Hi -
For my current tube amp project, I want to use a CT plate choke for the input stage, but since it will be loading a pentode LTP, I need extra high inductance. So I was thinking of buying a couple of Hammond Mfg. 124E's and swapping its silicon steel core for an 80% nickel one to get the highest possible incremental permeability with this material and by also interleaving the EI laminations. Due to its balanced nature, the DC component should mostly cancel out, so I expect less than 1mA imbalance. With the CT primary properly seriesed up with the split secondary winding a nominal impedance of over 200K Ohm CT should be attainable, and the high mu core material I figure should extend the flat LF response to 20 hz or lower at this impedance.
Has anyone here ever done something like this, and if they have, have they run into problems removing the bobbin from the core once the outer core portions are cut due to the varnish or tight fit? I contacted Hammond about selling the 124E wound bobbin separately, but they appeared inflexible here.
Thanks in advance for any replies.
For my current tube amp project, I want to use a CT plate choke for the input stage, but since it will be loading a pentode LTP, I need extra high inductance. So I was thinking of buying a couple of Hammond Mfg. 124E's and swapping its silicon steel core for an 80% nickel one to get the highest possible incremental permeability with this material and by also interleaving the EI laminations. Due to its balanced nature, the DC component should mostly cancel out, so I expect less than 1mA imbalance. With the CT primary properly seriesed up with the split secondary winding a nominal impedance of over 200K Ohm CT should be attainable, and the high mu core material I figure should extend the flat LF response to 20 hz or lower at this impedance.
Has anyone here ever done something like this, and if they have, have they run into problems removing the bobbin from the core once the outer core portions are cut due to the varnish or tight fit? I contacted Hammond about selling the 124E wound bobbin separately, but they appeared inflexible here.
Thanks in advance for any replies.
Permalloy has higher inductivity but the same flux density. Therefore, in a saturation-limited winding (they all are, at 20Hz), you have no advantage in using permalloy over, say, a rusty grate (mu ~ 800).
If you must split the core up, the best way I've found is to use a blade of some sort to wedge the lams apart. Somehow or another, you have to jam them out of the bobbin, which usually involves a hammer and screwdriver for the first few lams. Once you have space to work in, they should split off easily. Discard the first couple of warped lams, you'll never get them back in anyway.
Tim
If you must split the core up, the best way I've found is to use a blade of some sort to wedge the lams apart. Somehow or another, you have to jam them out of the bobbin, which usually involves a hammer and screwdriver for the first few lams. Once you have space to work in, they should split off easily. Discard the first couple of warped lams, you'll never get them back in anyway.
Tim
Hi, Tim -
Thanks for the info. Well, I ran some numbers (probably should have done that before) and with an estimated ~10K turns total and 0.4mA max imbalance (which I believe is doable since I plan to run less than 10mA DC per side), I get about 1000H end to end. With no DC imbalance, this increases about an order of magnitude.
At 10Hz and 100Vrms (no imbalance), I calculate 6000 Gauss (or about 7100 Gauss with the 0.4mA DC imbalance). These numbers are close to the saturation flux of this material. This application should normally see no more than 40Vrms across the total winding where I plan to series up the split secondary of the 124E with the CT primary to get a nominal "240K Ohm" CT working impedance, going by Hammond's online specs.
These seem like workable, if less than optimal numbers for my application, although it would probably be worth it to have some kind of DC balancing scheme to maximize the available permeability.
I'm not sure how much I gain by gapping here, except perhaps reducing some LF distortion. Any gap over a couple mils would probably reduce the permeability below the minimum figure I mentioned above and from what I gather, even an interleaved EI stack will have close to a mil residual air gap due to less than 100% stacking factor and EI lamination butt gaps.
My biggest concern at this point is to ensure that I have sufficient phase margin worst case in circuit to guarantee stability when feedback is applied. I am considering using both local and global feedback for this application.
Thanks for the info. Well, I ran some numbers (probably should have done that before) and with an estimated ~10K turns total and 0.4mA max imbalance (which I believe is doable since I plan to run less than 10mA DC per side), I get about 1000H end to end. With no DC imbalance, this increases about an order of magnitude.
