Foreplay goes to my Headphones

The following is mostly a diary of what I tried. The oldest experiments are first and the newest experiments are last. At the end of this page I have a table of Headphones and a few transformer measurements.

Uses the capacitor coupled cathode follower of the Foreplay to drive a transformer that is plugged into the output jack of the Foreplay.
<- Version 1 with minor changes (still usable as a preamp.)
Version 1 to 4 look the same.

Uses the capacitor coupled cathode follower of the Foreplay, but the cathode follower current is increased.

Uses the capacitor coupled cathode follower of the Foreplay, the cathode follower current is increased and it introduces a Jan Meier Bass Boost Crossfeed Circuit. The Bass boost is just a bit more than 1 dB so don't worry about it.

Uses both plates of the 12AU7 in parallel, sets the amp up as a CCS loaded parafeed amp and it uses the Jan Meier Bass Boost Crossfeed Circuit. It works fine with a high level input  (1.5 to 2.0V) with efficient headphones but does not play loud enough with a 0.5V input.

Version 5 uses one plate of the 12AU7 for gain, the other plate is used as a CCS loaded parafeed amp. Ver 5 still uses the Jan Meier Bass Boost Crossfeed Circuit. This circuit has slightly different tuning than the Ver 4 design because of the need for grid resistors to ground.  I've got plenty of voltage gain now and plenty of volume, but I think I like Version 4's sound better than Ver 5A. I believe this is because the transformer was being driven from a lower Rplate.  Version 5B sounds better than 5A and the only difference is the bias point. It has been too long since I listened to Version 4 to compare it to Version 5B.

I updated my  Load Line (click me) calculator just to help make this version work better. The 115V socket in back is so that a wall wart to power a portable CD player can be plugged in.

I couldn't wait for the Paramours to come to try a 70V line transformer to couple the Foreplay's output to my Grados. So I went out and bought a pair of  $4.89 Philmore MT75 70V 10W transformers at Fry's electronics.

I hooked the 0.625W primaries (BLK-BRN) of the 10W transformer up to a pair of cheap nickel plated RCA jacks and the 8 ohm outputs to a 1/4 inch headphone plug. The black leads going to ground on both primary and secondary. This was not an autoformer hook up, but it was quick and not too dirty.

Note 1A: To make this an autoformer, the primary's black lead would go the the 8 ohm's white lead and then go off to the headphone output. The 8 ohm's grounds and the RCA grounds would be star grounded at the headphone jack.

Note 1B: A second configuration would be to leave the 8 ohm output not hooked up, hook the primary's Black lead to ground, drive the brown lead and use the red lead (6.56 turns ratio referenced to 8 ohms) for a 600 ohm headphone's hook up connection.

Note 1C: In my opinion, when there is a large step down in voltage, an autoformer hook up does not buy you much performance improvement. When going from 11:1 autoformer to 10:1 transformer is only a 20% impedance difference. Not much in my book. I'd rather be able to isolate the primary and secondary ground hook ups for this a 20% impedance difference.

10 minutes after the solder had cooled on the cheap Philmore transformer, there were sweet sounds and great articulation of voices and instruments coming from my headphones. However, the bass was weird and there was a bit of brightness/hardness/breakup not normally present with my Grado 325 headphones. The transformer coupled headphone driver showed promise, but this transformer was not the right tool for the job. I couldn't listen to this for very long before saying "no more."

A 26.3:1 turns ratios gives me a 27.7K reflected primary. With 12H primary inductance, 30 Hz adds an inductive 2.3 kohm load across the reflected primary impedance. This is an effective 2.25 kohm load at 30 Hz. With 2 mA peak available to drive this, we run out of primary drive at 3.19V RMS on the primary. This is 0.121 mV on the output. If we ignore losses, we get about 92 dB peak at full volume from the Grados. No wonder the bass was weird, I need at least 10X more primary inductance.

In the Foreplay preamp, I used micro-clips to short out the resistor from the selector switch to the stepped attenuator. This changed the Sweet Whispers volume control  from a -20 to -50 attenuator to a -0 to -30 attenuator. I was running the Foreplay at near full volume. So tube gain is an issue when using headphone amps.

