New Project: El Mighty Cacahuate


The bones of a new project write-up are posted!  This is for a ~2W EL84 SET amplifier. the amp is a modest project in cost and complexity, but I tried to make this write-up more in-depth than usual. Hopefully it’s a good glimpse into single-ended amps and general tube design for some aspiring hobbyists out there.

Click here for El Mighty Cacahuate



Class A FET power buffers

I’ve been kicking the hybrid amplifier can down the road for quite a while. In essence, I’m looking to do a bigger EL Estudiante. An amp capable of driving speakers to a dozen or two watts using a MOSFET follower output stage for current gain and a tube handling the voltage gain. While this is not especially difficult on paper, making something that is an interesting and practical alternative to tube output stages is not necessarily so straightforward.

On one hand, one should consider the target user. Most tube enthusiasts do not need so much power, so we can bias in Class A and avoid a whole lot of AB headaches and worry about bias adjustment, crossover distortion, etc. This has to be balanced against heatsinking and thermal considerations, of course. Coming from the world of tubes, our audiophile anti-bodies have already pretty well encysted any commonsensical tendencies we may once have harbored, but the smoke point of drywall hasn’t changed. That is to say solid state does not magically make Class A cool, efficient, or pragmatic, but why make something hotter, more wasteful, and more burdensome than vacuum tech? We’re hoping to make something that is more than the sum of its parts.

ccs load se

The Aikido Hybrid 16W SE by John Broskie has about as much power as most probably need. Sixteen is two to the fourth watts, so four times three decibels per doubling of power for a 12db increase over nominal speaker rating (eg your 90db speakers peak at 102db). The quiescent current is a very serious 2A. It is a single-ended MOSFET source-follower loaded by a current source. There’s little not to like other than the coupling caps (contrast this with output transformers for AC coupling in full tube amps). See also Rod Elliott and Pavel Macura’s CCS-loaded follower here.

inductor load se

We can make single-ended more efficient with an inductive load.  Just like a choke load with tubes (eg Luciernaga), an inductive load on a MOSFET lets it swing voltage past the power rails.  The MoFo is an example of a single-ended source-follower MOSFET with a simple passive choke load (50-150mH and very low DCR). This is much more efficient than an active CCS in an absolute watts dissipated sense (note the much lower supply voltage), but chokes ain’t cheap and you still need a good dose of current.

diode bias pp

If you want to lower the quiescent current needed, but stay in Class A, push-pull source-followers are the way to go. Papa Pass’s F4 power buffer does exactly this to cut the quiescent current needed in half. The need to match FETs may be unappealing, but compared to his single-ended F3, F2, or Aleph J, the power delivery into 4 ohm loads is much improved. Note Broskie suggests the same push-pull MOSFET approach (with a different biasing scheme) in the Moskido amp design.

Any of the above would probably sound pretty good: like a tube feeding a very transparent solid state amp. If the amount of power you need is modest (and it probably is if you’re a tube enthusiast), these approaches have made very nice speaker amps. Hopefully I’ll have my own design to contribute soon. MOSFET followers would also make a great multi-watt amp for low sensitivity and low impedance headphones, like the HIFIMAN HE-6.

The HE-6 are rated at a sensitivity of 83.5db at 1 mW.  With 1000x more power (1W), we’d make 103.5db (a 20db increase). Around 5W input makes it a cool 110db. Coincidentally, this is the power rating of the amp HIFIMAN recommends as a pairing.  The HE-6’s 50 ohm impedance lowers the quiescent current needed in a MOSFET output stage, though still requires a voltage rail high enough to prevent clipping. A quick approximation for voltage would be:

power = Vrms²  / impedance

5W x 50 ohms = 16 Vrms²

16 Vrms x √2 x 2 = 46 Vptp

So about a 48V power rail (or +/- 24V) gets us in the neighborhood. Doing the same for current:

power = Irms² x impedance

5W / 50 ohms = 0.32 Irms²

0.32 Irms x √2 = 0.5 Ipeak

We only need about a 48V power rail and 0.5A quiescent current per channel to get us 5W into a 50 ohm load if using a CCS loaded MOSFET. If we choke load, cut the 48V in half. If we use Class A push pull, cut the current in half. The heatsinks aren’t going to be tiny, but a desktop size amp isn’t out of the question.





The difference between a headphone amp and a preamp

This is a question that, as a beginner builder, confused me quite a bit. While it isn’t too hard to understand why a preamp cannot drive power-hungry low-impedance headphones, it’s less obvious what separates an amp that can drive headphones from a low gain line stage. Headphone amps and preamps often share the same small signal tubes, usually Class A, and often single-ended.

