Jump to content

The Amazing Autotransformer


Langston

Recommended Posts

Moving on from inductors in my insane pursuit of a passive crossover that is audibly indistinguishable from active, I turned my attention toward mid/high frequency gain reduction that horn loaded systems can't live without.

 

About 6 months before I was born, PWK published this paper supporting his use of auto-(Greek for self or one)-transformers instead of resistor pads. Automatic transmissions and washing machines! What will they think of next?! : )

 

The autotransformer (AX) route is interesting in part because it's WAY more expensive than a resistor - so why suffer the expense when they worked so hard to cut costs in every conceivable way to survive in the market? There must have been some choice words to PWK from whoever kept the books back then...

 

Be thankful that the engineer won this one. From my measurements, a well made AX behaves almost identically to active gain control, while the voltage division provided by resistors do not and can not. Is the difference audible? Probably not. IF you drive the loudspeaker at very low levels so the voice coil remains cool enough not to increase its impedance.

 

PWK's article claimed two AX advantages:
1. They allow user adjustable gain changes without screwing up crossover behavior.
2. They reduce series resistance from the amp to the driver, thus improving electrical damping with measurable improvement in the acoustic domain.

 

Claim #1:
Is true and false.

 

Is not true unless you parallel a 10Ω (or thereabouts) resistor across the input of the AX, as recommended by @Al Klappenberger and others. The problem with this is that the lower impedance seen by the crossover will increase costs (larger caps) to maintain the same high pass behavior. Maybe the guy that kept the books won this round.

 

The great news is that this isn't a problem for us and we can use a 10Ω "swamping" resistor to give us a nearly flat 7Ω compression driver - thus claim #1 is fulfilled! BTW, you can't just add the resistor - you'll need to redesign the crossover, after which you'll be able to adjust AX gain reduction at will without compromising crossover behavior.

 

Claim #2:
Is true and false.

 

It is true that the AX results in almost zero added series impedance to the driver, and not just because of low DCR, but the electrical impedance transformation (the greater the gain reduction selected, the lower the impedance the driver sees). Remember this is an AC circuit. The swamping resistor improves things a bit more. BUT does this increase in electrical damping really make a difference to a horn loaded driver that gets the majority of its damping through acoustic impedance? BTW, acoustic impedance is the reason to fall in love with horns if you haven't already.

 

PWK's paper shows a measurement from an experiment he did in a shootout between the AX and a resistor based pad. I repeated the experiment this morning. For the most part. Unfortunately, PWK did not provide full details of his measurement setup, but he did employ a two different passive crossovers, one designed for the AX (values shown) and the other for the resistive pad (values not shown1). I don't like the smell of that and prefer to hold all variables constant excepting, in this case, the gain reduction method.

 

Even if PWK had revealed all crossover components used, the stuff available today wouldn't be the same, thus it's impossible to recreate his experiment exactly, which is fine with me as I already implied. The other items that aren't available to me are the driver/horn combination and AX model he used, but what I used instead is more relevant to us - a Crites A-55-G on a K-400 horn with the Crites 3636 AX. The quality of this setup and the possibly additional acoustic impedance of my '74 vintage horn could be the reason I saw no improvement I could attribute to the higher electrical damping of the AX method.

 

Thus I'll assume PWK's setup used a lower quality (by today's standards) driver/horn/AX system than my stuff and as a result, the significantly higher electrical damping that played a role in reducing driver ringing for him, didn't help in my system.

 

All is not lost, the AX remains a superior method of passive gain reduction for additional reasons that'll be shown later.

 

My measurement setup used a 14.7Ω series resistance that resulted in a mid-passband 6dB SPL reduction from the driver/horn system. The 3636 AX -6dB gain setting nailed that SPL reduction exactly (0-5 input, 0-3 output).

 

1599017781_AutoXfrmrvsPad1.thumb.jpg.2d854873a136b64bd6c9b8cd2dc26e2b.jpg

 

Is it ringing yet?

 

832040492_AutoXfrmrvsPad2.thumb.jpg.5b656f20c822516ad0170a13992ce745.jpg

 

Where's the meat? In the 3636 of course.