At 10Hz and 100Vrms (no imbalance), I calculate 6000 Gauss (or about 7100 Gauss with the 0.4mA DC imbalance). These numbers are close to the saturation flux of this material. This application should normally see no more than 40Vrms across the total winding where I plan to series up the split secondary of the 124E with the CT primary to get a nominal "240K Ohm" CT working impedance, going by Hammond's online specs.
These seem like workable, if less than optimal numbers for my application, although it would probably be worth it to have some kind of DC balancing scheme to maximize the available permeability.
I'm not sure how much I gain by gapping here, except perhaps reducing some LF distortion. Any gap over a couple mils would probably reduce the permeability below the minimum figure I mentioned above and from what I gather, even an interleaved EI stack will have close to a mil residual air gap due to less than 100% stacking factor and EI lamination butt gaps.
My biggest concern at this point is to ensure that I have sufficient phase margin worst case in circuit to guarantee stability when feedback is applied. I am considering using both local and global feedback for this application.
Last edited:
What is inductivity?
Whether or not you need high inductance to load a pentode is debatable. You should look over the forum at Intact Audio, Dave Slagle's website for a more comprehensive discussion.
John
In plain English,...I'm quite confused about what you are trying to do. I've been building amps and designing and modifying them for about 40 years and also built special inductors and rebuilt tube output transformers, interstage transformers, and filter inductors.
As far as taking transformers or inductors apart, I agree with Jlsem above. I've used razor blades and hammers for that purpose. Quite a tedious and thankless job. It is truely difficult. Hammond transformers are really tightly wound and well varnished. I have only taken apart burned transformers this way, in which case destoying windings wasn't important.
I really do not understand what you mean by loading an input of an a pentode (I don't know what (LTP) means,? not Long Tail Phase-inverter?) with an inductor. As far as I know one can load a device with a resistor or transformer because they approximate a power factor of one (power disapates because voltage, and current draw are only a few degrees out of phase with each other,... but you can not truely load a circuit with a capacitance or inductance alone because of the near perfect 90 lead or lag of voltage and current with respect to one another.
As far as increasing inductance enormously by using a large core (E and I separated to decrease saturation) and using bucking windings (series but in opposition) I did try something like that once (to increase inductance vastly), and it worked very well, but was loaded into a low inpedance transformer input (balanced load) and I used it for filtering a signal. So I guess I'm just not familiar with this theory you are using, unless you are hoping to drive an inductive load and perhaps permeate it to increase eddy currents and lnearize load???? Somehow the theory seems to be at cross purposes with itself???,,, or perhaps I just have no idea what you are trying to do. Probably the latter?? Once you completely permeate and (saturate) an inductors core, it stops behaving like a linear inductor. You would get high inductance for smalll signals and hysteresis distortion for larger signals and after saturation, it would become a much lower inductance, with something approaching resistive load characteristics without frequency related inductive reactance. Or perhaps you are filtering the power supply???? Still confused. Maybe you are trying to load the center tap of an output transformer so that it is a much higher impedance at 20Hz for a small signal pentode????
GeoReed
As far as taking transformers or inductors apart, I agree with Jlsem above. I've used razor blades and hammers for that purpose. Quite a tedious and thankless job. It is truely difficult. Hammond transformers are really tightly wound and well varnished. I have only taken apart burned transformers this way, in which case destoying windings wasn't important.
I really do not understand what you mean by loading an input of an a pentode (I don't know what (LTP) means,? not Long Tail Phase-inverter?) with an inductor. As far as I know one can load a device with a resistor or transformer because they approximate a power factor of one (power disapates because voltage, and current draw are only a few degrees out of phase with each other,... but you can not truely load a circuit with a capacitance or inductance alone because of the near perfect 90 lead or lag of voltage and current with respect to one another.
