[AudioCats]

xbox brick is a switching supply.

I kind of knew the battery pack would end up winning, just didn't know by how much margin; but I was also expecting the transformer box to do better than it actually did. Maybe the xbox brick's extra power (12A rated, driving a 0.5A load) did help the brick to sound better, or maybe the transformer box just didn't have quite enough extra reserve, it might take at least 20x capacity to make an non-regulated supply sing.

I think a good linear regulated supply will be close to the battery pack. My sigma-22 is in pieces (in the process of a re-casing) so I can not do a comparison right now. I don't know how "super" a linear supply has to be to show the benefits, though. I am sure a Sigma-22 will provide good sound, don't know if the other stuff, such as the $100 teralab supply, can do as well.

But if it is only to run a E90, for the $100 it might be better to do a battery box: get some AA Eneloops, put into a box with a charger, add a switch to select between two banks (so there is always one set of fresh batteries available, as long as the charger is plugged in)

https://www.amazon.com/Panasonic-BK...hild=1&keywords=eneloop&qid=1613938469&sr=8-5
https://www.amazon.com/Tenergy-RC-A...eywords=nimh+charger+rc&qid=1613938594&sr=8-8

then again, the sonic differences were clearly audible when using the test rig, but will it be worth the trouble to do a battery box for a stock E90..... maybe not. I did try a battery pack with the stock E90 when I first got the 95x, and the stock sound lacks emotion even with the battery pack.

 

[AudioCats]

a quick listen comparing stock E90 to the mod (with AD8397+RTX)

Both using the Xbox power brick as source, voltage goes into the KZ capacitor box before going into the power jack.

The stock is again very clean and clear, every word in the song was clearly heard; there is some kind of brightness /etchyness, started to bother me after a couple of songs, kind of a “burn” sensation. Bass details are not very defined, when turned up the bass is like a haze pressing forward, a “sound pressure cloud” if you will.

And again, no emotion came out of the stock E90, the feel I came away with after listening to Eva Cassidy’s songbird was totally different from the two setups.

 

[AudioCats]

E90 Bias voltage line:
bias for the phone is taken from the middle of a voltage devider ( R129 / R130 ).

Both of my E90's are massdrop version so are probably from more recent productions, R129=10M, R130=12M, ==> the bias voltage going to the phones is about 320V.

R130 is an unmarked blue resistor, my previous measurement of R130 was wrong: the 12M was from my in-circuit measurement. After desoldering the blue resistor and measured it by itself, it was close to open. So the bias voltage to the phone is basically at whatever the +HV rails is, around 600v.

Adding a resistor bank and a switch to the face plate can allow the user to change bias voltage. This might be useful if one wants to run 230v bias phones on an E90.

effect of different bias voltage
*** test was done on the E90 test rig, bias voltage was changed bia a switch, at about 220v/480v; driving ESP950
-- flipping the switch lever is quick but the sonic change took at least 1 minute to fully take effect.
-- at 220v bias the sound was much quieter;
-- switch over to 480v, the sound got A LOT (!) louder, it was like going from back-ground music listening to "rocking out".
-- the quality difference is not obvious, at 220v the sound was (maybe) slightly rounder. so unless you plan to run low bias phones there is no reason to make any changes to the existing bias circuit.

 

[AudioCats]

Compared the test rig to my other electrostatic amps. The result was quite favorable. RTX is a bit sharp sounding though, which works well with ESP950 but might be too bright for something like the 009. (I don't have the 009, maybe I should get a set just to find out..... )

The main constrain for an E90 mod is the space available for the upgrade caps. There just aren't that many 600+v audiophile caps with small enough diameter.

I got some MiFlex copper foil polyprop caps coming, they should be small enough to fit, at least according to the spec sheet. Lets see if they can darken the sound without compromising other areas.

 

[vmiguel]

... (I don't have the 009, maybe I should get a set just to find out..... )

...

So, just for science... that's a good enough excuse reason to buy new gear...

On a more serious note, thanks for the topic, been following with interest.

 

[AudioCats]

just realized I do have a pair of Lambda 303, which is bright, probably similar to the 009.

SR-303 driven by E90 test rig ( AD8397 in bottom board, OPA637 in top board, signal caps RTX 0.022uf/600v)
-- slightly sibilant, but not overly bright or etchy. Tolerable.
-- bass is a little funny, the mid-bass hump of lambda make it sounds like it "attempts" to make impactful bass but was not successful; bass texture is not visible, much worse than esp950; (the 950 doesn't make such attempt, what it shows is good quality and I don't feel the "want" when listing to the 950's)
-- vocal not as real as esp950.

These 303's have 404 pads installed. On this setup the ESP950 is much better sounding.

