gallery/2 ecpsoftware articles diy audio links contact

Less-pressivo* Parafeed Modular Headphone+ Amplifier

A low cost, autoformer or transformer coupled, parafeed, single ended, modular headphone amplifier based on the Espressivo, or How to build a real tube headphone amplifier for about $100.

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.

When I was a kid, one of my father's favorite activities was trying to reproduce "restaurant" food at home. He did this largely out of necessity as my mother was a health food nut, which meant that we ate a lot of things with no flavor, being that fat, sugar, flavoring, coloring, and salt were all banned. So, my father's attempts were not geared toward high end food, but rather were in the more pedestrian vein of reproducing fast food. Indeed, the reproduced Egg McMuffinTM was perhaps his greatest triumph. It is common for foodies to dismiss fast food as horrible, but the truth, of course, is that it tastes good. It should be avoided for cultural, environmental, heath, and moral reasons, but not taste necessarilly.

Anyhow, with that as background, and with my own interests being more audio minded, I became curious about the Espressivo headphone amplifier. It is a basic parafeed design by Gary Dahl that, I assume, sounds great. The only problem with it is that Gary used prohibitively expensive parts everywhere. For instance, the tubes cost upward of $500 per pair, if you can even find them, and being high mu, high Gm tubes, one would certainly need more than two to come up with a reasonably matched pair. Moreover, the output transformers need to be custom ordered from Magnequest and cost about $400 for the pair. The amp uses a Dact stepped attenuator which will set you back a couple hundred more, and due to the pentode based CCS, the power supply needs to be overbuilt raising the price even more.

The Less-pressivo, click for a more readable version.

Having recently sunk too much money into my own megabuck tube headphone amp, I couldn't justify the cost, so I took a page from my father's cookbook, so to speak, and I decided to try to reproduce the Espressivo with what would be allowed at home, which meant a total bill under $100. Additionally, since I can never make up my mind about what tubes to use, I decided to make it modular so that different tubes could be used to suit my mood.

An early breadboard version using 6N6p tubes, 10m45 CCS and Speco T-7010 transformers.

To understand what this would entail, a little background into the design is perhaps in order. Most tube amplifiers use a small signal tube (something like a 6SN7 or a 12AU7) for voltage gain and a second tube for power output (an EL34 or 300B). The Espressivo, however, is a "spud" amp, meaning it uses a single tube capable of both functions. The number of tubes that can do this is limited, but since it is a headphone amplifier and actual necessary power output is minimal, there are a number of good choices (touched on below).

Additionally, the Espressivo is a parafeed design using a constant current source (CCS) on the plate of the tube. For more background on how this works, see the MEHA page on this site.

Assembled with 119DA transformers and Bottlehead C4S boards for Grado use -- Click for a larger version

To understand this design, you can think of a tube as an adjustable diode (it is an adjustable diode which is why you can think of it this way). When you pass a certain current through a diode, it drops a certain amount of voltage. The relationship between the current and the voltage drop is governed by the current, but also by where the tube is adjusted to. In the case of a triode (as used here), the tube is adjusted by "biasing" the grid -- which means by altering the voltage between the grid and the cathode. In a standard tube design, you start by thinking about the voltage being applied to the plate. Then, by using a resistor betwen the cathode and ground, and by connecting the grid to ground through a resistor, the tube is biased.

In other words, current flows across the tube between the cathode to the anode. The current thus must also flow across the resistor between the cathode and groud. This drops a certain amount of volts across the resistor (V = IR). Since the grid is at ground, the tube is biased by the amount dropped across the resistor.

Thus, the voltage on the plate of the tube and the resistor from cathode to ground determins the tube's operating point. In the case of the Less-pressivo, however, this works almost backwards. Here, one applies a CCS to the plate of the tube. Instead of the current flowing being a function of the bias and the plate voltage, the plate voltage is a function of the current and the bias.

So, assume that you have a current of X through the tube, and you bias the tube up Y number of volts. This then will point to a specific spot on the plate curves (where the bias curve crosses the horizontal current line.) If you then draw a line down, this will cross the voltage axis. What this means is that the tube will operate with this voltage (+ the bias voltage) on the plate. If the power supply of the tube is higher than this plate voltage (and this is the most important sentence on this page, so read it twice if it doesn't make sense), then the CCS will automagically drop the necessary voltage as heat, which is what it means to be a constant surrent source.