 

AutoXfrmrs.thumb.jpg.63959e67c37256b9a5d2f39d389b6b95.jpg

 

From an email exchange with Bob Crites about the 3636:
On 30, Dec, 2016, at 3:14 PM, Bob Crites <...> wrote:

When I had the autotransformers made, I went to the same place that built them for Klipsch.  I just asked if there were any improvements they could suggest over how they made them for Klipsch,  I was told that Klipsch had them sized for a maximum continuous rating of 40 watts.  That saved them a buck or so each compared to the standard 100 watt rating.  So I just told them to make them as good as they know how to make them and rated at the 100 watts.

 

I am a fanboy of the Crites. Bob was a gentle member of the No BS club. Salt what I say about them as you wish, but (proper) measurements don't lie (includes the T2A) in another post. I'm planning on three: My version of PWK's experiment, an impedance analysis attached to a driver/horn system and a transfer function analysis that includes wideband low-level and max. wattage, both into 10Ω.

 

God bless you and your precious family - Langston

 

- - -

 

1 I think. If he used the same crossover for both the AX and resistor pad, then the rising output at the impedance resonance would have been a simple voltage division thing, not a damping thing - a kind of variable EQ due to the high source impedance of the resistor pad. I would love to get yelled at by PWK about this - maybe somebody at Klipsch will put me straight. : )

Edited by Langston
  • Like 2
  • Thanks 2
Link to comment
Share on other sites

First measurement group: PWK's autotransformer experiment, revised.

 

The GREEN measurement traces are the -6dB resistive pad setup. The BLUE traces are the raw driver/horn and RED traces are the 3636 AX setup for 6dB gain reduction. I've included THD at this low drive level, but there's nothing to see there, both gain reduction methods were the same.

 

You'll notice a ±0.5dB departure of the GREEN resistive pad trace relative to the RED AX trace, but the departure is lock step with the driver impedance curve (second plot), thus a result of voltage division - not electrical damping.

 

The A-55-G driver has a very smooth impedance curve, thus the voltage division induced SPL variations are slight. In PWK's case, his driver/horn probably had impedance peaks at 1.5kHz and 2kHz, thus the apparent "ringing" of the driver was actually caused by increased voltage being supplied to the driver, not a reduction in electrical damping.

 

Now you can shoot me. : )

 

God bless you and your precious family - Langston

 

5897641_A55GAutoXfrmrLevelandTHD.thumb.png.ec6acad71a583d9b03c01eaeb9adac71.png

 

1759485904_A55GAutoXfrmrRelativeLevel.thumb.png.b8999ea98730180395a65cca2ac78b3f.png

Edited by Langston
  • Like 1
Link to comment
Share on other sites

That was the funniest thread I've read in a very long time. I don't understand the zip tie thing, but I can guess it's related to duct tape, which is related to.. Nevermind, I'll look it up.. Well, maybe not. : )

 

My purpose in this thread is identical to the inductor thread, my walk from ignorance to somewhere else sans preconceptions as much as possible. I stole the PWK article link from one of Bob Crites' posts - I'd stolen it from you if I'd seen it earlier!

 

There's so much talk and so little measurement around here. : )

 

God bless you and your precious family - Langston

 

One accurate measurement is worth a thousand expert opinions. - Grace Hopper

Link to comment
Share on other sites

5 hours ago, Langston said:

The great news is that this isn't a problem for us and we can use a 10Ω "swamping" resistor to give us a nearly flat 7Ω compression driver

As an old friend used to say, “there is no free lunch.” You don’t gain something without losing something. I’m sure PK was aware that JBL was swamping their autotransformers for on the fly attenuation, but was apparently unimpressed. 
 

You say there isn’t much measuring here. Maybe you are just late. 

  • Like 1
Link to comment
Share on other sites

16 minutes ago, Deang said:

You say there isn’t much measuring here. Maybe you are just late.

 

You're right - this forum attracted me largely due to that fact. I should have said that in reference to the thread @RandyH001 pointed me to, that's what I was thinking about.