As far as increasing inductance enormously by using a large core (E and I separated to decrease saturation) and using bucking windings (series but in opposition) I did try something like that once (to increase inductance vastly), and it worked very well, but was loaded into a low inpedance transformer input (balanced load) and I used it for filtering a signal. So I guess I'm just not familiar with this theory you are using, unless you are hoping to drive an inductive load and perhaps permeate it to increase eddy currents and lnearize load???? Somehow the theory seems to be at cross purposes with itself???,,, or perhaps I just have no idea what you are trying to do. Probably the latter?? Once you completely permeate and (saturate) an inductors core, it stops behaving like a linear inductor. You would get high inductance for smalll signals and hysteresis distortion for larger signals and after saturation, it would become a much lower inductance, with something approaching resistive load characteristics without frequency related inductive reactance. Or perhaps you are filtering the power supply???? Still confused. Maybe you are trying to load the center tap of an output transformer so that it is a much higher impedance at 20Hz for a small signal pentode????
GeoReed
Hi,
I think "inductivity" is permeability or mu.
What are the reasons for using the bobbin and windings from a ready made transformer?
Do you know what inductance you need at what current per half?
and will the DC current in each half be in opposition or not?
Sorry for all the questions but if you know what you need then it should be quite easy for you to wind one.
Cheers Matt.
I think "inductivity" is permeability or mu.
What are the reasons for using the bobbin and windings from a ready made transformer?
Do you know what inductance you need at what current per half?
and will the DC current in each half be in opposition or not?
Sorry for all the questions but if you know what you need then it should be quite easy for you to wind one.
Cheers Matt.
Permeability is the bulk inductance, independent of geometry, exactly analogous to resistivity, which tells you the resistance of a bulk materia. In fact, A_L = mu*Ae / le, exactly the same as with resistivity, R = rho*A / l.
The difference is, you get to put on N turns, so L = A_L * N^2, where A_L is what I'm calling the core inductivity (dependent on geometry).
One disadvantage to high mu materials is DC offset saturates it so easily. You need almost exactly zero to use a perfectly closed permalloy core -- indeed, fluxgate magnetometers use the Earth's magnetic field to saturate permalloy -- this stuff is frickin sensitive!
Gapping reduces the effective permeability of a core. A slightly gapped permalloy core has exactly the same response as an ungapped core of lesser material, like stripwound GOSS, which might have mu_r = 10k or so. It takes an extraordinarily small air gap to do that, so there really isn't even any point in using permalloy unless you can be very, very careful of how that core is assembled. Mere fingerprints will have enough thickness to reduce effective permeability noticably!
One approach that might be reasonable is to have saturation occur at approximately the same point as inductance cuts into the bandwidth. A high permeability core will have excessive inductance, exposing saturation first; a low permeability core will be inductance limited, and very hard to saturate (simply because your tubes are "shorted out" by the inductance before you can even deliver enough flux to saturate it).
Let's say you have 6SN7 driving a winding in PP. Let's say it's supplied from +300V, so you get maybe 180V peak swing, and you want this down to 20Hz. Rp = 8k per anode, so the total is 4kohms, referenced to one half the primary winding. 4k/(2*pi*20Hz) = 32H, or 128H end to end. Also, Phi = Vrms / 4.44F = 2.03 Wb, so we need at least N*Ae = 2.03Wb / 1.2T = 1.69 m^2.t (that's meters squared-turns, so put on, say, 10k turns and you only need 1.69 cm^2 cross sectional area).
Now, since L = N^2 * mu*Ae/le, and Ae = Vrms / (4.44*F*N), L = N*mu*Vrms / (4.44*le*F).
Since we know Ae, we can also eliminate le if typical proportions are known. For a "wasteless" iron core, guess le = 19 * sqrt(Ae/4) = 19/2 * sqrt(Vrms/(4.44*F*N)).
Then, L = N^3/2 * mu*sqrt(Vrms) / (20*sqrt(F)).
Now all that remains is N, which depends on current density, and practical considerations, like not wanting to use a square meter core for an interstage transformer.
If we guess N = 5k, then 128H*20*sqrt(20Hz) / (5000^3/2 *sqrt(180V) = mu = 2.87 mH/m. mu_0 = 1.256 uH/m, so this case would need mu_r = 2285, a quite reasonable value, and quite realistic.
