What’s the deal with hybrid amps?

hybrids.png

All books on audio design that stoop to cover the archaic and backwards idea of vacuum tube amplification begrudgingly admit tubes are wonderful open-loop voltage amplification devices. They’re very linear (much more so than transistors without feedback), tolerant of high voltage, and forgiving of approximated parts values. Tubes do not make great current gain devices though. Therein lies the problem for us glow bulb fanatics.  To make power, we need both voltage and current. We usually side-step the current-handling weakness of tubes by developing large voltage signals with multiple stages and then using an output transformer to turn the big voltage at modest current into modest voltage at big current.

Let’s look at an example.

A somewhat classic single-ended triode uses two halves of a 6SN7 and a 300B in cascaded stages followed by a 3.5k to 8 ohm output transformer:

6SN7-300B-Single-Ended-Tube-Amp-Schematic

We know that the voltage gain of a grounded cathode with a bypassed cathode resistor is the Mu multiplied by the plate load divided by the sum of the plate load and the plate resistance. Accordingly, the amp above develops voltage gain of about 18x in the first stage, 16x in the second stage, and 3.2x in the final stage. This is an overall voltage gain of about 900x, meaning a 1V signal at the input becomes a 900V signal at the output. In reality, the 300B runs into grid or current cutoff before it gets anywhere near that much voltage swing at its plate and a more likely figure is about half this or 450V peak to peak.

This 450V peak to peak is still quite a lot of voltage. If you could directly drive an 8 ohm load with it [narrator: you can’t] you’d produce thousands of watts. To produce the thousands of watts, you’d use dozens of amps. You have about 0.06 amps [sad trombone]. We use an output transformer to step down the voltage and step up the current. We know that the voltage ratio of an output transformer is the square root of the impedance ratio. In the case of a 3.5k to 8 ohm transformer, that is the square root of 3,500/8 or about 21. Divide 450V by 21 and we get the voltage swing that the 8 ohm speaker is seeing. It’s about 22V peak to peak (seven and a half watts).

We created a hell of a lot of voltage just to step it down to a measly 22V peak to peak. This is where hybrids might come in. Solid state is quite happy driving amps of current into an 8 ohm load and only need a supply voltage of a couple dozen volts. They do away with the multiple voltage gain stages and output transformer. If you can create 22V of signal with a single tube stage, a transistor doesn’t need to make it any bigger; it just needs to provide enough current to drive a low impedance load like a speaker or headphone. Let the tube do what it does best (voltage gain) and let the transistor do what it does best (source lots of current).

So why don’t we see more hybrid designs? For one thing, the power supplies get complicated. You often want a bipolar (plus and minus) supply for the solid state section, a low voltage heater supply, and a high voltage supply for the tube’s plate. Although you rid yourself of an output transformer, you probably added a power transformer (and rectification, filter, etc). Another reason we don’t see more hybrid designs is that many designs which do exist don’t use the devices to their strengths and so cast doubt on the concept. When you see a single tube in an integrated amp, it’s often there as a simple cathode follower. I’ve got nothing against cathode followers, but that implementation is about as much a hybrid design as a burger with lettuce and tomato is a salad.

But by far the most likely reason we don’t see more hybrids (in my opinion) is that devotees of the objective/subjective, transistor/tube, modern/traditional design school are too human. If modern politics hasn’t sufficiently convinced you, the state of the audio market should. We’re kind of a bunch of tribal-minded, technocentric, get-off-my-lawn jerks. If you build a hybrid, you piss off both sides.

So yeah. Screw that. This was the long way of saying I’m building a hybrid amp.

Advertisements

Listening on someone else’s system

Like many audio enthusiasts, I have a general philosophy for audio that guides me when designing (or shopping for) new gear. In a nutshell, I value an objective and empirical approach to design, but this is tempered by the notion that music is art. At its core, art appreciation is a subjective, and often situational, experience. Objective design for subjective ends reads like a paradox; designers have egos too and so maybe conflict between engineering and ‘the feels’ is inescapable. If you’ve been on audio forums or blogs long enough, you know that objectivity and subjectivity do not usually mix in the audiophile hobby. I’ll steer clear of that morass, save to add one recently encountered perspective.

Last weekend I delivered a preamp (design write-up on the way) to its new owner, J. We spent a couple of hours listening to his system with and without the new piece. J’s system is different than mine and the music it makes sounds different, too (including recordings I know). We both had fun going through albums and cranking up the tunes. In a way, it was a little like seeing a favorite group perform live. You know the music but can appreciate fresh nuance all the same. That we got to do so together, on a social level, only added to the enjoyment.