 

[AudioCats]

effect of removing C3/13/23/33
these are 47pf ceramic caps, from output lines to the ground, parallel with the phone, probably to provide some load to the circuit when the phone cable is not plugged in.

according to my previous notes the 950's cable+driver capacitance is 80pF.

removing these four caps:
-- basically no change to the bass. This is kind of surprising, I was hoping more energy available to the phone drivers will help at least the impact... did not happen.
-- sound is a little brighter. still ok with the SR303. More "clear" sounding.
-- sound feels a bit more "opened up", especially with the SR303 (not as noticeable with esp950)

 

[AudioCats]

upgrade caps for a budget mod.....
CDM 0.015uf/1000v silver mica

^^ available from ebay, used, for just $2 a piece. The "new" price was over $30 from mouser, so should be pretty high quality stuff.
I could not find the dimension spec on these, but the 0.012uf thickness spec is 11mm, so 0.015uf shouldn't be that much thicker. Likely about 1"x 1"x 1/2". You will still need to raise or drop the top board to gain enough space to mount them, but doable.

I have tried Russian SGM silver-mica caps, they sound similar to the RTX, maybe not as "detailed" but pretty close.

 

[ursdiego]

Hey, @AudioCats thanks a million for all that efforts in tuning this Koss-Setup. I got mine yesterday from Drop. It feels like they have saved money wherever it was possible in order to sell an electrostatic headphone at this price tag. Indeed it sounds good for the cost, no doubt about that. However, I join you with the assumption that not only the housing of the headphones, but also the amp-unit might have been built with cost-optimization in mind. So there is certainly some potential.

Now, it looks that you have found out a lot - just that I must admit that for a electro technically non-knowledgeable guy like me, it is impossible to draw a conclusion out of all this information. Might I ask you if you have come to some conclusion about which parts could be upgraded, and that might be described in a way, that an amateur like me could follow the steps? I guess I can do decent soldering on a board and replace parts, but I am unable to understand the meaning of all that parts or the symbols on a diagram.

 

[AudioCats]

The "mod" is basically just replacing the 6 opamps and the 8 caps in the signal path, very straight forward. I guess what you are looking for is some kind of parts-list.

unfortunately, at this point I can not recommand a "fail proof" parts selection. The main reason being the heat dissipation problem when using ad8397.
So far the sonic winner is ad8397 (AD8599 is warm and dark and intimate but too closed-in; OPA627 is fine and open sounding but expensive and too lean sounding, I am not liking it more than AD8397).but 8397 only comes in SOIC package and it burns twice as much power as anybody else and can overheat very easily, requires special attention to thermal management (adapter with heatpad, glued-on mini heatsink, reduced operating voltage, etc). So I can not recommend the ad8397 either.....
But there might be hope. I just went though some of my stax 001 mod notes and saw somebody mentioned AD8620 being a nice opamp for the 001, he also said it is a "warm" sounding opamp and that is what we are looking for here (the lesser expensive signal caps tend to be colder and sharp sounding, a warm opamp can help offsetting that problem).
AD8620 specs are very good, fast slew rate at 50v/uS, and more importantly the power consumption is only 3.5mA per amp at +/-13v, which is about 1/3 of ad8397 consumption, so it should not have heat build-up problem. The price is high, though, at $16 each, and the E90 needs 6.
so I guess AD8620 might be a good potential option, but I don't have 6 of them and have not tried them in the E90.

as far as signal caps go.... the only two options that should be "trouble-free" (as in "will fit physically without any extra fiddling"), are theta 0.1uf/600v (diameter spec is 0.44" ) and rel-cap TFT 0.047uf/600v (OD spec is also 0.44"). I have not tried either. The ones I have, RTX 0.022uf/600v and Miflex KFPM 0.022uf/600v are both border-line too large, it will be tricky to fit them into the Koss case.

Therefore, no part-list type recommendations, though the likely-good options are AD8620 x6 ($16 each) + theta 0.1uf/600v x8 ($10 each).

somebody should try out the above combo and report back.

 

[AudioCats]

using DC/DC converter to provide +/- LV to run opamps
all of my tests above were done on the test rig, which uses a DC/DC converter to provide the +/- 12v to run opamps. This reduces the stress of the E90 on-board power supply so the little transformer only need to do HV+/-. The DC/DC converter also provides a more stable voltage.

Power to the DC/DC converter is taken at U1 pin3.

The converter (495-TMR4-1222, available from Mouser) outputs +/-12v; the output lines go to R9/R10 locations (R9/R10 removed).
in the picture below: orange wire is the +12v line ; brown -12v line

 

[AudioCats]

But you don't HAVE TO use a DC/DC to do the LV +/-.
For a simpler, more straight forward mod, simply leave R9/R10 on the board. (R9/R10 are 39 ohms.)The voltage to the opamps will be lower though, due to upgrade opamps higher current consumption.

voltage test, using stock E90 on-board supply to provide opamp power.
** opamps = LME49720 x6
LV: +6.3v, -6.8v
HV: +586v, -602v
power draw: 0.3A at 12v
sound: sounds like $h1t

** opamps = opa627 x12
LV: +5.7v, -6.1v
HV: +594v, -614v
sound: sounds nice. Softer sound (compare to using DC/DC converter), but not bad at all.

** opamps = AD8397 x4 (bottom), opa637 x4 (top)
LV: +5.7v, -6.2v
HV: +608v, -634v
sound: very very very soft and rounded sound. Not good.

