| Tweeker |
Does anyone have any suggestions for learning about transformer and choke design, either online or off?
Im also interested in commentary on iron and things such as magnet wire insulation choice, potting, and so on. |
|
|
| Brian Donaldson |
Great, Another brave soul to dive into the black art of iron spinning with me. I have a few links on my computer at home that I'll post this evening. I'm ordering the Radiotron Designer whatever book that has a chapter on iron design.
The cost and availibility of 200w SE transformers for 833 transmitting tubes has forced me into this
I may have found a supplier of EI laminations, but they were closed for the holiday Monday. Any other sources will be helpfull |
|
|
| Brian Donaldson |
Here's 2 sites about output transformers.
http://www.geofex.com/Article_Folde...des/xformer.htm
http://www.diyaudio.8m.com/Ot/out_tran.html
I also found a 100+ page manual on audio transformers that is basicly a cookbook with little theory, but explanations of what makes one not work as it should, But it's on my computer at work:xeye: Also it's $45.00 and home made and bound and I have no idea if the info is at all acurate. It may be $0.50/sheet toilet paper that's not very aborbant! We'll see what kind of infor the radiotron designer's handbook has when it comes. |
|
|
| tubelab.com |
Forum member "GEEK" has the entire Radiotron Designers Handbook 4rth ed. posted on his web site.
http://geek.scorpiorising.ca/RDH4.html
Pete Millett has several electronics books posted on his web site including the RD3 and a book called Electronic Transformers and circuits.
http://www.pmillett.com/tecnical_books_online.htm
When you figure out how to make a big SE transformer that works, let us know. I have decided that my success rate is much higher making amplifiers, so I have given up on transformers for now.
Did your 100 page cookbook come from David Lucas? |
|
|
| Brian Donaldson |
It was this
http://www.rgwdesign.com/page0003.htm
Sure sound easy.
I just spent $15.00 for a cd PDF of Radiotron Designer' Handbook 4th ed. Wish Google would have pulled up Geek's site. |
|
|
| Brian Donaldson |
Looks like I'm building a 50 lb transformer. 40 lbs of iron and 10 lbs of copper.
Tell me what's worng with this idea. For testing, I want to put a resistor parallel to a choke + current source across the primary to model the staning bias, and drive the secondary with a SS amp measure some results. I'm trying to avoid playing with full tube amp and 1500 v B+ supply to check the transformers extreme bandwidths. I'f the tranformer doesn't measure up, I may abort the whole mess and go to my 6C33C OTL plans. |
|
|
| anatech |
Hi Brian,
You know, when something goes wrong with an OTL, it goes really, really badly wrong. I wish you nothing but good luck with this.
-Chris |
|
|
| Tweeker |
First problem I see is iron loss, that much iron:copper is going to mean more distortion.
From what I can tell, there are two ready approaches to the problem of a 200 watt plus SE transfomer:
1.) Have it wound by a wizard using spendlon wire on an amorphous unobtanite core.
2.) Accept lower bandwidth. It shouldnt take a wizard or exotic materiel to wind 200 watt SE transformers with 8 octaves bandwidth. On the bass xformer accept some HF rolloff due to capacitance. Have the other transformer have it wound more like a 400hz power transformer, sacrifice inductance and accept some LF rolloff. Use less turns of heavier copper taking care to minimize capacitance and core losses. Of course, biamping with 2 833As would be way over the top. |
|
|
| Brian Donaldson |
Not for me:D
What about bi wired and bi transformered? Don't bother shooting this one down, I'll do it as I try to fall asleep. |
|
|
| Brian Donaldson |
OK, the raw goods are on the way. One more auction on Ebay and I'll have my wire. Time to start cutting PC board stock to make my bobbin ends and oak for the mandrel.
There are a ton of conflicting variables to get right to make it happen. I'll likely rewind it a few times before I pass the 20K at -1db.
There wil be 9 interleaves total and the primary will be vertically sectioned in each horizontal section as well...like in rf chokes. This should keep the leakage capacitance in check. I estimate by Hanna's method in RTDH4 75H primary, 5kohm. |
|
|
| tubelab.com |
Have you inquired into the cost and minimum order amount for their laminations? This info is not on their web site. If the minimum order is enough for several transformers, I might be convinced to try making transformers again. We could split the cost of iron. I have wire from my previous transformer experiments.
I still have all of the parts to make the "big one" just no output transformers. |
|
|
| smoking-amp |
Ahhh... Long E cores. Finally some decent iron. Me too! How much mulla for these lams? The EI-175L looks more to my size requirements though.
I still have to get around to winding something with my last find at the scrapyard, but they are a little big for my current projects:
http://www.diyaudio.com/forums/show...3917#post633917
Don |
|
|
| Brian Donaldson |
They're giving me samples enough for 2 square stacks, I just need to pay for the 64 lbs of shipping. I'm also waiting for a quote for 280 lbs more.
I was planning on 3" stacks, but if I drop to 2.25", I could make the 2 transformers from the samples alone.
I thought the long E's looked cooler, then I found that the long coil is better for low inductance leakage, but not so good for capacitance, but I can make it up with vertical sectioning in the primary layers.
Laminations are the hard part to get. Companies the make laminations only sell to large manufacturers. There is no distribution network set up for tube amp diy'ers. I'll have to butter the salesman a bit I guess. If I get set up with him, I could be the goto guy.
I'll keep you posted.
Cool scrap yard, Don! I once bought 2 huge sheets of 1/4" aluminum at the scrap yard in Houston for $0.40 / pound. Made a lot of nice amp enclosures. Sadly I'm out and haven't seen such a find in a long time. They had 40 sheets or so I wish I would have bought.
Tubelab, I'm making the transformers for my SE 833. I'll let you know how they come out. If all works well, I will share the design. I spoke with a guy at a recent Bottlehead meeting in Houston who knows Jack with Electoprint who said the transformer can't be made. I think he just didn't want me to top his 211 amp. |
|
|
| Tweeker |
| Ive been looking at standardized (EU) SU series C cores for power transformers. They are available in good electrical steel- M2H and better to 12,900 VA. |
|
|
| Brian Donaldson |
OK, O got the quotes for the laminates.
Wait for it
wait for it
$3.30 / pound.( plus shipping and tax) So a square stack will be about $100/ transformer for the 30 lbs of iron. Add another $50 for copper, a little time winding and I'll have a $500 transformer. (multiply all number x2 for stereo)
I need to think about building slightly smaller transformers so I can make two with the free Iron or plunking down $150 to get enough iron to build the full beast |
|
|
| smoking-amp |
Hmmm....., a DISTRIBUTED gap toroid would preserve the magnetic field symmetry of a normal toroid, to still give very low leakage inductance for a SE gapped design. (the steel strip gets co-wound with a non-magnetic, insulating strip) This is likely the BEST way to beat the bandwidth problem of a LARGE gapped SE xfmr. And since one is trying to reduce the permeability anyway, one could wind the core as a large diameter ring that would be easy to wind the copper on. (ie, a big hole in the center like a big current xfmr core)
Haven't heard of anyone trying this yet.
I myself don't see the need for gapped xfmrs (or inductors) for SE at all anymore:
http://www.diyaudio.com/forums/show...7967#post637967
But for winding convenience, the long E laminations are appealing and they get one half way to the low leakage L performance of toroids.
How about the EI-175L laminations in the Tempel catalog? Would be half the weight of the EI-225L lams for a square cross section. Even better if they listed EI-150L lams, but I don't see them there.
Don |
|
|
| smoking-amp |
Doin a little thinkin about the distributed gap toroid. One could just use a LARGE powdered iron toroid (or stack of them) for SE to get high bandwidth at high power. These have a Mu between 20 and 200. The gap is effectively already built in. The toroid shape gives low leakage L to get bandwidth.
Where's Valveitude and his ground up brake drum shavings when you need him! However, a powdered iron pot core is not suitable for SE or audio in general. You need a long thin winding to get low leakage L. Pot cores have the worst leakage L specs.
Valveitude,
just pour that "stuff" into a big ring Jello mold and you will have the best large SE xfmr core on the planet! (need a plastic one to avoid a shorted turn of course.)