Now, it bears mention that you need to supply a voltage higher than what is on the plate, but it doesn't really matter how much higher -- so long as it is over a certain minimum differential. So long as the CCS can dissipate the heat and stand the voltage, it can be literally as high as you want. It is probably a good idea to shoot for a plate supply between 40V and 60V higher than what you expect on the plate. All of this means that the tube will find its own operating point, drop the amount of voltage it wants, and leave the CCS to drop the rest.

Additionally, there is another wrinkle to the Less-pressivo which is different from the Espressivo. The Less-pressivo uses an LED to bias the cathode whereas the original uses a resistor bypassed by a capacitor. There are a number of reason for this (lower cost, better sound, easier build) but for now the important takeaway is that the LED will always drop the same voltage, regardless of current (this is not strictly true, but close enough for here). Most LEDs will drop ~2V. This means that the operating point of the tube can be read by simply finding the point on the plate curves where the 2V curve crosses the current set by the CCS.

So, with that as background, here is a rundown of where I managed to cut costs as well as some notes on how to go about assembling one of these things ...

Output Transformers
As mentioned above, the output transformers on the original Espressivo are very expensive, and said to be worth every penny. However, there are some very inexpensive transformers that can be used. For instance, Bottlehead uses Speco T-7010 line matching transformers as autoformers on their S.E.X. headphone amp. These cost $4.13 at Allied Electronics. If you use these, you simply hook up the primaries as shown on the espressivo page and tie off the secondaries. With Grados, the bass performance seems to suffer a little with these, so they may be better suited to higher impedance phones. But, you can also use the secondaries for Grados, in which case youmight want to try paralleling a resistor with the phones -- something in the 32R range is fine. There are lots of other cheap line matching transformers available (though the Specos are the cheapest I've seen) but one can also use a more traditional parafeed transformer with secondaries on the output. Indeed, for safety reasons, this might not be a bad idea at all.

One option for this is the Hammond 119DA, available from Antique Electronic Supply for about $35 per pair. They are used on the output of the Mappletree amps, and they are of very high quality for the money. These are for Grados (and other low impedance phones) only, and you really need a tube with an ra of under 1K2 with them (see below for more on this), but if you meet these criteria, these are really really good. Indeed, they are slightly less detailed than my Electra-Prints, maybe have a little less depth in the bass, and cost 1/7th as much. Sowter also makes some, but they are quite expensive, and thus not within the $100 budget. Total transformer cost, $8.26, or $35.

Finally, there are a number of ways to hookup parafeed transformers. They are sometimes connected to the plate with the parafeed cap coupling the transformer to ground, or to the cathode. Here, the capacitor is connected to the plate with the transformer gounded. This is esential for safety reasons as otherwise there will be high voltages on the output -- don't be tempted to change this.

First, as shown on the schematic, if using the Speco transformers, the output can be taken from either the primary (as in the Espressivo) or from the secondary. If taken from the primary, the higher the tube's plate resistance (ra), the higher the output impedance. ra is simply mu (which is gain) divided by Gm (transconductance). As mu gets higher, or Gm gets lower, ra gets higher. If you are using high impedance phones, this does not matter much -- so long as the trubes are within reason. However, with low impedance phones like Grados, it makes a big difference.

There are a lot of options here. The original uses WE437A's which are very expensive. Less expensive options include the 6C45pi which is a high mu high Gm. Gary himself suggests this one. It has a similar mu and Gm to the 437A, but being new production does not cost nearly so much. However, it is still a little more expensive than I wanted -- usually about $40 for a pair, and I don't think it actually sounds that good. Other disagree on this point, but they are my ears. If you do use it, 10 to 15mA is probably as high as it should be run.

There are a number of lower Gm tubes that are still high Gm and will work fine. Chief among these is the venerable 5842, sold by Western Electric as the 417A. This is about my favorite tube for just about everything. Again, however, they can be a little more expensive than I wanted (though if you look around enough $25 for a pair is common, though the WE versions can bring up to $150 per pair), but since I already own a stash I was willing to make an exception. Another option is to use the 6688, also called the E180F. It is a pentode, but when triode strapped has a similar mu and Gm to the 5842. I used it triode strapped i nmine with good results. They can be found for about $15 for a pair which is getting quite reasonable for a budget setup. ra for these is under 2K.

Additionally, there are a number of dual triodes that are promising. These generally have lower Gm, but if both sides of the tube are run in parallel, this doubles the Gm making them good choices. Perhaps the best option is the 6N1p. These are available all over for about $12 for a matched pair, and sound quite good -- in fact, better than quite good. They are often compared to 6DJ6's, but they have a higher mu, lower Gm, and draw more heater current. Also, the 6DJ8 family works well so long as you don't need too much gain, and truthfully, you probably don't. These also want lower V on the plate than some others, which might be a good thing. Other things like the 12AT7 or the 5687, or the 6H30pi, or 6N6 are also good. In fact, there are tons of other good options that can be used.

