MPB's Purifi Amplifier Project
Spring 2020
Spring 2020
This project started long ago when I came across an Audio Research D-110 amplifier chassis and meters that had been parted out. The faceplate was not in great shape, so it sat in my parts bin until recently when a New Old Stock (NOS) spare ARC D-79 faceplate showed up on ebay. Although the chassis didn't match all the faceplate cutouts, it was good start, and matches the size of the D-79 (which is BIG).
I've always been a fan of Audio Research gear, and have owned a number of pieces over the years. I especially like their industrial design from the 1970s and 1980s, so my plan was to build a newer amplifier circuit into this vintage chassis, keeping the original D-79 cosmetics, including the 5" meters, as original as possible.
The chassis is very large (10.5" by 19" faceplate and 18.5" deep), and could accommodate a variety of amplifier circuits, but I was interested in a high efficiency design, which quickly led me to look into class D amps. The IcePower and NCore Hypex amps are quite popular and well reviewed, and the designer of those amps, Bruno Putzeys, recently started a new company called Purifi Audio to design a next generation class D amp.
Herb Reichert, in the May 2020 issue of STEREOPHILE magazine (p. 97) said "Looking at the bigger picture, it's fair to say: Putzeys's inventions are currently reshaping the landscape of high-fidelity audio."
Putzeys's latest amplifier design is the Purifi 1ET400A, and is available in a package called the EVAL1 with 2 channels of amplification and an I/O buffer board. That formed the basis for this project. Although Purifi is apparently designing their own power supply, it's not yet available. However, a suitable power supply is the Hypex SMPS1200A400. In fact, Purifi used that Hypex supply in the prototype amps sent out to reviewers to demonstrate the 1ET400A amplifier.
Purifi's website: https://purifi-audio.com/
One of the evolutionary changes in the Purifi amps is the attention to reducing hysteresis distortion in the amp output coil, which supposedly causes the granularity some attribute to class D . Purifi recently published a white paper (for those technically inclined) on the subject: https://purifi-audio.com/2020/04/28/this-thing-we-have-about-hysteresis-distortion/
One of the reviews that got me interested in the Purifi amplifiers was this one, from AudioScienceReview.com, which gave the prototype setup from Purifi a very positive review.
This is a single 1ET400A amplification channel of the Purifi amp. Its actually quite small, only 3.25" by 2.5". It has an integral heat sink plate on the bottom which must be bolted to a larger heatsink or amplifier chassis for adequate heat dissipation.
These units are rated by Purifi as 425 watts into 4 ohms (at 1% THD) and 227 watts into 8 ohms.
They need a power supply of about plus and minus 63 VDC, for which the Hypex (see below) fills the bill quite nicely.
This is the EVAL1 setup from Purifi, with two 1ET400A amp channels and an input/output PCB which includes a bypass-able buffer using the OPA 1612 IC. The amp itself provides 12.8 dB of gain, and the buffer board can supply another 14dB for a total of 27 dB gain, which is pretty typical for an audio amplifier.
The I/O PCB includes both balanced XLR inputs and well as banana jack speaker outputs, but in both cases, I'll be bypassing those for chassis mounted XLRs and binding posts.
This is the Hypex SMPS1200A400 power supply. It's a 1200 watt switching supply that supplies +/- 63VDC to power the amp along with a 20V low voltage supply for the EVAL1 buffer PCB. This power supply also includes circuitry to couple with the 1ET400A to shut itself down if a problem is detected with the amp.
Like the Purifi amplifier, this assembly is also mounted on a metal plate used to dissipate heat to the chassis it's mounted on.
The bottom plate on my Hypex unit wasn't quite as smooth as I'd like, so I lightly sanded it down to remove some of the marks on the base.
Here's the final result - smoother, which will make much better contact to my mounting plate and allow better heat dissipation. Cooler operation = longer life for electronics.
Note that Hypex does not take a strong position on the use of thermal/heatsink compound when mounting the power supply (see their application note labelled "Thermal Design" on the Hypex website), but warns that uneven or excess application results in WORSE performance than no heatsink compound with the large mounting plates of the power supplies compared to smaller components. At this point I have no thermal compound on this amp, but thats subject to change.