The Foreplay used to drive this transformer had the Anticipation active loads installed, both HV and filament snubbers installed and 4.7 uF SCR output capacitors. I used my 2 meter AudioQuest Turquoise to connect the Foreplay to the transformer's RCA jacks.

I don't think you can pull a headphone driver off with the Foreplay unless it has the Anticipation mod (C4S) installed. You would have enough clean current swing.

I think the brightness is from the Foreplay not liking the 2 meters of Turquoise interconnect hooked to it. If I put 200 ohms in series with my 4.7 uF output cap inside my Foreplay, that may fix the brightness, but why bother when the transformer is messing the sound up this much.

Note: The AudioQuest Turquoise was also very bright and edgy when I use them to hook the Foreplay into my Dynaco Mk IIIs. I usually use DNM Audio solid silver interconnects for this hook up. Your own cable experience will be different than mine. However, cathode and emitter followers do not like capacitive loads. The 2 meters of interconnect and step down transformer could be pushing the Foreplay's cathode follower "over the edge" so to speak. To fix that, the cathode follower will need a cathode and/or a grid stop resistor.

This transformer is designed to be a headphone transformer. The turns ratios on the web are low by a factor of 3. I hope they don't change the design and they just fix the web page. I prefer the turns ratio they shipped me of 36:1, 36:2 and 36:3 (36:1, 18:1, 12:1).

With the 18:1 turns ratio the 8665-X/ Foreplay/ Grado combination sounds pretty good. It just needs the headroom crossfeed circuit or something similar installed.

My first listen was with a -8 to -38 dB shunt mode Sweet Whispers. The 18:1 turns ratio just barely plays loud enough with an Aiwa XP-500 CD player's line out for drive. So I put the transformers away for a while until I could do some soldering in my Foreplay.

As an experiment, I changed two things at once (bad). I undid the shunt mod on my sweet whispers and changed the 18:1 transformer hook up to a 12:1 hook up. The 12:1 hook up played louder, but it did not sounded better. So I switched back to a 18:1 hook up. This sounded pretty good. It still seemed a little strained, but the transformers were never broken in. It definitely needs a crossfeed circuit. I'm spoiled by my HeadRoom's crossfeed circuit.

With the transformer wired for 18:1 and driving a 40 ohm load on the secondary, the tube sees a 13K ohm reflected load on the primary. 105 dB from the Grados equals a 7.94 mA / 0.564 V drive on the secondary of the transformer. From the Foreplay we'll need 7.94 mA/18 = 0.441 mA RMS (0.624 mA peak.) and a voltage swing of 0.564 V * 18 = 10V RMS (14.4V peak.)  A Foreplay with an anticipation mod should be able to drive this. I had to convert back to a standard 0-30 dB Sweet Whispers attenuator to get it loud enough when using a portable CD player as a signal source.

At 20 Hz, 200H is j25K ohm. So a 25 K reactive impedance in parallel with 13K ohm is 11.50K. We'll need about 11 % more current to be able to drive the transformer at 20 Hz than just to drive a Grado headphones. At 30 Hz, the inductive loading is even less and the reactive impedance in parallel with the reflected headphone load is 12.3K.

With a 12:1 turns ratio, the reflected primary impedance is 5.76 kohm. This is fairly low for a Foreplay and may be why it sounded bad.

With a 36:1 turns ratio, the reflected primary impedance is 51.8K. With 200H primary, the tube will see 22.6K at 20 Hz and 31.7 K at 32 Hz. This is a fairly easy load, I forgot to try it and I doubt it would play loud enough.

1. On the Sowter 8665X, I measured a turns ratio of 36:1, 36:2, 36:3 (36:1, 18:1, 12:1) instead of the ratios posted on the Sowter web page. These ratios are a good thing, I hope they don't change. At VSAC 2003 I was told by someone who measure the 8665 with no X and he got the advertised turns ratios. Scary.
2. I have not measured the rated output power of the Sowter. If the power is respectable, at 36:1 this could make a nice 10K:8 low power parafeed transformer.