 

Anyway, I'd love to hear about the lunch expenses. So far it's looking like a cost issue alone - I'm really impressed with this gizmo. Other than the ≈35Hz impedance bump in the Klipschorn, the autotransformer will allow a virtually resistive load to 20kHz. That would be a huge plus for tube amps and other higher output impedance amps. Zobel type networks are a partial solution, but like any other part of a passive network they're dependent on load impedance and when that changes (driver voice coil heating), it changes. The autotransformer just doesn't care. Apparently. TBD. : )

 

God bless you and your precious family - Langston

Edited by Langston
Link to comment
Share on other sites

I for one have always wondered if the autoformer's inductance is as irrelevant as  some have claimed.  The spec sheet for the T2A states that the inductance between
taps 0 - 3 is 11.4mH +- 15% (at 1000Hz?).  My meter measures 8.5mH on one of my T2A's, taps 0 - 3.

 

I'm sure your equipment is better that mine.  What is the inductance across taps 0 - 3 on the 3636?  Do you have a T2A to measure?

 

I assume your tests were conducted without a series capacitor, therefore the autoformer's inductance would have no affect on the voltage across the driver.
It would be interesting to run similar tests with appropriate series capacitors to see how the shunt inductance affects the frequency curve.

 

Thanks, Mike

Link to comment
Share on other sites

11 hours ago, Langston said:

 

You're right - this forum attracted me largely due to that fact. I should have said that in reference to the thread @RandyH001 pointed me to, that's what I was thinking about.

 

Anyway, I'd love to hear about the lunch expenses. So far it's looking like a cost issue alone - I'm really impressed with this gizmo. Other than the ≈35Hz impedance bump in the Klipschorn, the autotransformer will allow a virtually resistive load to 20kHz. That would be a huge plus for tube amps and other higher output impedance amps. Zobel type networks are a partial solution, but like any other part of a passive network they're dependent on load impedance and when that changes (driver voice coil heating), it changes. The autotransformer just doesn't care. Apparently. TBD. : )

 

God bless you and your precious family - Langston

Great you took the time to do so much research and testing of autotransformers. PWK was very opinionated and a proficient writer. At the time of his choosing an autotransformer tube amplifiers ruled with low output so it is my belief, and many others, that the main reason for choosing an autotransformer is low to minimum power loss. As Dean stated there are negatives of using an autotransformer, the main one I do not like is the phase shift that naturally occurs using iron. Most all if not all speaker manufacturers have ditched the autotransformer in passive crossovers with discrete l-pad for attenuation. Does the job much cheaper and with amplifiers these day's watt savings is a mute point. Klipsch as well. Many many years since the went to discrete l-pad for attenuation over an autotransformer.  

 

There is one famous company that uses an autotransformer in their SS amplifiers, not speakers, is McIntosh. 

  • Like 1
Link to comment
Share on other sites

Second measurement group: The autotransformer's impedance effect.

 

With passive crossovers and horns, significant mid/high frequency attenuation relative to the low frequency passband is required. My taste at present for the Klipschorn is about 7dB less mid/high output after equalizing for room acoustics under 200Hz or so. Before playing bop-a-mole with the room modes, I needed a 10-11dB mid/high reduction for balance but the bass quality remained unacceptable, of course.

 

The two huge benefits of the AX over resistive pads are the ability to change mid/high attenuation without screwing up crossover response (when used with a swamping resistor), and immunity from voice coil temperature changes. Both benefits result from the AX's presentation of a constant impedance to the passive crossover network. This also allows simpler crossover design (imagine building a network for an 8Ω resistor) and the ability to use any genre of amplifier without tonal changes.

 

Downsides? Phase? Group delay? Inductance? Distortion? Frequency response? That's coming up in the third measurement group. Very helpful post by @henry4841 on the AX background, thank you. I am a newbie at home audio and it's turning out to be as hard to get right as professional audio - maybe harder. I thought it might be boring at first and so far it's kicking my butt. Thanks also to @mboxler for the series capacitor idea - will do, but given that I'd lose most of the impedance compensation benefits without a swamping resistor, it too will be part of the setup. Without the swamping resistor the AX doesn't interest me.

