dsavitsk

I'll add a little clarification to Ryan's post. His amp is a Moth 2A3/45 with a 6J5 driver (the 6J5 being 1/2 of a 6SN7). The driver is direct coupled to the output tube meaning that the voltage on the plate of the driver sets the grid bias for the output tube. This isn't necessarilly my favorite topology (every design has trade offs, but some of the trade offs in this direct coupled design are ones I wouldn't necessarilly make), but rather than tear it all up we decided to try to get as much out of it as possible. As Ryan said, most of the work was on the drivers. On Ryan's, the drivers were biased at ~-3.5V and ran at ~4mA which put the plates at about 115V. The current seemed low to me as I think the 6J5 needs a bit more current to sound its best, and rp drops a bunch with just a few more mA. But, increasing the current changes the plate voltage which changes the output tube bias which all has to be accounted for. So, we also lowered the bias on the driver. By lowering this to 1.8V and increasing the current to ~7mA, we got the plate at about the same spot. Also, 1.8V is conveniently the voltage dropped by a red LED which we used for bias replacing the resistor and cap which is a good thing. The LED provides more or less fixed bias, very low impedance, and lower distortion than a standard capacitor. We also used a CCS to set the current. For a parts outlay of about $15, these two changes effectively get 2 (electrolytic) capacitors out of the signal path and increase how linearly the tube performs. The downside is that since the bias on the driver is lower, it probably can't drive the output tube to full output, but this is hardly an issue with headphones. For now, we added some bypasses to the output tube's cathode bypass caps. We are talking about maybe changing to ultrapaths at some point, or something else, but it is a weak part of the amp. There are also a few other modifications such as snubbers to the filament supplies, etc.

Am I the only one to whom this is new? Parts Connexion

Look at that -- it's hard to tell just how much voltage it needs across it at 250mA, but it looks promising. I am, for now, just using LM317 based CCSes (this is for the L'espressivo 3 filament supply). I had a trimmer in there for adjusting current, but the heat was causing the resistance to change, and they were turning themselves off after about 20 minutes of use. The build is such that adjusting them while on was problematic. I tried this circuit diyAudio Forums but it ended up not working at all, or rather, not working as expected -- which I am still a bit confused about. Colin thinks, I think, that the transistor cannot drive the mosfet properly as the circuit simmed properly. I replaced the trimmers in the originals (LM317 based) with some fixed resistors and all seems OK now -- but a better CCS would certainly be a good thing.

Me either. I made up those boards and then got swamped with work. The intention has been to use it as a driver with a phase splitting choke for a PP amp. I have the tubes, the OPTs, the chokes, and just about everything else necessary for the project except for the time to do it. Maybe this summer ... For those who have not seen it (it isn't linked from my site for some reason), the D3a project is an experimental breadboard for experimenting with lots of different circuits. Details are at ecp.cc

I'm late to my own party More marketing than reality. A CCS, if it's any good, should not add any "transistory" colorations, but instead just keeps the power supply caps out of the signal loop, thus removing "capacitory" colorations which are far worse offenders. The IXYS parts are the 10M45 and the 10M90 for higher voltage. The DN2540 is a Supertex part. They are essentially the same, though the biasing resistor values will differ.

Now that this thread is thoroughly off topic ... do you have any thoughts on where to get a quad, or even a pair, of unmatched LU1014D's? I am looking for a decent high current low dropout CCS to use for a filament supply. None of the depletion mode mosfets that I can find are high enough current (I need ~250mA), and everything else I have tried has either not worked, or worked poorly for one reason or another.

I appreciate the offer. I probably have enough close matches, but if I don't, I'll take you up on it.

DC offset. Depending upon the OPT used, the offset between sides has to be really small. For instance, I am working on the version below as an IV to a differential NOS AD1865N DAC. Talking with K&K, offset between sides needs to be in the less than 1mV range, with lower being better. I went through 100 2sk170's, and while I found reasonably match parts, I am still going to need trimmers to be able to make adjustments to get things in that range making choke loading impractical. In fact, I may need to pick up more fets to find better matches. With the Cinemag on the headphone driver, offset issues are slightly relaxed, but not really by much -- it is still a fully interleaved 80% nickel core. More than a few microamps of current will still saturate it. However, barring the offset issue, I wonder if dual chokes on the same core might work for both sides ... By the way, the topology is in use here: http://www.diyaudio.com/forums/showthread.php?s=&threadid=141865 TDA1543's, cheap OPT's (Cinemags that cost $18) and it sounds great -- one of my favorite things I've built.

