[John Chapman]

Hello!

There was a thread posted over at HD that was interesting and since there is more activity over here these days I thought I'd move it over here as well.

It was about using a transformer as an attenuator for headphones. The idea was to connect directly to your source and then to the transformer attenuator and after that to the headphones. This would eliminate the headphone amp but as the thread shows the TX102 is not a candidate for this - but it may be worth exploring a different unit that would work in this place.

Thread below....

Thanks!

John

[John Chapman]

Czilla9000 wrote:


Hello, I am looking to build a device which will allow me to passively attunuate a signal for headphones. Since you guys tend to be the leaders in this stuff, I was wondering how I would go about wiring a transformer (like the B&S one) for this use.



Thank you.

[John Chapman]

I wrote:

Hello!

Thanks for the post! I hate to answer with a question but need a bit more info first. What is the source driving the headphones?

The TX102 transformer is used for level (volume) control and I have a few in headphone amps but they attenuate the signal at line level before the amp's output circuits. They can handle a pretty high signal (like >7 Vrms) but I don't think they'd be the trick to kick down the headphone signal in-line with the phones - which have a much lower impedance than the TX102 would normally see.

Send a bit more info about the setup you have and we'll see what might work.

Many Thanks!

John Chapman
www.bentaudio.com

[John Chapman]

Czilla9000 wrote:

I am trying to build, eccentually, a passive headphone amplifier. The idea of building a passive headamp is very contraversial with headphone audiophiles, despite the fact most sources give out more than enough power to power any phone (leaving only attenuation neccesary). I don't know much about my source, but I have heard that using a transformer would be better than a stepped atennuator.


The reason I want a passive headamp is because I don't want coloration. To me simplicity is better, and an active amp is certainley not simpler than a simple resistor or transformer.

[John Chapman]

I wrote:

Hello!

It's true that a transformer would do a better job of using the energy from a source to drive headphones directly but I have to question if most sources give out enough total power. They may give out enough voltage (2V typical and many higher than that) but most would not handle the low impedance of headphones well at all. Most sources might have an output impedance of say 100 or 200 ohms for example. This would be a typical solid state CD players output impedance. Some would be lower but much below 100 ohms would be rare. If you hook up a 10K to 50K amp load most of the signal will get to the amp with a very small amount lost via the source/transformer output impedance (which can be thought of as a resistor in series). With a headphone load fo maybe 40 ohms or so (I don't follow this stuff that close so correct me if I am off on this) most of the signal will be lost across the sources output impedance.

As you can tell from the products I carry I love passive stuff and I appreciate what you are trying to do. I think you'd get some sound out but I have to think there would be improved quality with a buffer stage to drive the headphones easier - even though you'd be adding additional stuff to the signal path. I agree that simpler is better but this setup may be pushing it a bit far.

If you want to give it a go look for a source with a very high voltage out and - most important - an extremely low output impedance. Some early Theta DAC's and as I recall some Wadia's had killer output stages and level controls. Basically in these cases the buffer stage is built into the source. A transformer after an output stage like this would lower the sources output impedance even more - which is a good thing! and we could play with the TX102.

If you can let me know what the impedance of the headphones you have in mind are and what the sources you'd think of using are I'll try and re-create the setup here with my AP system 2 analyser and see how the TX102 would behave with those conditions.

Many Thanks!

John Chapman
www.bentaudio.com

[John Chapman]

Czilla9000 Wrote:

To be honest I am not an expert yet on this impedence mismatching stuff. Could you be so kind to explain what all of this impedence stuff is?



So I take it you believe that a transformer would be better than a ladder type stepped attenuator?

Headphones have a wide variety of output impedences. Grados, for instance, have an impedence of 32 ohms, while the most popular audiophile phone, the famed Sennheiser HD600 has an impedence of 300 ohms.


PS - Thank you for all of your help.

[John Chapman]

I wrote:

Hello!

While I would not call myself an expert either the work with the TX102 (and the way it 'translates' impedance) has got me thinking about it quite a lot. I'll run through it as best I can but please post more questions as they come up.

Everytime you connect one component to another you have 2 impedances to think about - the output imp of the 'source' and the input impedance of the downstream component. An exapmple would be the output impedance of a pre-amp deefing into the input impedance of an amp. Another would be the output imp of that amp feeding into the input impedance of a speaker. In every case it's desireable to have a low output impedance feeding a higher input impedance. Here are some typical values:

CD Player with a 100 ohm output Imp feeding a pre-amp with a 40K input Imp.

