Linear Systems Releases JFETs Into The Wild

The silicon wizards at Linear Systems have been cooking up a special batch of JFETs in their secret underground laboratory.  The LSJ74, a replacement for the Toshiba 2SJ74, will be sold exclusively to the DIY community through www.diyaudio.com.  This is a nice nod to the audio enthusiasts who have been pushing for these parts, and will no doubt put them through their paces before they are released to OEMs.

Linear Audio Volume Six Out Now!

I've written an article titled "A Concise Model for Static Induction Transistor IV Characteristics", which will appear in Volume 6 of Linear Audio.  Linear Audio is a book-zine published twice yearly by Jan Didden.  Each issue is packed with quality technical articles you're sure to enjoy.  Copies of Linear Audio can be purchased at www.linearaudio.net.

The new article details a new model for SIT characteristics and includes software you can use to make your own models.  I hope you find it useful.

Fun With PASS-SIT-1

FUN WITH PASS-SIT-1

by Michael Rothacher

Unobtanium Obtained

This isn’t really a technical article, and it isn’t really a review, it’s more of a techni-gonzo piece about Nelson Pass’ new Static Induction Transistors.

Okay, here’s the deal.  A couple of months ago I wrote an article about a really simple SIT amplifier that uses Sony 2SK82s and light bulbs.  I did this because curiosity got the better of me after reading that Pass had his own special batch of Static Induction Transistors custom made for him by SemiSouth.  I really didn’t know much about Static Induction Transistors, but I figured if Nelson was spending his Ferrari money on them they must be pretty sweet.

I built up a few amps with the Sony parts and was pleased to find they performed well and sounded really good.   Writing-up the project was fun and educational, but things got a lot more interesting when Nelson offered to send me a pair of his PASS-SIT-1 transistors.

Let me type that again.

Nelson Pass offered to send me a pair of his PASS-SIT-1 transistors.

Maybe Nelson bumped his head or something; I don’t know why he’d make me this offer, but believe me, I don’t question it.  For me, this was like George Lucas inviting a random fan to Skywalker Ranch to preview one of the Star Wars films.  And I don’t mean Phantom Menace.

A few weeks later they arrived in my mailbox and I went to work.  Here’s a picture of me holding one.

The anti-static gloves probably weren’t necessary, but I thought perhaps they shouldn’t be handled by human hands.

The PASS-SIT-1 is a silicon carbide static induction transistor.  It shouldn’t be confused with other SemiSouth power JFETs Pass has used in his designs.  The PASS-SIT-1 has a characteristic curve which looks a lot like a triode vacuum tube and has been scaled to work specifically with speaker loads.  Here’s what it looks like.

The first thing I did was build up a couple of modules for testing.  These allowed me to quickly switch the SITs in and out of different circuits without un-mounting or de-soldering; after all, these things are rare unobtanium.  Here’s a picture of the modules.

My goal was to try these in a simple amp like L’Amp/DeLite and report my findings to you.  I chose this goal for several reasons:

2. 2.   I’m not very interested in cloning or reverse-engineering.
3. 3.   I really dig the light bulb amps.
4. 4.   It just so happens that these parts are ideal for simple circuits.

Pass sagely realized that last point some time ago; SITs aren’t new, but they’ve generally been used in more complex circuits even though their unique properties make them especially suitable for use in minimalist amplifiers.

The Circuit

I used the same circuit from the first L’Amp article, but I lowered the input resistor value to adjust for my sample SIT’s 100uA gate leakage current.  Everything else is the same.

I’ve gotten a few emails about Gate leakage and what one should do about it, after all, it sounds like an awful thing.  But, being aware of it is half the battle, so it needn’t keep you awake at night.

Gate leakage current is a small current that flows between the Gate and channel.  It is common to JFETs operated at certain values of Vds and Id.

One option is to hand select devices for low leakage current.  This must be done at the current, voltage, and temperature we expect to expose the device to.   This is, no doubt, what Nelson does for his commercial amps.  If we can’t hand select for low leakage current, we can account for it in our circuit, providing it is sufficiently low.