I'm sure it would work great, but I hate hunting down obscure solid state parts. Someone should try it, though. So far, my experiments with the topology have been really promising.

Here's another variation I am in the middle of:

For cheap and good sound with Grados, I think you'd be hard pressed to beat a simple MOSFET follower. Even with top quality parts it will be cheaper and dare I say better sounding, than any of the Millett Hybrids. http://www.ecp.cc/SZ.html

The RS-1s are a good option since you won't need any gain to power them from your DAC, just a buffer. Here is some detail on what they need. Why don't you just build something? A couple of back to back source followers into a 150:600 transformer will sound great, and should match the character of the DAC pretty well.

Just for the record, these are not topologically similar to the DNA, nor to each other. The DNA is a standard transformer coupled single ended design. The Wheatfield is a cathode follower that is cap coupled on the output. The Decware has a single ended driver stage and a push-pull output cap coupled to an impedance matching transformer. I've no doubt that the DNA sounds better than either of those based on just a peek at the schematics.

I think that's too bad. All someone will need is a sine wave and a multimeter to uncover this great secret, and it would be easier if you just said what you used. I guess the issue is that, unless you have tremendous core and copper losses, the numbers just don't seem to add up. But, maybe I'm missing something -- wouldn't be the first time I know you are a speaker designer, so the emphasis on matching to the speaker is reasonable. Maybe you are on to something. But, matching the load to the tube also matters, and with what little we know, this seems to not be optimal here.

I'm not much of an OTL guy, so I really know very little about how to optimize a circuit. But, I guess you can keep paralleling tubes and keep adding feedback and whatnot to get the Zout sufficiently low. As for transformer coupling, there are probably practical limits as to the load that can be presented as too high of an impedance leads to high frequency issues. But again, you can keep paralleling tubes to lower the Z out or use very low rp tubes. You might need very high mu tubes to get sufficient voltage gain to overcome the impedance ration, which means using the super high Gm stuff like the 7788 or 437a, or a driver. Also, an autoformer can change that a little as the winding is not as capacitive. As an example, the L'espressivo uses a 71a which has a plate resistance of a little under 2K. On the lowest setting (which is actually plenty loud in most cases) the impedance ratio is 1:16000 which results in an impedance of about 1/8th of an ohm. If you used a 300B it would be less than half of that. One might use a similar design but with a fixed autotransformer, a standard pot and a driver tube to drive higher Z phones, or to get the same damping with higher output. The particular autoformers I used are limited in voltage handling due to their small size, so you'd need a bigger core.

The 120 is supposed to be 6db, the 28 less. An 8:1 is an impedance ratio of 64:1. This gets you a Z out in about the right range but a voltage gain that is a bit high. The problem here is that with an 8:1, the reflected load from low Z phones is too low for the tube -- you'll get more power but you'll also get higher distortion. For the 120, a 4:1 is about right, but again, the load on the tube is too low. My guess would be that the changing sound with different Z settings has more to do with the load the tube is seeing than the interaction of impedance and phones. Maybe the distortion is all 2nd harmonic, and maybe the designer likes this. Or, maybe I am totally wrong that damping factor matters and I have been barking up the wrong tree for years. But, when someone shows up and says that, notwithstanding that fact that in nearly every case nearly everybody thinks lower Zout means better sound (with the exception noted above), that he thinks the opposite is true, I think there is reason to be skeptical. This is easy enough to test, or course, with an opamp and some resistors. I'd also like to see some distortion numbers from his amp with various loads.