Now what if we were all uncompromising in our objectivity? What if all systems and designers pursued the same goal and weighed compromises equally? Worse yet, what if compromises did not have to be made and all playback was “perfect?” While I know this is ostensibly what many of us seek in the audiophile hobby, where would it leave the hobby on a social and experiential level? I would visit J and hear the same songs in the same way that I always do.

The art in music is not a one-way street. The lenses and filters we use to experience and share art enrich both the art itself and culture as a whole. The process of internalization, expression, and rebirth keeps music relevant and vital. I’m off into abstraction, but there is a kernel of truth for audio here, too: how terrible the tyranny of ‘exactly as the artist intended’ could be if we took it too literally.

You are the artist of your listening, the world is your mixing console, seek out new stereos, and all that jazz.

 

Letters to WTF: Why do “neutral” amps sound different?

One of the issues is that although we can measure a heck of a lot about an amplifier, we can’t express those measurements in ways that are easy to digest and interpret for the average reader. We end up with data that is easy to misinterpret or take out of context. Part of the blame lies with consumers, who want to compare and contrast specs as a replacement for firsthand experience, and part lies with marketing departments that are being deliberately selective or unintentionally incomplete with what they publish.

THD is one of the classic examples of this. The THD as a percent often appears in marketing specifications, but it is not nearly so often accompanied by the all-important context. To make practical sense of a THD specification, you need to know what kind of signal was applied during measurement (frequency, level, single sine wave, multiples for IMD, program material, etc) and at what level the output was measured (voltage or wattage). Even when that information is provided, judging % THD across devices doesn’t give you a good comparative idea of the sound unless you know the harmonic makeup of the distortion spectrum (very rarely provided by manufacturers). All 0.X% THD measurements are not created equal in terms of your listening perception.

Power output is another example of specs that tell less than they should if we want to make a judgement on numbers alone. Like THD, power output requires context in the form of % THD at the measured level, the load used to measure (reactive load, resistive load, etc), and the signal used to measure. This information is not often given and so apples-to-apples comparisons that would allow someone to say with confidence that “these are equally neutral amplifiers” really is not possible. To make matters worse, most of these measurements are interactive, making expressing the results in a practical way more difficult.

While we hope manufacturers would include this kind of information in specs as we shop, it just does not happen. A big reason for that is the average person just doesn’t care. The responsibility falls to the niche press that reviews products and is able to measure them with a consistent method. But there’s a diminishing return on the amount of work it takes to measure and share all the data needed to give a clear picture of how something sounds. Even if the measurement work is done, it takes a certain amount of technical knowledge and experience to interpret it, so we end up with catch-all terms like ‘warm’ or ‘neutral’ to paint the picture. This audiophile language is as much a shorthand for all the measurements that aren’t being done as it is flowery jargon.

TL;DR It takes a huge amount of work to measure and express the findings if you want to capture all there is to say about how something sounds. Few are willing to do it, so we have words like ‘neutral’ that are part subjectivity and part experienced evaluation by trained ears. Even then, it’s not a perfect substitute for hearing something yourself.

New 300B kit by Elekit (via diyaudio.com)

Note I have no affiliation with Elekit other than being an admirer of what they do for DIY audio hobbyists.

Elekit’s kits seem to hit attractive price points for what’s included (tubes, transformers, components, and a chassis) and the quality of the documentation. VK Music (Canadian importer of Elekit) just announced a new 300B amplifier on diyaudio.com here. Although details are sparse, we can maybe glean some ideas from the specs and the previous incarnation (Elekit TU-8300).

Specs for TU-8600:

• Tube Set : 12AU7 X 2 + 12AX7 X 1 + 300B X 2
• Now compatible with low to high impedance headphones
• Frequency response (-3dB) : 10Hz – 80kHz
• Max. output (THD 10%) : 8.3W + 8.3W (Input voltage : 250mV r.m.s)
• Residual noise : 42uV rms (IHF-A)
• Power consumption : 80W when no signal; 80W at max. output

The previous TU-8300 used MOSFET regulation for a B+ of 375V and because the new amp is rated for the same power, I think we can assume the new version operates at about the same voltage with a similar bias (-60V). Rather than two stages of 12AT7 on the input, the new design uses two 12AU7 and one 12AX7 triode per channel. So how are these arranged?

