gallery/2 ecpsoftware articles diy audio links contact

L'espressivo Parafeed Headphone Amplifier

This is still a work in progress, and I'll be updating it as I remember things that I forgot, and as I add some more specific build instructions.

Last Update: Monday, 18-Feb-2008 15:37:51 EST

Topics discussed in this article deal with high voltages that can and will kill you. Do not attempt to use this information unless you are qualified to work with high voltages. I am not responsible for your lack of knowledge or mistakes. Nor am I responsible for your utilization of my incorrect or misleading information. This is the internet, not an engineering text book.

Since the original L'espressivo page is a bit of a kludge, and since I have gotten several requests to do so, I thought I'd put together an actual BOM on how to actually build one of these things. Basically, this page is a schematic, a parts list, and a suggestion on what upgrades might be worth doing. The goal of the original L'espressivo was to create a very inexpensive, transformer coupled, tube headphone amplifier. (If you did not see it, you can read about the original story/justification/experiment here.)

The parts list, before shipping, and not including things like screws and solder, comes in just shy of $120. The upgraded versions will cost somewhat more, ranging from about $170 for modest changes, to upwards of $500 for silly things that likely won't matter. The parts suggested are not the only ones that will work. You should feel free to modify and change anything that you like, within reason of course. And, as always, feel free to email me, or to post to Head-Fi or Headwize with questions.

On the schematic and parts list, for any part number followed by an "a" (C5a for example), you need two -- one for each side.

Part Number Allied Electronics Antique Electronic Supply Digikey $
T1 230-0-230 @ 50mA, 6.3V
6K49VG
    $20.66
T2a Speco T-7010     2x $4.17
P1     EVJ-C20F02d54
Any 10K to 100K linear taper pot is fine
$2.66
D1, D2     UF4007
Use your favorite high voltage diodes here. Also, it might be worth putting a resistor (300K or so) and a small ceramic cap (10nF or so) in parallel with each.
2x $0.30
LED1a     HLMP-6000
You want red, and you want inefficient -- no super brights. You can use a resistor bypassed by a capacitor if you like (such as in the original Espressivo) but the LED is much less expensive, and sounds better -- even than an expensive Blackgate. If you do use a resistor, it should be sized to drop about 1.8V to 2V at the current the CCS is set at (~25mA) which means about 70R to 80R. The cap must be sized accordingly -- at least 470uF if not larger. (A non-bypassed resistor won't work here unless you connect the primary of the transformer to the cathode rather than ground -- a so-called Western Electric connection. But, if you do this, you can only use the secondary transformer output, not the autoformer output. Further, the benefit of doing this here is minimal, and it is really not recommended for safety reasons.)
2x $0.60
C1     493-1461-ND
47uF - 450V
$1.87
C2, C3     493-1462-ND
100uF - 450V
2x $3.46
C4     565-1601-ND
100uF - 50V
$0.24
C5a   Solen 4.7uF - 630V
Use your favorite caps here -- 350V minimum.
  2x $4.45
R1, R2     560R - 5W
560W-5-ND
As a small upgrade, replacing R1 with a choke is probably not a bad idea. It not only reduces ripple, but due to the limited bandwidth also filters some high frequency noise caused by the solid state rectifier. I used a Hammond 158L (15H, 75mA) but only because I had it around. Anything above 5H or so that can take the current draw is fine.
2x $0.33
R3     200K - 1W
200KW-1-ND
$0.15
R4     24K - 1W
24KW-1-ND
$0.15
R5a     OD514JE-ND
Carbon Comp. Resistor, 510K - 0.25W
2x $0.42
R6a     OD471JE-ND
Carbon Comp. Resistor, 470R - 0.25W
2x $0.42
R7a     OD101JE-ND
Carbon Comp. Resistor, 100R - 0.25W
2x $0.42
R8a     OD101JE-ND
Carbon Comp. Resistor, 100R - 0.25W
This sets the CCS current at about 25mA per side. Don't be tempted to increase the current without changing the power transformer for a larger one -- this is the limit of what the transformer can handle. Also, these chips are not perfect, so it is worth experimenting with values to get this close to right. A trimmer and a piece of protoboard may prove helpful.
2x $0.42
R9a     OD102JE-ND
Carbon Comp. Resistor, 1K - 0.25W
2x $0.42
V1a   2x 7KY6
There are many other options, but these are a good and inexpensive choice. Here are some plate curves for the triode strapped 7KY6. For other options, check out Pete Millett's page on high Gm pentodes. Notice that these tubes have 7V heaters, but running them low won't hurt them, and since we are under utilizing the heater winding, it will run a little above 6.3V anyway -- mine run at 7.2VAC which is just about right. Note also that AES lists the 9KX6 as the same tube which is slightly unfortunate as a) it isn't, and b) they will send it to you when you order the other one. Actually, the only difference is the heater voltage which is 8.6V or so. It should probably still work okay, but it might be worth requesting that they send you the 7V version.
  2x $2.10
Q1a     10M45 -- As mentioned above, these are set to run at about 25mA. Different tubes will operate better at different currents, so it is worth consulting the datasheets as well as experimenting. But, due to the power transformer's limitations, keep it at 25mA per side, or lower, or get a bigger power transformer. 2x $1.76
F1     Fuse size is (VA/117)*1.33 or about 500mA
F2698-ND
$4.03
Chassis   12" x 10" x 2" (Hammond 1444-29)
Get the bottom cover too (Hammond 1434-29)
  $19.50 and $8.35
RCA Jacks   SH268B and SH268R
One red and one black
  2x $2.95
Phone Jack     Switchcraft 1/4" Phone jack
SC1094-ND
$2.49
Power Inlet     Q210-ND $1.00
Power Swith     EG1892-ND
SPST
$2.75
Optional Impedance Switch     DPDT (Optional, so not included in the price)
360-1832-ND
The original uses a rotary switch, but really there probably isn't much point to the lower impedance settings. But, try them if you like.
$6.25
Knob   PK345
Ugly and cheap -- choose you favorite knob
  $1.55
Fuse Holder     F1494-ND $2.80
Tube Sockets   You need 2 9-pin mini sockets
P-ST9-511
  2x $1.10
Terminal Strips   P-0500H   $2.90
Heat Sinks (for the 10M45's)     HS370-ND
These need to dissipate about 2W each
$2x $1.11
Total $29.00 $53.50 $37.46 $119.96

