Stereomour II 45 Conversion - Anyone Do It Yet?

[Paul Birkeland]

You could perform the 4K vs. 8K experiment with our iron and draw your conclusions.  This doesn't require buying new transformers. 

I do not personally consider Lundahl iron to be an upgrade over what we provide.

[Deke609]

You could perform the 4K vs. 8K experiment with our iron and draw your conclusions.

The OT-2 can be wired for 8K?

I do not personally consider Lundahl iron to be an upgrade over what we provide.

Huh. I would have thought that at the very least a larger OPT, other things being equal, and up to a point, would have better bass response.  Is the Kaiju's better bass just a function of the 60H plate choke - and the larger OPT has no role?

many thanks, Derek

[Paul Birkeland]

The OT-2 can be wired for 8K?

Yes, in the SEX amp it is setup as an 8K transformer.  You just wire the transformer for 4 ohms when using an 8 ohm load.  The design splits the difference a bit in terms of 4K and 8K performance, though in practice this isn't particularly obvious.  I would mention, however, that if you tried to use it as a 12K or 2K transformer, I would expect some frequency response issues that were audible and easily measured.

Huh. I would have thought that at the very least a larger OPT, other things being equal, and up to a point, would have better bass response. 

I had a pair of 45 amps in my workshop earlier this year with Hammond 1628SE output transformers.  They are huge 30W series feed output transformers.  The bass was decent, but the high frequency response was a mess among other issues.

Is the Kaiju's better bass just a function of the 60H plate choke - and the larger OPT has no role?

They both work together.  The Kaiju uses 3K output transformers, so the demands placed upon the plate choke aren't as crazy as they would be with a 4K or 5K load.  Even with that relief, the Kaiju plate choke has more inductance anyway.  The Kaiju OT is larger, which also helps with low frequency response at higher power.  For a #45 that makes 2W at most, going larger than the  iron we provide is a difficult decision to rationalize. 

[Deke609]

Many thanks PB.

For a #45 that makes 2W at most, going larger than the  iron we provide is a difficult decision to rationalize.

Ah. I should have stated that I'll be going as high as 4.5W, maybe even 5W, since the SII-45 rebuild will be purpose built for the EML 45B (which is the only tube I use with the amp in any event).  The 45B has a max output of 5.2W, although Jac has hinted that this may be a conservative estimate.  I think the higher power output configuration should make a better comparator for the Kaiju's 8 wpc.  Some of the better bass I hear in the Kaiju may be power related. Dunno. But I'd like to find out.

Edit: And I should also clarify that the Lundahl OPTs I am considering are parafeed, not series or push-pull. They are zero gapped.

cheers, Derek

[Paul Joppa]

There are a lot of things going on with this, too many to do more than mention in a single post. Here are (I think) the main ones:

* OPT inductance, which is much smaller at very low signal levels and usually sets a LF frequency response limit. Nickel cores have higher small-signal inductance.

* The frequency response at mid levels extends much lower than the power bandwidth.

* OPT saturation at high levels sets the power band lower limit. Nickel cores saturate at much smaller signal levels.

* Plate choke inductance

* Parafeed capacitor value, which if done right maintains a resistive load to a lower frequency than the choke alone, reducing tube distortion (but not transformer effects) at the lowest frequencies

The trick is to balance all these effects, which interact - they are not independent of each other. I've been working on this (in my own desultory fashion) for 20 years and I'm still learning.

The high frequency performance is a whole 'nother issue  :^)

[Deke609]

Many thanks PJ. This is a very helpful summary.  In playing around on paper with different possible configurations, I have been guided by three things, mostly taken from guidance you have posted previously:

(1) Other things being equal (which they likely never are), more inductance in the plate choke and OPT = more better. 

 (2) The tradeoff (inverse relationship) between distortion and power from either (a) increasing OPT impedance, and/or (b) moving to mumetal (and presumably "amorphous") cores. As you stated:

Output transformer - keep the stock 4K, it is already overkill for a 45 in terms of bass handling capabillity. You might consider a Magnequest EXO-45/46 if you can get one with an all-nickel core, but it's a tradeoff - the 5K impedance reduces tube distortion and the nickel core reduces transformer distortion, but the physically smaller core and nickel laminations both reduce deep-bass power capability and the 5K impedance reduces available power full-range. (These effects would be very subtle.)