 

The first of several iterations to my '74 Klipschorn top hats used a Crites sourced DE120 high frequency driver with their horn and A-55-G mid frequency driver for the existing K-400 horn. The latter was a big improvement across the board, but the little horn provided with the DE120 resulted in a rough rolloff right around the crossover frequency, thus I got Dave Harris's (Fastlane Audio) drop-in wood Tractrix horn and the DE120 became nearly flat down to 1kHz with a rolloff that looked just like the B&C literature. Brilliant.

 

The following impedance measurements use these two driver/horn combinations without any passive components other than the 3636 AX and a 10Ω swamping resistor across its input.

 

Raw measurements of the DE120/Fastlane Tractrix and A-55-G/K-400:

 

1197529146_DE120A55GRawZ.thumb.png.ef73aba53543ab3499ae0abb6e636e7c.png

 

The A-55-G/K-400 adding the 3636 at -3dB and -5dB attenuations:

 

914678435_A55G3636Z.thumb.png.9cd8bbee5fa65e5070d0a6817ca6f9cf.png

 

The same, but this time with the magic swamping resistor:

 

207907689_A55G3636Z10R.thumb.png.2c4dc73f23c39aa461f3b8a0644ac42e.png

 

Finally, the DE120/Fastlane Tractrix and 3636 at -3dB with and without swamping resistor:

 

1922324348_DE1203636Z10R.thumb.png.140ab198dc55923de1b2bf019813b4d7.png

 

God bless you and your precious family - Langston

Edited by Langston
  • Like 1
Link to comment
Share on other sites

2 hours ago, henry4841 said:

As Dean stated there are negatives of using an autotransformer, the main one I do not like is the phase shift that naturally occurs using iron.

 

No, I meant using a swamping resistor as opposed to not using one, but I suppose the same applies to the use of resistors in general to an autotransformer. 
 

The “cost” of using the swamping resistor is higher insertion loss, since quite a bit of power is being dumped into the resistor to maintain that load. Bob and me finally convinced Al to go to a 20 watt resistor because people kept burning them up. I had one customer where it got so hot it melted into the capacitor below it. 

 

Kerry Geist of Klipsch had the following to say:


 

 

600ED990-FE02-46D3-8E72-7E8A463D3EB2.jpeg

Link to comment
Share on other sites

I’m not an engineer, I just build these things. I was a licensed builder for Al for many years and I can argue this one both ways. 
 

You should do an acoustic measurement. One with the resistor, and one without - both tapped out on 0 and 4. Now that one would settle an old argument. 
 

I have most people drop down a tap and halve the primary capacitance value. This allows for the use of high quality film and foils - which you aren’t going to find in the higher values.

 

@mboxler Yeah, the inductance. I just always figured that at the frequencies it’s operating at, what could it possibly be doing? DCR is about .3

 

Link to comment
Share on other sites

So turn off your air conditioning if you don't want to "waste" power. You'll live. Alone maybe, but you'll live.

 

You buy comfort with that power usage, I'm gonna buy a resistive loudspeaker over the lion's share of its output for the sake of rock solid passive network behavior regardless of voice coil temps and the ability to use tube or constant current amps without having to measure the system and apply inverse EQ to it to flatten it back out AND the ability to change mid/high attenuation levels at will, and driving the system with a single amplifier.

 

Cranky soldermeister indeed! : )

 

BTW, thanks for the suggestions for other measurements. I also genuinely enjoy and learn from your posts - it's hard to interpret weird people's writing like my own at times, thus I'm stating that for the record. : )

Edited by Langston
Link to comment
Share on other sites

Ha ha. All true. I get it.

 

A resistor is just another tool. I used to get pretty worked up over this stuff, but no more. 

 

I spend most of my time in front of a pair of ATC scm20psl with two SVS subs. Impedance is rather benign, and I push them with 60wpc. Sensitivity is terrible, PK would have hated them. 

 

Some people want Klipsch sensitivity and output levels but aren’t exactly enamored with the house sound or voicing. Others love the forward and aggressive presentation. You can’t please everyone. For the former, you can perform surgery. 🙂

  • Like 1
Link to comment
Share on other sites

Third measurement group: Transfer function analysis.

 

Since I've already prepared the electrical domain measurements, I'm going to post them here and add a 4th post to cover questions. So far I have (2) from @mboxler and (1) from @Deang. Everyone is invited to fire away with more questions and I'll answer them (with measurements) if I'm able.