Pre-amp with 50 ohm output impedance feeding an amp with a 20K input Imp.

A solid state amp with a .05 ohm output Imp feeding an 8 ohm speaker.

A single ended tube amp with a 2 ohm output imp feeding a 16 ohm speaker.

These are kinda typical values - if there is such a thing!

In each case you can see that the input impedance is many times more than the impedance feeding it. I spotted a good article on impedance here:

http://www.transcendentsound.com/amplifier_output_impedance.htm

The ratios range quite a bit but 100 to 1 is kinda the middle ground - with many cases typically higher. The triode amp is included in the list on purpose. It's ratio is much lower (although still 8 to 1) and many folks would think this would not work. In practice if the following impedance is somewhat constant that can work well. The article referenced above is a good description of this from the amps point of view. Basically a high ratio makes the system more immune to variations in impedance of the downstream component and also has other benifit's to the signal transfer as well.

Finally - lets look at the headphone case. If you have a source with a 200 ohm output impedance (typical CD player but impedance figures vary wildly) and headphones like the HD600's with a 300 ohm impedance we have a 1.5 to 1 ratio - many many times lower than the triode amp to speaker ratio that would freak out a lot of 'measurement types'. These amps can work very well and I am a bit of a fan. I included them in the list to show that exteme ratio's like 400 to 1 are not needed to get good results.

Our 1 to 1.5 ratio would be pushing it too far I think and so if you were to want to do a good job of driving the phones I'd guess a ratio of more like 20 or 30 to 1 would be a good target. This would lead us to be looking for a source with a 10 or 15 ohm output impedance. THis is low and it's rare to find a source with this low an output impedance.

Now on to transformers as a level control.... We'd need to test if the TX102 would behave at all into such a low load - I have no idea what will happen! You have me curious now so the next time I crank up the test gear I am going to play with it and see.

The TX102 steps the voltage down to lower level but the bonus is how it lowers the output impedance as you step down further and further. The cool thing is that the impdance goes down really fast because it is related to the square of the turns ratio. Once you get down below about half way it is way way lower than the source feeding it. This - if it would work - might be the savior of this passive headphone plan. A source with an output Imp of 10 ohms is rare - very rare. If we can use a transfomer to lower voltage we could get away with a higher output impedance at the source - as long as the normal listenning was done at a level well down from the full level the source puts out. We'd still be looking for a low output impedance and a higher than typical output voltage from our source - so that we can step it down and still have sufficient level. You can see from this that it looks like it might just work! - but only if we pay careful attention to the source and the phones impedances and voltage requirements.


Long post - sorry! Next steps:

1- Could you dig around and see what a voltage would be required accross a pair of HD600's for a typical listening level.

2- I'll test the TX102 here and when I next talk to S&B bounce it off him and see what his thought are....

Many Thanks!

John Chapman
www.bentaudio.com

[John Chapman]

I wrote again....


Hello!

Moving too fast again. Lots of typo's in the post above and I missed the 'edit window'. Sorry guys....


Thanks!

John

[John Chapman]

Czilla9000 wrote:

Thank you.....do you mind if I relay your posts on this issue to another disscussion forum. I know a headphone forum which would be interested in this.

[John Chapman]

I wrote:

Hello!

No problem relaying the posts - but please let then know I am not a headphone guy so I have been making a lot of guesses when it comes to headphones!

I spoke to S&B today and I had asked Jonathan about this. His first reaction was like mine - not likely to work well. With very high impedance phones like the HD600's 300 ohms he started to think about it a bit more. I have leasured the TX102 into a 600 ohm load and it stays well behaved.

I had missed one thing that he brought up - resistive losses. When a TX102 is connected to a high impedance load (like it usually is) the resistive losses are not a factor as the TX102's internal resistance is only a small fraction of the resistance of the load. When we move the load inpedance down then the losses at the transformer become a factor.

What this means is that we'd loose a big chunk of the signal accross the transformer - this makes the source output voltage requirement even higher than we'd thought. I think we may be reaching the point where you have added so much restriction on the source (likely needing a source with extra gain stages and buffers inside) that it may defeat the purpose of the passive level control - which is to simplify the circuit! I am thinking you'll find that a simple output stage from the source with a level control followed by a gain stage / buffer circuit (ie. a nice headphone amp!) would perform better than the gain stage / buffer in the source followed by the level control.