A current flowing toward the gate will drop some voltage across our input resistor, whose value is often 47K Ohms or higher, since we like to have a high input impedance.  Now, if our leakage current is 1uA, the voltage dropped is pretty small (I x R = .000001 x 47000 = .047V).  But, if our leakage current is 100uA, we’re looking at 4.7V.  An easy way to lower this voltage is to lower the value of our input resistor, providing our preamp can drive the lower impedance.  In this case, I’ve lowered the input resistor to 10K, so we’re dropping about a Volt across it.

It’s important to note the polarity of this voltage, because your voltage reference will need to supply the required Vgs, plus the voltage dropped across the input resistor.  For example, my SITs needed about -8V from Gate to Source, so my voltage reference was adjusted to -9V, since -9V + 1V = -8V.  Now perhaps you can see how this could causes biasing problems if you’re not aware of it.

Finding the Sweet Spot

Once I had the channels built, my next task was to find the sweet spot for the PASS-SIT-1s.  I tried a number of different power supply voltages and currents along with different drain and speaker loads.  Here are my findings:

1 300W Bulb, 8 Ohm Speaker

 V(supply) V(ds) I(d) V(bias) V(drain load) 1W THD% 40 13.35 1.18 7.6 26.36 0.259 45 14.86 1.26 7.8 29.7 0.213 50 15.71 1.35 7.9 33.86 0.189 55 17.2 1.42 8.1 37.2 0.175 60 17.95 1.5 8.2 41.4 0.155 65 18.79 1.57 8.3 45.4 0.146 70 19.44 1.66 8.4 49.8 0.133 75 20.37 1.72 8.5 53.7 0.115 80 20.73 1.8 8.5 58.6 0.105

2 300W Bulbs, 8 Ohm Speaker

 V(supply) V(ds) I(d) V(bias) V(drain load) 1W THD% 35 15.14 2 7.6 18.96 0.177 40 17.1 2.15 7.8 21.94 0.144 45 18.6 2.3 7.9 25.22 0.14 50 19.57 2.49 7.9 29.47 0.108 55 20.6 2.65 7.9 33.2 0.098 60 21.7 2.8 7.9 36.9 0.106

2 300W Bulbs, 4 Ohm Speaker

 V(supply) V(ds) I(d) V(bias) V(drain load) 1W THD% 40 12.34 2.36 7.00 26.56 0.264 45 12.97 2.54 7.00 30.72 0.231 50 13.71 2.72 7.00 35 0.201 55 14.42 2.87 7.00 39 0.163 60 15.21 3.03 7.00 43.1 0.148

1 300W Bulb, 12 Ohm Speaker

 V(supply) V(ds) I(d) V(bias) V(drain load) 1W THD% 40 16.14 1.1 8.10 23.22 0.239 45 18.31 1.17 8.40 25.95 0.199 50 19.87 1.25 8.60 29.58 0.176 55 20.61 1.33 8.70 33.5 0.155 60 21.54 1.42 8.80 37.5 0.154

I settled on a 50-55V supply at approximately 2A bias, with two 300W bulbs.  That keeps the dissipation below the recommended 50W maximum and offers good performance with varied loads.

The Measurements

This amp measures really well.  Firstly, the PASS-SIT-1 offers respectable gain at around 17dB.  This is a big improvement over the Sony 2SK82s.  Figure 1 shows distortion vs. power.

FIGURE 1

17dB gain and 1 watt distortion in the vicinity of .2% in a zero-feedback amplifier is pretty impressive.

Figure 2 is the distortion vs. frequency at 1 watt and it’s amazingly flat.

FIGURE 2

Figure 3 is our square wave at 40 kHz.

FIGURE 3

The Sound

So how do these things sound?  Well, first I’ll say I think these are great amplifiers.  In fact, I think they’re among the best amps I’ve ever listened to DIY or commercial.  The usual low-power and speaker matching caveats apply, but beyond that I don’t feel any need to qualify my assessment.

The soundstage is particularly impressive.  The images are life-sized and really seem to occupy physical space in the listening room.  Everything seems vivid and realistic.  This quality is often associated with zero-feedback SE triode amps, and solid state amps rarely pull this off well.  These PASS-SIT-1 amps not only pull it off, they pull it off with sharp focus and great specificity, without the fuzzy edges and image drift that vacuum tubes sometimes impart.