Don |
|
|
| Brian Donaldson |
| Use a bunt cake pan! 14" OD:bigeyes: |
|
|
| Brian Donaldson |
I'm chasing the long E because it's do-able without sinking thousands into a toroid winder. I also think there is magic in not crossing the 0 flux of the transformer.
If you're putting a tube and power on the other leg of a PP transformer, I can't justify not driving the tube with signal...then you have a plain old PP amp. Some things like 100W filaments plus 300W plate dissapation must be considered with an 833 amp.
Parafeed has issues with cap in the signal and choke, cap and transformer interaction that I'm not ready to study. |
|
|
| smoking-amp |
Truck tire inner tube core! :cool:
I have seen some pretty good size powdered iron cores around, like maybe 8 inch OD. Somebody makes them.
Don |
|
|
| valveitude |
Don,
| quote: | | just pour that "stuff" into a big ring Jello mold and you will have the best large SE xfmr core on the planet! (need a plastic one to avoid a shorted turn of course.) |
:headshot:
:D
This was pretty much the plan from the beginning with the whole "roll yer own" powder core, only I was thinking a "cup" core. When I finally put the idea on the shelf (miserable-hard work), I had achieved around 20mu by tamping in a mold, giving it a good soak in H2O, and then baking dry...it would oxidize into a solid block with a resistance of several megaohms per inch.
I still have 20 or so pounds of the stuff...I'll find a use(s) for it.
Casey |
|
|
| tubelab.com |
| quote: | | I myself don't see the need for gapped xfmrs (or inductors) for SE at all anymore: |
Well I don't either except for two items. 1) I haven't been able to make an SE amplifier that uses balanced DC or no DC in the transformer sound as good as a conventional SE amplifier. 2) Most of these gapless experiments have been carried out at low power levels, and we are talking about big SE power here, 100W plus.
I have had no luck finding a suitable big SE transformer, so I am going to rethink the gapless approach. I have successfully built a 100 mA (max) CCS using an 813 it will work at 1200 volts. The next step is an 833A. I have the parts for a 2500 to 3000 volt 1/2 amp supply. Maybe it's time to build a big chokeless parafeed experiment again. Along with Brian's transformer experiments we have two chances of sucess.
I followed the Alpha core link provided by Tony. They have off the shelf toroids that are good for big transformers. If you have ever experimented with winding these you will know what a pain it is. They offer gapped toroids and large "C" cores but they are on a custom basis. Two "C" cores would make a nice SE OPT. |
|
|
| Gluca |
Tubelab ... you are going to blow the entire block away the very moment you'll plug your guitar into that ominous amp!
:o
Gianluca (I luv your site!) |
|
|
| tubelab.com |
| Yeah, maybe, but I will try it anyway. I believe in stress testing my amps. If an amp survives that type of testing, it should take anything. |
|
|
| Looneytunes |
Hey Tweeker,
To respond to your original question, have you checked out Pete Millett's site lately? He has a pdf of "Electronic Transformers and Circuits, Reuben Lee, 1955" available for download on his technical books page (http://www.pmillett.com/tecnical_books_online.htm). There seems to be a fair amount of information on transformer design, etc. (This from the perspective of a NewB.)
I know this is a little late, and the thread has taken a new direction, but hey, it's new to me! :)
Looney
:confused: |
|
|
| Brian Donaldson |
Iron is in the mail. I need to order wire, but I'm still scratching my head about a few things. I'm guessing....
8500 turns #25 primary 5K imp. 300ohm resistance
320 turns #18 doubled sec 8ohm imp .8ohm resistance
75 H 250 ma bias
I made the bobbin out of manilla folder...it's huge
7" x 4.75" x 5.75"
Any comments? |
|
|
| Brian Donaldson |
I'm sitting at my desk thinking... this thing is way over the top huge. Are any of my specs out of line with reality? Losses are a percentage of something, you don't simply loose the first 5v of a signal. With a 150W, I can afford a few percent loss as long as I have the bandwidth I want. My bias current may be a little high, but I wanted a little room to use higher current and less voltage If it sounded better. I'll rerun the canls @ 180ma and see how different it is.
Edit
I see that 250 ma at 1200V is pushing the max plate dissapation with convection cooling. But even dropping to 180 ma only decreases primary to 7200 turns...just a little more room for teflon tape...still fkn huge.:devilr: |
|
|
| smoking-amp |
"7" x 4.75" x 5.75" Any comments?" Brian D.
Ummm, well, going to a bigger xfmr in general makes the leakage inductance and distributed capacitance go up, hence the lowered bandwidth for larger SE transformers that everyone has run into.
However, the long E core should be helpful at reducing the leakage L. If you don't need all that winding window for copper, or can sacrifice some efficiency, then a "leaner" winding will improve leakage L further yet. (ie, a long thin winding cross section is what really reduces leakage L) Sectioning the winding efficiently now becomes paramount to get distributed capacitance down.
The more the core area, the lower the number of turns to get the minimum primary L. And the fewer the turns, the smaller the core air gap can be (ampere turns going down). At some point in this process one should be able to remove the gap completely (hopefully at only welding transformer size and not utility xfmr size, this optimization is an accelerating one since the effective Mu goes up as the gap shrinks).
Unfortunately this "optimization" procedure increases the core losses since max flux density is staying the same. Carried to an extreme, one might then consider low loss material like "amorphous", permalloy or ferrite, but will be costly at this size.
Don |
|
|
| Brian Donaldson |
Less copper also means less money. :clown:
I have better then 1:10 resistance to impeadance which I think is very good. I could drop a wire size maybe and add more insulation and decrease capacitance. I planned to vertically setion the primary windings to help with capacitance. Also, I'm calculating around .07 V/turn, so the voltage graient is small making the capacitance less of an issue.
I'll have 5 horizontal interleaves, and vertical sectioning in the middle and upper primary interleaves. can anyone measure the capacitance between anode lead and speaker lead with a DMV on a SE OPT for me so I can see what I'm shooting to beat. |
|
|
| tubelab.com |
I borrowed a digital capacitance meter from work and tested a few transformers. The results are below. Oddly the best sounding transformer that I have has the highest capacitance. The meter was a Tenma 72-040. I verified the readings with a 1000pF and a 4700pF cap. The reagings were taken from the primary lead that gave the lowest reading to the secondary lead that gave the lowest reading. In most cases it didn't matter much. In the case of the Electra-Print, the B+ lead reads 10000 pF to either of the speaker leads.
The 200 watt transformer 5170 pF
A 300B transformer from the same vendor 5630 pF
One Electron UBT-3 3870 pF
Edcor XSE-15-8-5K 1300 pF
Hammond 125CSE 890 pF
Electra-Print custom 5K for 45's 6980 pF
Generic P-P guitar amp transformer 960 pF
Handwound Transformers 6K P-P 310 pF
Clearly the capacitance is not the only HF response killer. Unfortunately I don't have an inductance bridge, and I don't think that I can borrow one, but I will check. |
|
|
| Brian Donaldson |
10 uF, Christ. no wonder it rolled off the highs.
I'm building a winding jig now. Hope to have the wire before the weekend so I can wind. I may wind 2 layers and check capacitance and try a few things before I put $50 worth of wire in the trash. I need to buy an a ssload of teflon tape. |
|
|
| tubelab.com |
10,000pF is .01uF, still way too much for good HF response. Yet this is an Electra-Print transformer that sounds really good. I am at work now, so I don't have any frequency response plots for this transformer, but if I remember correctly it had a 3db point around 25KHz. This is with almost 7000 pF being fed by a 45.
The big transformer has 5170 pF. the Frequency response plots had 3 obvious dips in the 10Khz and above region from some sort of resonant condition. I suspect that there is some sort of interaction between the capacitance and the parasitic inductances present inside the transformer. The plot is on my web site.
http://www.tubelab.com/833SE.htm
I tested this transformer with 45's, 300B's, 845's, a pair of 845's and the 833A's at currents from 30 mA to 300 mA. The HF response did not change. The LF response was better at lower currents. |
|
|
| cerrem |
Just trying to be helpfull...
Once you decided on a plate load and that is set in stone....then you decide on a "reasonable" bandwidth to start with.... Then you calculate the minumum Inductance, maximum Leakage inductance and maximum total winding capacitance, based on the plate load, plate resistance and bandwidth desired...
Once armed with these specified numbers, you set out to calculate and arrange the winding geometry to meet or exceed these specifications...