Since we've come this far I'd still like to find out what the voltage level at the phones would be for a typical 'high' listenning level and maybe the folks at the other forum could help with that.


Many Thanks!

John Chapman
www.bentaudio.com

[John Chapman]

Czilla9000 wrote:

Thank you for all of your help. So basically it is impractical and worse than using an active headamp? Correct?

[John Chapman]

I wrote:

Hello!

I think it comes down the the fact the source you use (cd player / dac, etc) would have to have an output stage that was the equivilent of a headphone amp in it! My thinking is at that point why not just use the source you want (instead of having to choose it primarily on it's output stage) and then use a heaphone amp after that. There is such a variety of headphone amp circuits that you could pick anything from a solid state unit to a tube unit depending on what would best match your phones and your listenning preference.

I was starting to think it may work but the TX102 is just not made for this low an impedance. I think there may be ways to get the passive control just ahead of the phones to work. An autoformer along the lines of the 'zero' autoformer for speakers (but scaled down) with many taps like a TX102 so you could select levels might work - again given the right source. If a unit were made especially for this application then the transformer losses could be minimized by keeping the winding resistance's low. The transformer (or autoformer) could solve the impedance mis-match issue - much like the 'zero autoformers' that are popular with OTL amp guys solve the amp to speaker impedance mis-match.

Did you ever spot what a typical voltage would be to get the HD600's to a somewhat loud listenning level?

Thanks!

John

[John Chapman]

Czilla900 wrote:

Yes, about 1 volt (I have had people at another forum try this and say 1.2 volts is very loud, so I assume 1 volt would be nessecary.

[John Chapman]

Jazz (good handle!) wrote:

...

john chapman wrote:
...Finally - let's look at the headphone case. If you have a source with a 200 ohm output impedance (typical CD player but impedance figures vary wildly) and headphones like the HD600's with a 300 ohm impedance we have a 1.5 to 1 ratio... Our 1 to 1.5 ratio would be pushing it too far I think and so if you were to want to do a good job of driving the phones I'd guess a ratio of more like 20 or 30 to 1 would be a good target. This would lead us to be looking for a source with a 10 or 15 ohm output impedance. This is low and it's rare to find a source with this low an output impedance.

...I'm coming from the same forum (Head-Fi) as Czilla9000. We have discussed a lot - basically with other people who rate the direct-path idea as obsolete and impracticable, roughly spoken with merely the thought in mind that «line-out amps aren't designed to drive complex and low-impedance loads». Of course this argument has to be taken into consideration. Otherwise it's easy to try it, in the first stage by just plugging a headphone into your soundcard's line-out and thus avoiding any additional resistors in the signal path. This shows that it principally works, at least in most cases, I guess. The other question is about the impedance ratios. I have to agree that the majority of the common CDPs and DACs have output impedances around 100 or 200 ohm. Which is a bad thing in view of adequately driving headphones without serious colorations due to the interaction with their impedance curves and also due to the low amount of remaining current.

But in my case it's a Theta Pro basic II and a Bel Canto DAC2 with 6 and 20 ohm output impedance. Whereas it's not a really great value looking at the 250-300 ohm of the concerning headphones (Beyerdynamic DT 880 and Sennheiser HD 600) with their bass-resonance induced impedance amplitudes towards 500 ohm or so, it seems to work flawlessly! I come to this conclusion after some tests: how does the complex, low-impedance load affect the line-out amp's signal? For normal listening I use dedicated headphone amps. I had one switched to the DAC's line out, serving as a monitor. Additionally I switched a 500-ohm potentiometer with a plugged-in headphone onto the line out, parallel to the amp. Then I listened to some music samples through the monitor amp with my reference headphone, alternatingly plugging the potentiometer/headphone system out and in... to verify to what extent the sound would change. Result: there was barely any sound change noticeable. Even a Philips DVD 963 SA with its 200 ohm caused only negligible sound changes (hard to rate because of the considerable volume change in its case).

So my conclusion is: At least some low-impedance line-out stages can drive headphones directly (or via potentiometer, resp.) in an adequate manner, without causing any serious sound degradation. This in contrast to all headphone amplifiers I have tried - and that's the actual motivation behind my direct-path idea. The main reason for being bashed therefor by the majority of the forum members is my experience - and theirs, particularly - that the direct path sounds somewhat less appealing (mostly called «worse») than ordinary headphone amps, although it excels in the main criteria: dynamic, transient speed, focus, resolution, transparency, clarity, cleanness... The main concern from my side is a certain lack of coherency and color(ations).