Using both sides of a 6DJ8 in parallel (ra ~1K2) there was no bass for the Grados taking output from the primaries. Using both sides of a 6N6p, with ra closer to 600R, bass was quite a bit better, but still clearly not great. This suggests that even with the 6c45 with its high Gm (and thus low ra) bass may still be not great. As mentioned, the solution to this is to use the secondaries instead. The down sides of doing this is that the transformer has more of an impact on the sound, which, for a $4 transformer can't be good. And indeed, this does impact the midrange clarity a bit. Anyhow, that said, the lower the ra (and the higher the Gm) the better for tubes. CCS
In order for a circuit to be a parallel feed, there need to be two feeds, in parallel. The first is the output transformer. The second, in this case, is the constant current source, or CCS. For more on using a CCS, see my MEHA page. The Espressivo uses a fancy CCS designed by Gary Pimm that uses EL84 pentodes and Mosfets. Besides the added complication and expense of lots of circuit parts (not to mention the complication that Mr. Pimm has taken down his web page), these also require one to more than double the voltage of the power supply. To keep things simple and cheap, I simply used a single chip CCS, the IXYS 10M45. These cost less than $2 and are used in a lot of good sounding circuits. You also only need to drop about 10 to 20 volts across them easing the PS requirements. Another good CCS option is the C4S kit from Bottlehead.

Power Supply
One of the joys of using a CCS is that the power supply is much less important. So long as the ripple is reasonably low, the CCS will take care of the rest. I initially used a rectifier tube (I had them so they were free to me), a CRCRCRC filter using a bunch of cheap electrolytics, and a very cheap power transformer. There are a bunch of cheapies out there, but the 6K3Vg from allied for $22 is clearly the best deal. It has a 6.3V tap for the tube heaters, a 5V tap for the rectifier heater, and a 650Vct winding for B+. This voltage is a little high, so you need to use some beefy resistors in the PS to bring it down a bit -- or else a large heatsink on the CCS. I use AC for the heaters, and just bias them up to about 40V via the center tap for low noise and hum. In fact, when I first turned mine on, I thought it wasn't working because it was so quiet.

Thus, all that really matters is getting the ripple low enough for the CCS to get rid of the rest. I searched Digikey for the cheapest 100uF/450V caps I could find. They were about $2.28 per cap. For the resistors, I used 3W metal oxides. You could also use a choke if you like. Tube or solid state rectifier is fine. If you do not use a tube, it will allow you to use a transformer without a 5V winding which can save significant money. Allied has a transformer (6K49VG, which is a Hammond in disguise) with a 460Vct winding, and a 6.3Vct winding that is about perfect for just over $20. Just make sure that the current rating is high enough. As a last PS consideration, using a couple of resistors (220K and 50K), build a voltage divider to bias up the heater supply. Bypass the lower resistor with a small (~100uF) electrolytic.

Other Stuff
There are a few other places I cut some costs. First, I used cheaper parafeed caps. The Espressivo uses Hovland film and foil caps that cost $50 to $60 for the pair. You can see some recommendations from me for cheap caps here. I went with some Obbligato caps from DIY Hifi Supply which are about $13.50 for the pair. There are plenty of cheaper options, though, that would be fine.
I also cut costs by LED biasing the tubes rather than by using a resistor and a cap. For about $0.25 you get better performance and save the cost of an expensive resistor and an even more expensive Blackgate cap.

So, those are the basics of how to build a Less-pressivo for about 1/20th the cost of the original. Here is a parts list if you want to build your own. Oh, and a note on circuit boards -- for now they are not available. If you really want one, I might be convinced to pass on a Gerber file that could be sent to Olimex so you could make a set. But otherwise, this is a simple enough project that point to point should work fine.

+With something like a 6c45pi as the tube, the Less-pressivo should produce about 1.5 to 2W to power speakers.
* The initial thought was that this should be called the Espressivito, playing on the diminuitive suffix. However, a little bit of (shoddy and incomplete) internet research suggests that ito may actually be a Spanish diminuitive, which does not match the Italian origin of Espressivo, making it a bit nonsensical. On the other hand, other sources suggest that ito is indeed Italian, and I seem to remember from a few Italian classes many years ago that it popped up occasionally. At any rate, this seemed a little too cryptic, with Less-pressivo being the more obvious joke.