Remember that you have two flux densities to keep track of and the sum of these two Flux densities should add up to be somewhere between 12,000 to 15,000 Gauss in a Single Ended OPT with M6 core..you choose this based on the core distortion you can tolerate, now the flux excursion is a bit more than a P-P OPT, due to the gap linearization..... You have the DC flux density and the AC flux density... DC flux density is basically set up from the DC bias point and is proportional to GAP and turns..area..ect.. The AC flux is simply signal excursion at full power output.... As you can see the DC flux density should not be less than .5 the entire AC+DC flux sum, to avoid flattening waveform at the origin of BH loop...
The neat thing is that the more you push the DC flux up the curve, the smaller the AC flux excursion will be...due to the inverse relationship between these FLUX densities, with respect to turns... Set these equations = to each other and you will see the inter-relationships...
Chris |
|
|
| Brian Donaldson |
| Thanks for the info Chris. I've read a few books (including the RHD 4 chapter over and over again) now about transformer design and think I'm on the track. Now I'm only waiting for the wire. I'm ready to wind a coil and see where the leakage inductance and capacitance come out and see which one need optimizing and which one can be sacrificed a little. I'm using all of the tools I've seen to minimize both. Check out my winder I made. I'll power the drill with a foot switch and my power supply. There is also a micro switch and counter to keep tabs on the number of turns. The mandel is 7" wide :bigeyes: |
|
|
| rcavictim |
Tubelab, All,
I would be inclined, when measuring the inter winding capacitances, or coil to transformer core of a transformer like these to short the two outer ends of each winding to be tested and connect the cap meter to this junction. I fear that a digital capacitance meter may become confused by the series L that is progressively present in the capacitance that is being measured, since measuring capacitance relies on pulsing the capacitor at a voltage and rep rate dependant on the meter and range. |
|
|
| tubelab.com |
I tried this when I measured the capacitance of the transformers listed above. It made no difference on any transformer except the Electra-Print. I can't beleive the numbers that I measured though. I can't see how you would get any HF response with 6000 pF connected from the plate to ground. I will try sticking a few capacitors from plate to ground the next time I have my equipment set up and see what happens.
Currently I am preoccupied with the screen driven P-P amp design, and updating my web site. I am still listening to the triode connected 6AV5's through the Edcor transformers. They still sound good to me, although I have found a few musical selections that will demonstrate their weaknesses. |
|
|
| cerrem |
A couple of things concerning measuring transformer capacitance... It can't be done with a bridge meter..ect.. And can't be done in a static environment..
Since the capacitance varies greatly over applied power level...it actually varies according to the voltage gradient of the specific di-electric being used... It also changes with applied DC voltage...
One way to accuratlye measure the transformer, is to first measure the leakage inductance, since that number is fixed by the geometry of the windings...The short circuit method of measuring leakage has big error involved..
You need to drive the secondary windings to the equivelent current of full power output and vary the frequency till you find the resonant frequency of the transformer...then you can extract the Capacitance once the leakage is known..... DO this at various power levels and you will see the capacitance start to increase as the power level increases...ie, the upper bandwidth reduces with increased power output...
Also...you have two major capacitances to look at...Primary to Secondary....and ..... Self Primary winding...... The sum of thee two makes up the total transformer capacitance..
Chris |
|
|
| Brian Donaldson |
What a huge PITA. I wound 2.5 of the 7 interleaves and noticed the counter quit working, so now I'm having to estmate turns, and I used 12 rolls of teflon tape, and can't find the other 20. Glad I went into this thinking the first coil would be junk. I also found out the nice teflon wire that I planned to use was only rated 300V, so I sleaved it in teflon tubing. But on the up side, with 2 of the 6 coil boundrys wired up, my meter is reading 1.5nf, so my pri to sec capacitance is looking good, and the teflon tape and winding geometry should keep the prim interlayer capacitance and leakage inductance within range.
I just need to practice.
Hope to have it somewhat together in 2 days so I can hook it to my old Kenwood and a 5K resistor and see if it works. |
|
|
| Brian Donaldson |
No Bueno... I'm reading 16 nf way too much capacitance. And I'm sure it is due to poor winding. I'll plug in the iron and put some juice to it and see what I get. I may try the calcs with much less copper.....
Double no bueno...I destroyed it trying to get it off the mandrel. Something must be done to releive the presure netx try so It"ll slide off intact
:bawling: |
|
|
| Brian Donaldson |
I couldn't see to type for the tears. And it won't let me edit.
My skill level has really increased. I unscrewed the mandrel back into 3 pieces so I can hammer the center section out and relieve the tension so the other sections can be removed without damaging the first layer's fish paper and mylar. Also, I added a sheet of laser transparency stock between every second layer, and use teflon tape to between all layers. Now I'm able to make every layer perfect.
I'll post some pics and new capacitance measurements tomorrow. |
|
|
| smoking-amp |
A couple of wedges cut from one piece of wood works good too for tightening the mandrel.
Another nice approach is to use a bunch of narrow bobbins that stack on the iron, "sectioning by bobbin". That way you only have to wind a fraction of the total to get measurements and alter course if necessary, and lots of tap connections for configuring CFB or floating UL windings too. One can get double or triple enamaled wire too if capacitance is problematic, but that will increase leakage L some.
I have just been reading H. A. Hartley's book "Audio Design Handbook" and he mentions the "ultimate" design is done using thin disk like winding sections that get stacked on the core with insulation wafers between them (only two wire thick sections that wind up from the center for each end of the wire). These only have a primary or secondary winding on each disk. Next runner up is a stack of pi - wound sections. These approaches are quite similar to the chevron sectioning by dithering used on high bandwidth toroids. Hartley mentions that these were laboratory techniques, too expensive for commercial products.
Don |
|
|
| Brian Donaldson |
Second try sucks.
P-s capacitance 14 nf
-6 bd @ 14000 hz
90 deg phase shift @ 25000 hz
Hey George, this may be your rocking SET guitar amp OPT.
But at least I have some hard data to know how many orders of magnitude off I am.:smash:
Time to buy more wire.
Edit
I spoke too soon. I do have 90 phase shift at 25Khz, but I didn't have the second seconary connected (there are 2 secondarys that are paralleled)
I have a -3 bd (or so by eye) dip at 8500hz (an annoying frequency anyway) and 0 phase shift
The response is back up to 0db by 9500 hz and climbs to + 6 db at 13000 hz
back to 0db at 25000 hz then falls off to -6 db at 50000hz
May be worth investigating.
|
|
|
| tubelab.com |
Sorry to hear about your transformer luck. I tried this a few years ago, got similar results, and I started with working transformers! I took apart some 80 watt P-P transformers, rewound them with the idea of building a 10 watt SE transformer. The results were not good. Then I found out that the P-P transformers worked OK as is. They were a lot better than anything that I coul do. I could at least rip out all of the laminations, and re-stuff them with a gap. The gap was determined by how many layers of masking tape I used.
| quote: | | Hey George, this may be your rocking SET guitar amp OPT. |
I still have the one that I paid $200 for. |
|
|
| tubelab.com |
| Didn't see the edit. This one might work. The dips (notches) in the transformer that I have seem immune to any manipulation. Load resistance, or primary current have no effect on the frequency of the dips. Changing the load impedance only affects the amplitude of the notch, and only by a couple of db. |
|
|
| Brian Donaldson |
| Check it for size |
|
|
| Brian Donaldson |
I was editting as you posted
I have iron, 833, 1200V psu and a victim reciever to act as input and driver... I'm about to mock up a amp. The wife and the kid will be home any minute, so I gotta get to work
I still think I can improve the next one. These numbers are all running the thing backwards with no bias, but that shouldn't matter more than 10% |
|
|
| smoking-amp |
Looks like the transformer out of my 200A Lincoln welder.
Are you designing for DC current at 1/2 the max AC current peak to peak, to get the minimum gap and least number of turns?
The peaks and valleys are probably from layer to layer interaction, maybe need more layer insulation, but that will likely hurt the top end then. Appears that you have no sectioning so far, so still room to manuever the design. Sectioning will probably get rid of the peaks and valleys too.
Might also be helpful to leave a 1/4 inch gap or so at the end of the bobbin, where the E-I gap is, to avoid turns in the gap fringing field. If you do the sectioning by shorter bobbins, you can wind just one section and test it to save on wire filling up the trash bin. Good luck!