Here are the modifications I would make to the El Estudiante headphone amp to make it better suited to line stage duty. While a purposely designed line stage might perform better, I can’t think of a way to do a halfway decent tube line stage any cheaper or simpler. If you don’t go mad on caps, this costs less than the headphone version.

Output Stage

Power requires both voltage and current. How much voltage or current required for a given amount of power depends on the load you intend to drive. Remember:

Power = Voltage x Current

But also:

Power = Voltage² / Impedance


Power = Current² x Impedance

To create power into low impedance headphones, we need current. This drives a lot of design decisions in tube headphone amplifiers. Common approaches to create power are push-pull output stages (eg SRPP, White Cathode Follower), output transformers, and solid state power buffering. The Estudiante creates the power required for low impedance headphones using the latter approach: a single-ended CCS-loaded MOSFET buffer. At a 100mA quiescent current, it can make about 150mW into 32 ohms:

0.1A² x 32 ohms x 1/2 = 150mW

(note RMS = Peak / √2)

On the other hand, with a 10,000 ohm input impedance on an amplifier, this current is unnecessary because the maximum ‘power’ is limited by the voltage, not the current:

24V² / (10,000 ohms x 2) = 25 mW

Now we don’t really look at power output per se in line stages and we’re rounding up the peak output voltage as half the power rail voltage, but it’s obvious that we don’t need all the current to drive the input impedance of an amplifier because we’re limited by voltage anyways. Consequently, we can lower the current in the MOSFET output stage to something that doesn’t even require a heatsink, making a preamp build that much simpler and cheaper.

With the LM317 CCS, we calculate the needed set resistor as 1.25V / Iq (where Iq is the idle current). A resistor of 100 ohms will give us 12.5mA idle current, which should be plenty for a reasonably low output impedance, but not enough to need a heatsink (I would probably still bolt my TO220 parts to the chassis though).

linestage estudiante

In addition to lowering the idle current in the MOSFETs, we can change the big nasty electrolytic cap found in the headphone amplifier to a higher quality film cap. Electrolytics are great where you need a large capacitance in a small and affordable package, like the output coupling cap in a headphone amplifier, but electrolytic capacitors have been shown to create distortion at low frequencies (see Douglas Self’s Small Signal Audio Design) and exhibit leakage current that creates a thump on power down (which may just be annoying on headphones, but potentially damaging on a high power speaker amplifier).

For an input impedance of 10,000 ohms and a -3db point of 5 hertz, Our new cap size in microfarads (uF) is calculated as:

1,000,000 / (2 x Pi x 10,000 ohms x 5 hz) = ~ 3uF

A film cap of this size at a rating of only 63V+ is not hard to come by. I’d probably buy an assortment just to see if I could hear a difference. We should also increase the size of the loading resistor on the output from the 1k in the headphone amplifier to something like 100k or 1M so that we aren’t rolling off the bass or unnecessarily loading down the MOSFET output stage.

Finally, because we’re reducing the current in the output stage, our power supply requirement is relaxed, maybe opening up more wall-wart options to power the project. So if you’re looking for a simple, low-voltage, and cheap tube preamp option, modifying a headphone amplifier like the El Estudiante may be a good option. I’ve even used the headphone amp to feed power amplifiers in a pinch and it sounds surprisingly good.


DIY amplifier top plate easel

I wanted a better way to wire my amps and had been lusting after Decware’s amazing assembly room for a while.  While I can’t afford the custom extrusions and equipment that Decware has, I can get creative with common materials.  I decided to try a simple easel based on a couple of rails of t track and some standard angle and rod extrusion.

The basis of my easel is two 2ft sections of t track. Finding a 4ft section with bolts and knobs on sale at Rockler was what pushed me over the edge to build this daydream. The t track is 3/4″ wide by 3/8″ deep. Rather than trying to route a channel in a thick board, I sandwiched the t track in scrap with 3/4″ thick scrap under the track and 3/8″ to form the outside. These were glued and clamped face down so that I could be sure the t track would be flush with the top edge. 

My bench is cantilevered from basement joists with 2×6 vertical supports. I ran a 1/2 aluminum rod between the supports to provide lateral movement and adjustment to the t track rails. The rails simply have a 1/2″ hole through which the rod passes. By mounting the rod 12″ from the surface, the 24″ rails give a 30 degree angle. This leaves plenty of clearance for transformers and is comfortable to work on while standing or sitting). I purchased some 1/2″ washers and collars, but they may not really be necessary (I’ll find a use eventually). 