Based on my experiments my standpoint now - in contrast to others - is that amps (be it tubes or solid-state; I own both) generally color the sound in a euphonic manner. This idea is broadly and passionately rejected. Preferably by just ignoring my tests...

After my experiences I'm rather optimistic that the theoretically superior transformer operation will offer similar results, besides avoiding the damage potential provided by the high current flow with the potentiometer operation (could you tell something to this?). Since your theoretical considerations to the subject, cited by Czilla in the Head-Fi forum, already have lead to the proclamation of the death of the direct-path idea, I hope to have given you some useful information in view of your own standpoint. I'm not an electronics expert, BTW, and my findings are solely based on my listening experiences.

JaZZ

[John Chapman]

I wrote:

Hello!

Thanks for the post! I have far from given up on the idea and sure don't want to be responsible for it's death! I had thought in my last post about it that I was stating that given the right transformer (or autoformer) and given a low output impedance source it could work well. Most of the comments about it not working were with reference to the TX102 - which was designed to feed loads more like 10K and above (typical amp inputs). This would however never be the kind of thing you could just plug-n-play without considdering the source and the impedance of the phones.

To answer your question about the potentiometer:

If you introduce a potentiometer into the signal path between the source and the phones you wildly change the impedance situation - for the most part for the worse! Even a low value pot (what value did you use?) will introduce a series resistance in all but the ful volume setting. As you adjust the level of the pot you are changing this series resistance - from 0 ohms when wide open to a quite high value as the pot lowers the volume. This is the area where the transformer/autoformer may show promise. With a transformer as you lower the level the output impedance drops - and it drops with a realtionship following the square of the turns ratio. Every time the turns ratio is changed down 2:1 (6db drop in volume level) then the output impedance is 4x lower than before. The winding restance and some other things come into play but that's the general idea. There is a table at S&B's TX102 page showing output impedance for a 1K source here:

http://www.stevens-billington.co.uk/page102.htm

This table (and page) are for the TX102 transformer but they do show how as the level is lowered the output impedance goes down and after a few steps down from the top it goes down drastically fast. The TX102 is not made for driving the low impedance headphone load but that does not mean the same effect can not be used to advantage is a new transformer or autoformer were made that was designed for this. The 'zero autoformer' that the OTL amp guys use for lower impedance speakers uses the same principle to translate impedance. Both these impedance translations are after the same thing you are with the direct headphone drive - to take advantage of the sources output impedance and lower it even more so it can better cope with lower impedance loads. This is pretty much the oposite effect that a pot has - where you lower the pot's level the output impdance is going up!

Good thread and thanks for your comments!

John Chapman
www.bentaudio.com

[JaZZ]

[Copied from the Harmonic Discord Forum]

John...

I'm glad the direct-path idea isn't dead in your eyes...  

john chapman wrote:
If you introduce a potentiometer into the signal path between the source and the phones you wildly change the impedance situation - for the most part for the worse!

I'm aware of this. BTW the one I use is a 500-ohm type. Fortunately the sound is virtually unaltered by the direct-path system it is part of. This is a surprise to me too. (I hope you have noticed my test procedure and configuration.) So the conclusion that amps cause clearly more important sound alteration is justified from my point of view.

JaZZ

...I should have mentioned that I've also tried the "direct" direct connection without the potentiometer (with low-level music passages) to verify the change of sonic character by the serial resistance.
 
 --------------------------------------------------------------------------------
 
Hello!

500 ohms is a lowish value - and that's what I hoped you'd played with. Do you find you have the level turned down quite a bit for 'normal' listening levels? What kind of phones are you using and what is their impedance.

John Chapman

 --------------------------------------------------------------------------------

John...

...first of all: thanks for your replies!

My previous potentiometer was a 1 kOhm type. It worked equally well, but sounded somewhat darker than the 500 ohm type I'm using now - a logical consequence of the then mainly used HD 600's impedance curve. The 500-ohm type provides a sonic balance comparable to the headphone amps I use (Earmax Pro and Corda HA-2). Anyway, I wouldn't call the sonic signature between 500 and 1000 ohm principally different. The commonness they share is the relative dry, unvarnished sound compared to the more colorful presentation from the amps.