Don |
|
|
| Brian Donaldson |
No vertical sectioning. I tied that on the first one that was still born, and the exicution fell far short of the concept. Sort of like your suggestion for 2 wire wide discs. (not meant to be snotty. I know you were talking optimum design) The first few layers went within 1/8" of the extremes, each following layer a little shorter so it wouldn't fall off into the hole.
It's playing music now and doesn't sound half bad for the POS speaker it's pushing and it has NO bias on the grid...total A2 900V at 120 ma. A pair of these thing will warm your room if not your heart.
Where can you find the Hartley book you recomended? Is it still in print or free on the net somewhere? I want to read some more before I burn up another $75 worth of wire.
I'll post pics in a minute |
|
|
| Brian Donaldson |
| Playing Eric Johnson through POS sony ht speaker |
|
|
| smoking-amp |
The Hartley book (1958, 224 pages) was one of the Gernsback Library book series like N. Crowhurst's "Understanding Hi-Fi Circuits". Its not an easy read unfortunately, it needed some serious editing to make it readable by us mere mortals. If you already know the stuff then maybe it reads OK. Too many places he discusses circuit subtleties without bothering to mention the operating class of the circuit like class A, AB, .... , have to check details in the schematic given. The mention of transformer design is relatively brief with just pointers on how problems can be overcome. There no doubt was some research paper somewhere about the disk sectioned design, but no bibliography or references are offered.
Abebooks.com and Addall.com used book sites have copies of the book listed at 50 to 60 bucks. I myself just did an interlibrary loan at the local library and scanned it onto CD and printed out a copy. Maybe I could send the scan to Pete Millet for posting on his site, needs conversion from 220 .tif files to a PDF file yet. Too big to email from my dialup connection.
Don |
|
|
| Brian Donaldson |
OK, i've been very busy with work the last few weeks, and I've also been working on a PC board for the input and drive for the beast. Since my last post I bought a LC meter (crappy Velleman, but it gives more accurate readings than my syliva wetted index finger)
I shorted the outputs of the prototype transformer, and it reads 120 mH across the primary (15 kohm reactance at 20khz. sucks)
With the prim shorted together and the sec shorted together, the capacitance across the pri and sec reads 6 nf and if i leave one end of prim open and one end of sec open it reads .8 nf between the two. So guessing (with emf added so the prim voltage varies fom full swing to B+, I'll guess the real seen capacitance is around 2 nf.( 4 kohm reactance at 20 khz. sucks, but not too bad)
Clearly, I need less copper. More interleaves to reduce the leakage inductance will give more leakage capacitance.
I could try vertical sectioning. That will practically eliminate the capacitance, but will it dog the leakage inductance? Both are equal problems. I now have 3 primary layers and 2 seconary layers. So there is around 300 in^2 of primary - secondary exposure. I have the room, I could use more interleaves with more insulation space between. I'm assuming that most of the flux is running around the iron and very little through the air where the coil is exposed, so I could put some large insulator in that area and almost half the Lk Cap and not affect the Lk Ind. Then I could add another interleave.
Any hints what way to go befor I use this iron for a big arss choke and parafeed
This endeavor is half learning experience and half trying to fill a need. If I must, I'll get a nice push pull transformer and make a choke or maybe the power transformer with this iron. |
|
|
| smoking-amp |
Hi Brian,
The leakage L and distributed C ( Ll and Cd ) should make for an observable resonance at the top freq. end. 2*pi*f*Ll = 1/ (2*pi*f*Cd) = Zd
If the primary Z (actually, the effective loading Z, composed of R plate and transformed Load ) is equal to this Zd, then the resonance is critically damped, giving the smoothest top end. This should be a guide as to how to trade off Ll versus Cd.
Putting more insulation between layers or more enamal on the wire will generally just trade off Cd for Ll, giving the same top end freq. but changing the damping Zd. Check RDH on this for sure.
The initial sectioning should drop Cd considerably without increasing Ll much, as long as not too much insulation is put in the wafer between sections. But this is a diminishing return process, more sections eventually yielding not that much drop in Cd but increasing Ll more so. Trick is finding the optimum number of sections, but this should be a broadly peaked curve, so just go with the least effort design.
The RDH xfmr chapter should have some calculations on this sectioning. Another very good reference is "Transformers for Electronic Circuits" by Nathan Grossner, 2nd edition, 1983. Also Reuben Lee's book "Electronic Transformers and Circuits" (download available on Pete Milletts site)
Usually standard lamination sizes have nylon bobbins available for them. Maybe you can find a short bobbin that can be stacked for sectioning (or can cut one down), otherwise you will have to make some insulating wafers and wafer clamps. Getting the last section in will be a problem unless the 2nd to last section is varnished by layer so it doesn't collapse outward, or can wind all sections incrementally at once but wire ends will be a nuisance.
Don |
|
|
| astouffer |
Not to rain on anyones parade but theres probably a good reason why companies that wind transformers for a living won't do large designs like this. Some things weren't meant to scale. Anyhow this may be slightly off topic but its been sitting on my hard drive for about 10 years now.
From: gilmore@casbah.acns.nwu.edu (Kevin Gilmore)
Subject: high power single ended class A tube amp
Date: 16 Jun 94 16:58:48 GMT
I have been reading and seeing numerous all triode, single
ended pure class A power amps show up, but they all have the
same problem. Not enough power. So I set out to design an
amplifier that is pure triode, single ended, pure class A
and have 200 watts or more of power output. Impossible you
say, well not really, its taken me almost 3 months to get
it together, and the sound is fabulous.
WARNING: Ultra high voltages that will not only kill you
but will also set your dead corpse on fire if you are not
careful.
Input tube is a 12ax7. Middle gain stage is a 12au7. The
final output tube is a EIMAC (varian) Y831. You will also
need 3 100 watt 10k power resistors, a 50 uF/3000V output
capacitor, A suitable output transformer rated for atleast
200 watts, and that can withstand 1500V. You will also need
the special socket for the Y831, and a chassis where you
can force air over the tube at the rate of 300CFM.
The power supply needs to supply the following voltages.
3000V at .25 amp, .5 amp peak (for each channel)
450V at .05 amp
150V at .02 amp
Building this is not for the faint of heart. I had access
to virtually all the parts by scraping an NRC induction
heating device plus other parts from a high power AM modulator.
The reason I am using the voltages above is thats what
the power supply chassis I have generates, I did not
want to have to modify it. It weighs almost 200 lbs and
requires 220V.
Each amplifier chassis weighs in at about 120 lbs.
The completed amplifier with power supply takes up an entire
5 foot rack panel.
Current richardson price on the output tube is $1700 each. You
can probably find some at a hamfest for a lot less.
When i finish the electronic version of the schematic it
will be in HPGL format, anyone who wants it can mail me.
gilmore at casbah.acns.nwu.edu |
|
|
| Brian Donaldson |
I'm inches from throwing in the towel and using the iron and wire to make a hella pair of 200H 200ma chokes and go parafeed, but I'm only off by a factor of 2 with my first try. I just have little faith in my capacitance measurements with DVM and don't really know what would be the correct adjustment to make because of that.
I've also seen that most SE designs over 40W use parafeed, and I don't think I'm the smartest guy ever to try to wind a tranny, so I'm not very confident that I'll be the only man to successfully make a 200W SE OPT that is less than -1 db at 25Khz.
What I'm doing is so far off the charts in RDHB chapter on transformers that I don't have a lot of faith in them as well, So I'm off in theory land trying to decide what way to throw $75 worth of wire to get closer to the goal.
I'm also thinking, maybe more iron, less copper. I could have fewer layers closer together. Lc could be a wash and Li be reduced. I think the Li is the problem. I need to fire up the mock amp again and add a small resistance between the plate and OPT and see if I can calculate current change at high frequency. If the problem is more Lc, seems the current would go up and If Li the current would drop. I also maybe need to make a good graph of output response.(PITA by hand though) |
|
|
| smoking-amp |
Hi Brian,
If you can detect some kind of resonance peak in the transfer curve before it falls off big time at the HF end, you should be able to check your leakage L and distributed C measurements by putting them in the formula for a resonant circuit.
