-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
DIY Purifi Amp builds
- Thread starter JimB
- Start date
230v where I live, thank you for the info!
OP
- Thread Starter
- #484
I finally got around to doing operation validations on several sets of cables and boards. I find the 5-wire interface between J3 of the FE02B, and J4 of the SMPS1200A400, to be problem free. With no input signal, there is a hint of a click at power on/off, but since the outputs have ~5 mV DC offset, that is expected. At power off, it mutes audio output, instantly. When using the EVAL1 and the SMPS1200 with the signals available on J4, I see no benefit in any special adaptations to manage power on/off behavior (unless you are adding 12V trigger control). J5 of the SMPS remains free to drive a power-on LED, such as with the Ghent cases.
And I can confirm that the Standby switch, S1 on the FE02B, serves no useful purpose with this arrangement. I'll be removing it in my builds once I have proper cases (no extra hole).
Please don't ask how it sounds. While I did drive some speakers, this wasn't that kind of test. I can just say, "it works".
And I can confirm that the Standby switch, S1 on the FE02B, serves no useful purpose with this arrangement. I'll be removing it in my builds once I have proper cases (no extra hole).
Please don't ask how it sounds. While I did drive some speakers, this wasn't that kind of test. I can just say, "it works".
Last edited:
It's been close to a month since I got my build up and running, have used it with both Magnepan 1.7 and 3 way direct drivers, of all my builds it is still one of my favorite amps. Compared to my Mark Levison 585 integrated it has more detail in mid range and the high frequencies are smother. But then I am driving the Purifi with a benchmark Dac and using it as a pre amp which has a more detailed sound than the Mark Levison DAC. The Mark Levison may have a little better base, but once again that may be the DAC as it has a warmer sound than the Bench Mark. Using my tube phono amp and records the differences are between are even less. Once again it appears to be a detailed very neutral amplifier and I believe it will reproduce whatever sound signature that's feeding it, as I said one of my favorite, If I compare it with a amp I have owned then it would be the Benchmark AHB2 with more power.
OP
- Thread Starter
- #486
FYI, in testing all combinations of my hardware, I found that some channels have 'substantial' DC offset when there is no input applied. This will also occur with sources that are AC coupled. My values ranged from +, or -, 1 to 165 mV, using the input buffer. The output DC offset dropped to within +10mV when bypassing the buffer, and when using a source that is DC coupled (and which has a low offset).
I've checked with Purifi about it, and this (and more!) has to be considered normal - within the range of possible input offset currents of the OPA1612, given the differential input resistance of 94K (2*47K) and the total system gain of ~27 dB.
NOTE. The values that follow are calculated - I have not tested and measured all these.
IF you have an AC coupled source and this is a concern to you, you can improve it in these two ways:
1. Reduce the buffer gain, by increasing the value of R17, as described here: https://www.audiosciencereview.com/...iy-purifi-amp-builds.10478/page-8#post-296652 The output offset will reduce proportionally with the reduction in gain. In my case, using 5.36K there, drops the total gain by 7dB, making the 165mV value drop to 74. The extreme of this would be to by-pass the buffer, but then you also a low input resistance of 4.4K. IF you bypass the buffer, the next method will not apply.
2. Reduce the differential input resistance. We can do this the easy way, or the hard way. The easy way is to add a parallel resistance across the input of the buffer, such as with a leaded resistor on the back side of the PCB, between pins 2 and 3 of the XLR connector. The hard way is to replace R14&19 with values less than 47K (say, 5.1K). The following is from Purifi, with the locations of R14 and R19 highlighted in yellow:
Either way, reducing the input resistance improves the output offset, proportionally. I prefer the easy way. Adding 11K there drops the differential input resistance from 94K to 10K, and reduces a 165 mV output offset to 18mV, with the full system gain of 27 dB (assumes only differential input offset current for the OPA1612, not differential input offset V). You can use other values to get more, or less reduction, but be sure of the capability of your AC coupled source to drive the lower resistance value you choose. And, with an AC coupled source, lowering input resistance means raising the low frequency cutoff. If 10K ohms is too low, using 24.9K, will net 20K input resistance, in which case an offset of 165mV becomes 35. The resistors can be low power, 'only' 1% tolerance, and any reasonable 'quality', including thick films (which is what R14&19 are, now).
