!) causes higher-than-nominal current to flow when tube filaments are cold. That can, and should be accounted for. A simple gate-to-source shorted MOSFET (depletion-mode) can act as a $0.75 current limiter. Cheap enough that each tube can have its own!, yet all still driven by a constant-voltage supply.
[3] Pure FUD (fear, uncertainty, doubt for the newbies). There is nothing at all about semiconductor voltage regulators, with the usual "follow the manufacturer's casebook design" recommendations for both pre-regulator and post-regulator small-value electrolytic capacitors ... that would lead to such a conclusion. EVEN if used "naked" (without capacitors), the high-resolution oscilloscope readings show only a few dozen millivolts of noise at frequencies above 100 kHz. Hence... why even a small ceramic disk capacitor is enough to eliminate. Don't oversell the RFI thing, dude. Its disingenuous.
Folks... its simple: if you want to use a regulated DC power source for your heaters ... by all means it is harmless enough to do, and will make your circuit happy. Remember though: the inrush current on a 12AX7 which nominally likes 150 milliamps when warmed up, may be as high as 500 milliamps (and some higher!) ... this is significant enough that when quite a few are placed in parallel, that one really does need to make sure that the regulator(s) are stout enough, and well-heat-sinked enough to handle the additional load.
Use of depletion mode MOSFETS-as-current-regulators is an excellent way to keep the inrush current contained, as well as having the benefits (if any) of voltage regulation. Finally though ... be aware that since tube filaments all heat up at different rates (due to the mass of both the filament windings and the surrounding cathode cylinder), that if you're going to use current-limiting, it should be "per tube". It can be by "class of tube" if you're being that scrimping in cost... but at something like 50 cents apiece... depletion-mode MOSFETS are amongst the cheapest add-ons you can buy.
GoatGuy
first factor is thermal noise from the series regulator transistor, second is internal switching noise from the internal zener diode.
this noise can be injected into the cathode due to the heater to cathode diode effect. then any noise that is injected on a typical grounded cathode amp will be amplified in reverse phase on the plate.
all high gain tubes have natural oscillations on the plate due to the acceleration of electrons hitting the plate. when the rf energy gets too great it causes a decrease in gain.
historically the case of the 417A preamp tube when used as a high gain amp like a phono stage has a natural tendency to cause poor gain because of this phenomenon. It is usually fixed by injecting positive feedback from plate to grid by a small value capacitor.
this noise can be injected into the cathode due to the heater to cathode diode effect. then any noise that is injected on a typical grounded cathode amp will be amplified in reverse phase on the plate.
all high gain tubes have natural oscillations on the plate due to the acceleration of electrons hitting the plate. when the rf energy gets too great it causes a decrease in gain.
historically the case of the 417A preamp tube when used as a high gain amp like a phono stage has a natural tendency to cause poor gain because of this phenomenon. It is usually fixed by injecting positive feedback from plate to grid by a small value capacitor.
Not buying it, Dave, sorry.
You are technically correct - there must be thermal noise from a regulator transistor, and internal zener-breakdown noise from the voltage reference. Vnoise(z) is on the order of low millivolts, typically. Schott noise from the transistor similarly low. A capacitor on the output squelches both very effectively, if it is of concern. Alternately, small series inductors would do. However the whole idea really would benefit from a simple A/B switch test. I would put money on "can't be heard" by anyone, no matter how golden the ear.
Now, whether high-gain tubes have their own noisy plate dynamics is a different matter. One nominally tries to avoid the effects of Miller capacitance, so the idea of ADDING plate-to-grid low-value capacitors is a bit ... counter-intuitive. Yet, if the goal is just to ensure that the tube has a natural roll-off above say "50 kHz" or something, then I can see the utility of clamping the higher frequencies through such an added capacitor. Its one thing to have to deal with a parasitic capacitance that one can't do much about (Miller), and its another to take it by the horns, and decide to engineer a particular solution.
PS: you've made a good argument for using multiple stages instead of fewer, with the problems of very-hi-mu tubes being stated thus. Moreover, using multiple stages diminishes the distortion effect of non-cascode connected triodes in the front stages. The only reason to use fewer tubes are [cost] and [appeal to audiophiles who desire simplicity]. I prefer more tubes, lower amplification factors. I must be an old fashioned fuddy-duddy, I guess.
GoatGuy
You are technically correct - there must be thermal noise from a regulator transistor, and internal zener-breakdown noise from the voltage reference. Vnoise(z) is on the order of low millivolts, typically. Schott noise from the transistor similarly low. A capacitor on the output squelches both very effectively, if it is of concern. Alternately, small series inductors would do. However the whole idea really would benefit from a simple A/B switch test. I would put money on "can't be heard" by anyone, no matter how golden the ear.
Now, whether high-gain tubes have their own noisy plate dynamics is a different matter. One nominally tries to avoid the effects of Miller capacitance, so the idea of ADDING plate-to-grid low-value capacitors is a bit ... counter-intuitive. Yet, if the goal is just to ensure that the tube has a natural roll-off above say "50 kHz" or something, then I can see the utility of clamping the higher frequencies through such an added capacitor. Its one thing to have to deal with a parasitic capacitance that one can't do much about (Miller), and its another to take it by the horns, and decide to engineer a particular solution.
PS: you've made a good argument for using multiple stages instead of fewer, with the problems of very-hi-mu tubes being stated thus. Moreover, using multiple stages diminishes the distortion effect of non-cascode connected triodes in the front stages. The only reason to use fewer tubes are [cost] and [appeal to audiophiles who desire simplicity]. I prefer more tubes, lower amplification factors. I must be an old fashioned fuddy-duddy, I guess.
GoatGuy
- Status
- Not open for further replies.