Tonally, I’d place them slightly on the warm side of neutral, which is what I like.  The highs are extended and have a slight sweetness; the lows were nicely taut and tuneful, free from one-note bloat.  They do a beautiful job of rendering timbre with fine detail and bell-like clarity.

Speed is also remarkable.  This isn’t the kind of amp you’d expect to sound fast, but it certainly does.  Attack transients are nicely energetic and have a very real feel.  The excellent speed and soundstaging combine to create the fantastically convincing illusion that the instruments are in the room with you.

Dynamically speaking, they’re terrific within their power limitations, but despite their low power, they don’t seem to collapse into a jumbled mess on complex passages or singe your eardrums on a clip.

All of these qualities make for a very musically engaging and emotional experience.  I could live with these amps for a long time.

So, has Nelson Pass captured lightning in a bottle?

You bet he has.

Outro

The high-end audio industry is ostensibly all about innovation, but in actual fact it has its fair share of taboos and groupthink.  Nelson Pass brings much more to our hobby than his knowledge; he also brings a healthy bit of irreverence and a “think different” outlook that gives us all permission to try crazy stuff even if it bends the rules of conventional wisdom.  I’ve called Pass a paradigm shifter elsewhere and I really think it’s a fitting title; from dynamic bias, to single-ended solid state, to amps with one transistor, Pass does stuff others dismiss and in doing so advances the art.  And now, we have a cool new gain device made just for audio that blurs the line between solid-state and tube sound for real.

The PASS-SIT-1 is currently only available in First Watt products, and I’m sure they’re due to receive a lot of attention.  Like me you’re probably wondering if these parts will ever be released to DIY.  Honestly, I have no idea.  But who knows what tomorrow brings?  We’ll just have to wait and see how things unfold.  In the meantime there are old stock parts available to experiment with and new ones on the horizon.  As for PASS-SIT-1, it can’t hurt to keep your fingers crossed.

I do.

Raspberry Pi for Audiophiles

Here's how to use a high-end DAC and a Raspberry Pi to make an audiophile-quality music player. With this project, you can play high-resolution music files through your stereo from the Raspberry Pi's local disk or from a network share. You could drop multi-kilobucks on a device to do this, but we’ll put together a great sounding rig for less than $150. I’ll present a “minimum-steps, basic implementation” to get you up and running in less than an hour. From there, you can tweak yourself silly if you want to. The Hardware: Raspberry Pi I confess I have a soft spot for things British: Chuchill, Doctor Who, Elite, Jaffa Cakes, and now Raspberry Pi. The Raspberry Pi is a credit card sized computer developed for education. It is available in the US for around$35. Buying one is good Karma.

First you'll need a Raspberry Pi that’s already set up and running with the latest Raspian “Wheezy” disk image. I recommend you have your keyboard and time zone correctly configured, your root file system expanded to allow you to use all the space on your SD card (if you plan to store your music there), and you’ll need to have wi-fi setup if you’re connecting to your network wirelessly.

If you're new to Raspberry Pi, here's a great tutorial on getting started:

The Software: MPD

MPD is a music player daemon for Linux. You can install it on your Pi like this:

$sudo apt-get install mpd You may get some errors like these: Setting up mpd (0.16.7-2) ... [....] Starting Music Player Daemon: mpd listen: bind to '[::1]:6600' failed: Failed to create socket: Address family not supported by protocol (continuing anyway, because binding to '127.0.0.1:6600' succeeded) Failed to load database: Failed to open database file "/var/lib/mpd/tag_cache": No such file or directory Not to worry, everything will be fine momentarily. If you plan to copy music to your SD card, the default location for music files is /var/lib/mpd/music/. You’ll need to fix permissions so members of the audio group (which includes the pi user) can copy files into that directory.$ sudo chmod g+w /var/lib/mpd/music/
$sudo chgrp audio /var/lib/mpd/music/ Once you’ve done that, you may want to copy some music over for testing. You can use a utility like WinSCP or CyberDuck for this. Now, modify your mpd.conf file:$ sudo nano /etc/mpd.conf

And change the string:

to

Reboot

$sudo reboot MPD is now running and you can connect to it from a client application and play music, but we’re not finished yet. The DAC: HiFimeDIY Sabre USB DAC As currently, configured, music will play through the Raspberry Pi’s analog output, which is not audiophile-approved. We want to connect a spiffy USB DAC. Feel free to spend thousands of dollars here, but since you’re reading this, I’m guessing you love a bargain. If you’re the thrifty sort, you may want to try the HifimeDIY USB Sabre DAC. At a mere$42 this is a real gem.

