Question on taps of T2A autoformer

[WMcD]

I have now posted a corrected version.

Please read my CAUTION in my edit of the original post. Three of the cells need an extra zero.

I'll work on including some diagrams which will clear up part of the confusion.

Gil

[Cut-Throat]

Please read my CAUTION in my edit of the original post. Three of the cells need an extra zero.

I'll work on including some diagrams which will clear up part of the confusion.

Gil

OK, Thanks!

[Deang]

There are three ways you can attenuate:

1) No resistors: You can drop down on the taps, which gives you 3db increments. However, the reflected impedance through the autoformer back to the amplifier doubles with each 'step'. We are only dealing with the T2A here. If you use Bob's 3636, you can do 1dB increments. Keep in mind here that with this method you HAVE to change the capacitor value with each tap setting. Honest, if don't know how to do the basic calculations or use a crossover calculator so you can scale the parts properly, you shouldn't be messing with this stuff.

There are two "sides" to an autoformer: The input side (tap 5), and the output side (taps 1-4). On the input side, between tap 5 and the amp -- is the primary high pass cap. Everytime you go down a tap on the output side, the impedance back through the autoformer to tap 5, doubles. This is the reflected impedance.

Tap 4, -3dB, 2x the impedance. You go from 15 ohms to 30 ohms. For a 400Hz crossover at 30 ohms you have to use a 13uF capacitor. Tap 3 = 4x the impedance.

Tap 3, -6dB, 4x the impedance. You are now at 60 ohms, so the cap value gets cut in half, which is 6.5uF but you can use 6.8uF -- close enough.

Tap 2, -9dB, 8x the impedance, 120 ohms. Halve the cap size again -- 3.3uF works good..

Tap 1, -12dB, 16x the impedance, 240 ohms. Halve it again - you are now down to a 1.6uF cap!

Those are the cap values for 400Hz, If you use a different crossover point -- they all change!

RESISTORS:

Bob Crites hates resistors. He is in good company, PWK didn't care much for them either. Since they change power into heat, they reduce efficiency, Since I like being in good company, I'd like to hate them too -- but it's hard! They are convenient, inexpensive, and do bring a few things to the table. Using them means you don't have to keep changing the cap size, they bring down the impedance which helps amplifiers with a high output impedance deliver flat response, and regardless of amplifier ...

..."The very small mouth on the K700 causes reflections down the throat that show up as impedance peaks. When the impedance peaks at these frequencies the crossover has very limited attenuation compared to driving a pure resistor. The addition of the parallel resistor makes the loading of the network closer to ideal. 6dB networks are the most sensitive to this. Sometimes the Klipschorn and LaScala can 'buzz' at this frequency (around 250hz on the K400, and the extra couple of dB attenuation on this one note can clean up the 'buzz' without going to the expense of rebuilding the K55 (a notch filter can be added if the resistor is not enough)." -- DJK

For the purpose of this discussion, we'll just talk about why don't have to change the cap size.

You want to drop another 3dB, so you go to the #3 tap (-6dB). When you do it, the reflected impedance doubles and you're at 60 ohms. This means 6.5uF to keep the crossover point the same. If you don't want to change the 13uF cap, you have to get the reflected impedance back to 30 ohms. So, you put a 15 ohm resistor in parallel with the squawker.

A 15 ohm resistor in parallel with a 15 ohm driver = 7.5 ohms. On tap 3 you have 4x the impedance, or 7.5 x 4 = 30 ohms. There you go -- you get to keep your 13uF cap.

Say you want 3dB more of attenuation (-9dB) because you're using a different driver that's more efficient than a K-55 (and big honking horn to go with it). So, you have to drop to tap 2, which will double the reflected impedance again (8x). 120 ohms now, and you need to get it back to 30 ohms if you don't want to change the cap. Using your trusty Parallel Resistor Calculator, you figure out that if you parallel a 5 ohm resistor with your 15 ohm squawker, you get 3.75 ohms -- and when you multiply by 8 -- you get 30 ohms. At this point you should probably go to a 25 watt resistor, but I think if you need this much attenuation, you're better off swamping the autoformer and adjusting the capacitor value.

SWAMPING THE AUTOFORMER (or ALKing your network, "Constant Impedance")

We'll get into this later -- unless someone feels like saving me the trouble.