I can't find my notes on this, but I remember the SOWTER sounded much cleaner. The EDCOR had much lower primary inductance because of the M6 core. Now if it had a high nickel core. . .

I needed more bias current in my Foreplay's cathode follower. Since I have read that the Foreplay transformer does not have much head room when it comes to current, I was worried about increasing the load on B+.

So let's measure the temperature rise of the transformer.

What determines the current rating for a transformer besides load regulation is the allowed temperature rise of the transformer and the rated temperature of the insulation in the transformer. Modern power transformer temperature ratings start at 105C and go up from there. I am used to seeing 130C rated parts off the shelf. Remember: we don't want to be any where near the rated temperature.

I normally use the fact that resistance of copper changes 0.393%/C to determine the average temperature of a coil. This method does not work well on coils with just a few number of layers, but it works well on standard 60 Hz EI transformers. The temperature rise needs to be measured with a few seconds of removing the primary power.

The resistance change temperature rise method measures average temperature rise. When using this method there needs to be a correction for the hot spot temperature. The hot spot correction normally adds 10 to 60% more temperature rise to what the average rise was. Because the primary and secondary measure so close to each other, I'd use a 10% correction on this design.

I used two DVMs to do the readings. One on the primary and one on the secondary. The meters were set to voltage while power was applied and after power was removed, I switched them to ohms. I switched back to voltage before I reapplied power. If you forget to switch back to voltage before you apply power, you can blow some meters up. Don't try this on higher voltage transformers. The spark gaps in the meter could arc and possibly set your amp or meter on fire. Check the rated voltage of the DVM before you hook it on a transformer secondary.

I left my Foreplay off overnight before I measured the temperature rise to make sure it was cold.

After 1.5 hours of warm up on my PDMPS modified Foreplay (with the Anticipation mod), I measured the resistance change of the primary to be 7.92% (20C average rise) and the secondary to be 7.95% (20C average rise). After increasing the cathode follower current by 3 more mA per channel, I measured the resistance change of the primary to be 7.52% (19C average rise) and the secondary to be 8.55% (22C average rise). This is not a big enough change to worry about.

Notes:

1. The filament draws much more power than the B+, so a small change in the B+ power can be tolerated.
2. I increased the cathode follower current by adding a parallel resistor across the current set resistor in the CF C4S.

Cons:
There still is some solid state like glare left in the sound. This could be from the AIWA XP-500 CD player. The sound is still "headphone like" and I get tired listening after a CD or two. I wear out batteries in the CD player with my Headroom before I get tired of the sound. The sound is still not good enough.

Plan for improvement:
I'll recalculate the crossfeed circuit values for use in a tube amp. I'll use the one with moderate crossfeed and bass boost first. After I get the optimum values using ideal amplifiers, I'll correct the circuit for the output impedance of the voltage amplifier tube. I'll assume the values posted on Headwize already have the correct frequency shaping.
 The Jan Meier Circuit

This mod requires removing the jumper between the VA and CF stage in the Foreplay and adding the circuit below. The jumper is replaced by two resistors and one cap per channel. Two more resistors and one cap are shared by the right and left channels. The fit is extremely tight. I had sleeving on the leads of the parts to make sure there weren't any oops followed by smoke. Once you do this mod, your Foreplay becomes a headphone amp only.

I used values I had in my junk box pile. From reading on the web, I quickly picked up that there are many ways to setup the crossfeed ratios. Because of this, I decided not to sweat the exact resistor and cap values.

(The 1 nF is a 1000 pF 1600V polypropylene from Roderstein. I sometimes amaze myself on what I find in my boxes of parts.)

Pros:
Without the 1 dB bass boost cap installed the sound is wonderful! This the highs actually sound a little rolled off now instead of being harsh. I could listen to this for long duration's with out getting tired. This is write home to Mom good (but put it on a post card.)