Let’s reverse engineer the numbers for an educated guess. We get full output of 8W from a -60V biased tube with an input of 250mVrms. If the 250mVrms (0.7V peak to peak) fully drives the output tube’s -60V bias (120V peak to peak), we have a gain of about 160x. That could be two successive stages of 12AU7, but then what is the 12AX7 doing? We appear to be using one half of the 12AX7 dual triode per channel. It would be an odd choice for a buffer stage, but it’s a plentiful tube and would make sense in that regard.

The other likely explanation is that we have some feedback at work and the 12AX7s are used for voltage gain. Maybe this is a grounded cathode 12AX7 into a 12AU7 SRPP. In terms of driving the Miller Capacitance of a 300B, this seems like a plausible arrangement. Gain would be in the 400-500x neighborhood, but feedback would knock this back down to the 160x overall and lower the output impedance.

Lastly, this amp includes a headphone output. Maybe the extra triodes are employed in some kind of follower specifically for the headphone section. A 12AU7 white cathode follower seems like a potential candidate. Whatever it is, I’ll be anxiously awaiting the manual and schematics to see what Mr. Fujita has come up with!

If you’re headed to the LA Audio Show (June 2nd-4th), the amp will be on display in the VK Music booth.

UPDATE 9/15/17: Here’s the first review I’ve seen, courtesy of Wall of Sound

Speaker sensitivity ratings and amplifier power

10 * log (power) = decibel

10 ^ (decibel / 10) = power

10db increase (10x the power) is perceived as twice as loud

Most desktop-size speakers are in the mid 80s db/W @ 1m sensitivity wise. We’ll call it 85db for the sake of calculating stuff. The sensitivity rating means that with one watt of power, you’ll get 85db of sound at one meter away. For reference, 80db is pretty loud. It’s about the level of a running garbage disposal or an alarm clock. You can listen at 85db for eight hours before you start risking hearing loss; this is also the sound level at which OSHA will fuck your shit up.

For nearfield listening, there may be less than a meter between you and the speakers. If you halve the distance, you can add 6db to the sensitivity rating. Now with the same speakers you’re getting 91db at half a meter with one watt of power. You should probably turn it down a touch to protect your hearing (2 hours at 91db is the maximum recommended duration). Every halving of the power deducts 3db, so one quarter of the power (0.25W) gets you back down 6db to a non-litigious 85db. If you want to listen at 80db (which is comfortably loud, believe me) you only need around 100mW.

Aren’t decibels fun?

This goes to say that you do not need a whole bunch of power for nearfield listening, even if the speakers have a low sensitivity rating. And if you have high sensitivity speakers in your “main rig”, a single-ended low wattage amplifier works there, too. Say you have speakers rated at 95 db/W @ 1m and like to listen around 80db. If you listen at one meter, you only need 32mW. If you listen at two meters, you need just 125mW of power.

The above discussion of power and decibels does not take into account dynamic headroom. It’s always good to have some power in reserve for music dynamics. Or for cranking it when OSHA isn’t paying attention. I try to have at least 10db to spare (10x the power) over what I expect my average listening levels to be. If you didn’t fall asleep while I fapped around with decibels and logarithmic math, you noticed that average, safe listening levels (80-85db) need only a fraction of a watt with average sensitivity speakers nearfield or high sensitivity speakers at a regular distance.  Ten times more power is just a couple watts and will often get you pretty comfortable listening levels with headroom to spare.

Just make them some high quality watts.

Back at it finally!  This is an excerpt from the first speaker amp write-up I’m doing for the site.  Happy Cinco de Mayo!

Something for beginners

Pete Millett’s Starving Student was one of the first amps I ever built completely from scratch. Unfortunately, the 19J6 tubes have become rare (or at least no longer dirt cheap) due to all the bright eyed DIYers scooping them up to build amps. I think the world needs another <50V tube amp for beginners, so I’m designing one. Like the original, it’s an oddball tube with a MOSFET buffer and an off-the-shelf power brick (same brick, in fact).

Millett is one of my personal tube heroes. This is a tribute.  Full write up coming soon (and parts values subject to change once tested).

 

I love you litte Weller choo choo

IMG_20170110_195626453.jpg

Sometimes people ask about how much money they need to invest to start playing around with DIY tubes and audio. The answer is probably less that you think.

I have been using the puny 23W Weller pencil iron for several years. You can buy this thing at literally every hardware store on the face of the earth. The tip is replaceable. Sometimes the little iron struggles with larger buss bars or binding posts, but its never given up. It’s my little orange choo choo iron.

Why am I thinking about this now?  I bought a 45W iron recently and immediately found it to be almost too hot for small sockets and parts. Little choo choo is not going to the tool box graveyard yet.