My objective was to replicate the industrial design of the Audio Research D-79 as closely as possible, but with a modern high efficiency amplifier inside.
This is a photo of an ARC D-79 from the ARC sales flyer, photo cribbed from the great ARCDB.WS website. This is what I was hoping to replicate for the look of the finished amp.
Since I had the D-79 faceplate and the meters from an ARC D-110 (same meters as the D-79), the most critical parts were in hand.
This is the ARC D-79 faceplate as received. The three upside down Mickey Mouse shaped cutouts are for the meters, with the bottom circles (the mouse ears) used for lamps to illuminate the meters. The larger holes below are for specialized industrial fuse holders and lamps to indicate fuses are out.
The rack handles used by Audio Research for many years were available from a number of sources. I found matching surplus handles made by Midland Ross, part number 415-2385-02-14-00. They are 4-5/8" mounting centers for this amp, but ARC used a number of sizes.
The skirted knobs ARC originally used were military types by Raytheon, still in production by a number of manufacturers. The part number is MS91528-2F2B.
The lamps originally used by Audio Research to indicate fuse failures were NE2J neon lamps, but since I'm only trying to replicate the look of the lamps, these Radio Shack lamps (part number 272-0331) come very close. Since I wasn't using them as lamps, they backs were cut off and the front portion was superglued to the sub-panel behind the faceplate.
The most difficult parts to identify and locate were these fuse holders. They are Bussman HPM type, shown here. These are large industrial fuse holders for 13/32" by 1.5" fuses, used on both the ARC D-79 and the D-150 before it. They are relatively rare as they aren't used in consumer electronics today, but I found some surplus.
These were trimmed down (the rear removed) and were superglued to the sub-panel, like the red lamps above, as they contain no fuses and are decorative only.
Skipping ahead a bit, this is the completed front panel assembly bolted to the chassis, looking very much like the original ARC D-79. Although the fuse holders and associated indicator lamps aren't needed for the Purifi amp and are decorative only, my intention was to re-purpose the bias meter switches (more below) and make all 3 meters work as originally intended.
This shows the Audio Research D-110 amplifier chassis, with the top of the amp (at the top of this photo) having a large cutout for the meters. The square boxes on each side of the face originally held large power transformers. Because the bias adjustment switches for the D-79 were in those positions, I cut square holes for those switches to pass through.
Notice the bottom of the chassis is screened (for ventilation), and no top plate came with this chassis, so I made one of 1/8 aluminum stock.
I used a pair of CentraLab PSA-205 rotary switches for the D-79 bias setting control. These are adjustable, so they are set to 3 positions to match the D-79 faceplate, however their function will be different.
Since no bias settings will be needed in this amp, I wired the switches to turn off the meter lamps (actually LEDs in my unit) as well as to switch off the meters completely if desired.
Test assembly of the faceplate to the chassis. You can see the switches (no knobs yet) fitted in the Bias adjust positions. Also note a number of flat head screws attach the panel to the chassis at the top, which are later covered by the 5" meters. The rack handles and three 1/4" bolts secure the faceplate at the bottom.
Top view of the amp chassis, assembled with the faceplate and meters. You can clearly see the two internal aluminum boxes that held the two (dual mono) original power transformers in the D-110. An aluminum plate was bolted atop those boxes onto which mounted a huge capacitor bank for the power supply. Neither large transformers nor a large cap bank will be required for the Purifi amp, so I decided to simply add a deep aluminum shelf atop the two boxes to mount the new amp and power supply.
You can also see a circuit board at the bottom rear of the chassis - thats a power relay board for 12 volt trigger for the amp, more later.
This is the aluminum shelf - actually just a 10.5" by 19" aluminum rack panel (1/8" thick) cut down with a non-ferrous metal blade on the table saw, and sanded smooth with an orbital sander for good mating with the heatsink plates at the bottom of the amp and power supply modules. I printed diagrams of the Purifi EVAL1 and the Hypex power supply to scale and used them for layout and drilling.
Here's the amp mounting plate in place in the chassis. The larger hole on the right will allow the 120V AC line to be dropped down and away from the audio circuitry.