 

One of the questions has to do with inductance values of the AX between the 0-3 output. A note on this: I've decided not to try to figure out how to accurately add the AX to my crossover optimization software, instead I include the swamping resistor plus AX in my driver impedance measurements as if the driver came with that stuff built inside of it. The 3636 AX does not change acoustic magnitude or phase, it only applies attenuation, thus no need to include the AX in acoustic measurements (unless you want to do maximum output testing).

 

Thus I avoid screw ups characterizing the AX as a simple pair of inductors in L-Pad configuration, which I assume it is (nope, it's a continuous coil sharing the same core with taps). If someone has successfully modeled an AX in Spice type software such that the predictions proved correct, please let me know. : )

 

- - -

 

I show two measurements of each AX; the 3636 and a '74 vintage T2A/3110A. The first is low-level, wide bandwidth and the second is the most the AX can handle at a given attenuation level between 100Hz and 10kHz. Even though the 3636 can do 1dB attenuation steps to -12dB, the T2A only has four output taps at -3dB, -6dB, -9dB and -12dB. Thus I decided to measure each AX at those attenuation levels. Every measurement included a 10Ω load resistor across the AX output to simulate a driver.

 

First up is the reference measurement of the AHB2 amp in bridged mode into 10Ω alone. I only took it up to 200W because that was the max that either AX could handle.

 

1159001076_AHB2Reference.thumb.png.905b5781ac32424745b9449e6e95938d.png

 

Next is phase and group delay (which is just another way to look at phase). I include every measurement made just to get it out of the way because both AX units were effectively perfect through the audio range. You'll see a small bump at the beginning of the high level 100Hz-10kHz sweeps. That's due to suddenly hitting the AX inputs with high power. If you happen to do that to your loudspeakers at some point, that little bump will be the least of your worries.

 

1717005107_AXPhaseGD.thumb.png.48e457c3dfe9f23666485596ea84fe72.png

 

Next is low level wide bandwidth output and THD. The dashed traces are the T2A, the solid traces are the 3636 (as on every plot).

 

711710790_AXLowLevelOutput.thumb.png.20ffe02661f3d3e5e468a4354bcd9c28.png

 

Next are the maximum outputs at each nominal attenuation setting that remain under 1% THD. I adjusted the generator in 1dB steps. Once the AX exceeded 1%, I dropped the generator level 1dB and recorded that as the max. This wattage level is across the output loaded by 10Ω. To figure out the much higher wattage absorbed by the 10Ω swamping resistor on the input, multiply the -3dB output by 2, -6dB by 4, -9dB by 8 and -12dB by 16!

 

There is a surprise in this one. The T2A for the first time outperforms the 3636 at something. At the -9dB attenuation setting the T2A's max output is as expected, mid level between the -6dB and -12dB settings, but the 3636 is low. It must be a design issue (due to the 1dB steps?) because I got the same results with a 2nd unit that is 3 years newer in mfg. date. It's fine for our use, but it shouldn't do this and I'm going to look into it.

 

1827888245_AXMaxOutput.thumb.png.b1b932b9450bfc21cc8ca9adbe1db5e9.png

 

Next are the actual attenuations provided at the four nominal level settings. This includes both low and max levels. The 3636 nails it, but the T2A is only off by 0.6dB worst case.

 

710555330_AXGain.thumb.png.242cbf5fbc204185a8a9b636a152f46a.png

 

God bless you and your precious family - Langston

 

Link to comment
Share on other sites

42 minutes ago, Langston said:

If someone has successfully modeled an AX in Spice type software such that the predictions proved correct, please let me know. : )

 

 

I don't fully understand your predictions, but here are the 3636 and T2A.  The 3636 is based on 76mH taps 0-5.  The T2A on 44.8mH.

 

Mike

 

3636.asc T2A.asc

  • Like 1
Link to comment
Share on other sites

Third measurement group: Transfer function analysis. (Addendum)

 

We found out from my previous post's Max Output plot that both the 3636 and T2A have attenuation settings (taps) that shouldn't be there. The 3636's -9dB (5.4W) and -12dB (2.7W) taps are unusable due to the very low output available before clipping (saturation). The T2A's -12dB tap shares the same 2.7W clipping threshold.