Plug in the DAC, then edit MPD config file.

$sudo nano /etc/mpd.conf Find this section: # An example of an ALSA output: # audio_output { type "alsa" name "My ALSA Device" device "hw:0,0" # optional format "44100:16:2" # optional mixer_device "default" # optional mixer_control "PCM" # optional mixer_index "0" # optional } Comment out all of those lines (enter a # symbol at the start of each line). Now, we'll add two audio_output entries: audio_output { type "alsa" name "HiFimeDIY DAC" device "hw:1,0" format "44100:16:2" } audio_output { type "alsa" name "HiFimeDIY DAC Direct" device "hw:1,0" } Okay, we’re almost there. The Clients: MPC, MPoD MPC is a command line client for MPD, you don't strictly need to install it, but it's handy to have available. Here's how you install it:$ sudo apt-get install mpc

I installed an mpd client called MPod on my iPhone. It makes navigating my music library and playing files super easy.

To configure the client, open settings and tap "Add player manually".

Tap "Update database"

Tap "Refresh local cache"

You'll see two outputs:

Here's why we created two outputs:

Most of my music has been ripped from CD's with Exact Audio Copy, but I do have a few 96khz/24bit flac albums. All my CD quality files play just fine, but the high-resolution flacs have some occasional pops and clicks. This isn't the DAC's fault, it's a software issue. The general consensus is that the current Debian build for Raspberry Pi has a bug either in the ALSA, or USB device drivers. Some very brainy boffins are working hard at fixing this for the next release, but in the meantime, I'm working around it using two audio inputs, one which streams bit-perfect audio directly to the DAC and another which downsamples the audio prior to sending it to the DAC. I just select the correct one depending on what material I'm listening to.

Playing Music From a Network Share

First, on my NAS, I created a user called music, assigned a password to the account, and gave it permissions for a folder called audiophile.

Creating mount directory:

mkdir /mnt/nasmusic
chmod 777 /mnt/nas
$sudo mount –t cifs –o username=music,password=musicpassword //192.168.1.2/audiophile /mnt/nasmusic Now, we'll make it so that our drive gets mounted everytime we boot up:$sudo nano /etc/fstab

Append the following to the /etc/fstab file

Now, we need to tell MPD to look for music on the network share

$sudo nano /etc/mpd.conf Change: music_directory “/var/lib/mpd/music” To: music_directory “/mnt/nas” Reboot. Don't forget to spend at least several hundred dollars on a fancy Mini-to-RCA cable for hooking it all up to your preamp. Actually, I bought a spiffy one from Blue Jeans Cable for around$30.

Have fun.

P.S. If fiddling with all this isn't your cup of tea, there is a terrific distribution out there called The Raspyfi Project which is already configured and ready to go. Just go to www.raspyfi.com, download the image, and follow the configuration instructions. There's lots of information there to help you get started. Thank you, Michelangelo G.

Simulating SITs

NOTE* The model described here has been replaced by an improved version which appears in Linear Audio Volume 6.

Circuit simulation can be pretty handy, but I try to use it in moderation, because I'd rather be in my lab than staring at a computer monitor. However, when working with SIT transistors at \$25 a pop, I can understand why you might want to simulate your circuits before applying real electricity. To my knowledge, there are no readily available spice models for SITs, but with a little jiggery pokery, we can build our own. Here's my current model for the Sony 2SK82:

** 2SK82 KD-33 ************************************************************
*M. ROTHACHER
*--------------------------------------------------
.SUBCKT 2SK82 1 2 3 ; Drain Gate Source
+ PARAMS: MU=4.9140 EX=2.352 KG1=101.25 KP=75.0 KVB=24.0 VCT=7.04 RGI=2MEG
*--------------------------------------------------
E1 7 0 VALUE={V(1,3)/KP*LN(1+EXP(KP*(1/MU+(VCT+V(2,3))/SQRT(KVB+V(1,3)*V(1,3)))))}
RE1 7 0 1G
G1 1 3 VALUE={(PWR(V(7),EX)+PWRS(V(7),EX))/KG1}
RDS 1 3 1G   ; TO AVOID FLOATING NODES
D1 5 2 DX
R1 5 3 {RGI}
.MODEL DX D(IS=1N RS=1 CJO=10PF TT=1N)
.ENDS
*--------------------------------------------------

You'll note this doesn't look like a MOSFET or JFET. This model actually uses Norman Koren's improved model for triodes.