 Since I'll be using the EML 45B at >2W output (possibly as high as 5W), I figured a bigger OPT should be a good thing, provided it is a well made parafeed type -- particularly as I will be (a) using either a Sowter "mumetal sandwich" OPT or a Lundahl "amorphous core" OPT, and (b) if I get the Lundahl, experimenting with going as high as 9.7K primary impedance (with higher plate voltage and a difference PT to get me there).

(3) Your rule of thumb that the optimal parafeed cap value = 2*L/R*R (where L is plate choke inductance and R is OPT impedance), and that any value twice or half as big will work. I plan to use the midpoint - i.e., what your formula suggests.

There are a lot of things going on with this, too many to do more than mention in a single post. Here are (I think) the main ones:

* OPT inductance, which is much smaller at very low signal levels and usually sets a LF frequency response limit. Nickel cores have higher small-signal inductance.

[...]

* OPT saturation at high levels sets the power band lower limit. Nickel cores saturate at much smaller signal levels.

From what I've read, it sounds like mumetal and amorphous (glass metal) cores have similar effects as compared to silicon steel. So I have assumed that if one switches from M6 to one of these fancy core materials, inductance goes up but the core will saturate twice as easily (just a pure guess that I hope is good for estimating worst case scenarios). So in making the switch I need to double the size of the core to keep saturation levels the same. And if I want to decrease saturation levels, I need to go even bigger still.

* Plate choke inductance

I've ordered Lundahl plate chokes spec'ed as 70H @ 60 mA, with 160 DCR. 

* Parafeed capacitor value, which if done right maintains a resistive load to a lower frequency than the choke alone, reducing tube distortion (but not transformer effects) at the lowest frequencies

As stated above, I will be following your 2L/R*R formula.

The high frequency performance is a whole 'nother issue  :^)

The effect of different iron and different OPT impedances on treble will be interesting to hear. The Kaiju treble is superb and has really "opened my ears" to how wonderful a treble-centered (to my ears) presentation can be, whereas I usually gravitate to a more mid-to-low low centered presentation - which the SII and the EML 45Bs provide (but lacking the detail and punch of the Kaiju - again, to my ears).

Edit: And I almost forgot, I would very much appreciate any guidance you can provide about what if any questions I should ask about the Lundahl OPTs before committing. For example, I read an older post from you (PJ) on AA to the effect that PP OPTs are configured with a center-tap that is meant to be at AC ground potential, whereas parafeed requires the B+ connection to be at AC ground potential. So I assume that the parafeed version of the Lundahl OPT I am considering should have no center tap, and that maybe I should inquire about whether this is the case?

cheers and thanks, Derek

[Paul Birkeland]

If I am remembering correctly, you're using these with headphones right?

[Deke609]

If I am remembering correctly, you're using these with headphones right?

Yeah, my Audeze LCD4 heaphones. With BeePre feeding Kaiju, I do most of my normal (but admittedly, loud-ish) listening with BeeQuiet set to -18db.  The extra power isn't overkill. When I only had the SII-2A3 I could run out of volume/headroom on some tracks - and this was when wired for 16 ohms.

I imagine the same setup would blow the drivers in most other headphones.

cheers, Derek

[2wo]

Regarding the center tap, Lundah uses a number of separate windings on both the primary and secondary that you connect in various series and parallel combinations to get the  impedance that you're looking for.  Theoretically you can add a tap  wherever two windings join, center or otherwise but there is no built in CT...John

[Paul Birkeland]

The manufacturer indicates that about 10V from the amp will be adequate for the LCD-4.  You won't get this from a #45 amp or a #45B amp unless you get custom wound output transformers with higher impedance taps to make more power available.  You will get this from a BeePre, and probably even on the 8 ohm OT configuration because the load is light.  The huge downside to the custom transformer taps is that the noise floor will absolutely come up when you do this.  I just yesterday finished a #45 shunt regulated headphone amp with 125H Magnequest chokes and TL-404 5K autoformers that are tapped at a ton of different spots for lots of impedance choices.  While this amp is very quiet for a #45 amp, the noise floor comes up pretty quickly.