 

If I didn't err in my measurements, the 3636 is unusable from -9dB to -12dB and the T2A is unusable at -12dB.

 

Because I'm so interested in using the 3636 in my passive crossover designs I went back and performed all the measurements again, and this time I measured each 1dB step of the 3636. Turns out my prior measurements were correct, but I want to spill the beans on what I learned. Somebody (Universal Transformer or the Crites or both) should have warned us about the 3636 limitation. I suppose it's very rare for anyone to use those things at greater than 8dB attenuation, but the additional taps really shouldn't be there.

 

First, the good news. Each gain step is dead-on. The thick traces from 100Hz to 10kHz are the Max Output measurements. They overlap the low level, wideband measurements because the 3636 (and T2A) attenuations are constant regardless of drive level (within their linear regions).

 

631185683_3636Gain.thumb.png.4a4defd75cd11d1e78258aca18e0b505.png

 

Next is the Max Output (watts into 10Ω) plot at each of the 12 attenuation steps.

 

1706417768_3636MaxOutput.thumb.png.8731e6e03c747b7f3a5023b7ec55ba9f.png

 

Finally, I made a cheat sheet for the 3636 that should be helpful. Red is bad. There's a download link for a PDF version below the picture.

816705114_3636Autotransformer.thumb.png.0bb1933374a5e9be05088f11f8b8bebe.png

 

3636 Autotransformer.pdf

 

God bless you and your precious family - Langston

Edited by Langston
  • Thanks 1
Link to comment
Share on other sites

First everyone hears differently so there is no one size fits all. This is from my own personal experience. Years ago I built all the published schematics I could find for the LaScala speakers along with some designs of my own. We are talking 30 or more with many hours listening to each one. No extreme measurements as Langston has done which is an enjoyable endeavor I have done with other electronic equipment charting measurements. To me ditching the autotransformer made for better sound with my equipment and room. I contributed to better sound of not using iron for attenuation as being the phase shift. There was one LaScala crossover, cannot remember the exact one, that the engineers at Klipsch could not decide if reversing the leads at the crossover was the correct thing to do. The were sent out one way and a few years later it was determined the leads should be reversed to compensate for the phase shift. After doing all that work I decided to just build an AA network with the autotransformer using decent parts to leave in my LaScala's. Good enough is the way I figured it and when they are passed down my sons will not have trouble selling them being stock. L-pad discrete resistor attenuation or autotransformer? The fact is a resistor is the less offensive electronic component.  

 

PS: I do not listen myself with the stock LaScala network, mid horn or tweeter. What I have made sits on top of my speakers only using the bass bin of the LaScala. Easily switchable to stock for selling at a later date. That is if one of my sons do not keep them. The crossover network does not include an autotransformer. 

  • Like 1
Link to comment
Share on other sites

This thread is extremely interesting.  Thank you @Langston for sharing your thoughts and experiments.  

 

With the assistance of Bob Crites and Dennis, aka djk, I incorporated 3636 autotransformers into La Scalas (with the djk bass mod) used in a high school band practice room, KP 201s used in a middle school band practice room, and several iterations of “Super” Heresys for home use.  The ability to adjust the squawker outputs in 1db increments made fine tuning the speakers for the vastly different spaces relatively easy.

 

As an aside, this post makes me feel old when “the late” could refer to Bob, Dennis, and PWK.

 

 

  • Like 1
  • Thanks 1
Link to comment
Share on other sites

11 hours ago, Langston said:

We found out from my previous post's Max Output plot that both the 3636 and T2A have attenuation settings (taps) that shouldn't be there. The 3636's -9dB (5.4W) and -12dB (2.7W) taps are unusable due to the very low output available before clipping (saturation). The T2A's -12dB tap shares the same 2.7W clipping threshold.

 

If I didn't err in my measurements, the 3636 is unusable from -9dB to -12dB and the T2A is unusable at -12dB.


If you take into account the efficiency of the compression driver/horn (example a TAD TD4002 is good for 110db with 1w input) and the maximum listening level expected in the listening room then 2.7w might be plenty for many home listening rooms. 
 

miketn

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...