Koren's triode model is a two-equation set:

Looking at the model, we see E1 which is a voltage controlled voltage source:

E1 7 0 VALUE={V(1,3)/KP*LN(1+EXP(KP*(1/MU+(VCT+V(2,3))/SQRT(KVB+V(1,3)*V(1,3)))))}

and G1, which is a voltage controlled current source:

G1 1 3 VALUE={(PWR(V(7),EX)+PWRS(V(7),EX))/KG1}

In the PARAMS section, we set the various parameters used in these equations:

+ PARAMS: MU=4.9140 EX=2.352 KG1=101.25 KP=75.0 KVB=24.0 VCT=7.04

Finding the appropriate values for these parameters is part art and part science. There are many tools available for curve fitting which you will find useful. A lot of trial and error may be necessary, so be prepared to experiment.

D1 and R1, a diode and resistor, model gate leakage, and the value RGI in the PARAMS sets the current:

D1 5 2 DX
R1 5 3 {RGI}
.MODEL DX D(IS=1N RS=1 CJO=10PF TT=1N)

Once I found some curves that matched-up pretty well, I wired up a little circuit in Multisim to generate some IV curves. Here's what my virtual curve tracer looks like:

And here are the virtual curves:

Once I compared them to my real curves and deemed them "close enough for government work", I built up a little virtual amp to check them out. Here's what it looked like running a simulation:

Not too shabby, really. If you use these models, you must be aware that you do so at your own risk, I can't promise they'll make you happy and you're welcome to build on my work if you'd like to improve them. They're not perfect, but few models are. If someone knows the interelectrode capacitances, I'd be happy to put them in.

As a special bonus, here's my model for the unobtanium PASS-SIT-1:

** PASS-SIT-1 ************************************************************
* M. ROTHACHER
*--------------------------------------------------
.SUBCKT PASS1 1 2 3 ; Drain Gate Source
+ PARAMS: MU=19.4880 EX=1.204 KG1=0.7031 KP=81.0 KVB=42.0 VCT=3.616 RGI=1MEG
*--------------------------------------------------
E1 7 0 VALUE={V(1,3)/KP*LN(1+EXP(KP*(1/MU+(VCT+V(2,3))/SQRT(KVB+V(1,3)*V(1,3)))))}
RE1 7 0 1G
G1 1 3 VALUE={(PWR(V(7),EX)+PWRS(V(7),EX))/KG1}
RDS 1 3 1G   ; TO AVOID FLOATING NODES
D1 5 2 DX
R1 5 3 {RGI}
.MODEL DX D(IS=1N RS=1 CJO=10PF TT=1N)
.ENDS
*--------------------------------------------------

Now go have fun, and don't forget to build something real once in awhile.

P.S. We can make a p-channel part by reversing a few connections. Here's a 2SJ28 if you like:

** 2SJ28 ************************************************************
*M. ROTHACHER
*--------------------------------------------------
.SUBCKT 2SJ28 1 2 3 ; Drain Gate Source
+ PARAMS: MU=4.9140 EX=2.352 KG1=101.25 KP=75.0 KVB=24.0 VCT=7.04 RGI=2MEG
*--------------------------------------------------
E1 7 0 VALUE={V(3,1)/KP*LN(1+EXP(KP*(1/MU+(VCT+V(3,2))/SQRT(KVB+V(3,1)*V(3,1)))))}
RE1 7 0 1G
G1 3 1 VALUE={(PWR(V(7),EX)+PWRS(V(7),EX))/KG1}
RDS 1 3 1G   ; TO AVOID FLOATING NODES
D1 2 5 DX
R1 5 3 {RGI}
.MODEL DX D(IS=1N RS=1 CJO=10PF TT=1N)
.ENDS
*--------------------------------------------------