EBFA-565 versus BFA-565 (Validating a new design)

Launching a new product is a lot of work! I’ve been working on these new boards for the Adcom GFA-565 for about six months now. (Now available here.)

This is the last step before I consider it fit for sale—design validation. It has to perform at least as good as the previous version, and it has to deliver on the usability improvements. I need to install it into an amp, and not only test its electrical operation, but the user experience of installing it.

A little background on my boards for the Adcom GFA-565

In 2016, I had a customer who wanted a pair of Adcom GFA-565’s restored. The input boards were so ruined from capacitor leakage, that it motivated me to learn PCB design in Eagle CAD, and make new boards. This was the product that launched Hoppe’s Brain circuit boards!

Adcom fans know about the issues these amps have with exploding capacitors.

Long story short: Elna—a good quality brand—made a bad batch of these little brown capacitors, ironically labeled “Long-Life”, LOL. Adcom put them in the GFA-565, and they destroyed nearly every one of these beautiful amplifiers, and destroyed many speakers with them. This random component failure put an undeserved dent in Adcom’s reputation. They were using a quality supplier, and they used very nice quality parts throughout the amp.

The boards get ruined. The electrolyte soaks into the fiberglass and causes it to become slightly conductive. This all happens around the sensitive high-impedance circuitry of the input section, where a tiny leakage can cause a DC offset to form, which is amplified by the circuit.

To fix the original board, one must fly leads above the board, or run bodge wires beneath the board.

Eww.

I could fix it but I don’t want to. It’s not up to my quality standards. I’d be worried about it breaking sometime in the future, and once I fix something… I set it free, and I never want to see it again. :^}

The new board was called the BFA-565. (Better F–cking Amplifier) It’s gone through a number of improvements over the years, as I learn more and more about PCB design. (Thanks to authors and YouTubers Eric Bogatin, Henry Ott, Dave Jones, Susy Webb, Robert Feranec, Rick Hartley and Zach Peterson)

This has been a best-selling product for me. Many GFA-565’s have been saved from the landfill and are playing music right now. So I thought it was time for a re-design, using what I’ve learned about PCB design to improve it’s electrical performance, and implementing features that make it easier for customers to install and troubleshoot.

The layout is an almost complete departure from OEM, so of course everything needs to be thoroughly tested.

I don’t have a working original GFA-565 board to compare this new EBFA-565 to, so I will instead compare it to my current BFA-565 board, which has been field-proven since 2016.

I want to do a before-and-after comparison, so the test subject is this GFA-565 with one of my BFA-565 boards installed.

I ran this amp through a battery of tests to collect as much data as I could, before swapping in the new EBFA-565 board. Then I ran it through the exact same tests with the new board.

Testing:

I love my HP 8903B Audio Analyzer, however being from the late 1980’s, its lower measurement limit is around 0.005% THD+N. (Published spec is <0.01%) Differences in performance between the BFA-565 and the EBFA-565 are subtle, and close to the lower limit of this instrument. Still, I do believe there are statistically significant improvements, as observed by the HP. A more modern analyzer like an Audio Precision system would reveal more.

Pasted below are my notes from the process. The most notable improvement, was lower distortion at high power levels, below 1KHz.

Overall I’m delighted! The EBFA-565 board shows some real improvements. Stability is excellent as before.

Graphs generated automatically with Pete Millet’s HP 803 Software.

BFA-565
Frequency Response, 0.12V input, 2.8V output 8 ohms

EBFA-565
Frequency Response, 0.12V input, 2.8V output 8 ohms

BFA-565
Frequency Response, 2.0V input, 46.2V output 8 ohms

EBFA-565
Frequency Response, 2.0V input, 46.2V output 8 ohms

No change in frequency response.

