Hi everyone! One of my customers sent me a particularly nice GFA-555 MKI to be refurbished. It’s never been tampered with, and has very few scratches on the case… just beautiful. So I thought I would take lots of pictures along the way as I refurbish it, and show you the latest iteration of my GFA-555 rebuild process. This is the full monty upgrade package including the Hoppe’s Brain soft-start power supply board, with all sound quality and reliability improvements. $1025 in total.
Starting out, it was pretty dusty inside, but everything was intact and no one had attempted any bodgey repairs. The right channel worked fine, and the left had a pair of blown driver transistors. (The ones that drive the big output transistors) GFA-555’s are famously reliable amps, but when they do blow out, this is the most common cause. The originals driver transistors have a breakdown voltage rating that’s only barely sufficient, and sometimes they pop on big transient spikes in the music. I replace them with 250V transistors from On Semiconductor. I suspect these drivers also contribute significantly to the better sound of the finished product.
And after refurbishing…
Improvements to the input board:
- 47uF local power supply bypass capacitors are added to provide a local source loop for high-frequency currents. (The main power supply capacitors are separated from the input circuitry by several inches of wire that necessarily has some inductance.) These bypasses are again bypassed with parallel 0.1uF polypropylene capacitors mounted beneath the board. These handle frequencies too high for the electrolytics.
- Hand-matched input transistors are installed and thermally bonded in heat-shrink for better thermal gain tracking.
- Dale RN55C mil-spec 0.1% tolerance low thermal coefficient resistors are installed in the signal path and feedback loop. The gains of the left and right channels of this particular amp are matched within 0.01db.
- Heat sinks are attached to the four TO-216 transistors to prevent heat stress cracks in their solder joints. (A known issue)
- Bourns sealed trim pots installed
- For the feedback to ground, the original generic cap is replaced with a Nichicon Muse 47uF. This cap exists in parallel with the original silver-mica 1uF, which I leave alone, as it’s a fantastic capacitor.
- Nichicon FG 4.7uF for the bias circuit.
- R22 and C6—the zobel network—are moved off the input board and instead mounted directly to the speaker binding posts. This shortens the total current path length for the zobel by about 6 inches, and thus lowers its total inductance, improving its effectiveness. A separate wire is run from the positive binding post to the negative feedback return, so that the feedback signal does not ride on the same conductor as zobel currents. Adcom made this arrangement standard on the MKII.
- Wire-wrapping posts are removed in favor of directly soldering the wires.
- Teflon signal wires are run to the new Tiffany-style RCA input jacks.
- Solder joints are checked and touched-up wherever necessary, and the board cleaned of flux.
The output modules get new 250V On Semi MJE15032/33 driver transistors to prevent the most common cause of GFA-555 failure. Local power supply bypass capacitors are added to the output modules. (As stock on GFA-555 MKII) I use high-endurance Panasonic ED 47uF/250V.
The amp is completely torn to bits, and the chassis cleaned.
I start laying out the power supply.
PC-board stand-offs are installed and the 35A bridge rectifiers mounted to the chassis floor for heat-sinking, with their leads pointing skyward for connection through the bottom of the board. A razor blade is used to check for flatness where the bridges will be bolted to the chassis. This one was slightly bulged around the original transformer bolt hole, but a couple of whacks with a rubber hammer flattens it. The board is lowered down onto the pins of the bridge rectifiers and screwed into place, and then the bridge rectifiers can then be soldered into their final alignment. The board is removed with the bridges now installed, and the big caps installed to complete the power supply board.
Now that the amp is back together, I power it up slowly on the variac in case I did something wrong, or in case there is still a fault somewhere. They come up perfectly nearly every time, as I’ve inspected everything so closely by this point.
The amp is checked for correct operating parameters, including but not limited to:
- Low distortion, closely matched between channels
- Low distortion at full power into 8 and 4 ohms.
- Clean square-wave performance into 2 ohms, with minimal ringing, indicative of good damping, stability, and frequency compensation.
- Low noise from the speaker output terminals. I use a sound-card based FFT analyzer, as well as simply connecting a high-efficiency PA horn speaker to the speaker terminals to actually listen to the noise. This amp is stone quiet compared to the OEM configuration. The FFT shows noise spikes from 6-20db lower than stock at various harmonics. With the PA speaker hooked to this amp, I have to put my ear right against the horn to hear an extremely faint buzz. Through normal efficiency hi-fi speakers, the noise is almost imperceptible. It’s actually quieter than a stock GFA-555 MKII. This is due to the re-location of the power transformer, and the optimized design of the power supply board.
- Correct bias operating points at all gain stages
- Close matching of output transistors. Power dissipation of each output transistor is checked at idle as well as 1/3, 1/2, 2/3, and full power. If matching is poor, or if some particular transistor is running too hot or too cold compared to its friends, I can usually find a good match from my collection of good original spares, or the customer can choose to replace them all with new matched sets. ($130 per side on a GFA-555)
- Low DC offset from the speaker terminals. This indicates good matching of the input transistors, and is generally a sign of good health. Incidentally, the GFA-555 MKII has a DC servo, so will show very low DC offset, even with poorly matched input transistors. In such an amp, the output current from the DC servo can tell you how well balanced the amp is.
- Close gain matching between channels.
Assuming all that checks out, it’s time to make sure the amp is reliable. It’s hooked to a dummy load, and run through a battery of stress tests. The amp is driven hard into the dummy load and allowed to heat up to about 75C and back to cool again at least three times. Bias should track as expected throughout this temperature range. It’s briefly tested at full power at 2 ohms, and 1/4 power with square waves. The amp should survive all this with no complaints.
Electronics failure statistics follow a “bathtub reliability curve“. Many failures occur early in life of the product—mostly due to manufacturing defects—after which there is a long steady period of reliable operation. Then after some years, failures start to increase with age. What I’m trying to do is provoke the early failures.
Furthermore, the use of high-endurance components is meant to delay aging failures. There are many OEM GFA-555’s that have been running fine for 30 years now, and I aim to make mine last even longer than that.
So now that I’m confident the amp is operating as expected, I check the sound quality by having a listen. I usually use the B&W 602’s at my workbench for this. They sound pretty great, and while not the most high-end speaker in the world, they are very good, and I find they are very revealing of bad sounding amps. Also, I am very accustomed to their sound, as I listen to them while I work.
If the sound is good, then it’s time to produce a performance report and email it to the owner as a PDF. I use a Hewlett Packard 8903B Audio analyzer connected to my laptop to produce performance charts and measurements. Pete Millet, thanks for the cool HP 8903 automated performance charting software!
Charts and measurements are produced for:
- Power output at 8 and 4 ohms at less than 0.1% THD+Noise
- Gain in decibels for each channel. (They should match closely.)
- DC offset at the speaker terminals with inputs shorted
- Frequency response at 8 ohms, 1W
- Power versus THD+Noise at 8 and 4 ohms
- THD+Noise vs. Frequency at 8 ohms, 1W
That’s it, really. The amp is shipped to the customer after payment, and I eagerly await the owner’s comments on the sound.
Thanks for reading!