The Mainline on the high impedance setting should be quite satisfying with these headphones.

[Deke609]

there is no built in CT...John

Many thanks John. That clear that up. The funny thing is: I had looked over the datasheet for the PP version of the transformer a bunch of times and it clearly showed what you describe: no wires, just a large number of winding taps that you connect together, for different configurations - but despite the answer staring me in the face, I didn't put 2 and 2 together.  Amazing (to me, anyways, probably not to others) how often I do things like this. 

Thanks for pointing this out.

cheers, Derek

[Deke609]

The manufacturer indicates that about 10V from the amp will be adequate for the LCD-4. 

Never thought to look at that since voltage requirements isn't a quoted spec for the headphones. But i think I see how you got there: P = I-sqr*R, so with Audeze's recommended 500 mW, I = 0.05 A, and so V = 0.05A*200R = 10V. 

I'll have to think about this more. I only recently clued into how output impedance and load interact: high load to output impedance ratio should maximize voltage dropped across the load; but setting R-load = output impedance maximizes power transfer (I haven't actually demonstrated the latter principle for myself - b/c I can't remember the calculus needed to do so, and so am simply taking this on faith until I find the time and gumption to relearn some math and probably newly learn a bunch more (complex numbers and phasors are somewhere in my future)). If my foregoing understanding is correct, then that means the SII-45B isn't capable of 10V and is instead geared for high current?  And my headphones are power-starved? Need to think/read about this more.

And I'm not sure what to make of the manufacturer's stated specs. They keep changing. When I decided to buy the LCD4s (used for the same price as a new pair of LCD3s), Inner Fidelity had done measurements and stated than the LCD4s needed about 500 mV (Volts not Watts) to reach 90dB. Around the same time, Audeze was recommending amps capable of "at least 4W" (without specifying more) for the LCD4s and stated that the phones could withstand up to 15W of momentary power (that stat has since disappeared - and they now claim that all of their over-ear phones are good for 5W rms).  And now (I just looked) Audeze states that all their over-ears, except the LCD4, which range from 15 ohms to 110 ohms input impedance, work best with ">250 mW". But they recommend >500mW for the LCD4s. 

I'm not sure how reliable or clear their specs are. When I first got the LCD4s, I picked up a SS headamp that on paper looked more than capable of driving them that cost about $425.   It was 4 wpc  with high, medium and low impedance, and capable of delivering 1500mW into 300Ω. Result: sounded like hell - lots of sharp treble, not enough volume and it sounded bogged down and muted if there was lots of bass. I was very disappointed. And that thing got good reviews (still does). One reviewer, the guy I bought my LCD4s from, even recommended the amp for LCD2, LCD3 and LCD4z (but not, conspicuously, the LCD4 - which is interesting, b/c he still owned my LCD4s at the time and was using them as his main "reference" set -- since replaced by the Abyss 1266).  Better sounding, amazingly, was the Massdrop "CTH" designed by Cavali -- it was too quiet, but it sounded good.  So I started fussing over finding amp for them (after the miserable failure of the SS amp that on paper looked ideally suited) I came across a bunch of comments on various forums suggesting that Audeze's LCD4 recommendations were way off and were incorrectly based on the specs for their other phones.  So I dunno. 

So the search was on for amp to pair them with. And that's what landed me on this forum 2 or so years ago. I had already built a Crack w/ SB for my Massdrop HD-6XX and really liked the sound, so I figured maybe a higher power BH amp would do the trick. And it did: the headphones sound amazing with BeePre and either SII or Kaiju.  Better than I've heard. I've not listened to many high end systems, but I have a friend with a top-of-the-line Linn system that costs a small fortune - it sounds great (a little too analytical and dry for my tastes, but with great detail, punch, etc.). I'd take my headphone system over his 2 channel system any day (except for resale value).

The Mainline on the high impedance setting should be quite satisfying with these headphones.

Well, one of those might be in my future.  But based on my past experience with that SS amp that could put way more than 500mW into the LCD4, I don't think picking an amp based on the manufacturer's specs is the way to go. I don't trust their specs - they've always been kinda unclear.