Cons:
The sound is now a just little too laid back and a little bit dark. (OUCH!, I'm hard to please.)

Plans for improvement:
Try adding the bass boost capacitor. Not only does this boost the bass by 1 dB, it increases the amount and linearity of the delay between the direct and crossfeed signal.

Try trimming the crossfeed capacitors to the exact scaled Meier values. This shouldn't make a difference, but you never know.

Investigate the John Conover crossfeed ratios. Mr. Conover  recommends a higher crossover frequency than Mr. Meier.
 John Conover's calculations on headphone crossfeed.

Investigate if the high impedance into the cathode follower is causing a high frequency roll off.

I hooked C2 up with an alligator clip to see what it changed.

I discovered that with out C2 there was a little phasiness to the sound that adding the capacitor took out. The volume and position of solo acoustic guitars near the center of the sound stage would slightly jump with different pitch notes.

Note: The bass did not seem that much louder.

Surpassingly, the sound was now a little bit brighter! The bass boost capacitor added apparent high frequencies?  The brightness change could be from the alligator clip. So many circuits, so little time!

Summary: The added parts for the crossfeed make the sound very listenable when compared to normal headphone listening. Next, I think I'll trim the 1000 pF capacitors by adding 220 pF capacitors across them.

I'd like to build a parafeed headphone amp. The Foreplay has a much lower output impedance than the plate of a small signal tube. I wanted to know what my headphones sounded like with a lower (worse) damping factor than what the Foreplay provided.

My plan was to rebuild this using a 6SN7 tube. The 6SN7 tube's the drive impedance from the plate to the transformer will be about 7k.. To see what this damping factor sounded like I added a 7.3k resistor between the Foreplay and the headphone transformer. (7.3k is the closest value I had in my junk box. 6.2k would be a better value to use.)

The volume was cut about 3 dB with the resistor, but the sound was a smidgen cleaner even with the gain turned up one click on the Sweet Whispers to keep the volume the same.

Once I saw that I could get about enough gain from the Foreplay 12AU7 voltage gain stage, I thought I'd try driving the Headphone transformer directly from the plate of the 12AU7s. So I did some major surgery on the Foreplay.

I basically removed every wire on the A and B tube but the wires on 4,5 and 9 (The filament connections).

* The CF C4S was completely removed from the Foreplay.
* The VA C4S was changed to have a 120k 1/2W bias resistor and the current set resistor to be 180 ohms (5.3 mA).
* I put a short #4 screw and nut through the MJE transistor in the C4S to add just a little more heatsinking.
* The 3.3 uF cap could be dropped to 1u, but I already had a 3.3 uF.

Presently I do not have the DPDT impedance switch installed. This is a planned addition to the design.

I wanted to play with the transformer connections so I wired each transformer primary to a pair of RCA plugs.

I made the normal Foreplay output jack the C3/ C5 connection point to the output transformer.

I made the normal Foreplay input #1 the 3.8V cathode mid point connection to the output transformer.

By pulling the mid point connection and holding it to the RCA round, I can compare having a normal Parafeed connection and the advanced Parafeed connection where the transformer returns to the cathode of the tube.

Notes:
1. I am normally listening to this with the my Sweet Whispers volume set 1 to 2 clicks on from full volume. I am using the -20 to -50 dB Sweet Whispers with the resistor that goes to the selector switch shorted out so that it is a 0 dB to - 30 dB 10k stepped attenuator. I needed a 10k stepped attenuator to drive the crossfeed circuit so the resistance change with volume changes is small when seen by the 60 k R2/ R9 resistor.

2. It was too long between hearing the cathode follower version and this version of a headphone amp to do a fair comparison. I do think that the new configuration plays the loud parts cleaner. That is about all I can do with out having two amps side by side.

3. As a point of interest: I have 0.96 and 0.97 V across each of my C4S current set resistors. Changing the current into the C4S LED by 0.75 mA showed the dynamic impedance of the LED to be about 61 ohms. This tells me one aluminum cap that goes across both LEDs may help the C4S perform better. I'm thinking >470 uF at 6.3V.

