L'Fake - A Cheap- A** Expedient to Replace the Elusive SIT

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
OK, Michael Rothacher has an excellent thread detailing a series of SE amps made using a Yamaha power JFET/SIT. What if you're too lazy/cussed/cheap to pony up for said part?

There's hope, but not if you're too lazy (the folks that inquired regarding a kit for the Delite ( (one active part. for chrissakes!), should just buzz on, as they'll get no sympathy here. This is a DIY thread in the purest sense of the word.

At any rate, this thread will talk about a simple amp (3 active parts + some passives) that makes a SIT-ish amp using readily available power mosfets. There may be a preferred part for this particular application, but I'm too lazy to try and find it, so I'll make do with what I have on hand, as should the rest of you. You'll need a big N-Channel part of some sort in a TO-247 package. Lord knows there's enough of those out in the world...
I'll start with a really simple Delite-ish amp using the classic light bulb drain load as detailed by Papa Pass in his Zen series and the Delite amp introduced at the Burning Amp festival a few years back. I'm not going to even talk about the amp circuitry just yet, just the light bulb load.

I'm a path-of-least-resistance person, if you get my drift, so the first place I looked for an appropriate light bulb load was the local hardware store - in this case, Orchard Supply, a local (N. California) hardware store chain that's been around long before the other Big-Box stores reared their ugly heads. Right on the shelves there was a 120V/300W clear light bulb with standard Edison base made by Feit Electric (home of cheap (yep)-and-cheerful??? lighting products from China). The part number is 300M - I assume that the M might stand for "mogul" socket. At any rate, they were there, and reasonably cheap, so I nabbed four of them.

At home, I hooked one up to a DC bench supply. At 25V input, the bulb passes a steady-state current of ~1A, at which point the filament is glowing a cheap-and- cheerful orange (this means long, long bulb life). 1A is a reasonable starting bias current for a MOSFET class A SE amp. For a symmetrical swing capability, that means you would need a 50V supply, and you would set the mosfet to drop 25V at 1A. This is 25W dissipation, or 50W dissipation for 2 channels, not too bad for a reasonably big heat sink sans fan.
If you wanted more bias current, you could use more supply voltage, or use two bulbs in parallel and adjust the total supply voltage to taste. For an unknown bulb, you'd do a similar exercise with a bench supply to figure out the necessary supply voltage for a given bias current. For European types, this is likely to work better for 115V and lower bulbs than for the standard 230V bulbs available there - be warned. You may need to use foreign bulbs with a 115/120V rating or higher wattage bulbs - some have tried the linear high-wattage Quartz-Iodine bulbs. Automotive bulbs in series might be another option.

At any rate, you're on your own, but light bulbs are cheap, so potter around until you find a combination that will yield a reasonable bias current for a reasonable voltage. Assume that the bulb/s will drop ~1/2 the supply voltage, and also assume somewhere between 1-2A bias current. And no, I'm not going to find your "sweet spot" for you - that's your problem. Go away and do some homework - I won't be back for a while.
 
Last edited:
Been doing some thinking since the first post and two paralleled light bulbs and a ~40V supply might yield the most power for a simple setup. This would be at the cost of ~40W dissipation in yer mosfet of choice, so you'll need one with reasonably large die area for a good junction-to-case thermal resistance. More on that theme later...

If you do a Google search for "Feit Electric 300M", there are loads of hits - you can even buy the blasted things from Amazon...
 
Here's a prelim schematic that I'll most likely be firing up this weekend. There are 4 active devices rather than 3, as I used a ring-of-two current source to bias the J175 jfet rather than a Supertex DN2530N3 depletion mode mosfet, my usual choice. I've been having a lot of trouble with bad Supertex parts lately - I suspect it's because I've been using an ungrounded cheap-a** soldering iron in my upstairs assembly location. I just purchased a new Hakko soldering station (the venerable "888") with proper grounding, so this may fix the problem - more on that later. The Fairchild HUF***** mosfet I'm using in the output stage was from an old Goldmine Electronics buy and is not critical to the operation of the amp - remember this circuit is based on feedback. Possible substitutions are many - IRFP140, IRFP240, IRFP044A, IRFP150, IRFP250, IRFP350, IRFP360, etc. I suspect that most any mosfet in a TO-247 package with voltage rating of 500V or less and >50V will have a low enough junction to case thermal resistance to properly dissipate the power required in this design.

That's enough for now...
 