My horizontal ledges are 1/2″ by 3/4″ aluminum angle (3/4″ side is flat against the t track rails). I used self adhesive cork sheet to protect plates in the easel from scratching on the support. Depending on your knobs or wing nuts, you may have to trim some of the horizontal rails so that they can be completely tightened. I cut my rails to fit a 18″ wide top plate, but aluminum extrusion is cheap if I ever want to build something wider. 

All in all, this is a handy and relatively simple addition to my tube amp building bench. And it is a lot cheaper than custom extrusions or lab fixtures!



New page on headphone output stages


I posted a new page on power output stages with a headphone focus (though it can be applied to speakers as well). My aim is to eventually try all the variants.  Between the Estudiante MOSFET hybrid, Papa Rusa parafeed SE, Bad Hombre differential, and Luciernaga series-feed SE, I’m part way there. Been working on ideas for an all-tube OTL lately.  Hopefully I’ll tick that box soon!


Another Muchedumbre lives

I finished another Muchedumbre build with some slight variations.  This has two outputs and two inputs (easily switchable back to the 1+3 arrangement).  The power supply CLC filter uses all motor run caps instead of a mix of motor run and electrolytic. Other than these small tweaks, it is built as designed.

The wood apron is a very nice piece of walnut with a lot of prominent grain motion and color variation and the panel is inset rather than sitting on separate interior spacer boards. This is going to live a very happy life in Madison, WI.


WTF amplifier badges

logo plate white

Currently getting some quotes on badges for my builds.  I’ll be ordering extras if any other builders would like one to adorn their own DIY build!

Size standard is based on Neutrik D-series panel mount jacks.


USB interfaces for DIY audio measurements

USB interfaces

Looking for a better way to measure my line-level and amplifier projects, I decided to investigate some USB prosumer interfaces. Rather than options like the QA401 with its required software suite, the Keithley 2015 with uninspiring THD specs, or HP 8903 with a footprint and compatibility penalty, I wanted something small, flexible, and with performance good enough for tube audio. A USB audio interface will require voltage dividers for many measurements (whereas the lab equipment usually allows a higher Vrms input), but recording interfaces are inexpensive and flexible with software. Also, I don’t have the play money for an AP or dScope rig.

RMAA interfaces screenshot

These were all measured back to back on the same laptop with latest drivers and the same unbalanced cables. The same -1db level was used for all interfaces to get a relative distortion/noise baseline. RMAA doesn’t necessarily give an absolute and repeatable spec, but it is good enough for relative comparisons. All interfaces were measured several times; the displayed specs capture the “average” performance (calculated by eye).

  • The MBox 3rd Gen is an obvious winner in just about every regard. It was also a much more expensive interface when it was new.  MBoxes are no longer produced, but used interfaces aren’t difficult to find. This would be a decent basis for THD measurements of amplifiers (my intended use).
  • The M-Track 2×2 did rather well (as much as I hate to admit it) but doesn’t have two identical channels for this kind of thing (one TRS and one combo jack with mic pre). Wouldn’t recommend it for measurements for that reason.
  • The old Fast Track Duo (Avid branded bu made by M-Audio) blew chunks. Can’t rule out that my unit has some kind of issue.
  • The AudioBox USB looks good but there is a cross-talk issue. Possibly grounding with the unbalanced cables. The knobs are also too fiddly for fine adjustment in my opinion.
  • I really wanted the iConnectivity to perform the best here. In my opinion it’s the nicest piece of hardware. Unfortunately, the relative measurements don’t make it the best choice. It can be run from a 9V supply rather than the USB bus though and I may try that to see if there’s any improvement. No supply handy for this test.
  • The AudioBox 44VSL does pretty well (this is what I had been using for measurements). It also requires a 12V external supply and is a larger 4 mic pre interface, making it a little less convenient for a bench-top test setup. The 22VSL is smaller and may measure just as well (don’t have one to play with).

Here’s after some fine tuning the MBox levels in REW (sampling rate set to 96khz):


This is close to the -110db THD Avid spec’d. All in all, I think I can live with the MBox for a while for my testing. Although all the caveats of RMAA and testing conditions/methodology apply, performance is on par with some specs I’ve seen on the cheaper audio analyzers and definitely a cut above the other USB interfaces here.

Link to RMAA software

Link to REW software


Something DHT this way comes…