Note: I tried this with the 36:1 turns and I like how the sound sweetened up, but the gain was way too low with an RS3400 CD player. I also tried a few other signal sources and often could not get the volume loud enough. So here I am, good sound, but not loud enough. So what do I do? SOLDER!

At VSAC 2003 I exhibited Version 5 with a RS3400 driving a modified ART DI/O. I used the DI/O used a 2X setting because I couldn't get it to work on any other setting. It turns out some of the sound I didn't like in the Craftsman's room was from the 2X setting on the ART DI/O.  I figured this out after I went home and I got the ext. sync to work. [ I'll bet you're saying "Ya Ya that's just Volt making excuses." ;-) ] The amp only had a few hours of burn-in on it when it was placed in the Craftsman's room Friday afternoon. The sound is still getting better; however, my memory says Version 4 was better when it played loud enough.

My apologies to those at the show: some where along the line Friday or Saturday, my schematics disappeared. I brought a spare, but cosmetically damaged set down late Saturday. One of the OTS guys asked for DIBs on the second set of schematics first, so I gave them to him.

So here's the Power supply schematic. +V_B is 165V. C1, C2, R1 and R2 are the RRSF I had in the Foreplay. If I were building this from scratch, I would not use R1, R2 and C2. C1 alone does the trick. J2 is a plug I added to power a wall wart so a standard portable CD player could be powered up without batteries:

Here is the gain stage. I chose to put the crossfeed between the two tube stages so I could play with the volume control and not change the effective value of R24/ R34. The effective crossfeed value of R24 and R34 is the actual resistor value (49.9k) plus the plate impedance of the tube (24k) that feeds the resistor for a total of 74K. I did try to set the second CCS to 2.5 mA, but it ended up measuring 2.4 mA.

During the build process I first tried biasing the tube with a 45K resistor on the 1st tube and 30K resistors on the second tube before using the CCSs. It sounded distant, hissy, lacking in detail and lacking in power. So in went a two terminal Hawksford CCS I'm playing with. It now sounded like it had a chance. A C4S will work fine if you build one of these at home. I still like Version 4 better when it was loud enough, but Version 5 is still young and is no slouch.

This circuit plays plenty loud from many sources and with a flick of the output impedance switch I could drive just about any headphone I wanted between 30 and 600 ohms. The crossfeed disconnect switch causes about a 3 dB increase in gain when the crossfeed is switched out, so the volume needs to be turned down one click when the crossfeed is kick out for a fair comparison. I really like the crossfeed. It fills in the center of the sound stage so that it sounds more natural.

If I had an output transformer with taps, instead of windings that need to be put in parallel series, I'd probably put in three headphone jacks instead of a switch. The headphone jacks cost less and are easier to install and clean.

Before I increase the bias current from 2.5 mA to 4 mA, I wanted to measure the 12.6V Foreplay's power transformer's average winding temperature rises. The primary temperature rise was 20C average and the HV secondary temperature rise was 17C average. This tells me I have margin to increase the bias current. Increasing the bias current will also drop and linearize the plate impedance of the 12AU7. This change became version 5B.

The lossy parafeed also removes a peak in the loaded and unloaded output gain. Low frequency gain peaking is a bad thing because it causes the transformer to saturate earlier than necessary. This increased core drive from the peaking increases the output distortion. Lossy parafeed lets us control damping separately from where we set the -3 dB point. This means we can set the -3 dB point at a higher frequency for similar damping as compared to just using a larger value parafeed capacitor. Typically the lossy parafeed cap quality can be a 1/2 step down in quality from the main parafeed cap. For example: In the version 5 headphone amp, I used polypropylene and tin foil for the main cap and metalized polypropylene for the lossy parafeed cap. This is because the resistor in series with the lossy parafeed cap hides the second cap from the output at higher frequencies.

The following is the unloaded DC coupled gain of the output stage only with and without lossy parafeed. Notice the peak is significantly attenuated. This plot was with the primary inductance at 300H.