This value uses the restriction that inductance is the given value, and that saturation occurs at some frequency, which just happens to be the R/L cutoff frequency. Thus, saturation and cutoff will occur at the same time. Since turns are still a factor in the final expression, even after factoring out core proportions (that is, if you use fewer turns, you'll have to use a proportionally larger standard-proportion laminated iron core, and if you specify too many turns, the wire may not fit on the smaller core), you can see that the relationship between saturation, cutoff frequency and permeability is neither simple nor absolute.
It's interesting that, despite a guess, permeability came out to a standard number: 2000-3000 mu ferrites are very common. Lightly gapped iron cores (i.e., assembled with no explicit gap) will do as well, I expect.
Tim
The difference is, you get to put on N turns, so L = A_L * N^2, where A_L is what I'm calling the core inductivity (dependent on geometry).
One disadvantage to high mu materials is DC offset saturates it so easily. You need almost exactly zero to use a perfectly closed permalloy core -- indeed, fluxgate magnetometers use the Earth's magnetic field to saturate permalloy -- this stuff is frickin sensitive!
Gapping reduces the effective permeability of a core. A slightly gapped permalloy core has exactly the same response as an ungapped core of lesser material, like stripwound GOSS, which might have mu_r = 10k or so. It takes an extraordinarily small air gap to do that, so there really isn't even any point in using permalloy unless you can be very, very careful of how that core is assembled. Mere fingerprints will have enough thickness to reduce effective permeability noticably!
One approach that might be reasonable is to have saturation occur at approximately the same point as inductance cuts into the bandwidth. A high permeability core will have excessive inductance, exposing saturation first; a low permeability core will be inductance limited, and very hard to saturate (simply because your tubes are "shorted out" by the inductance before you can even deliver enough flux to saturate it).
Let's say you have 6SN7 driving a winding in PP. Let's say it's supplied from +300V, so you get maybe 180V peak swing, and you want this down to 20Hz. Rp = 8k per anode, so the total is 4kohms, referenced to one half the primary winding. 4k/(2*pi*20Hz) = 32H, or 128H end to end. Also, Phi = Vrms / 4.44F = 2.03 Wb, so we need at least N*Ae = 2.03Wb / 1.2T = 1.69 m^2.t (that's meters squared-turns, so put on, say, 10k turns and you only need 1.69 cm^2 cross sectional area).
Now, since L = N^2 * mu*Ae/le, and Ae = Vrms / (4.44*F*N), L = N*mu*Vrms / (4.44*le*F).
Since we know Ae, we can also eliminate le if typical proportions are known. For a "wasteless" iron core, guess le = 19 * sqrt(Ae/4) = 19/2 * sqrt(Vrms/(4.44*F*N)).
Then, L = N^3/2 * mu*sqrt(Vrms) / (20*sqrt(F)).
Now all that remains is N, which depends on current density, and practical considerations, like not wanting to use a square meter core for an interstage transformer.
If we guess N = 5k, then 128H*20*sqrt(20Hz) / (5000^3/2 *sqrt(180V) = mu = 2.87 mH/m. mu_0 = 1.256 uH/m, so this case would need mu_r = 2285, a quite reasonable value, and quite realistic.
This value uses the restriction that inductance is the given value, and that saturation occurs at some frequency, which just happens to be the R/L cutoff frequency. Thus, saturation and cutoff will occur at the same time. Since turns are still a factor in the final expression, even after factoring out core proportions (that is, if you use fewer turns, you'll have to use a proportionally larger standard-proportion laminated iron core, and if you specify too many turns, the wire may not fit on the smaller core), you can see that the relationship between saturation, cutoff frequency and permeability is neither simple nor absolute.
It's interesting that, despite a guess, permeability came out to a standard number: 2000-3000 mu ferrites are very common. Lightly gapped iron cores (i.e., assembled with no explicit gap) will do as well, I expect.
Tim
Here are DC perm curves for common commercial, loss less E/I core materials. These curves start with a perm you will find when finished transformers are tested at 1 volt AC/120 Hz. Ac perm, which only applies for the first half cycle of any given signal is about 4 times the DC perm, but you must design with the DC perm or you will only have bass thumps, without any information after that.