The connections in this Purifi/Hypex setup are mostly JST (Japan Solderless Technology) quick connect terminals.
The primary cable set was bought from Purifi with the EVAL1 setup, shown here. Not included in this photo, but included with the cable kit, are an unconnected VLP-06V and an EHR-7 connector.
This photo shows the main wiring harness from the power supply to the amp as well as low voltage cables, all with JST connections. Note that although the EVAL1 end of the main harness has the VLP-06V connector attached, the power supply end of the harness has only the crimped connectors as shown - they are not inserted into the included VLP-06V connector. The wires are color coded, and are simply inserted in the same positions onto the VLP-06V at the other end. The JST connectors lock into position when inserted, pretty easy.
Additionally an EHR-7 connector for the low voltage supplies from the amp are included, along with a second EHR-7 for the crimped ends without a connector. This setup can be combined with the two additional control wires (for nAMPON and nFATAL) that Purifi includes to complete the connections between the EVAL1 and the Hypex power supply. More below...
There were 3 additional cable related items I purchased from GhentAudio.com (they also have an ebay site) for this project:
The Hypex power supply by default is set up for 240VAC power, and requires a jumper for 120VAC. This is a JST VHR-2N housing from Ghent. It comes with the contacts (which just slide into the housing and clip in), and I simply soldered a jumper between the contacts. It mounts onto connector J3 on the Hypex board, enabling it to work with 120VAC.
2. The 120VAC power supply cable to the Hypex board was also bought from Ghent, shown here. It includes a VHR-3n connector to mate with the Hypex board J2 connector, and comes pre-wired with a 2 foot pair of 16 gauge cables which worked well in this larger amplifier chassis.
3. The Hypex low voltage (+/-20vDC) supplies to the EVAL1 board and the control circuits for the power supply are on Hypex terminal J4, which is a ribbon cable type connector. However, there is an extra EHR-7 connector included with the Purifi wiring set, and the cables they include are already terminated with JST connectors. So the easiest connection method is to add another EHR-7 (from Ghent), and I was able to make up a suitable connector that worked well with the JST connections on the pre-terminated Purifi cables.
The edges of two EHR-7 are trimmed, and the flat side of the connector must be sanded down slightly to fit, but then you can glue the two EHR-7s together and use the JST Connectors on the pre-terminated Purifi cables for this connection.
Note this custom connector has spots for 14 connections, but each outer pair are not used in this instance. There are only 10 pins on the Hypex J4, but the trimmed double EHR-7s fit quite well, and allow the +/- low voltage connections, ground, and two control connections.
Thanks to JimB on the AudioScienceReview forum for this technique. See this thread for more details and photos, scroll down to post 466:
Closeup of JST pins. To inset into the housing, simply align the tab (pointed out here) into the similarly sized slot on one side of the opening. The tab snaps into place in the housing and keeps the connector locked in the housing.
This shows my custom double EHR-7 connector in the Hypex J4 position for low voltage and control connections on the Hypex board. The low voltage power supply connections are on the left and the two control wires (with the blue heat shrink indicating the corresponding EVAL1 pin) are at the right.
Note that the Hypex J4 connector is 10 pins, and the outside housing sockets on my custom connector have no pins/connections below them.
The J2 120VAC input to the power supply is at the bottom of the photo with the black and red wires.
This table shows the pin connections from the EVAL1 board J3 connector to the corresponding pin connection points on the Hypex J4 connector shown above.
Thanks again to JimB on the AudioScienceReview forum for this helpful table. See his original message here, scroll down to post #171:
https://www.audiosciencereview.com/forum/index.php?threads/diy-purifi-amp-builds.10478/page-9
This photo shows the J3 connector on the EVAL1 board, with connections as above. Pin 1 (leftmost): nAMPON, pin 3: nFATAL, pin 5: V+, pin 6: ground, pin 7: V-.
Since I wanted to use the original ARC back panel layout, I didn't use either the XLR inputs or the speaker outputs on the EVAL1 board.
For the inputs, I used Audioquest Viper cable (triple conductors plus shield) for the input wiring. I simply soldered the connections to the pads on the bottom of the board.