Below is the input to output gain changes using both tubes gain stages with the crossfeed turned off and lossy parafeed both in and out of the circuit. C20 is set to 20 nF to raise the lower -3 dB frequency point to further reduce the wasted gauss in the output transformer. I changed the primary inductance to 200H for this plot. Notice how the frequency response looks nicer with the lossy parafeed.

I played with the lossy parafeed while playing 20 Hz sinewaves. With the lossy parafeed in, the sinewave was visibly less distorted on the scope trace than with it out. The loss parafeed isn't very sensitive to resistor value, you can change the resistor value 20% and the response only changes a little bit. This is because near optimum, lossy parafeed is a low Q circuit. Low Q circuits aren't very sensitive to component value.

It looks like a smaller parafeed capacitor (0.47 uF) with lossy parafeed could be a good thing to try. However, the capacitors are GLUED DOWN in the chassis so the headphone amp would survive the ride to SeaTac in my suitcase.  This experiment won't happen any time soon.

I changed R27 and R37 from 1.65K to 933 ohms and I changed the output tube's bias was changed from 2.4 mA to 4.0 mA.
The output plate voltages measured 105 and 107V where the driver's plate voltage measured 66V and 67V.
B+ dropped from 165V to 161V from the increased current flow.

The hardest three headphones in each category are highlighted in bold. This table assumes a Sowter 8665X (36:1, 18:1, 12:1) output transformer is used. The tube gain calculations do not include a factor of about 1.5 lost in the crossfeed circuit. So if the needed tube gain is 20, it turns into a needed gain of 30 because of the crossfeed. The AKG K1000 was designed to be driven by a power amp, so we can throw it out of the requirements table. If left in, it will only make it to a loudness of about 91 dB with the Version 5 Headphone amp. Once the K1000 is eliminated, we see that only 16 mW is needed to generate 100 dB in the headphones.
 

HeadPhone data mostly from 1993 October Audio	Ohm	dB/ 1mW	mW 100 dB	mV for 100 dB	mA 100 dB	Est Pri ratio	Est Pri Z	Est Pri Volts	Est Pri mA	Tube Gain re 0.7V	dB at 2.4 mA / 30V rms Pri
Aiwa HP-J9	16	105	0.3	71	4.45	36	20736	2.56	0.12	3.7	121
Aiwa HP-X1000	40	104	0.4	126	3.15	36	51840	4.54	0.09	6.5	116
AKG K1000	120	74	398.1	6912	57.60	12	17280	82.94	4.80	118.5	91
AKG K240M	600	88	15.8	3084	5.14	12	86400	37.00	0.43	52.9	98
AKG K271 Studio/ K240S	55	91	7.9	661	12.02	36	71280	23.79	0.33	34.0	102
AKG K280	75	92	6.3	688	9.17	18	24300	12.38	0.51	17.7	108
Audio-Technica ATH-M40fs	60	100	1.0	245	4.08	36	77760	8.82	0.11	12.6	111
Beyerdynamic DT 770 Pro DT 990 Pro	250	96	2.5	792	3.17	18	81000	14.26	0.18	20.4	106
Beyerdynamic DT211	40	118	0.0	25	0.63	36	51840	0.91	0.02	1.3	130
Beyerdynamic DT911	250	115	0.0	89	0.36	12	36000	1.07	0.03	1.5	129
Denon AH-D950	30	106	0.3	87	2.89	36	38880	3.13	0.08	4.5	120
Denon AH-D950	30	106	0.3	87	2.89	36	38880	3.13	0.08	4.5	120
Etymotic ER-4S	100	98	1.6	398	3.98	18	32400	7.17	0.22	10.2	112
Grado HP1	40	96	2.5	317	7.92	36	51840	11.41	0.22	16.3	108
Grado SR325	40	96	2.5	317	7.92	36	51840	11.41	0.22	16.3	108
Grado SR80	32	94	4.0	357	11.15	36	41472	12.85	0.31	18.4	107
Koss PRO/4AA	230	94	4.0	957	4.16	12	33120	11.48	0.35	16.4	108
Sennheiser HD-280 Pro	64	95	3.2	455	7.11	18	20736	8.19	0.40	11.7	111
Sennheiser HD490II	70	94	4.0	528	7.54	18	22680	9.50	0.42	13.6	110
Sennheiser HD540II	300	94	4.0	1093	3.64	12	43200	13.11	0.30	18.7	107
Sennheiser HD-600/ 580	300	97	2.0	774	2.58	12	43200	9.28	0.21	13.3	110
Sennheiser HD-650	300	98	1.7	708	2.36	12	43200	8.49	0.20	12.1	111
Sony MDR-CD1000	32	104	0.4	113	3.53	36	41472	4.06	0.10	5.8	117
Stanton SRS-275	100	101	0.8	282	2.82	18	32400	5.07	0.16	7.2	115