The average effective serial impedance resulting from the 500-ohm potentiometer's voltage-divider function in cooperation with the 300-ohm Sennheiser or the 250-ohm Beyer DT 880 should be in the range of OTL tube amps.

For normal listening levels the settings are between 1/3 and 2/3 (BTW it's logarithmic) with the DAC2 (max. 2 volt); with the Theta (max 5 or 6 volt) it was a bit less.

BTW, what do you think: do such low load impedances (= such high currents) represent a serious damage risk for normal (CDP/DAC) line-out stages?

 JaZZ

[John Chapman]

Hello!

Thanks again to you Jazz and to Czilla9000 for all the info - this has been a really fun thread!

With settings typically from 1/3 to 2/3 on a log taper pot the way the transformer or autoformer would be in the range where it would significantly lower output impedance - which would be a big help in finding sources that'd drive the phones directly. I would suggest that the 250 or 300 ohm phones you have been using would also be a requirements. 32 ohm Grado's would be out!

As far as damage to output stages I think we are pretty safe. As long as you were picking a source with a fairly low output impedance to start with - say less than 100 ohms you should be fine. At 2V out feeding into 300 ohms (or accross 400 ohms if we include the sources output impedance) we'd be asking the source to feed about 6mA or so - certainly no chance of damaging a typical op-amp based output stage at that level. Most could easily feed 5 to 10 times that level and an output stage like the Theta (which i think used the BUF03 chip as an output buffer) would be happy feeding even more current. These calc's assume the 300 ohm impedance is flat and does not dip. If phones behave as speakers (like I'd guess they do) this is not likely the case! I have to think though that there would not be any significant dips down to a really low impedance - has anyone measured the impedance curve of headphones?

You guys have me curious now so I am going to get a pair of HD600's to mess with and test. Maybe we'll wind up something in a while to try this out!

Thanks!

John Chapman
www.bentaudio.com

[John Chapman]

Hello!

I spotted the impedance curve for the HD600's here:

http://www.headphone.com/layout.php?topicID=3&subTopicID=26&productID=0020080600

The curve does not get down to 300 ohms and so it never goes lower.

Thanks!

John

[JaZZ]

John...

...yeah, you got the impedance curve I otherwise would have picked up. Of course a transformer would do a better job than my 500-ohm potentiometer in view of the 540-Ohm resonance peak. But believe me, even though  there may be a certain predictable hump in the electrical frequency response, about 1/2 or 1 dB I guess (whereas a dedicated amp would produce a perfect straight line), this is negligible (to my ears!) compared to the sound degradation/coloration caused by amps. You just can't measure the latter. BTW, the 32-ohm Grados do work quite well, surprisingly, or less surprisingly when you look at their impedance curves:



The loss of efficiency caused by the line-out's output impedance and the potentiometer is undisputed, though.

Unfortunately Czilla9000 is out of play. He has completely misinterpreted your comment concerning the TX102's possible inability to work with headphones and now thinks that the direct-connection idea must be buried, not least due to the massive pressure from the majority of the Head-Fi people who feel attacked by it – because for them amps are like a holy cow and nobody should dare to call them effect devices... BTW: I also think that headphones sound better through amps than through direct connection (haven't I already mentioned this?) – but not because they're more accurate, but because they're more euphonic.

Do you know this line-stage transformer[/url] specially designed for headphones?



I'm looking forward to your headphone direct-drive experiments and am very curious. Are you also considering to try a (500-ohm) potentiometer? And finally: Be warned: you may first be shocked by the unvarnished direct-path sound. I hope you don't confuse it with inaccuracy. IMO it's as accurate as can be.

JaZZ

[John Chapman]

Hello!

I am quite familiar with the ASL stuff - I own 2 of their amps and kind of act as a dealer locally here now. I had not seen that device - not a lot of info but I will be talking to Divergent and will try and get the scoop on it. Does it do just what we are talking about here?

I spoke with S&B today and he was going to mess around with some things like maybe an autoformer with low winding resistance (to minimize losses) and maybe come up with something we might offer.

Your experience is great and a transformer  or autoformer should do a good job of doing the impedance translation. If we are down at say  -12 db for a typical listenning level then that would turn a 200 ohm CD players output impedance into a about a 20 ohm output impedance driving the phones.

Thanks again for all the info.


John Chapman
www.bentaudio.com