The sectioning process will reduce distributed capacitance within a winding but not between windings. So important to have good insulation between primary and secondary interleaves ( also to the core itself).
Can also think of sectioning as just connecting a bunch of smaller xfmrs in series, so if a smaller xfmr can do 50 watts with acceptable bandwidth, then four of them in series should be able to do 200 watts at higher Z. (providing primary to secondary interleave and to core capacitance are kept under control.) Primary inductance and Z here is of course going up faster than just their sum, so there is an advantage in putting them on the same core.
So can look at the design of each section as a Z_final/(Number of sections squared) xfmr design. But must allow N * extra insulation between winding interleaves and core too.
Another possibility could be to use double long E laminations (end to end) instead of the E-I config. This should give you about 1/4 the leakage inductance with the doubled length and halved winding thickness. Also can reduce the air gap due to doubled magnetic path length. Convenient to do at least a two section design that way too.
That would double the high end frequency, even more with sectioning. So your goal should be reachable.
Since you saw bumpy peaks in the response earlier, you are probably getting sub-resonances from interleave to interleave lumping of distributed capacitance and leakage, so thicker insulation between interleaves would be indicated if sectioning is not done.
My guess is that xfmr manufacturers do a bunch of trials before they reach an acceptable design. But no one uses long E lams anymore, so you have a better chance of making it where others failed.
Then again, a simple inductor for parafeed is a lot easier, but not trivial either.
Another, VERY esoteric approach is possible, which I have explored in another thread ( taken to its extreme implementation as magnetic film coated bifilar cable networks). This technique can easily blow the doors off any other approach. But gets complicated fast.
Lets say we want to section the xfmr primary into 4 sections. We wind the primary as a quadrifilar winding. Now we would like to connect the four wires in series, but we would face an enormous wire to wire distributed capacitance. So what we do is put a common mode choke on the feed wire ends to each wire sub-winding. ( So at least 3, better yet 4 common mode chokes required here.) This allows the difference in AC voltage to collapse between the quadrifilar series connected windings, eliminating their distributed capacitance effectively. But current still goes thru them in series.
The common mode chokes are wound as heavy enameled or teflon insulated bifilar windings (in isolated sections) on another part of the core or another core altogether. (the DC currents cancel in the bifilar common mode chokes, so no air gap is needed in their core, a big plus for getting CM inductance up enough, and the total CM series inductances do appear across the primary terminals driven by the tube)
The more sections we divide the primary up into, the lower the differential voltage across the bifilar common mode chokes (as well as the common mode voltage from end to end). And the less the differential voltage on the bifilar CM chokes, the less their effective distributed capacitance becomes (between the bifilar wires).
We can even include the secondary into this n-filar scheme with enough sections to get the required N to 1 turns ratio. One can almost arbitrarily reduce leakage inductance between primary and secondary this way. (Wind the whole thing with Litz wire and sort out the strands 50 -50 for primary and secondary, secondary strands paralleled, primary strands in series using the CM chokes.)
More sections leads to less voltage differential across the bifilar common mode chokes, so distributed capacitance comes down in them too (provided good isolation is maintained between the CM chokes and core too). And less CM voltage drop per choke too. (One ends up with total CM chokes turns equivalent to the original primary required turns) One still needs sufficient insulation between the n-filar strands to handle any DC voltage difference between primary and secondary, circuit design can eliminate this DC though by floating the HV B+ supply.
This scheme works even better for a non SE xfmr since the CM bifilar chokes can be put on the same non air gapped core as the n-filar windings and the CM voltages are maintained by xfmr action using the right number of turns on them (rather than just high CM inductance)
Well, its complicated though. :bigeyes:
(almost DC to light bandwidth when implemented in the mag. film coated bifilar transmission line extreme approach.)
Don |
|
|
| smoking-amp |
Was getting late last night when I posted the "Esoteric xfmr" option. The "Correct" way to implement this is to put the Common Mode inductors on the secondary side of the xfmr. (rather than the primary side)
Ie., for an N to 1 turns ratio xfmr, we use N, 1 to 1 bifilar xfmrs (just isolated sections on the same core) and connect the primaries in series and the secondaries in parallel using the bifilar common mode chokes to spread out the common mode voltages of each secondary winding to match that of its associated primary winding. (This is now formally exactly equivalent to my earlier thread posting on using magnetic film wrapped bifilar transmission lines.) Each bifilar CM inductor is put on as an isolated secton too, either on the same core, or, for SE design we would use a separate non-gapped core for the CM inductors.
Putting the CM inductors on the secondary side only requires half as much CM inductor turn winding as putting them on the primary side. (Each CM inductor just spans from one secondary winding to the next to parallel them, so they are cumulative in inductance for reaching the primary voltage extremes.) So this is the more "Correct" way to do things. The CM inductors still have balanced currents even for a SE xfmr, so they can be wound on a non gapped core. (the 1 to 1 bifilars though still have to be put on a gapped core for a SE design)
Good low distributed capacitance winding practice must still be observed for the winding of all isolated CM or 1 to 1 xfmr sections. Minimizing insulation between these sections is no longer a problem however (as in normal sectionalized xfmr designs) since the leakage flux still largely wraps around each section, still producing tight coupling from primary to secondary of the 1 to 1 bifilars. (We have reduced the usual N to 1 xfmr problem to N, 1 to 1, xfmr problems ) So thick insulation wafers or bobbins can be used to isolate all sections.
The total CM inductance of all the CM inductors in series still appears across the primary, so their inductances must be sufficient to reduce magnetization current load on the tube driver(s).
The differential capacitance of all the CM inductors is paralleled across the secondary, so the effective transmission line impedance of these bifilar inductors must be a minimum of N * the 8 Ohm output impedance. (just use teflon or kynar insulated twisted wire to wind them)
Don |
|
|
| Brian Donaldson |
One thing I can't figure out... I understand that having the 2 windings thin and close together reduces Li, but is there any effect from how far the coil is from the core and how much of the window is open (the distance from the outer legs of the core to coil. Should I use larger wire to better fill the window an be near the outer core legs or is their only use completing the mag circuit?
Connecting the bifiliar windings together through a CM choke is a neat idea. It's making my head spin. I'll need to digest that one for a few days. Also using double E core is not a bad idea, but I really don't want a transformer that's 16" long :bigeyes: Before I do anything, I need to spend some time getting reliable data, not just 5 min with a LC meter. I have little faith in my lumped capacitance measures, But I think shorting the sec and measuring the I of the pri should be a decent approximation of the Li. If so, I would say Li is my main problem. I wanted to make this amp full range, but truth is it will be crossed out below 100hz, so I could always drop my target Primary Inductance from 75H to 15H. That would greatly reduce the number of windings, so I could reduce insulation between P-S, reduce # of interleaves and drop Li and keep Lc same.
The parafeed is looking better every day because with SE transformer, I'll still need a hella choke in the B+ to get it perfectly clean where the choke in the parafeed takes care of that.
Thanks for all your help, Don. I'll buy you a beer or 6 if you're ever in Houston. |
|
|
| poobah |
Brian,
Window fill does effect the result. Under fill will cheat you on inductance and creates excess resistance... all other things being equal. Not sure how it would affect leakage inductance. If there is a significant effect; it can't be good.
;) |
|
|
| smoking-amp |
Shorting the secondary and measuring the primary L should be OK for checking leakage L as long as your L meter is accurate. The leakage L generally is mostly air path so the issue of initial Mu of the iron doesn't cause a significant problem for leakage L measurement.
Distributed C is more tricky, since common mode capacitance to the core is involved too besides winding to winding and intra-winding C. Easiest way is to measure L leakage and the HF peak and calc the C. But measuring primary current at HF can do this too.
On winding very near the core surfaces, this generally should be avoided on capacitance issues alone. As far as magnetic issues go, if there are primary interleaves near both the inner core AND the outer core surfaces, their leakage Ls (that might otherwise go thru half core and half air) cancel out in the interior air (or winding actuallly), so flux stays in the core. But if only one surface has a primary interleave near it, then leakge will increase. So best to have symmetry of primary interleaves.
Other than this assymetric surface effect, leakage pathes will form symmetric ovals in the winding around primary sections. By making them long and thin, the air pathes have lengths comparable to the steel paths, so are not favored due to the steel's higher Mu. Short windows, or square winding cross sections make the leakage pathes more circular of smaller circumference than the steel path so become a more serious problem.