BFA-565
THD+N v power 20Hz, 8ohm

EBFA-565
THD+N v power 20Hz, 8ohm

BFA-565
THD+N v power 1KHz, 8ohm

EBFA-565
THD+N v power 1KHz, 8ohm

BFA-565
THD+N v power 10KHz, 8ohm (oops forgot to scale to 100mV)

EBFA-565
THD+N v power 10KHz, 8ohm

BFA-565
THD+N v frequency 4 ohms, 0.12V input 2.8V output, 20Hz- 60KHz (Ignore spike at 13KHz, it shows up in all my measurements and I haven’t figured out why yet.)

EBFA-565
THD+N v frequency 4 ohms, 0.12V input 2.8V output, 20Hz- 60KHz

Slight distortion reduction above 30KHz, nice!

BFA-565
THD+N v frequency 4 ohms, 1V input, 23.1V output, 20Hz- 60KHz

EBFA-565
THD+N v frequency 4 ohms, 1V input, 23.1V output, 20Hz- 60KHz
Very slightly less distortion above 20KHz

BFA-565
THD+N v frequency 4 ohms, 2V input, 46.2V output, 20Hz- 20KHz

EBFA-565
THD+N v frequency 4 ohms, 2V input, 46.2V output, 20Hz- 20KHz

Less distortion below 1KHz! Old board was about 0.025% below 1KHz near full power, new EBFA-565 board is 0.004% throughout this band! Again I should point out the caveat that this is down in the weeds of the resolution of the HP 8903B, and day-to-day sample variations might explain some of these results. But I think this is statistically significant. Even though a human being couldn’t reliably discern 0.025% THD from 0.004% THD, this result suggests something is right with the amplifier’s signal integrity.

BFA-565
Input voltage vs THD+N 20Hz 4 ohm

EBFA-565
Input voltage vs THD+N 20Hz 4 ohm

Slightly better overall

BFA-565
Input voltage vs THD+N 1KHz 4 ohm

EBFA-565
Input voltage vs THD+N 1KHz 4 ohm

Slightly better overall

BFA-565
Input voltage vs THD+N 10KHz 4 ohm

EBFA-565
Input voltage vs THD+N 10KHz 4 ohm

About the same

SQUARE WAVE TESTS

Open-circuit vs difficult load consisting of 1uF MKP cap in series with 2 ohm dummy load. (This my “pickled weiner dummy load”, which consists of four ceramic power resistors, suspended in a Ball jar filled with mineral oil.)

BFA-565

Input: Square wave generated by Rigol DG1022Z, 25MHz Function Generator, 1KHz, 830mV P-P.

GFA-565 Output: 20V P-P

50uS Horizontal

5V/div

Open circuit

With load 1uF in series with 2R

Zoomed to 1V/div

Open Circuit

With 1uF/2R load

About 0.4V P-P ringing, well-damped.

EBFA-565

(Spoiler: no difference)

Input: Square wave generated by Rigol DG1022Z, 25MHz Function Generator, 1KHz, 830mV P-P.

GFA-565 Output: 20V P-P

50uS Horizontal

5V/div

Open circuit

With load 1uF/2R

Zoomed to 1V/div

Open Circuit

With 1uF/2R load

Conclusion: No significant difference in square wave response determined.

Introducing the EBFA-565; the Even Better F–cking Amplifier (New version Adcom GFA-565 Circuit Boards)

Big news! I have completely re-designed circuit boards for the Adcom GFA-565. Based on customer feedback, I’ve made many improvements. These new boards are easier to install, harder to screw up, and they even perform a little better!

Buy them here. Hoppe’s Brain EBFA-565.

The GFA-565’s circuit—designed by the legendary Walt Jung—is a picture of perfect symmetry; A complementary mirror-image from positive to negative halves.

Unfortunately, the OEM circuit board layout doesn’t reflect this symmetry. Things are packed in kind of tight, and it’s kind of hard to follow.

I’m not trash-talking the original designers. The OEM board is a perfectly good design, but it’s single-sided, so compromises must be made in the layout, and they would have been under budgetary pressure to make it small, or to meet a certain size so they can fit a certain number of boards on a PCB panel. Thus, the OEM board is a bit cramped. The OEM board is designed for manufacturing. Hoppe’s Brain boards are designed for DIY.