My primary concern is 32 ohm Grados which sound best with a large damping factor. But, be that as it may, I am not sure your numbers add up here. With an rp of about 1500 and a mu of 20, what possible transformer ratio gets you less than 6db of gain and a Zout of 28? As for the overall design, I've built that amp (down to the JJ cap, the cerafine bypasses, the Edcor OPTs and the 5AR4 -- I think I used 6N6p's instead of the 6H30 ... basically the same tube) and it is fine, but there is plenty of room for easy improvement. I wouldn't take much from this one data point. Sounds more like a transformer with a too-low primary inductance for the load. ASL is not known for their parts quality, and it is doubtful that they invested in high quality OPTs for this.

You don't get a lot for $1200 these days. It is the single feed version of the now apparently discontinued parafeed Hagtech headphone amp (which I seem to recall sold for $1200 ... or rather, didn't sell for $1200), with a more rudimentary power supply. It's as basic as can be. I am all in favor of supporting local small shops, but they've got to try a little harder. There are a bunch of things that could be done with this circuit to improve it for only marginally more money. Also, either this is only meant for high impedance phones, or the load on the tube is awfully low.

There is some argument that DHT's have lower distortion than IDHT's. However, a small signal DHT preamp requires considerably more work and attention to detail than does an IDHT preamp, so it may be that were an IDHT preamp built with the same attention to detail, it would sound just as good. Also, people building with DHTs tend to go all out on parts, so again, perhaps building an IDHT preamp with the same attention to parts would sound just as good. Additionally, there is some contention that the big open sound people seem to like from DHTs is actually just a microphonic artifact from the filament resulting in a sort of reverb effect. Thus, the sound may be a sort of effect added to the signal. My own DHT headphone amp sounds pretty good to me -- certainly better than most other amps I've built. But, I can't say that this isn't because of the attention to detail, the parts, or the microphonics. On the other hand, it could be the tube.

I'd add outside of New Glarus, WI. Beautiful rolling hills, near enough to Madison for college town stuff and some city perks, near enough Chicago for actual city perks, lots of local organic farms, excellent beer, and cost of living is low.

I'm not really unhapppy. Had I spent another $100 I might be. The highs are not too bad, but there is a mid bass emphasis. The analogy was to suggest that there is a sort of flatness to them -- they seem to lack a little life. They are getting better with some break in, however, and at times sound quite good. I should also say that as closed phones, I was not expecting them to sound as good as the Grados. Maybe the better analogy is that the Grados are to Oris horns what the ATs are to B&W 803's. I really need some noise blocking in both directions (I share office space and don't want to annoy others, and don't want to listen to them either), so they are doing that part way better than the Grados could. However, based only on sound quality, for about the same price I think my wife's SR325's sound quite a bit better and I would be disappointed with these as my main phones. Maybe I just like the Grado sound. One other thought has occured to me which is that I always felt like the RS-1s were a little over priced (and I should say, I got mine during the Headroom/TTVJ spat when they were on a big sale). Now, I have not heard other phones that I like nearly as much (though comparisons do not go much beyond HD600's) but they did seem expensive for what they are. I think these AT's may have given me a new perspective on that; the Grados are actually a decent value in comparison.

I'm no wood worker and have only used a few different types, but for finish, I sand with up to 600 grit and then use 6 to 12 layers of good quality oil. For everything I have tried this on, walnut, padauk, maple, mahogany, it leaves the wood smooth and pleasant to touch, but still feeling like wood. I can't stand the plastic coats people use.

These showed up -- they appear to be real: real wood, well made, etc. I plugged them in and let them run for a little while, so their character may change a bit. Here's my first impression: is to what is to Oh, and they smell awful. I had to put all the packaging in a closet as the odor of industrial solvents is nauseating.

I got them from here: BuyDig.com, The Internet's Digital Superstore which was the best deal I could find (~$225). They seem legit, though the phones no longer appear on their site -- maybe I got the last pair? Their website is identical to BeachCamera.com (other than the name of course) although they cost something like $3 less at BuyDig and interestingly, Beach Camera also does not have these in stock now while they did last night. They sent me a FedEx tracking number, so I should have them by Friday or Monday. I suppose I can post a picture when they arrive and someone can tell me if they think they are fake. Any telltale signs of fakes?

Thanks again guys -- I decided that the ESW9A's looked like the best bet. Plus, since they are made from padauk, they'll match my padauk amp.