Don
Using less than the total window fill can reduce the leakage L some since it makes the winding cross-section thinner. But for a SE gapped design you are starved for primary inductance, as Poobah has just pointed out, so likely best to use the whole window for winding. However, I wouldn't just increase the wire size to fill the window if not needed, since bigger wire will have more distributed capacitance due to increased surface area. |
|
|
| astouffer |
| Brian, where are you getting your magnet wire from? I found that www.surplussales.com sells it for around $10/lb. |
|
|
| Brian Donaldson |
BAEWIRE on ebay has nice stuff for $6/ lb
Buy it now paypal and it's here in 3 days. they have different insulation thicknesses and all gauges 10 to 35 |
|
|
| tubelab.com |
I just got a chance to read this thread. Sorry to hear about the transformer difficulties, but we have all been told that this wouldn't be easy.
I got a bonus at work, so now I need to find some time, then decide whether to buy the big Hammond 1642SE's or go parafeed. Either way I am going to be buying some transformers.
I have been thinking about building a really big CCS using another 833A to eliminate the choke. Then maybe I could find a big toroid for thr OPT. Yes it would be big, and horribly inefficient, but it would be cool.
Nothing could top Brian Becks monster 833A amp. 4 833A's per channel 8000 volts, and NO OPT. His schematic is in post 210 of the 833A thread:
http://www.diyaudio.com/forums/show...4424&highlight= |
|
|
| Brian Donaldson |
His amp could be used as a cigarette lighter too. A hela choke would not be that difficult, and I think it may be the wizer way to go. My first thoughts that sent me off in this direction was the fact that the core will never cross the zero flux line. I thought there was magic to be had there, but lost of people put 3 and 4 transformer in the signal path that cross the zero flux line, so it can't be such a bad thing. Parafeed also greatly relaxes the need of flawlessly clean B+.
But I haven't yet thrown in the towel. I'll build at least 1 more. My first working prototype is off by less than a factor of 2, so I still think I can do it. The problem will come in making some sort of bell end. I may make a mold so I can fiberglass a few. |
|
|
| poobah |
Brian,
If you're using standard lams I can tell you where to buy end bells.
I don't recall if they produce monster sizes.
:xeye: |
|
|
| Brian Donaldson |
| They're long E, standard, but not common. I would be interested anyway because I may choose to wind most of my own transformers for this. |
|
|
| poobah |
Try looking here .
I far as I know... they are the last company in the states making these.
;) |
|
|
| Brian Donaldson |
| Nothing anywhere close to big enough. the lams are around 7.25"x9.75" |
|
|
| Brian Donaldson |
Here's some good data.
My method. Feed primary from sig gen through 466 ohm resistor. Then measured voltage drop across resistor holding sig gen at 2V peak. I read drop at 1/4 octave intervals from 1Khz to 32khz with the secondary open and with the secondary shorted.
I entered the data into a speadsheet to and computed voltage across the transformer (2V- Vr), Current through resistor, and reactance of transformer for each set. Next I computed what the equivanent capacitance of the open seconary data would be and the equivalent inductance of the shorted sec data would be. If I threw out the data around the resonant freq, 22khz, all of the data was within the error of my measurement from the scope and I have 420 pf leakage capacitance and 100mH of leakage inductance. The capacitance is not bad for 5K primary but 100 mH is way out. I would like to get this number down to 25 mH:D then I'll have a honking hunk of iron.
One thing I know is that I have too much insulation between layers, The top layer is almost round. Maybe that's part of the leakage problem. I saw a pic from a Japanese audio transformer winding shop, and the lady had the coil in a bench vise. I thought she was compressing it to make it fit into the window, but she amy have been squaring it up to reduce leakage I.
Upping the number of interleaves really drops the LI, so I may resort to 1 more interleave, vertically section prims, and reduce interlayer insulation. I'll read raiodtron handbook again and think. I'm incouraged. I can also always double up on the iron also. Or at least 50% ip to 3.5" stack. And I may double E. The transformer would then need to be pretty because there will be no way to hide its 16" of glory
edit- some of the redunudant descriptions of data points |
|
|
| Brian Donaldson |
| OK, yours truely is home alone and up to no good. |
|
|
| Brian Donaldson |
I made a board for the input and driver stage of the mother of all amps and took it upstairs to give it a real listen. The input is a 6SN7 led biased and loaded by CCS (second half of 6SN7). Driver is a 6EM7 small triode led biased and resistor loaded with the large triode as a resistor loaded cathode follower and a pot to wrap a little feedback around the driver to reduce output impeadance when you get heavy into A2 with the 833.
WARNING Kids don't try this at home. there's 1400V on the top of that tube. I stood back and plugged in the supplies one at a time. 833 heater, front end, Big Bad B+ |
|
|
| Brian Donaldson |
With the exeption of all of the background noise from AC heater and poor filtering with by bench supply, it really sounds good. The HF rolloff doesn't sound near as bad as I thought it would. There's still plenty of air. It sounds much like the Aleph clone it's playing with if you had the treble on my POS Denon pre turned down 2 db. (no, I keep the tone controls switched off.)
Sounds really musical and POWERFUL. I didn't feel like wagging my scope up there to measure power output, but it's as clip proof as the Aleph that does 200W @ 4ohm. My speakers are 90 db per 2.83V and I like it very loud. and the 833 had no problems giving that. I'm really stoked that I can make the next tranny work. I'm a lot closer than I thought.
Edit...saving another post
I should have put a CD case in the pic because all of those parts are large and it makes it hard to understand how big that tranny is. I was nervous of dropping it on me as I carried it up the stairs. |
|
|
| anatech |
Hi Brian,
What POS Denon pre are you using?
-Chris |
|
|
| Brian Donaldson |
| PRA-1500 It has a 10 mv 2 mhz oscillation on the outputs. It's all op amp. the balanced outputs are just RCA's through another chip. I'm glad I only gave $50 for it off of ebay. I thought it might be decent since it had balanced outs. NOT. |
|
|
| anatech |
Hi Brian,
$50.00. Not too bad. If I lived closer I'd happily give you that.
All balanced outs are op amp derived, unless it's xformer output. That includes most Tascam semi-pro stuff. The SSM chips are op amp based inside. I have seen much worse.
The 2 MHz osc. is solveable, op amp replacement would improve the sound. Try some power supply bypassing and check those old electrolytic caps. This still leaves you with an op amp based pre, but it should sound a little better. ;)
It's hard to find a really nice sounding preamp. Just tested another Carver Lightstar Direct. Good but not involving. Luxman C-05 was not nice (should be for the money). I'll have to dig up another Marantz 3650 to try.
Seems we can't win.
-Chris |
|
|
| Brian Donaldson |
I need to get back to this http://www.diyaudio.com/forums/show...3280&highlight=
I have a ground problem with the built in sub crossover. I got frustrated and benched it. I'm now thinking I will ditch the PSU, crossover and BOSOZ and insert a tube linestage of my own design. LTP and tube CCS are my present leaning. I'll see if there's still room for crossover.
I have an HK 825 that's really a lot nicer, but 15 yrs ago, beer was spilled in it at college. The vol pot went scratchy and the switches flakey and It's been apart numerous times since with bad solder joints. I still have it, but I'm tired of fixing it. So it sets in my closet.
When I'm home alone, I often plug the CD straight into the Alephs with Y's giong to the pre for the sub crossover. It's a bit loud, but that's how I like it. |
|
|
| Brian Donaldson |
| I know I could mod it and fix it's shortfalls, but after seeing inside, I have no pasion for it. It would be polishing a turd. It serves a purpose for now. |
|
|
| anatech |
Hi Brian,
Don't feel bad about the HK. The bad joints are perfectly normal. They tend to run warm, caps suffer in the heat. Beer doesn't help.
You sound like you have a good plan.
-Chris |
|
|
| Brian Donaldson |
Good news kids. I've wound the second trany adn here's what I have
I went with a 3" stack of Iron :bigeyes: 5000 turns 24ga primary 18 layers total
6 secondarys,18 ga, 73 turns each, 3 layers bifiliar.
core
3 layers prim
1 layer sec
6 layers prim
1 layer sec
6 layers prim
1 layer sec
3.4 layers prim
with laser transparency between prim and sec, and 3 layers of teflon tape between primary layers
Now the good stuff
leakage inductance is down to 35 mh:D
leakage capacitance is up to 1400 pf:xeye:
resonance is 23khz
and if the Rp is 5 kohms, we should be damped
We may have a winner.
