| johnferrier |
My "Boutique" Amp is a discrete class A design using super-matched complementary JFETs. Meaning devices matched closer than dual devices. I have no offset adjustment potentiometer and am not using a servo. However, because of careful matching the offsets are well below 1mV and as the batteries discharge to the dispose level the offset creeps up to 2mV. The only capacitors used (all polypropylene) are across the power supplies, made up of eight 9V alkaline batteries. The ground and power planes are constructed of 0.032” sheet copper insulated with electrical grade Teflon. All signal lines are air wired with silver (17AWG in most cases). Active devices are heat sunk with copper refrigerator tubing and solid 0AWG copper wire. Silver impregnated epoxy is used to permanently glue devices into heat sinks and for electrical connections to ground/power planes (a 350 watt soldering iron just didn’t work...but did that sheet copper get hot : ). Besides copper and silver, there is a lot of steel: the enclosure and a secondary lining. The top and bottom covers include vent holes (offset to permit only high angle penetration) that allow for a bit of a chimney effect. The amp weighs in at a hefty 10 pounds. I admit the excess (including the use of wires doubled back on themselves to avoid "cable directionality"...but that is another story).
An inside picture:
This design is a derivative of Erno Borbely’s Super Buffer. I departed from his design after being encouraged to try removing global feedback (credit Charles Hansen). I really hate listening tests, but I liked to listen to what I heard on the breadboard with global feedback removed. The immediate effect for me was that the music lay softer. It was less tense and less electronic sounding. Also, it seems to not have a problem with sibilance (maybe I'm saying the same thing another way). I'm not an expert at all so can't really provide much information, but an easy way to try an amp without global feedback is to not include the output stage in the feedback loop. I remember seeing a post of a two opamp headphone amplifier with the BUF634 outside the FB loop. Maybe that person had in mind to isolate the transducer load from the feedback loop.
I'm interested in finding somewhere I can have some measurements made: THD profiles and low level noise. If time permits, I'll try Magnolia HiFi or perhaps one of the local audio equipment mfgs. Because of another thread I was involved with, I’m bent on finding someplace with an Audio Precision System Two Cascade.
I'm glad to now be on the listening side of this project.
(This is a slightly altered version of my head-fi.org post…)
P.S. Class A is a great way to burn through batteries. I expect 24 hours on eight 9Vs. And I really like listening to this. I’m now considering external AC to DC converters…
JF |
|
|
| johnferrier |
Yep, saw that on another forum...and at $4.28 a piece at Digi-Key...it's certainly a very integrated world. I built a discrete amplifier, expecting if I didn't now, I probably never would.
However, even after reviewing the newest ICs with extremely low THD numbers, I believe that discretes offer lower noise (though in both cases likely inaudible). I haven't identified calibrated equipment to measure low level noise, so largely base my understanding of this using Borbely's number of 260nVrms (input referred) over the audio range for the Super Buffer. The one thing about noise, versus THD profiles (keep in mind some people's thinking about tube equipment), is that people agree that it should be low.
Also, FWIW with a 15mA supply current, it is not likely "pure" class A.
JF |
|
|
| jackinnj |
yeah, I ordered some samples this morning -- my problem is that I have been using Stax Lambda's for about 10 years, so I guess I'll have to go out and get some new Sennheiser's.
the Stax were purchased in an attempt to "ensure domestic tranquility". They use a transformer, however, and it doesn't work that well with my current setup.
Jack |
|
|
| johnferrier |
I've got several OPA627s, OPA637s, BUF634s, AD8106s, AD797s (may use some of these for low noise regulators), but still decided to go with discrete. Besides low noise, another idea that appealed to me is having a low (< 3) 10khz/1khz THD ratio. The TPA6120A2 looks okay in that department...
JF |
|
|
| tpenguin |
That looks like a very nice headphone amp. I'm interested in your construction methods. Could you post a close up picture of one of the JFETs and perhaps a schematic? What specific devices did you use? Exactly how tightly matched are the FETs, and about how many did you have to buy to get enough matched devices? I've never done any device matching before, so about how much time does it take to get a decent number of tightly matched devices? Are there noticable sonic benefits (or have you compared) between silver air-wiring and using a normal PCB?
Regarding the battery life, have you considered using rechargeable SLA batteries? With a few Ah, you could have a lot of listening time between recharges, and it would be fairly economical. It would also keep your amp completely isolated from the AC mains. |
|
|
| johnferrier |
| quote: | Originally posted by tpenguin
That looks like a very nice headphone amp. I'm interested in your construction methods. Could you post a close up picture of one of the JFETs and perhaps a schematic? What specific devices did you use? Exactly how tightly matched are the FETs, and about how many did you have to buy to get enough matched devices? I've never done any device matching before, so about how much time does it take to get a decent number of tightly matched devices? Are there noticable sonic benefits (or have you compared) between silver air-wiring and using a normal PCB?
Regarding the battery life, have you considered using rechargeable SLA batteries? With a few Ah, you could have a lot of listening time between recharges, and it would be fairly economical. It would also keep your amp completely isolated from the AC mains. |
From memory:
2sk170/2sj74
<2% idss (for parallel devices)
quantity: ~100 (That was always my question.)
Not much time required. I used tweezers to minimize transfer of body heat.
Air-wired silver: HUGE difference...kidding...uhm, I only did it because "technically" it would have lower resistance and higher isolation.
Noticed that for lowest offset that the p-channel and n-channel FETs match together differently. Example: and I may have this backwards, if the idss of the 2sk170 was 8.02 mA then the best match of the 2sj74s was 8.15 mA.
JF |
|
|
| johnferrier |
No noise or THD measurement yet. I have measured the frequency response: ~1hz-500khz…
And here is a schematic. It should be accurate (someone correct me if the battery symbol is reversed…hehehe…).
Besides the 22uF polypropylene caps, there are 0.1uF polypropylenes in parallel supplying each of JFETs (not shown on schematic…). Note: the capacitors are in the "right" part of the circuitry...heheheh... : ).
The 1k/333 ohm resistors are necessary because the circuit otherwise oscillated with the volume at the lowest levels. The two values differ because the input capacitance of the JFETs differ…The Elma stepped attenuators are not make before brake (and that makes sense to me) so the 1M ohm resistor.
The 2sk170/2sj74s are low noise parts (and low noise is the salient feature of this design). Paralleling four of them ought to divide the noise by two and, of course, provide more class A output. My circuit is running at ~35-40mA.
I enjoy listening to it a lot (and need to make another trip to Costco to purchase more eight packs of 9V batteries...my low noise power source...).
JF |
|
|
|