Attachments

  • L'fake1.gif
    L'fake1.gif
    17.5 KB · Views: 2,701
A valid point, Juma. A more reasonable gain with some negative feedback is 5-6, or 7-8 dB if you prefer. I'm lucky I have a line stage with gain to drive this thing. In the course of looking at the circuit further, I found other things to hate about it, mostly as a consequence of the resistive drain loading (how do I hate thee, let me count the ways...).

As a result, I'll only have the light bulb loaded incarnation of the L'Fake alive long enough to light up some bulbs with a bench supply and drive some speakers, just to say that I've done it (with a trophy photo, of course). I already have places on the boards for a front gain stage that helps the simulated characteristics quite a bit (more gain/lower distortion, while still yielding an attractive harmonic structure). An outboard current source load is easy to add. I'll post more about this as time permits.

An attractive/interesting variation on SIT emulation is the "Half-Nelson" amp built/discussed back in 2007 (search out the thread - it was a Schade feedback scheme/SIT emulation before I knew about the concept. It got its first public appearance at the first Burning Amp in 2007, driven by a bench supply.

I just finished up with revamping that amp in the basement - the last "improved" incarnation had some residual hum, which I managed to vanquish by moving the filter caps around. The DC fan cooling the heat sink was replaced with a 230V AC fan running at 115V (slow and quiet, with still enough airflow to keep the heat sink real cool). I'll be carting this puppy to Burning Amp this year along with everything else that I manage to finish. The HN sounded pretty fine down in the basement, driving a pair of bitty Celestion bookshelf speakers with some Hamiett Bluiett from a Discman.
 
Here's "L'Fake Lite" a brief transitional stage in the evolution of the amp, just to show it can be done if you so desire. I'll post a schematic later, as this beast turned out a trifle different from the first schematic. One of the important differences was using an IRFP260 rather that the original Fairchild output device, as the larger area die in the 260 has more transconductance than the original Fairchild fet I show in my first schematic.

With 2 bulbs per side in parallel as drain loads and a 40V Vcc from a bench supply, each output stage was running just under 2A bias current. The amp sounded ok(but not very loud) when driven by a Discman.

The bulb filaments glow a dull red - the digital camera exaggerates their brightness, as it sees more near-infrared than the human eye.

I'll tell you the reasons why I hate this amp (even though it came up without a lot of back-talk) before I move on to the next stage of the project.
 

Attachments

  • fake2.jpg
    fake2.jpg
    127.2 KB · Views: 1,298
Last edited:
Here's a bit more construction detail... I used fender washers (hardware store) to even out the pressure of the screw on the TO-247 output device packages for a little better thermal interface. I also used aluminum oxide washers (Mouser) for good thermal transfer from the output devices to the heat sink. I'll post some details on the heat sink I used when I can get down to the basement again with a tape measure, but you are going to basically be on your own in that respect, anyway. I picked up my heat sinks on Ebay from someone in Ohio that looked to be parting out old motor drives. Use the overall dimensions I post as a rough guide to what might be necessary.

At present, the heat sink is dissipating around 90W. If you bother to adjust the drop across the output devices to the proper level, this should go down somewhat, but expect to dissipate at least 80W for 2 channels. Putting each channel on a separate heat sink will help matters somewhat - that's up to you...

The amp is pretty listenable, even though I detest it from an engineering standpoint (more on that later). Right now, it's being driven by a Panasonic Discman into a pair of Celestion F15 bookshelf speakers, which are what I have in place in the basement for initial listenability tests. Ok, so far... The heat sink gets to about 50-55 degrees C (finger test) over the course of a CD's worth of listening - no doubt it would do better if I propped up the heat sink so that the fins and natural convection could do their thing.

I'm going to post a generalized schematic and some comments so that people who want to build this (or something similar with alternate parts) can do so, but I think I can do better. In fact, I already have - check my "Half Nelson" thread in the Pass Amps forum for one approach. I'm hoping that this thread will reveal an alternate approach, though at the cost of some more parts than what I've shown here so far..
 

Attachments

  • LFL_top.jpg
    LFL_top.jpg
    84.2 KB · Views: 1,197
Last edited:
Here's a schematic to get people started on the "L'Fake Lite"(if they want to start at all). The circuit is simple enough to build up on a piece of perforated board (a good thing - no PCBs, unless someone with more spare time than I have want to gin something up and offer it separately). The circuit is very tweakable, but you'll have to tweak on your own.

I chose Q2 as a mosfet for high output impedance, but a PNP bipolar transistor would also work. A device in the TO-92L or TO-126 package would be best, as the dissipation will be about 150mW with the specified output voltage - a little high for a TO-92 device.