Do note that the inductance number you were deriving from the impedance load is for - 3dB, at the frequency of interest. To obtain the inductance that will load the impedance at - 0.5 db, multiply the - 3 db inductance needed by 2.76.
Bud
Do note that the inductance number you were deriving from the impedance load is for - 3dB, at the frequency of interest. To obtain the inductance that will load the impedance at - 0.5 db, multiply the - 3 db inductance needed by 2.76.
Bud
Hi -
Sorry if some of my posting was confusing. I was interested in Permalloy because it exhibits a higher initial permeability, and barring considerable DC imbalance, seems to have a more constant permeability at lower flux levels without gapping than silicon steel.
My hope is that the high value of inductance would reduce LF phase shift in the audio band for my application and make it easier to maintain stability when applying feedback, considering that there is also the output transformer to take into account.
I considered winding on a bobbin, but first would need to locate a source for the bobbins (any suggestions appreciated), and then, for 10-20K turns and for at least two inductors, doing a good job of winding would turn out to be a somewhat time consuming affair, so I was thnking that at least for the initial prototype it might be easier to replace the core on an extant transformer, since I have already lined up a source for the permalloy laminations as a sample quantity.
However, if I can obtain a properly sized bobbin and mounting bracket, winding my own might still be a good option since I have also located a source for bifilar magnet wire at not too great a cost.
Sorry if some of my posting was confusing. I was interested in Permalloy because it exhibits a higher initial permeability, and barring considerable DC imbalance, seems to have a more constant permeability at lower flux levels without gapping than silicon steel.
My hope is that the high value of inductance would reduce LF phase shift in the audio band for my application and make it easier to maintain stability when applying feedback, considering that there is also the output transformer to take into account.
I considered winding on a bobbin, but first would need to locate a source for the bobbins (any suggestions appreciated), and then, for 10-20K turns and for at least two inductors, doing a good job of winding would turn out to be a somewhat time consuming affair, so I was thnking that at least for the initial prototype it might be easier to replace the core on an extant transformer, since I have already lined up a source for the permalloy laminations as a sample quantity.
However, if I can obtain a properly sized bobbin and mounting bracket, winding my own might still be a good option since I have also located a source for bifilar magnet wire at not too great a cost.
True. I was thinking in terms of the 45 degree phase shift at this frequency. All this makes me appreciate my OTL design
However, I am now considering an inductorless loading scheme for the pentodes that looks promising although it will significantly lower open circuit gain over most of the audio frequency range compared to the plate chokes. A lot of my indecision is due to the fact that I would like to run the output stage into mild AB2 while maintaining good bias and phase stability with, hopefully, a very simple two stage topology for the whole amplifier. A high transconductance twin pentode like the 6BN11 would seem to offer some possibilities here because I figure it should be able to give a single stage gain of over 100 at a plate current of several milliamperes as a LTP phase splitter, properly biased. If I can get it to drive the output stage (which I am using 6DZ7's for, which are basically two 7189A's in a single bulb) a little bit into AB2 while using some of Roger Modjeski's biasing tricks, I might be looking at a well performing 35-50 Wpc amplifier using only two tubes per channel.
Last edited:
When applying NFB to an amp with lots of stages, you're more than likely better off doing it locally around each stage. That way you can keep the phase shift small per stage, for most frequencies (with a maximum phase shift of 90 degrees around a single stage, you can apply as much NFB as you want, and the phase margin will only improve). Then you'll have a better margin to work with when applying global feedback.
It seems to me, adding local feedback to make a transformer (or inductor) work better is the same as reducing that stage's effective output impedance, which is the same effect as making the inductance larger. So if you would like to avoid phase shift, you are indeed just as well off using a high permeability core, which simplifies the circuit.
Saturation attenuates low frequencies, probably without much phase shift, while adding harmonics, so I would guess the amplifier will still be stable in saturation.
As far as appreciating designs with low phase shift, I prefer to simply direct-couple the whole damn thing. I've drawn a few such circuits now, I should build one some day.
Tim
It seems to me, adding local feedback to make a transformer (or inductor) work better is the same as reducing that stage's effective output impedance, which is the same effect as making the inductance larger. So if you would like to avoid phase shift, you are indeed just as well off using a high permeability core, which simplifies the circuit.