[Deang]

Gil,

When I checked to see if the file would open, I didn't really look at the data because I was in the middle of something else and just closed it after I confirmed it would open. I've been looking at it, and I must confess I don't really understand either. Between the driver and the autoformer -- impedance stays pretty much around 15 ohms. I mean, the value for the swamping resistor is based on the impedance of the driver, not the actual reflected impedance back through the input side. For those values to work, wouldn't they have to be strapped between Input neg and Input pos on the network. If so, that creates several problems!

[greg928gts]

Good post Dean, thanks.

Greg

[Cut-Throat]

Thanks for your post Dean. This is way more complicated than I originally thought. Someday I actually hope to know what you're talking about. I may read this about 10 times before it finally sinks in.

All I know so far is that when I moved my T2A from tap 4 to tap 3 and put a 15ohm resistor on my midrange it improved the sound imensely. It looks like if I want to move from -6db to -7db, I might have to have a PHD in electrical engineering![]

[WMcD]

I've corrected the spread sheet and added a few notations.

DeanG describes it as well as I could, if not better. Thanks DeanG.

I believe that the spread sheet gives accurate answers. But note that I'm starting with the assumption the dirver is 13 ohms. that is pretty much what you see on the Atlas data. [Edit, please see Dean G's post saying the value increases on a horn.]

When you change taps, you can use a new value of the cap, and not add a resistor. But note that when you open the spreadsheet, it show a cap value on the -3dB tap of just about 16 uF, and this is what Klispch uses. [Edit, please see Dean G, again, taking me to task - I should have checked the schematic -this is getting embarrassing - Klipsch uses a 13 uF as he says.], If you adjust crossover freq and attenuation, you should get values of the Cornwall and Heresy.

= = = =

Or, when you change taps, you can keep the same cap, and add a resistor in parallel. [Edit, This resistor is placed at the input to the autotransformer . . like MY DEAR FRIEND DEAN says Grrr, smile.] For example, if you START with the -3dB tap you have an input impedance of the autotransformer of 26 ohms. When you go to the next tap, or -6 dB, the input impedance is 52 ohms (doubled). So you have to put in a parallel cap [Edit, should read "resistor" as Dean says, this is really embarassing] of some value to get it back to 26 ohms. That means you should add a 52 ohm resistor.

This makes sense because the cap-filter sees the 52 ohm input of the autotransformer and the parallel 52 ohm resistor. Equal resistors in parallel result in half the individual values.

- - - -

Now, let's assume you want to use the -9 dB tap. The input impedance of the autotransformer is 104 ohms. This is four times what we started with. And we wish to add a resistor in parallel to bring it back to 26 ohms, so we can use the same cap value.

Dean used a handy dandy calculator. But we can do it in our head, with a little insight. The insight is that four resistors of equal value, in parallel give a result of one-fourth the individual values. Note this is where we want to get back to.

So, let's see. We've got one resistor in the form of the input to the autotransformer, which is 104 ohms. We could just put in three resistors (in parallel) of that same value, to bring it back to 26 ohms. And indeed you could do that. Essentially we've got four resistances in parallel, and again, one is the input to the autotransformer.

But suppose we want to use just one resistor, to replace those three 104 ohm resistors. Now we know that three resistors of equal value in parallel result in a value of one-third the individual values. Therefore our three extra parallel resistors (each 104 ohms) can be replaced by one of 104/3 = 34.67 ohms. If you look at the speadsheet formulas, I actually did a conductance type calculation, which gives the same result.

The "proof" column was just my check using the typical formula for two resistors of unequal value in parallel.

- - - - -

Some people out there might not be familiar with spreadsheets. The content of a cell can be text, or a number, or a forumula. The formula is similar to "BASIC" or even algebra. The variables are the names of of other cells. But what is displayed ("returned") is the text, or the number, or the result of the calculation. If you poke around you should see how it works.

- - - -

Final word via edit. Even though I messed up in a couple of places, I still believe the spreadsheet is accurate in caluculations.

Gil

[Deang]

"...note that I'm starting with the assumption the dirver is 13 ohms. that is pretty much what you see on the Atlas data.

Hi Gil,

13 ohms is the impedance on a plane wave tube. Attached to one of the horns and in-network -- the impedance jumps up a bit.