** opamps = AD8620 x6
hmm, I don't have these opamps. But the power consumption spec is similar to LME49720, so I suppose the rail voltages will be similar as well. AD8620 supposedly works well down to +/-5v, so maybe it really is the best balanced option for a simple/straight-forward upgrade?? somebody should try it out.

 

[AudioCats]

now what if R9/R10 is reduced, allowing more power to the +/- LV rails?

reduced resistance to half (~20 ohms)....

opamp=LME49720 x6
LV: +6.2v; -6.7v
HV: +520v; -540v
total current draw at power jack: 0.25A at 12v
sound: actually not super bad.

I don't know why the overall power consuption is lower. I was expecting it to increase due to more current available to opamps, LV rails should increase (and didn't, stayed the same). HV rails actually lowered.

so no, just leave the R9/R10 alone (unless if you want to add DC-DC converter)

 

[AudioCats]

I will put DC-DC converter back in, and continue the opa627 burn-in process. (they have less than 50hr at this point).

two more planned tests before I am done with the E90 experiments
-- check out the sound with opa627, after they have >150hr on them. Probably sometime next week.
-- check out the sound with "0.022uf miFlex copper // FT1 2200pf teflon", after the 500hr capacitor burn-in period. Probably get it done in May.

 

[AudioCats]

ok, a recommended parts list. Better write down the little details before I forget them.

Disclaimer: modding an E90 involves the risk of encountering high voltage. The info below is for casual reading and entertainment purpose only.

E90BM: basic upgrade: power input cap C33; opamps; opamp supply caps; signal caps; HV supply caps;

from Mouser: (about $125 +$10 shipping)
(1x) Nichicon UBX, 470uf/25v, (lower board C33 location)
https://www.mouser.com/ProductDetail/Nichicon/UBX1E471MHL?qs=wX5mvoJE4NEEhG5dfAyjwQ==

(2x) Nichicon UKZ, 100uf/25v, for LV +/- rails (lower board C6/C8 locations)
https://www.mouser.com/ProductDetail/Nichicon/UKZ1E101MPM1TD?qs=WJSLByB5Mu384kdNgzL0EQ==

(5x) Nichicon UBX, 15uf/400v (lower board C10/C11 and C26/C27 locations; hopefully you can get two closedly matched pairs out of 5 caps)
https://www.mouser.com/ProductDetail/Nichicon/UBX2G150MHL?qs=/ha2pyFaduhn2raJlu2/0E/tv6D1eG9C71jQnE%2Ba8t6hR9fck8tl4g==

(6x) DIP8 sockets for the opamps
https://www.mouser.com/ProductDetail/Mill-Max/110-41-308-41-001000?qs=dvxwXVM4mZVj1CY1TePuww==

(6x) AD8620
https://www.mouser.com/ProductDetail/Analog-Devices/AD8620ARZ?qs=/tpEQrCGXCwuXz/H22UVtg==

(10x) 5mm spacer (to raise the top board; you only need 4 but having some spares won't hurt)
https://www.mouser.com/ProductDetail/Bivar/SRS-1-5MM?qs=/ha2pyFaduj6mtQ5UcEqea2PB0fDO40qpyFOCmhI50wfqdWV6rUaUA==

(2 ft) 24awg 600v teflon insulated stranded wire, white (just in case you need to add wiring. Stranded wire is more flexible, 600v rating is necessary for the voltage encountered)
https://www.mouser.com/ProductDetail/TE-Connectivity-Raychem/55A0111-24-9?qs=/ha2pyFaduicEJE6GjTkr9wW3wYMdz9h3dCPvX1C%2B1PhWRXHcIdlQQ==

(2 ft) 24awg 600v wire, yellow
https://www.mouser.com/ProductDetail/TE-Connectivity-Raychem/55A0111-24-4?qs=/ha2pyFaduicEJE6GjTkr9qd1o2LPpGyExZUWyc3wzjDrH3ex5wwLw==

From SonicCraft (about $100 plus $10 shipping)
(16x) theta 0.01uf/600v (upper board C1, C10, C11, C20, C21, C30, C31, C40 )
basically parallel two 0.01uf/600v theta to get the 0.02uf. The outter diameter of 0.01uf theta is around 0.3" so there is plenty of room to fit these caps.
http://www.soniccraft.com/product_info.php/audiocap-theta-001uf-600vdc-p-63 )

(1 ft) clear heatshink, 3/8" (to cover the theta caps, for extra protection)
http://www.soniccraft.com/product_info.php/12-clear-21-polyolefin-p-4768

(2 ft) teflon tubing, 20awg (to cover theta leads)
http://www.soniccraft.com/product_info.php/20-gauge-teflon-tubing-p-4826


*** you will still need to take care of the SOIC8 -> DIP8 adapters. The ones available at Mouser are too wide, and pins too thick, probably won't work.
The "BrownDog" style adapters will work.
https://www.partsconnexion.com/BOARD-64950.html

To fancy up the mod, might as well change the RCA jacks to something better, so people can see this E90 is "different"
http://www.soniccraft.com/product_info.php/cardas-ctfa-rca-p-1452