Saturation attenuates low frequencies, probably without much phase shift, while adding harmonics, so I would guess the amplifier will still be stable in saturation.
As far as appreciating designs with low phase shift, I prefer to simply direct-couple the whole damn thing.
I've drawn a few such circuits now, I should build one some day.Tim
Here is the current most flexible bobbin molder in the US, with the largest range of bobbin sizes you can cadge a sample out of. Don't mislead them please, just tell them straight what you are doing and ask for a gift. Expect to pay for the freight.
Foremost Plastic | Bobbins | Modular Molding | Plastic Injection
I am not a fan of auto former anything for audio. They always remind me of 80's Japanese SS gear. Very nicely detailed and transparent, with an above 6 kHz hashy sound to transients and a garble of all information below -40 db or so. Some folks think they are great though.
What center leg tongue size have you found a source for? Where are you going to go to obtain an OPT that matches the nickle core in performance?
Do understand that 80% Nickle is a square anneal material, so there is no grace to it's saturation characteristics. Total saturation will occur across about a 40mv increase in signal above Bsat and distortion will change from 0.001% to 100% in that window. If you are using 80% nickle don't assume anything about it's characteristics in saturation and unlike M series core, which is all done with power transform at 400 Hz, 48% extends to 3.5 kHz and 80% to 10 kHz.
Means you MUST think about coil construction. So long as your voltages don't exceed a polar amount above 400 volts, winding wire to winding wire, bifillar winding is reasonably safe for opposite polarity windings. I would recommend vacuum impregnation of a polyester resin for anything above 240 vac just due to the corona that will be eating away at your wire insulation from 30 volts AC on up. Do note that a bifillar wound coil will have significant end turn capacitance and this could easily swamp your output above the power bandwidth of the core.
Don't want to squelch your enthusiasm, but don't hesitate to discuss isolated windings if the bifillar approach doesn't cut it.
I am working on a useful perm chart for Nickle materials and will post it when I finish it. Just as in the M materials graph I will expand the 1k to 10k window to a useful width, since that is where you will actually care about.
Bud
Foremost Plastic | Bobbins | Modular Molding | Plastic Injection
I am not a fan of auto former anything for audio. They always remind me of 80's Japanese SS gear. Very nicely detailed and transparent, with an above 6 kHz hashy sound to transients and a garble of all information below -40 db or so. Some folks think they are great though.
What center leg tongue size have you found a source for? Where are you going to go to obtain an OPT that matches the nickle core in performance?
Do understand that 80% Nickle is a square anneal material, so there is no grace to it's saturation characteristics. Total saturation will occur across about a 40mv increase in signal above Bsat and distortion will change from 0.001% to 100% in that window. If you are using 80% nickle don't assume anything about it's characteristics in saturation and unlike M series core, which is all done with power transform at 400 Hz, 48% extends to 3.5 kHz and 80% to 10 kHz.
Means you MUST think about coil construction. So long as your voltages don't exceed a polar amount above 400 volts, winding wire to winding wire, bifillar winding is reasonably safe for opposite polarity windings. I would recommend vacuum impregnation of a polyester resin for anything above 240 vac just due to the corona that will be eating away at your wire insulation from 30 volts AC on up. Do note that a bifillar wound coil will have significant end turn capacitance and this could easily swamp your output above the power bandwidth of the core.
Don't want to squelch your enthusiasm, but don't hesitate to discuss isolated windings if the bifillar approach doesn't cut it.
I am working on a useful perm chart for Nickle materials and will post it when I finish it. Just as in the M materials graph I will expand the 1k to 10k window to a useful width, since that is where you will actually care about.
Bud
Sch3mat1c
Only thing I didn't see in your very good suggestions was a comment about the time involved with local feedback as opposed to global feedback. The local feedback actually stands a chance of being phase correct across the 20 t0 20k range. You do provide really good information here!
Bud
Only thing I didn't see in your very good suggestions was a comment about the time involved with local feedback as opposed to global feedback. The local feedback actually stands a chance of being phase correct across the 20 t0 20k range. You do provide really good information here!
Bud
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.