"...when you open the spreadsheet, it shows a cap value on the -3dB tap of just about 16 uF, and this is what Klispch uses..."

26 ohms calculates out to 15uF -- 30 ohms calculates out to 13uF, which is what Klipsch uses.

"Or, when you change taps, you can keep the same cap, and add a resistor in parallel. For example, if you START with the -3dB tap you have an input impedance of the autotransformer of 26 ohms. When you go to the next tap, or -6 dB, the input impedance is 52 ohms (doubled). So you have to put in a parallel cap of some value to get it back to 26 ohms. That means you should add a 52 ohm resistor."

You said "parallel cap", but I'm pretty sure you meant to say "parallel resistor". Anyways, it appears we are in disagreement -- where are you putting the resistor? If you're putting it in parallel with the K-55, your resistor values won't work. We'll go with your 13 ohms here, which is the impedance between the driver and the ouput tap. We are swamping the drive down, not the whole network. It's not 52 ohms until it's reflected back through the autoformer and reaches the input side. IOW's, you're not paralleling with 52 ohms, you're paralleling with 13 ohms. If you want to get back to 26 ohms, 13R and 52R won't work! The driver is always 13-15 ohms unless you swamp it down, and to get back to 26 ohms, you would use a 13 ohm resistor. I better be right about this, or I'm going to have to initiate a massive recall.[]

[Deang]

Helpful links:

http://www.1728.com/resistrs.htm

http://www.kbapps.com/audio/speakerdesign/calculators/Default.htm

[Erik Mandaville]

Dean:

"Honest, if don't know how to do the basic calculations or use a crossover calculator so you can scale the parts properly, you shouldn't be messing with this stuff."

With all due respect: We can't learn anything unless we begin to work with whatever that thing may be -- learning to drive a car; change a spark plug in a lawn mower; install the diodes in an old Klipsch Heritage crossover correctly; or learn to understand and manipulate the way an autoformer works. You, I, Mark Deenen, Bob Crites, Craig, Shawn, LeoK -- everyone here that has some experience under our belts (some with obviosoly more years than others) all had to start somewhere. That 'somewhere' place naturally has a little confusion associated with it, as does anything when it's new.

Now: I haven't necessarily agreed with the Cheater plug approach Kevin is using, or mainly the public announcement concerning its safety (based on differing home electrical environments and wiring), but he has done a considerable amount of work in amplifiers and has the ability to carry any of the work out related to the the T2A question once he understands it. If he is instructed not to "mess around with it" He'll remain ignorant about it forever. A shame. Crossover calculators are EXTREMELY easy to use! It's simply punching in some numbers concerning crossover point, crossover slope, driver impedance, driver type (midrange, woofer, med, tweeter), etc., and the caculator does the rest.

As I mentioned with that network for which I rolled my own coils (which was easy to do too -- just required some patience), I completely abandonded the T2A for not the series resistor network, but the series/parallel type to maintain correct impedance. I liked it much better than the autoformer, although I know very well what P. Klipsch thought about them. I don't have his ears, though, but I do understand his theory and how the autoformer works. I'm sure Klipsch engineers are even more aware of what was behind the autoformer. McIntosh also employed them in output stages of amplifiers. It's a matter of preference, like anything. The far more common approach to driver attentuation in crossovers is done with resistors, either fixed or variable. You're right: they work well and are inexpensive. If I'm not mistaken, the most recent Heritage networks also use resistors rather than the autoformer.

The main point is that compared to some of the work Kevin has done in amplifiers -- both modifying his Moondogs and building his DRD -- the crossover work to be done is not only more simple (being an intelligent person he'll understand the why's and how's soon enough), but much safer: no B+ voltages to test. You and I both know what it feels like to be shocked by plate voltages. So does Mark, so does Craig, so does virtually every person I know who works with active electronics regularly. Not fun.

Kevin doesn't have to agree with me, but I have confidence in his ability to learn that changing taps on the autoformer will change the reflected impedance to the amplifier. That's simple enough. It's essentially like changing the impedance of the driver to which the amp is delivering power. Now, we just need to understand that, for a given crossover point, the capacitor needed will have to be altered depending on the impedance of the driver in question. It's actually very straightforward. People can learn, as you and I have.