For the speaker outputs, I ran Audioquest Type 4 (15 gauge star-quad) from the mounting holes on the EVAL1 to the binding posts on the rear panel.
This photo shows the bottom of the EVAL1 board with the input cables (center) and output cables (left and right) attached.
Keep in mind that if you REMOVE the speaker connectors on the EVAL1 (to use off-board connectors), you will have disabled the remote sense circuit. See the Eval1 manual for details.
At this point, I mounted the Purifi amp and Hypex power supply to the aluminum shelf, wired it up and connected it (using the EVAL1 connections) to my system to test and confirm everything worked as expected. So far, so good.
Power to the amplifier was supplied by a relay/remote trigger board I designed some time ago, using 10 amp Omron relays. It was originally designed to include a soft start circuit, with a 12 volt remote trigger circuit as well. Since the Hypex power supply includes an inrush thermister, the soft start was redundant and added unnecessary circuitry, so that portion was removed. There was originally a small Radio Shack 12.6 volt transformer on this board, as shown here, which powers the relays (and also eliminates the need for a high current front panel power switch). Once I mounted and tested this PCB, I found a slight hum from the Radio Shack transformer, so I replaced it with a small 10VA toroid from Antek.com, shown two photos down.
The rear panel with everything installed.
For the binding posts, I used Superior BP30-2BR, and the XLRs are Neutrik NC3FD-LX-B.
The IEC inlet is a 10 amp Qualtek line filter, #858-10/015.
I also added an 1/8" jack for the 12V remote trigger that connects to my relay PCB.
This shows the inside of the rear panel, wired up. I wanted to use the original ARC layout, and you can see the IEC inlet and fuse holder in the center with the speaker outputs and XLRs to the outside. The ARC chassis originally had RCA inputs, which I replaced with the XLRs.
The completed relay board is shown in place and wired up, including the new off-board toroidal 12v transformer.
You can see the Audioquest input wiring (multicolored) from the input XLRs to the EVAL1 board, and also the Audioquest Type 4 speaker cable (dark blue) from the EVAL1 to the speaker binding posts.
The D-79 originally had power level meters (just eye candy, but nice), and I wanted them to function with the new amp. One problem with the original ARC meter circuit is that at low listening levels, the meters barely move - you need a logarithmic circuit instead of a linear circuit in a power amplifier as levels change dramatically. I experimented with a number of circuits, but ultimately found a PCB on ebay that seems to work fine with a couple modifications, shown here. It is based on a BA6138 log driver IC, which is a square law compression amplifier. It converts the audio signal to its square root, and outputs that signal to the meter, so small signals move the meter proportionally more than large signals. That means that low signal levels will show up on the meters but very high level signals won't pin the meter, as if the meter had a log scale.
I modified the PCB as follows: First I replaced all the caps with Panasonic FCs for reliability reasons. I also removed the rectifier bridge and 12v regulator (both were simply jumpered) since this PCB would be powered from the regulated 12V supply on my relay board. I also experimented with the LED dropping resistors to get a relatively dim light on the meter LEDs, and eventually replaced the 430 ohm resistors with 2.2K RN60s.
This shows the meter PCB installed and connected. The audio inputs are fed from the XLR input jacks on the EVAL1 board, thru the ARC bias switches (to allow turning off the meters), then into this PCB. The 12v input to power the board is at the far side left in the photo, and the connections to the meters and the LEDs are across the rest of the far (front) side.
This is the completed interior from the top - the front of the amp is at the bottom in this photo.
They Hypex power supply is at the front center, with the Purifi 1ET400A modules above, connected to the EVAL buffer board.
The VU meter buffer/amplifier sits to the right of the power supply.
A similar shot, from the rear. You can see the back of the meters and the wiring for the LED lamps here. Not easy to see here, but the power meters have back to back 1N4007 diodes attached to their terminals for protection.
The center meter is an AC voltmeter to monitor line voltage, and has 120VAC on its terminals, so they are covered in heat-shrink for safety.
The completed amp, no top cover.
And with the top cover. I had planned to add a vented cutout in the top, but so far the amp runs quite cool in normal operation, thanks to the large aluminum chassis acting as a huge heatsink.