Last Updated on 10/11/03
By VoltSecond

A brief overview of the data:

Transformer impedances:
The 300 ohm Sennheiser headphones need about 3 times the voltage as the 40 ohm Grado headphones do to reach 100 dB. With three windings on a transformer secondary (Like the 8665X), all three could be in parallel for the Grados and all three in series for the Sennheiser. A 4 pole double throw switch could handle this or use two DPDT switches. With two DPDT switches we also have the option of two windings in parallel with one in series with the paralleled pair for 150 ohm headphones.

Another way to accomplish this impedance change is to wind the transformer with taps on the secondary. Properly wound, the unused taps don't hurt the sound. Poorly wound, the taps drastically increase the leakage inductance and kill the highs.

The AKG K240M is a potential problem child to drive with the Sowter and a 12AU7. A 10:1 transformer give us a 60K reflected primary impedance. A 5:1 transformer would give us a 15K primary. If you want to drive these real loud (>98 dB), look around for a 5:1 (15K : 600 ohm) nickel transformer. Pico or Magnequest may have something, or they may not. If 98 dB is loud enough, the Sowter will work.

I would like to see above a 16K primary impedance because I want to drive the headphone transformer with small signal tubes, not power tubes. Even with a 16K primary, we can still get reasonable performance with a 12AU7.

Again, remember, there are ways to wind transformers so that the taps don't add extra leakage inductance when they are not loaded. You just have to be smart enough to wind them that way. So if someone says taps make a transformer sound bad, it may be true for some transformers, but not for all transformers.

I occasionally have access to a good LC meter so I measured a few transformers. Sorry the drive levels were not constant, but I can't go back and remeasure the transformers any time soon.
 

Philmore	MT75 BLK-BRN	MT75 BLK-BRN	AES Nickel Intrastage Secondary	AES Nickel Intrastage Secondary	Hammond 125a	Hammond 125a	Sowter 8665X 1V drive	Sowter 8665-X 1V drive	Edcor WSB10K /150	Edcor WSB10K /150
Frequency	Inductance at 0.5V drive	Parallel Ohms	Inductance at 0.5V drive	Parallel Ohms	Inductance 0.5V Blu-Brn	Parallel Ohms	Full Primary	Parallel Ohm	Full Primary H 0.05V	Parallel Ohm
20	15.4	6.85k	180	141k	13.2	3.05k	283	121k	24.8	10.6k
50	11.95	20.8k	N/A	.	8.98	5.63k	224	184k	13.4	20k
100	10.3	43.4k	149	1.0 meg	6.57	8.89k	193	221k	8.7	47k
200	9.096	78.9k	N/A	.	n/a		163	256k	7.2	122k
500	8.33	118k	N/A	.	3.48	30.6k	122	324k	6.9	235k
1k	8.22	132k	547	4.3 meg	2.97	47.5k	106	424k	7.2	288k
2k	8.8	142k	-65	4.4 meg	2.67	62.2k	338	590k	9.62	320k
5k	72	169k	N/A	.	n/a	.	77p	904k	-6.2	356k
6k	-21	179k	N/A	.	n/a	.	.	.	.	.
10k	-2.2	218k	-1.75H	4.6 meg	7.7k	103k	97 pF	1160k	-893m	402k
20k	-0.403	286k	N/A		-738m	146k	.	.	.	.
20k	156 pF	286k	149 pF	1.6 meg	n/a	n/a	104 pF	1313k	-197m	420k

Notice how the core loss (parallel ohms) and inductance does not change near as much with drive level on the nickel cores as it does on the iron cores. This is one of the reasons, nickel cores can sound better than iron based cores. You'll notice some of the test points from the table below are different from the table above.  Mechanical handling, degaussing etc. causes differences in the readings. So look for at the trends, not the exact values.
 