It's too late to mess with this more. It'll have to wait until maybe next weekend. I may be able to mock up the amp and run a few quiet tests before then. (I hope. I'm stoked and would like to hear it) Only down side, I don't think there'll be much of a market. It took me 4 hours to wind the thing. I'll gladly share knowledge though.
edit. corrected wire size I transposed |
|
|
| rcavictim |
| Congratulations. That is a serious amount of work. I give you top marks for tenacity. I hope the xfmer performs as desired. I am curious. Have you measured the primary L with the amount of gap you think you will run and what is it if you know? |
|
|
| Brian Donaldson |
| Not yet, but I calculated for 75H with up to 180 ma bias. If I reduce the gap and bias, it could go much higher. My sig gen has gone flakey and won't go below 75 hz, so I'll have to get the amp mocked up and check bottom end response, but that's out of the question with kids around. With a 14.4V battery across the secondary, It'll throw a hella arc :hot: |
|
|
| schiller |
| quote: | leakage capacitance is up to 1400 pf:xeye:
|
Hi Brian, this is a very good value given the size and the interlave pattern. I have red your measuring procedure, but i couldn't figure out if you tie the "cold" ends of prim. and sec. together. If you float the speaker output there is no need to do it, but if you ground one end, you sould take it into account.
Greetings
Konstantinos |
|
|
| Brian Donaldson |
To calculate the leakage cap, I fed the prim with my sig gen with a 470ohm resistor in series. I held the sig gen output voltage to a constant 2V and measured the drop across the resistor. I calculated the impeadance at 1/4 octaves and discarding the 4 around the resonant peak, averaged the remaining data. The secondarys are floating.
I realize that this can be dangerous because the capacitive coupling could put a few hundred volts at the speaker terminal. I didn't concider this until you asked. As it is, it's like the CT of the secondary is connected to the CT of the primary and there is an increasing capacitance as you move to the extremes of the coil. If I ground one end of the primary, it'll be like (they are) one extreme is tied together and the other end will have an increasing cap, but swinging twice the voltage. I think this should double the percieved prim capacitance. :bawling: I may need speaker wires rated at 1000V :hot: because this will blow my capacitance out of the water.
When I have time, I'll check this. I haven't ran all of the tests on the second transformer I did on the first. I just checked the leakage L with LC meter and leakage C with sig gen at a few frequencys. |
|
|
| Brian Donaldson |
| Wow, I bet the floating secondary is why it sounded so much better that it measured. Maybe I need to add an input transformer and float the amp and ground the OPT sec. That would greatly help the leakage cap issue. Or is there some other method I've overlooked? |
|
|
| schiller |
I don't know how you can do it, i don't think that is possible. a couple of friends of mine do run the sec. floating, but this is risky, as you wrote.
Have you tried to measure the trans, simulating "real world" operation (generator,one resistor in series, so that the total resistance equals the Rp of the tube, secondary loaded with the intented resistive load)?
It would be very close to reality.
I wouldn't even dare to think winding something as big as your transformer, this makes your findings extreme valuable, at least for people like me, who wind transformers, or really want to know the true performance of the costly iron they intend to buy.
Audio Note und Sowter of U.K offer some big 10kOhm transformers with very good high frequency specs. They use split bobbins, combining both types of interlaving. If your current design doesn't satisfies you, you maybe wish to try the following:
The same interlave that you use now, but split equaly between the 2 halves of the bobbin. (14 sections, all in series prim. and sec) I' give it a try with a 2 inch stack, lams are 6in x 5in (10k-8ohm, 30w ). I have no clue if it is a good idea, but at least i shall have learned something after this.
Much smaller than yours, but if it works, it can be used with a proper anode choke ( easy to build with low shunt capacitance)
as parafeed to handle easily 150w.
One more question, if you don't mind: What quality of iron do you use?
Thanks for the tremendous amound of info you have provided so far.
Greetings
Konstantinos |
|
|
| Brian Donaldson |
My latest wind is a 3" stack of http://www.tempel.com/products/tld/Elkeepers/ei230.html . 29 ga M6. 3" stack is somewhere around 40 lbs. I can't find nylon bobbins, and I don't trust a 10 lb paper bobbin to stay together from the mandrell to the core. You have to do a little hammering to get the center peice out of the mandrel to collapse the 2 outer peices. Bu tI had a thought of making a bobbin with 2 or more square donut shapped PCboardsconnected together outside of the winding window with a few peices of #12 wire soldered to the outer PCBoards.
Because this wind is so thin and long, there's a lot of surface contact between P and S. This is not good for capacitance. I think I need fewer interleaves. The high leakage inductance in the first coil I believe was due to sloppy winding. I was using a sheet of transparency between every 2nd layer of primary, and sometimes the sheet didn't lay down tight on the first part of the layer. By the middle of the layer, the transparency was wrinkled. By the end of the transformer, the coil was almost round. I think this was the cause of the high leakage inductance. On the second transformer, I only used teflon tape between prim layers and I started cutting the transparency so it would conform better to the imperfections in the previous layer. I also filled the last little bit of the wind with electrical tape so the boundry turns wouldn't fall off into the gap.
Something I didn't consider the first go was that when I put my scope across the amp input and secondary to check freq response, I was grounding the sec. I need to go by Ace Electronics today and see if they have a small audio transformer so I can isolate the ground, or maybe they'll have an affordable differential probe that can handle 700V off ground.
I didn't check with resistor on the primary. I will check it with the whole breadboarded amp sometime in the next few days. My partner is out of town and I'm having to work double last week until next week. The primary will be swinging 1300V with 180 ma bias, so I don't totally trust the data had with 2V swing and no bias.
I know I can resort to parafeed if I can't make this work. The Iron was free samples but I've spent $250 in copper so far. And If will need to by more iron to make a pair (I was quoted somewhere around $4.00/lb in 64 lb increments). But if you consider that the transformer companys are selling transformers at half the wattage of this one at $400 each and up, it's not much money even if I waist a few hundred dollars in prototypes.
I had a wild notion that I could train one of my employees to wind these things after I get the kinks out of the design and sell them. I don't know now. The cost of materials to learn is a little high and I doubt that my guys would have my entusiasm to do it right. Maybe I need to buy a coulple of third world kids and lock in my garage:D |
|
|
| Tweeker |
| Are you planning on potting these lovely power plants? |
|
|
| Brian Donaldson |
I think potting is not an option. I've been thinking what I could do to doll them up. First, I'll semi reses into the chassis so I only have half to deal with. Then I was thinking get a friend that works in a metal fab shop to brake and weld a stainless bell with some holes punched in the transistion section between the parts in plane with the lams.
Now on to some results.
For this test, I hooked the sig gen to the primary and a 6 ohm dummy load to the sec. Scope is connected to the pri and through a small sig interstage transformer (to isolate the ground) to the secondary.
with voltages normalized at 2khz, output starts to droop slowly at 8 khz and is 2.8 db down (5.8v/8v) at 20 khz (3.5 db without isolating transformer or (read grounded sec)) then rising back to about 2 db down at 40 khz, then dropping again.
I need to rerun my complete Lc Lh test and make one more iteration on the design. I'm still learning with the winding tecnique and I think I'm reaching close to mastery. Next one, I'm thinking bigger wire to more fully fill the window and compress the windings with a vise to keep it tight. Also, more teflon and less air in the core.
I'm almost there. I'll try to listen tomorrow and report.
Edit to finish post. I hit the wrong button and submitted prematurely then correct error on 20khz output |
|
|
| Brian Donaldson |
I have an idea to make many vertical sections doable, but I want to run it by the physics experts before I run into the dark woods with no direction.
I have made ends for a ridgid bobbin out of PC board materials. The plan was to solder a 12 ga wire from 1 end to the other to hold the two ends a certain distance apart. This will slightly round the coil on the part that is outside the winding window.