If you decide not to do any selection for Q3, R11 will be necessary to tune in the drain voltage for Q4, which should be around 20V for a 40V supply voltage. Once the quiescent operating voltage is tweaked in, you can measure the pot and replace it with a fixed resistor. Using a "premium" device like the 2SJ74 for Q3 will likely yield inferior results due to high gate capacitance. NOS devices like the J175 or J271 look to b optimal - if there are any euro-equivalents, let us know here, as I don't do Euro-silicon.

I specify the Feit 300M as a load bulb, as it's available on Amazon. If you want to use something else, you'll have to characterize it with a bench supply in the manner I describe in the first post of this thread.
 

Attachments

  • L'Fake_Lite2.gif
    L'Fake_Lite2.gif
    19 KB · Views: 1,153
Last edited:
I've perhaps decided to keep the "L'Fake Lite" for the time being. I 'm building up the final incarnation of L'Fake, and If I can cram both sets of modules on the heat sink, I'll bring the ungainly chimera to this year's Burning Amp with some means of switching between the two amps.
 
It happens that I have a heat sink large enough to hold both the "L'Fake Lite" and "L'Fake" boards. It's also one of the filthiest heat sinks I've ever bought - maybe it came from a chain smoker who lived in a chicken coop (that's right, filth plus yellow stuff)... At any rate, the holes are mostly drilled & tapped, so I'm going to see what some Scotchbrite and orange cleaner/Bon Ami will do before I proceed. The construction will be slightly less rough than the original Nelson Pass "DeLite", but not by much. I'll post pictures, but people who want to experience it in all its sleazy glory will have to come to this year's Burning Amp.
 
The filthy heat hink got cleaned up last night, and I installed the L'Fake Lite and L'Fake boards on it. Attached is a picture of the beast on the bench while one channel is burning in. I'll talk about L'Fake when I've finally finished with L'Fake Lite. As a teaser, though, it's a "Half Nelson" amp with several nagging issues resolved.
 

Attachments

  • lfake+lite.jpg
    lfake+lite.jpg
    259.1 KB · Views: 1,011
After a couple of frustrating moments, both channels of "L'Fake" are up and running and burning in on the new, larger, (and cleaner) heat sink. It's standing up, so I'll see how simple natural convection copes with ~80W output device dissipation (both channels). I Think it'll do fine, from what I've seen (and felt) so far. The output devices are coupled to the heat sink via alumina washers (Mouser) with fender washers on top to even out the pressure from the mounting screw on the device packages. The fender washers can be seen in the picture embedded in the previous post.

For people looking to do this or similar things, the heat sink I'm using is 10" by 12". The web thickness is 1/4 ", and the fins are 2" long. Similar heat sinks aren't too hard to find on E-pay. Also, a little air flow from a DC fan with the input voltage dialed down can work wonders. I don't think it'll be necessary in this case, but people trying to make do with a smaller heat sink or higher power dissipation may want to consider such an option.
 
Last night, I burned in the L'Fake modules on the new heat sink for over an hour, and the equilibrium heat sink temperature appears to be around 40-45 degrees by finger test. How to tell? 40C feels hot, but you can keep your hand on it, 50C is 10 sec uncomfortable, 55C is ~5 sec uncomfortable, and with 60C it's one second or less. 70C is "immediate ouch". Your fingers are a pretty good non-linear thermometer.
So, the heat sink is a little large for dissipating the 80W required (40W per module), but the heat sink had to be kinda big to fit all four modules. Something between 2/3 and 3/4 the area would probably be ok. As only one pair of modules will be running at a given time, I can forget about the thermals and try to figure out how to package this baby with the power supply.
 
When I'm characterizing an amp, I like to whack its input with a 10kHz square wave adjusted in amplitude such that the amp doesn't clip. This gives me an idea of the amp slew rate and also helps to smoke out snarky oscillations and other antisocial behavior.

Attached is the square wave response of one of the channels of L'Fake (not L'Fake Light - I didn't want to drag the light bulbs in to work). There's no compensation cap. The top trace is the input, and the bottom is the output. Slew rate is decent, though much faster on the rising edge than on the falling. There is a bit of overshoot on the rising edge that will be cured with a small compensation cap. I used 470pF on the other channel, and it's heavily overcompensated. I think 100-150pF should do the trick. After I install these, it's time for listening tests with a bench supply.
 

Attachments

  • Lfake_res.gif
    Lfake_res.gif
    14 KB · Views: 763
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.