.	Philmore MT75 BLK-BRN	Philmore MT75 BLK-BRN	AES Nickel Intrastage Secondary	AES Nickel Intrastage Secondary	Hammond 125a Brn-Blu	Hammond 125a Brn-Blu	Sowter 8665X Primary	Sowter 8665X Primary	Edcor WSB10K /150	Edcor WSB10K /150
Voltage	Inductance at 20 Hz	Parallel Ohms	Inductance 20 Hz	Parallel Ohms	Inductance at 20 Hz	Parallel Ohms	Inductance at 50 Hz	Parallel Ohms	Inductance at 50 Hz	Parallel Ohms 50 Hz
5.000	29.9H	9.53k	233H	220k	25.3H	7.2k	394H	179k	98.1H	56k
0.500	15.4H	6.85k	179H	141k	13.2H	3.0k	222H	179k	46.6H	38k
0.050	10.1H	3.9k	142H	150k	6.55H	1.11k	179H	218k	15.7H	20k
0.005	8.92H	3.36k	0.133H	170k	4.60H	0.70k	170H	240k	7.3H	23k

 

Philmore	MT75
DC Bias mA BLK-BRN Primary	Inductance at 1 kHz
0	8.22
4	7.36
6	6.02
10	4.9
15	3.9

 

Philmore	MT75	AES Intrastage	Hammond 125a	Edcor WSB10K /150
Primary Leads	Turns Ratio referenced to 8 ohm output	Turns Ratio Secondary to Primary	Turns Ratio pri-sec	Turn Ratio Full Pri to Full Sec
BLK-BRN	26.3	0.332	1-2= 135:1	8.00:1.00
BLK-YEL	18.6	.	1-3=  60.1:1
BLK-ORG	13.2	.	1-4=  38.1:1
BLK-BLU	9.29	.	1-5=  25.9:1
BLK-RED	6.56	.	1-6=  19.3:1

 

Philmore	MT75	AES Intrastage	Hammond 125a	Sowter 8665X	Edcor WSB10K /150
Short leads listed below	Leakage inductance at 20 kHz	Measurements at 1 kHz	Measurements at 1 kHz	Measurements at 10 kHz	Measurements at 10 kHz
BLK-BRN	54 uH at 8 ohm out	Leakage sec = 4.46H 1.1k series	Leakage 1-6 sec = 45 uH	Leakage Primary = 5.81 mH	Leakage Full Sec (Pri short) 132 uH 24 ohm series
8 ohm	42 mH BLK-BRN	Leakage pri = 64 mH 564 ohm  series	Leakage pri = 20 mH	R_series  10 kHz = 533 ohms.	5-6 = 73 uH 12 ohm series
.	.	.	.	.	.
DCR 8 ohm out	0.611 ohms	DCR sec = 1.49K	DCR sec = 1.12 ohms	DCR Sec 1 0.614 ohm	6-8 = 60.1 uH 10.9 ohm series
DCR BLK-BRN	324 ohms	DCR pri = 399	DCR pri = 272 ohms.	DCR Sec 2 0.717 ohm	DCR Full Pri 174 ohm
	Capacitance at 20 kHz	Capacitance 1 kHz pri-sec	Capacitance pri-sec	DCR Sec 3 0.820 ohm	DCR Sec 5-6 11.3 ohm
BLK & BRN to 8 ohm	298 pF	184 pF	161 pF	DCR primary 179 ohm	DCR Sec 6-8 10.1 ohm