Erik

edit: with some other alterations to the network that are easy to do, you can also use a fixed L-pad, with the resistance values given to you by way of a calculator -- and put the T2A out to pasture. If you would like to try the actual math involved to fin R (parallel) and R (series), I can give you that formula. A true narrow band pass will the end result, and it works very, very well.

[Deang]

You're late.

I waited three days before I even posted into this thread to give you an opportunity to come in with your obligatory "it's all so easy, and anyone can learn it, and if my mom can weld a nulcear submarine together so can you, and by golly -- you don't even need to mess with those pesky autoformers -- you can just tie some resistors together and dial it all in by ear while you're taking a bath!" post. Hey, then I waited another three days -- and you dissapointed me yet again! So, I finally decided I needed to go ahead and post something of substance for those guys -- and here you are! Do I really need to point out that THEY are the ones that said they didn't understand? How about instead of telling everyone how easy it is, you now take the opportunity to teach everyone the ins and outs of this crossover stuff, I mean, it would take a lot of pressure off of me and I'm sure everyone is ready to hear it all from someone who really knows what they're doing! Why not pick up where I left off, and please don't forget the soldering lesson (pictures are helpful) -- it's your show!

[Guest " "]

no good dead goes unpunished. if you partly answer, you fall thru the trap door. if you provide a methodology for folks to understand it, you fall thru the trap door. if you offer alternative solutions, you fall thru the trap door. well that what is really all about....checking in once in a while to see who is falling thru trap doors.

[BEC]

no good dead goes unpunished. if you partly answer, you fall thru the trap door. if you provide a methodology for folks to understand it, you fall thru the trap door. if you offer alternative solutions, you fall thru the trap door. well that what is really all about....checking in once in a while to see who is falling thru trap doors.

Easier to avoid that trap door if you just stay off the scaffold.

[Deang]

The thread started asking for explantion. Explanation was provided, then expounded upon. Two are teaching, and one is a resounding gong. Telling people how easy things are and that they can learn isn't helpful -- telling them how it works and how to do it, is. I'm teaching my 7 year old how to read, it involves more than giving her the books and telling her she can do it.

[Deang]

Amen Bob, Amen.

[Erik Mandaville]

Dean:

Here's what I was responding to:

"Honest, if don't know how to do the basic calculations or use a crossover calculator so you can scale the parts properly, you shouldn't be messing with this stuff."

For the record, I have spent a considerable amount of time helping others on this forum. Kevin is one of those. In this case, my advice was to suggest the possibility of what I think is better solution than the autoformer approach. For me it was, at least.

However, consider your quote above and you tell me what the implications of it are?

That you are providing (some) solutions is helpful; telling people that if they don't yet know something they shouldn't try is NOT helpful. You didn't know what you now know a few years ago. If you simply avoided learning it for that reason, where would you be?

And you're right. Calculators make the work easy. It's simply a matter of plugging in some numbers. It really is very easy with what is now available to the average person. With cumputer based systems and calculators, the only real math required for this is basic. Kevin is MORE than able to make a very good crossover for himself, whether he uses the autoformer or not.

Soldering is a breeze.

It's true -- this place never changes.

Have fun,

Erik

edit: For those of you who are curious to do some of your own work -- DO SO! If you're interested enough and want to learn, it's within grasp. If you would rather leave the work to others (who also started out with training wheels early on), than by all means do what works best for you.

[Guest " "]

no good dead goes unpunished.  if you partly answer, you fall thru the trap door.  if you provide a methodology for folks to understand it, you fall thru the trap door.  if you offer alternative solutions, you fall thru the trap door.  well that what is really all about....checking in once in a while to see who is falling thru trap doors.

 

 

 

Easier to avoid that trap door if you just stay off the scaffold. 

I kind of do that. I exchange a lot of emails with folks who have questions. I do this to avoid the scaffold dancers. I know what I know and I know what I know how to do. I don't need to prove anything to anyone.

[Cut-Throat]

Easier to avoid that trap door if you just stay off the scaffold.

I kind of do that. I exchange a lot of emails with folks who have questions. I do this to avoid the scaffold dancers. I know what I know and I know what I know how to do. I don't need to prove anything to anyone.

Thank goodness for the ones that are willing to share their knowledge in public, so the rest of us can learn something.

[Deang]

However, consider your quote above and you tell me what the implications of it are?