Again—not complaining—Adcom’s circuit boards were better quality than similarly priced competition at the time.

The new layout reflects the symmetry of the circuit. It’s larger than the OEM board, and arranged in a more symmetrical fashion. Impedances seen by the circuit, that are caused by the board, are therefore more evenly balanced.

More features:

  • Solder-less installation: Something I kept hearing from customers, is that if something goes wrong and the magic smoke escapes, and the board needs to come out for repair… All the wires need to be de-soldered, and it’s hard to reach them without burning components, wires, or chewing up the PCB pads. It’s a little awkward to work with the older board with so many solder connections, and a major bummer to have to re-do them every time the board needs to come out.
    On the EBFA-565, all connections that used to be solder-pads, are now right up front on WAGO cage-clamp terminal blocks. These spring-loaded clamps grip the wire very, very firmly, and make an oxygen-tight connection that will never come loose. (Screw-terminal blocks can come loose over time as the copper squishes.) The board can be removed for troubleshooting or repair in a few minutes, and reinstalled with no damage to PCB pads. …so much better. And it makes the wiring neater too.
  • wp-1666225890475
  • Top-shelf components throughout: Kits and pre-assembled boards come with Dale RN and CMF resistors, 0.1% tolerance resistors in key locations, polypropylene capacitors, high-endurance electrolytic capacitors, silver-mica capacitors, etc… The good stuff.
  • Compatible with old parts from the original board, so if you are building up from a blank board, you may recycle many of the un-damaged original parts. (With a few exceptions, see below)
  • Two-layer, plated through-hole FR4 fiberglass circuit board.
  • Traces are routed on the top and bottom layers.
  • Ground planes cover the top and bottom layers.
  • High-impedance traces are given wider isolation from ground planes, other traces and pads.
  • Symmetrical layout makes it easier to build, and less error-prone, because it’s easier to locate the components.
  • More room to work: Things are less crowded. It’s easy to probe any point on the board, and it’s easier to assemble and re-work.
  • Clear labeling: Every component is labeled with its part number and value.
  • An annotated schematic, loaded with troubleshooting tips, DC operating points, etc, is included with the board, printed on 11×17 paper. Download it here.
  • Convenient test points for measuring the most important circuit parameters, referred to expected values on the schematic.
wp-1666225890516
Test points labeled TP_XXXX
https://www.wago.com/medias/cage-clamp2000x2000.jpg-1024?context=bWFzdGVyfGltYWdlc3w1MDkwN3xpbWFnZS9qcGVnfGltYWdlcy9oNzMvaDVhLzg5MzQ2NjA2MzY3MDIuanBnfDExYzJlYTc2NTRhYjdhYjM5OTUxYmRlODg5ZDVlYmI1MjBjODE1M2Q1OTFhNzhhMTEwYWI4MGRhNzE0MTNmMzI
Inside a WAGO Cage-Clamp
  • Improved heatsinking. The TO-126 VAS transistors in the original design run very hot. The heatsinking of this new design is much more effective and keeps temperatures below 60C.
  • Bypass caps are added to the +/-13.8V supplies
  • A 0.1uF polypropylene bypass cap has been added in parallel with the 10 ohm isolation resistance between amp ground and the input section. This improves high-frequency coupling between these ground potentials. This feature was added in response to a customer who experienced oscillation because of substituting a wire-wound resistor in this position. The added inductance was causing the oscillation, and a metal-film resistor fixed the issue. If a little bit of inductance can cause this issue, I figured that adding a bypass cap in parallel with prevent that potential issue, and it may improve performance a smidge.
  • Includes new cable assemblies for signal input, bias compensation transistors mounted on the heatsinks, and bias enable from the soft-start board. These cables are often corroded from capacitor leakage, or storage or operation in a humid environment. I had these cables custom-manufactured, as they are difficult to make by hand, and kind of essential for a good restore job.
    wp-1666297694638