What I was thinking to do was cut 10 L shaped peices of mdf and clamp them to the madrel as spacers for the secondary. Add a piece of cardstock to each side of the wood clamp/spacers and wind the primary in 6 verticle sections about 3/4" wide for the outside 2 and 1" wide for the middle 4. I'll enamel (or poly or silicone or something...any suggestions...something that is convienient to do on the fly between layers.) After the adhesive cures, remove the MDF spacer/clamps that are held together with zip clamps or tie wire. The cardstock will be glued to the primarys and I'll have a nice grove to wind the seconarys. This should reduce the pri-sec capacitance by a factor of 10 to 15 and the interwinding capacitance by almost as much, and I'm hoping the leakage inductance will also be reduced due to 1 more interleave, but I'm not sure what the difference will be going from horizontal sectioning to vertical sectioning. The windings will be closer to square as opposed to very thin rectangles.
The main thing is it is buildable, and maybe even less of a PITA than the one I'm doing now. When you get close to the end with all of the loose wires flopping about, it get's hairy.
Any thoughts, guys?
Edit. I lost my camera, so I can't post any pics until I buy a new one. Hope to next week if I get the time. |
|
|
| smoking-amp |
Hi Brian,
If I'm visualizing this right, you plan on one big bobbin with vertical sections along its length, with primary and secondary in separate sections. This will give near square winding cross sections as you say, and this won't be good for leakage inductance. (Going to a huge number of alternating sections would work OK since the sections get thin then) Better to put some primary and secondary layers in each section with an insulating layer between.
I myself would try to use a bunch of separate short bobbins that could be stacked endwise, so they could be wound separately, but I know this is a pain-ita if bobbins aren't readily available. Also, this requires very thin sides to the bobbins, so they have to be supported by the mandrel assembly during winding. And of course a bunch of wires have to be connected later too.
Going back to your long bobbin with vertical sections approach, do you have room with your winding rig to attach an idle wire spool to the bobbin/mandrel assembly while the other wire spool is being used? The idea would be to avoid cutting wires between primary and secondary windups, so that you can alternate back and forth readily between primary and secondary spools being actively used. That way it would be straightforward to do interleaved layers in each section. Although a bit awkward with a spool flying around during winding, this could be minimized if the bobbin/mandrel is suported on just one end and the idle spool affixed to the other bobbin/mandrel end to keep it centered on the rotation axis. A lathe with a pipe thru the mandrel comes to mind.
Don |
|
|
| Brian Donaldson |
I had the idea at work, and didn't have RDH4 to check. So it's the thin layers that help reduce leakage inductance.
Verical sectioning will only help my interwinding capacitance, not the prim-sec capacitance. I don't think that's my real problem. I need to reduce the leakage inductance and prim-sec capacitance. I think actual bobbins is also out because the thickness of the sides will add up if I need to go beyond 12 or 15 bobbins. I think I could clamp a stop and use a piece of paper with silicone on it to glue the paper ans winding edge together, then use it as the support for the next winding. The question is how many sections would I need to get below 20 mH leakage.
On the last itteration, I first thought to add iron, loose copper and sacrifice some bass and lower prim induc. But I thought maybe I can have my cake and eat it too.
Maybe I can try 3 skinny windings and see what I get. Would leakage inductance remain constant as I added windings if they were the same thickness? I could blow $5 worth of wire to see what I have. I'm booked solid this weekend with my daughters horse show and work, so Imay not get to listen to the one I have until this week. It's also a little hard to judge with the background noise of my HV B+. It's a voltage doubler with only the one cap per leg, so It hums/buzzes badly. |
|
|
| smoking-amp |
To estimate configuration effects on leakage inductance, think of the magnetic flux lines as rubber bands that like to wrap tightly around the driven winding (ie, the primary) but they like to avoid like the plague the load winding, due to its opposing load current. They like to shrink in length like elastic bands because that drops the magnetic path length "resistance". So by making the primary long and thin cross sectioned, any leakage pathes have to take a path that is nearly as long as the iron core path, and the iron core path has effectively a 5000 or so shorter equivalent magnetic length due to its high Mu. So this keeps the flux in the core which of course goes around the secondary like we want it to.
Breaking up the primary into smaller cross sections with gaps, or worse, secondary cross sections, gives the flux opportunity to find short pathes around just the primary sections. So the primary needs to look fairly homogeneous throughout the total winding cross section.
You need to keep symmetries in mind when picturing this, what might look like a short loop may not exist if nearby loops have opposite leakage flux direction, like close thin interleaved layers.
(for example, a long thin winding cross section does not end up with a bunch of small circular leakage fluxes along its cross sectional length, because the adjacent loops have opposite flux directions and cancel out, leaving just the long way around the outside)
As you pointed out, the vertical sectioning will only help the interwinding capacitance problem, and generally worsens leakage inductance, hence the usual tradeoff. So one needs to put insulation thickness in just the places where its really helpful since it makes space for leakage flux. So the vertical sectioning relieves the insulation requirements between intra winding layers, but not for primary to secondary layers, or winding to core interfaces.
On stacking up some thin sections, you should find that the leakage L comes out less than the sum of the individual section's leakage, since the local flux loops will get stretched longer with each section added.
I hate to waste a bunch of wire at your expense, but here is an idea that is moderate risk, but high gain, hopefully. You could try the progressive winding technique used on high bandwidth toroids. The risk is that it may not perform as well as a well thought out conventional design, but the gain is that its really easy to do once set up. This ONLY makes sense to try on a long E core like you have.
This uses no obvious vertical or horizontal sections, the primary and secondary get wound in one shot each (separate layers though) on the long bobbin. The trick is dithering the winding point back and forth a small amount regularly, so as to create the necessary thickness to the winding for all the required turns. The winding being done in a single pass from one end to the other. The result is many chevron shaped layers along the length of each winding that act as internal sectioning, so that adjacent turns have nearly the same voltage.
Of course, thick insulation will be required between the primary and secondary, and between the iron core and 1st winding. The way I would attempt this would be on a lathe with a thread turning attachment to handle the slow steady progression rate (maybe have to slow even more than the finest thread feed rate) and a small motorized teflon wobulator with a hole in it for the wire passing thru in place of the usual cutting tool or wire feed.
One would have to figure out the relative feed rates, wobulator oscillation rate and wire feed hole oscillation length to get the winding fitted in one pass down the bobbin. But once working, you could run these off like pancakes.
(you most likely need a linear rate of motion from the wobulator to get uniform chevron layers, maybe use an old 8 inch floppy disk head positioning assembly, some simple digital counters to control the stepper zig zagging)
Slightly off subject,
I just got some custom CFB xfmrs wound by Edcor. I just asked for their CXPP 100W 5K xfmr reconfigured for 20% UL and everything brought out as separate leads with good insulation between the 80% parts and 20% parts. Same price.
I'm going to use these for a bunch of experiments I have waiting on the shelf. Besides the usuall 20% UL (for horiz. defl. tubes with their lower screen voltage) (and can do separate lower screen voltage too), and also 20% CathFB distributed load designs, I plan to try SE with a DC balancing current source on the other P-P side, AND then my idea I posted a while back on complementary current SE simulation using a true complementary drive for the second P-P driver (no even harmonic cancellation like the usual P-P) When I get this up and running I will post some dist. spectrums to compare with the SE with current source. This may prove to be a workable way to get high power SE using off the shelf P-P xfmrs. A lot of work ahead!
Don |
|
|
| smoking-amp |
Ummm...
On that leakage inductance calc. for adding sections, I think that has to come out as near the sum of the individual sections leakages. This is becuase the number of turns are increasing too. So, for two sections, we double the turns for 4x (n squared) but double the leakage path length for 1/2x giving 2x net result.
If instead one put another thin section over the first thin section to double the turns, then the path length would be about the same but n squared would net a 4x increase. So going toward a square cross section is definately bad.
Keeping the winding cross-section the same shape, but increasing the turns by two, should give something in between like 2 to the 3/2 power. Lets see, n squared gives a 4x, and path length goes up by square root of 2 (for doubled cross section area), so checks out.
Don |
|
|
| Brian Donaldson |
| The winding you are sugesting would look like a dithered and interwoven thing, but would actually be slanted interleaves sort of like a new roll of kite string, but being multiple strands, they would lap each other in and out. That sounds like a great patend and mass produced idea that has great possiblities, But hardly acheavable with my cordless drill trigger in one hand and thumb/forefinger wire guide | | | |