It didn't have anything to do with learning, or the ability to learn -- it had to do with knowing. If you don't know what you're doing, don't do it until you learn how. If I say please don't do this if you don't understand, I don't need you informing me for the 1000th time that they can learn it -- and then once again not offering a single thing in helping them attain the goal. Conveniently, you always want to play the authority on the topic after the answers are provided -- which basically amounts to you muddying the waters as much as possible!

[Erik Mandaville]

Kevin:

It may be that the actual level of attenuation you quoted may ultimately not be what you prefer. If it is, that's fine. However, in my latest experiments with these networks, I found resistive attenuation much more to my liking than what's provided by the autoformer. That said, the results are most likely not due to the effect of the autoformer per se', because I don't think, nor have I found, that it offers any special sonic attribute. It attenuates, and does so by presenting an impedance mismatch. The horn changes you made on your K-horns very likely changed the sound much more than would replacing the autoformer with either fixed or variable resistance to attenuate the squawker.

What you can try, which is as easy if not easier than many of the connections you've made in amp mods and complete builds in the past, is to build another network that uses a variable resistive 16 ohm L-pad on the squawker (if you're using a K55, which I think is the case -- correct me if I'm wrong). You have to change the value of the capacitor needed, though, to offset the change in impedance brought about by its former connection to taps on the autoformer. Ideally it should make a difference in what you hear. I've done it both with the correct cap change, and with the stock value, and all I heard was the difference in attenuation (when using T2-A). Bob once confirmed the same thing.

With the autoformer gone, you connect the tweeter directly to the amp input, but will need to change the capacitor value if you do that. There is a reason for this that we can talk about. Actually, it was, at length, when I was going through the designing of the new band pass networks I made.

In any event, the type A design is absolute simplicity as far as crossovers go. Moreover, there is so much information available about the behavior of these drivers that it's not hard to use published tables (or even easier, a calculator) to make a new network that performs just as well as the type A, or in my case, one I really preferred. I very much liked the type A, by the way, but prefer this new one by far. It consists of a 4kHz crossover point, with an 18dB/octave slope on the tweeter for a slightly sharper cutoff. There is a capacitor and inductor in series with the squawker, the values for which were calculated for the upper and lower crossover limits -- thus the band pass. It's actually correctly called, in this case, a narrow band pass.

People speak so negatively about resistors, however I now consider the autoformer sort of an antique. One advantage to a variable resistor, which I was using on tweeters and midranges nearly 20 years ago (Altec, JBL, many others also use them), is that you can dial in JUST the amount of attenuation you want/need. The downside is deciding what's best and then balancing that with the other channel. It can be done using interstation hiss/white noise, as well as some other ways. Once you find what you want, you can build a fixed constant impedance design with one series and one parallel resistor. All the soldering for you will be easy.

Doing is a wonderful path to knowing. The most experienced here, among which I am not, also made mistakes along the way; but avoiding something is a very sure way of never learning and never knowing.....for yourself, once and for all.

Erik

Dean: I have seen your work, and complimented you on it in the past. You solder well, and the networks you have made for people are always built well. But, what I want to leave you with is the fact that both of us KNOW something about it because we took the time to learn, study, practice, and obviously make a few mistakes along the way. My premise is that there are many, many on this forum that could enjoy the fun of building something like this for themselves, and take pride in their work. It's fun to make something that works, but sometimes to get to that point takes work. When I started doing this, there weren't any forums for help like we have here. I got books and read, and bought cheap drivers and parts to learn. I put a lid on a hamster cage, and cut some speaker holes in it for drivers. I also blew one tweeter out, in what was it, eighteen years I think, but that's when I first started, and bought really cheap parts intentionally in case I made a mistake -- which I did.

And: I'm not trying to play the authority on anything. I have other things to do than worry about playing an authority. What I do think is important, though, is to not discourage others from something they might not only find useful and money-saving, but also incredibly fun, educational, and personally maybe very fulfilling. Because of their relative simplicity and lack of high voltages, the type A and AA networks are good places to start for that. I could care less if someone wants to use an autoformer rather than resistors. We all have the right to choose what we prefer. What I do care about is encouraging someone to try, even if he or she might have some initial doubts. I did when I first started! Schematics looked very strange to me, until I sort of learned to follow them like one does a road map. The symbols for resitance, inductance, and capacitance for such a network and so few parts is perfect for a beginner.