It would be great if someone posted the results of a listening test comparing the Mainline on hi-imp versus the SII with planars - say the LCD3s which are 110 ohms impedance (which looks to be the closest in the LCD series to the LCD4).  Maybe I will one day!  I would be absolutely shocked if the mainline could beat the combo of BeePre with Kaiju on the LCD4s - the sound is, as I stated above, literally the best I have ever heard. ... But maybe it does

cheers, Derek

[Paul Birkeland]

I'll have to think about this more. I only recently clued into how output impedance and load interact: high load to output impedance ratio should maximize voltage dropped across the load

The ratio establishes that the output of the amplifier won't vary too much as the load impedance changed (speakers rarely have a flat impedance curve).

but setting R-load = output impedance maximizes power transfer

IIRC, you get max power when the load impedance is about twice Rp. The loss of power for going to 3xRp is very much worth the trade-off of higher damping and lower distortion.

 If my foregoing understanding is correct, then that means the SII-45B isn't capable of 10V and is instead geared for high current?

Well, the output stage will swing about 100V RMS, so a 10:1 step-down will get you to about 10V.  It is, however, my suggestion that you wouldn't want to do this, especially if you have to spend a crap ton of money to perform the experiment. 

  And my headphones are power-starved? Need to think/read about this more.

If a Kaiju, especially wired for 16 ohms, is giving you a huge performance increase, it wouldn't be unreasonable to infer that this is a possibility.

 Inner Fidelity had done measurements and stated than the LCD4s needed about 500 mV (Volts not Watts) to reach 90dB.

That's a typo.

 Around the same time, Audeze was recommending amps capable of "at least 4W" (without specifying more) for the LCD4s and stated that the phones could withstand up to 15W of momentary power (that stat has since disappeared - and they now claim that all of their over-ear phones are good for 5W rms). 

They are trying to write specs that can match up with products.  A solid state amp that will put out 4W into an 8 ohm load will be a pure voltage source capable of 5.6W, which is 160mW.  A tube OTL amplifier that can put out 4W into 8 ohms will put out far more into 200 ohms.  In fact, a Crack could be modified to deliver this kind of power.

Well, one of those might be in my future.  But based on my past experience with that SS amp that could put way more than 500mW into the LCD4, I don't think picking an amp based on the manufacturer's specs is the way to go. I don't trust their specs - they've always been kinda unclear.

Well, we have decades of power amplifier development to make amps that work well into low impedance loads from say 1 ohm to 16 ohms.  Headphones are all over the place, and it's challenging to make an amp that will drive 16 ohm headphones really well and 600 ohm headphones really well, though with an output transformer it's not as tough. 

[Deke609]

Many thanks PB.

The ratio establishes that the output of the amplifier won't vary too much as the load impedance changed (speakers rarely have a flat impedance curve).

This I get (I *think*) - I understand (perhaps simplistically or even erroneously) the good regulation as a consequence of the voltage divider relationship where source resistance/impedance is the first resistor (R1) in series with the voltage source and the load is the second resistor. If R2 is much greater than R1, then small variations in R2 won't cause wild fluctuations in regulation; whereas if R2 is close to R1, the same magnitude of fluctuation will cause a big decrease/increase in voaltge dropping across the load. 

IIRC, you get max power when the load impedance is about twice Rp. The loss of power for going to 3xRp is very much worth the trade-off of higher damping and lower distortion.

I really have no clue here. I was just paraphrasing what I remembered reading in Horowitz and Hill, The Art of Electronics 3rd. ed.(my late night bedtime reading, along with Ott's latest edition on dealing with noise and interference - thanks Jamie!  ) I just looked it up again - from the end of p. 11: "Show that R-load = R-source maximizes the power in the load for a given source resistance. Note: skip this exercise if you don;t know calculus, and take it on faith that the answer is true." I can;t remember (assuming I ever knew) the calculus necessary to prove this, so I did as instructed and "took it on faith."

  If a Kaiju, especially wired for 16 ohms, is giving you a huge performance increase, it wouldn't be unreasonable to infer that this is a possibility.