Upgrades

Parafeed Capacitors
The above parts list will make a very basic amplifier which will perform well for the money and give you a nice introduction to transformer coupled tube amps. However, there are a small number of upgrades that can dramatically improve the amplifier's performance. The two most drastic upgrades are to the parafeed capacitors (C5a) and the output transformers (T2a). For the capacitors, any film caps that are about 4 to 5uF and over 350V will work. For suggestions, you can see my notes on output coupling caps. But, my default suggestion, unless you want to spend a lot of money, or buy vintage which is also a good option, would be Sonicap Gen 1's.

Output Transformers
Upgrading the transformers is a little trickier. For a modest price increase, a pair of Hammond 119DA's from AES will work well. These are about $60 per pair (they were half that a few months ago), and perform well at that price. For a slightly larger increase in performance, Electra-Print Audio will wind a pair of 5K:32 1W, (or 5K:300, or whatever you like) basic parafeed transformers for about $100. These offer a significant improivement over the Hammonds and are well worth the price and would be my real recommendation if you are going to upgrade. Another option in this price range (~$100) are the Lundahl LL1930's available from K&K Audio. These are really line output transformers (5.8+5.8:1+1), but can be used at ~4300:32 (primaries in series, secondaries in parallel), or about 9000:300 (primaries and secondaries both in parallel). You will be limited in power output with these, and I have not experimented enough to know the limits yet.

Beyond that, you are entering into exotic transformer territory. Also from Electra-Print, or from Magnequest, or Intact Audio, or Onetics, or any number of custom transformer winders, you can get parafeed transformers with nickel or amorphous cores, silver wires, etc. Expect to pay at least $300 for a pair, if not double that.

CCS
If you are going to bother upgrading the transformers and caps, then it is worth upgrading the CCS (Q1a). Simply using 2 IXYS chips per side (instead of one), and one additional resistor, will improve the CCS's performance significantly. The following schematic shows how to do this upgrade.

You must add Q2a (the same part as Q1a) and R10a (about 1K). This will do two things. The first is that it increases the load impedance by orders of magnitude which will improve performance. The second, is that by taking the output from the low impedance ("Low Z") output -- also known as the mu follower output -- (connecting "A" to Low Z) rather than the plate, the output impedance from the stage is lowered significantly. This not only should improve bass performance, but it will also allow the use of higher impedance tubes which should open up the tube possibilities somewhat. There is, however, a drawback. One of the major benefits of the parafeed design with a CCS plate load, as opposed to a single feed design with a resistor load, is that the CCS more or less keeps the power supply out of the signal path (as much as 100dB of isolation here). The input signal path, in the L'espressivo, consists of the tube, the LED, the pot, and the grid stopper. The output path is the tube, the parafeed capacitor, the transformer, and the LED (all in a loop.) By taking the output from the Low Z output the power supply is brought back in to an extent which can have an adverse influence on the amplifier's performance.

So, you can still use the High Z output, and if you are using a low impedance tube (like the recommended 7KY6) there is not much of a penalty in terms of impedance. Indeed, you may even find the sound preferable with the higher impedance output as a tradeoff for power supply isolation (I do!), so it is worth experimenting here. In the end, if you want to take advantage of the mu follower output, you probably should upgrade the power supply a little. For more detail and more background, Gary Pimm has a nice writeup of all of this at http://www.pacifier.com/~gpimm/Active_loads_and_signal_current_control.html.

One last note on the CCS is worth making: in both versions, the current is set by R8a. The recommended value for R8a in the parts list, with the single chip, will lead to a current of about 25mA per side. This is the absolute maximum current that the power transformer can support, and indeed, is too much current for many tubes. Moreover, the relationship between this resistor value and current will not be the same for the two chip CCS. It is recommended that if you either want to play with currents, or use different tubes, or use the high perfpormance CCS, that you replace R8a with a small resistor (~10R) in series with a trimmer (200R to 500R.) Put the trimmer in the midle position and adjust from there. Do remember that you are dealing with high voltages, and making adjustments should be done in a safe way. For more on this CCS circuit, see this thread (The DN2540 and the IXYS 10M45 are more or less interchangeable, but the operating points are likely different which would change R8a if you use that part instead.)

Pot and Power Supply
Finally, the volume pot is probably worth upgrading if you upgrade other parts. The listed panasonic is fine, but it can be bettered.

Also, the power supply does not make a big difference in this circuit, but at the point that you have upgraded everything else, using higher quality electrolytics or even film caps and a choke can make a subtle difference. As designed, with a 50mA draw, you will have about 3mV of ripple before the CCS. This is plenty clean as the CCS has good ripple rejection. However, substituting a $10 5H choke (Hammond 156L for instance) for R1 will lower the riple to under 0.5mV. A 15H choke (158L) is even better. However, because the DCR of the choke is lower than the specified resistor, the B+ voltage will increase a little. The net effect of this is minimal in this circuit, but you might want to use a slightly larger heatsink on the CCS, or increase R2 to make up the difference .