The amp on my shelf

Decent Amp for my Office


A Gain Clone Project
Fred M. Niell, III
9/11/2007
fred.niell@gmail.com
www.niell.org

[Home][Research]


Poor old stereo
The old broken amp
Nice speakers
Click for larger. The newly re-tweetered speakers
Why build an amp specifically for your office?

I have always loved analog audio projects, and have made several amps in the past, including some vacuum tube amps and a handful of solid-state amps based on chips from Apex Microtech. For a long time, I had a nice little JVC amp that was given to me in 1993 as a christmas gift for my office. The sound was great, and I eventually upgraded the speakers. The amp and the new speaker combo was great for years. But then there was some hissing from the little JVC amp. It was very quiet, and I simply assumed that it was a leaky capacitor or something causing some hissing. Actually, it was the little amp oscillating at 27kHz. At about 40W. Ouch. The tweeters went out, and the amp started sounding awful. So I had an impetus to do something.

I bought some new tweeters for my speakers (seen at left) to replace the old ones, and started to diagnose the problem with the little JVC UX-T1 amplifier. It turns out that the TD-series amp chip was oscillating all by itself due to some kind of internal fault (it oscillated when taken out of the system and powered up alone). So... time to ditch it for something else. I have been reading about gain clone amps for a while, and picked up some LM3886 amp chips from National a while ago. They sat languishing in my junk box for about 3 years while I waited for an excuse to do something with them.

Now that I had something to do, I read about the various things people have tried, and formulated my own plan:

Single chip non-inverting design
Shared heat-sink
Fixed gain
Dual Switching power supplies for DC Bus
Star grounding
Automatic anti-thump startup from mute RC
Point-to-point wiring
Stereo 2-channel amp
High pass rolloff at 10Hz, unity gain at DC
Stabilization pole on output


Office Amp Schematic
Click for larger. This is the Office Amp Schematic.
The bill of materials for this project is pathetically small, but here it is anyway.
Hardware Design

Anyone who's looked at these gainclone amp projects has seen that there are only a few basic variations on a general theme: inverting or not, rolloff at some defined frequency, exotic capacitors for various things, etc. All gain clones use the very nice single-chip solutions that National or TI have. The National chips are very nice, and I've used them before in a few designs. The Overture Series Amplifiers all use SPiKe protection, which is great in an uncontrolled audio signal environment, but sounds like crap if it actually kicks in. So, just design the amp to operate well within the SOA and within the limits of the SPiKe circuitry.

I decided to use a +/-20V supply, since I had a pair of 16V 60W switching supplies sitting around that I could tune up to 20V. Also, I knew I was driving a pair of 4 Ω speakers, so I didn't need much of a bus voltage to make a few watts. My computer's sound card wants to drive a few k, so I chose 3.4k input impedance. I wanted to get about 30W/channel, which is more than I needed for the office. The caps are all ultra-low ESR types. I chose to include a π filter on the input with a damper resistor to fix any oscillations at the pole. The point there was to add filtration to the somewhat noisy switching supply. This worked perfectly, as with no input loading impedance, the noise output is around 4mVpp centered around 120Hz.


Point-to-point wired office amp
Click for larger. Here's the Office Amp sitting on my shelf
Underside of the amp
Click for larger. This is the underside of the amp
Closeup of one channel
Click for larger. Closeup of 1 channel point-to-point wiring
Closeup of the input and star grounding
Click for larger. On the left you can see the input stereo jack and the star ground stud on the right. The 1 ohm resistors provide some impedance between the ground of the input jack and the ground of the chassis
Closeup of the DC filter bank
Click for larger. Closeup of the DC Filter bank
Hardware Photos

Here's some photos of the hardware. The first photo shows the amp as it sits on the shelf. These amp chips are actually plenty efficient to operate without a heatsink at office-volume levels. In fact, when everyone is gone and I'm left alone in the office, I can turn it up to clipping and run there without any heatsink for more than an hour with only minor heating of the top plate.

The schematic implementation is really just a version of the example shown on the datasheet for the LM3886. The only differences are really in the execution. The point-to-point wiring is not particularly neat, but the signal and power lines are kept separate, and a strict star-ground was wired.

The amp is very, very quiet with no signal, and only a faint 120Hz ripple is left. Since these are low-impedance, not particularly efficient speakers, the remaining 4mV ripple on the output is basically undetectable.

The construction of the amp was done in about 3 hours using crap we had lying about in the lab. The binding posts and project box top came from the junk box, and the only thing that cost any money at all were the LM3886 chips, which I brought from home anyway. Considering the time invested and the great sound I ended up with, I declare this project a great success. The sound is perfect for the office. I was struck at how tight the bass is, and how good the stereo separation is. I see more than 60dB separation, measurement limited by the scope probe. The noise figure of the zero-signal response is dominated by a 120Hz ramp that comes from the power supply. I had expected that the PSRR of the amp chip would have taken care of that, but in fact it didn't. I suppose that if I had bothered to dig up some larger filter caps, I could have taken care of that, but I can't really even hear a 4mVpp ripple in the office anyway.

Here's some scope shots of the amp in operation:

Channel separation experiment
Click for larger. This is 4.8mVpp pink noise out of the left channel with the right channel driven to clipping into 4 Ω
Noise figure experiment
Click for larger. This is 5mVpp almost-white noise (little 120Hz in there) out of the left channel with the right channel quiet
1kHz
Click for larger. 1kHz, gain = 20.2
1kHz XY Plot
Click for larger. 1kHz, gain = 20.2, note the phase envelope created by the 40Hz rolloff pole. This is sharply smaller with the 20Hz rolloff improvement.

I originally intentionally added a 40Hz rolloff for the amp with only 1k and 10uF in the DC null stage because the speakers should more or less be silent at that point. But, I realized after running for a bit that this was a little aggressive. The speakers can do 40Hz with only 3dB down, so I should extend the amp down farther. That's why the schematic calls for 22uF caps instead of the original 10uF. With 22uF, the amp makes room-shaking bass, taking advantage of the low frequency extension of the speaker design.

Design, text, photos, etc. copyright 9/2007 Fred Niell