The Kaiju just sounds better than the SII-45 (which sounds better than the SII-2A3). The K is cleaner and punchier and the treble is amazing. In comparison, the SII-45 sounds a little hazier and less decisive. Is it a huge step up in performance? Dunno. The difference is small, but appreciable. Kind of like a smeared mirror or window: that last fraction of a percent of dirt makes a huge difference; but quantitatively, as a percentage of the
amount of beginning dirt, it is insignificant.  At the same time, I imagine there would be no appreciable difference were I using a really cheap DAC, preamp and headphone set - they would add so much "dirt" that the improvement would be buried by the noise floor.

That's a typo.

Well I obviously have no basis to assert things are one way or the other and am just relaying what I read. So I looked it up again. IF repeats the less than 0.5V figure twice in the their measurements.  Here's the first instance:

"Volts RMS required to reach 90dB SPL: 0.482 Vrms
Impedance @ 1kHz: 206 Ohms
Power Needed for 90d BSPL 1.13 mW
Broadband Isolation in dB (100Hz to 10kHz): -2 dBr"

Second instance (same measurement page):

"The LCD-4 needs 482mVrms to reach 90dBspl at the ear. These headphones will run quite a bit more towards the top of your volume dial than most. But their 200 Ohm impedance means that they're only moderately less efficient than most."

(see http://www.innerfidelity.com/content/technologically-impressive-lcd-4-planar-magnetic-headphone-measurements#S6BkJ5yiLskokcbb.99)

 

... a Crack could be modified to deliver this kind of power.

That's sounds like a fun experiment. Could you elaborate on the required mods? I have a mostly intact Crack collecting dust that I will happily experiment with. 

But in the end: while I can't discount the possibility that I am missing out on what my LCD4s are capable of (and this is a concession I make based purely on logic applied to my acknowledged lack of both empirical experience and knowledge about the science of all this - but my ears say "no way"), I find this almost impossible to even imagine. The sound coming out of them when fed by the BP and K (or by the SII's, to a slightly lesser extent) is superb. Is better sound possible? Probably - I bet the K could be built with even better OPTs that cost thousands or even tens of thousands of dollars more that would make it sound better. Would the Mainline sound better than the BP and K? Maybe one day I'll get to find out. ... But I doubt it.  And if so, I'll have the makings for a helluva 2 channel system - I'll just need to buy Jagers, plus move to a place in the country where I won't have the cops at my door every night  )

cheers, Derek

[Paul Birkeland]

I suppose the 0.5V for 90dB will also depend on how far the microphone is from the driver, so perhaps there just isn't anything to be learned or gleaned from that measurement.

For the Crack, we would move from analysis in volts to analysis in mA.  We'll assume that you want 500mW into 200 ohms, which requires 70mA peak current swing.   If we assume about 100V at the cathode of the 6080, then that's 7W of dissipation through the Speedball if we set it to 70mA.  Case to junction thermal coefficient of the TIP50 is 3.125C/W, and the heatsink is 7.5C/W.  With 7W dissipated, that will be 75C rise, so you'll see a constant 100C during operation.  This is the absolute maximum dissipation that you could really run on these for very long, and putting the Crack over a laptop cooler to circulate air with fans will be a very good idea.

To set the large Speedball board for 70mA, R1 would need to be about 12.2 ohms 1/4W.  R2 would need to carry 7mA, so 25K/3W.

The 270 ohm resistors in the power supply will blow with 160mA flowing through them.  They would also drop a ridiculous amount of voltage.  I would use a pair of Triad C-14X chokes instead, which are rated for the current and have about half the DCR of the resistors, so the extra voltage drop won't be such a problem.  The power transformer will get hotter, though I suspect not harmfully so. 

The last thing I had to look at was to make sure the bias voltage of the 6080 doesn't get too low, as grid current can become a problem.  We are looking at 70mA of plate current and a plate voltage that will be around 80V, and there's still ample bias voltage available for that.  While staring at the curves, it also doesn't hurt to draw a 200 ohm load line to look at what kind of swing you get.  The problem you can run into here would be running into the 0V grid voltage line well before hitting 0mA on the other side.  With the proposed operating point, you'll hit 0mA of current with the grid driven about +35V, and on the other side you'll get to 140mA with a bit less signal.  It will work, but have some nice warm fuzz for you with asymmetrical clipping when pushed hard.