The two methods can be combined, so, for example, reducing the gain by 7dB (to a total of 20dB) and adding 24.7K across the inputs (to a net 20K input resistance) reduces an offset of 165mV to 16 mV.
I've checked with Purifi about it, and this (and more!) has to be considered normal - within the range of possible input offset currents of the OPA1612, given the differential input resistance of 94K (2*47K) and the total system gain of ~27 dB.
NOTE. The values that follow are calculated - I have not tested and measured all these.
IF you have an AC coupled source and this is a concern to you, you can improve it in these two ways:
1. Reduce the buffer gain, by increasing the value of R17, as described here: https://www.audiosciencereview.com/...iy-purifi-amp-builds.10478/page-8#post-296652 The output offset will reduce proportionally with the reduction in gain. In my case, using 5.36K there, drops the total gain by 7dB, making the 165mV value drop to 74. The extreme of this would be to by-pass the buffer, but then you also a low input resistance of 4.4K. IF you bypass the buffer, the next method will not apply.
2. Reduce the differential input resistance. We can do this the easy way, or the hard way. The easy way is to add a parallel resistance across the input of the buffer, such as with a leaded resistor on the back side of the PCB, between pins 2 and 3 of the XLR connector. The hard way is to replace R14&19 with values less than 47K (say, 5.1K). The following is from Purifi, with the locations of R14 and R19 highlighted in yellow:
Either way, reducing the input resistance improves the output offset, proportionally. I prefer the easy way. Adding 11K there drops the differential input resistance from 94K to 10K, and reduces a 165 mV output offset to 18mV, with the full system gain of 27 dB (assumes only differential input offset current for the OPA1612, not differential input offset V). You can use other values to get more, or less reduction, but be sure of the capability of your AC coupled source to drive the lower resistance value you choose. And, with an AC coupled source, lowering input resistance means raising the low frequency cutoff. If 10K ohms is too low, using 24.9K, will net 20K input resistance, in which case an offset of 165mV becomes 35. The resistors can be low power, 'only' 1% tolerance, and any reasonable 'quality', including thick films (which is what R14&19 are, now).
The two methods can be combined, so, for example, reducing the gain by 7dB (to a total of 20dB) and adding 24.7K across the inputs (to a net 20K input resistance) reduces an offset of 165mV to 16 mV.
Last edited:
OP
- Thread Starter
- #487
Purifi confirms that one can also just use a wire jumper from the feedback hole to any of the three output trace holes, rather than adding 0R SMD resistors. Two examples shown in green:
Also, when using remote sensing, off the board, one can install a 1 ohm resistor in that location as a backup against the amp going open-loop if ever the external sense connection was broken. There should be enough room in the feedback hole to accommodate a sense wire and a resistor lead. Alternatively, one can add such backup resistors as 0603 parts on the SMD pads for R8&9_A1&2.
Last edited:
OP
- Thread Starter
- #488
I've made a table summarizing total power dissipation for a Purifi EVAL1 plus Hypex SMPS1200A400 stereo amp, using the datasheets and making nominal allowances where there was ambiguity. I'm looking at this in preparation for choosing specific case metalwork. FWIW, for these modules, most of the heat is coupled to the bottom of the case, directly. The table lists waste heat (Watts) by module, at a few levels of continuous output power, into 8, 4 and 2 ohms (typically over estimating by a few watts). I have done limited modeling of some metal cases as well, but I will wait to add some actual measurements before committing. As expected, the heat to be dissipated varies, substantially, with output and load impedance, so deciding on a level of acceptable thermal performance requires specifying MY most demanding use scenario (OR using BIG-A__ cases, or fans, which I won't do). My most extreme target is to be able to run 2-channels, continuously, at 1/3 full rated output power into 4 ohms, with an outer case temperature no higher than 50C (assuming a 25C ambient). I doubt I will ever actually run even this hard, so I'll comprise on that a bit if it a matters to my preferred case selections. I color-coded the total amp dissipations for these levels of operation according to my degree of comfort with running them in the case I'd prefer to use. I'm leaning towards 50W as my design requirement. This should not impose any limit on peak output levels I will be able to use.
EDIT: Values refined, Mar 2, 2020
EDIT: Values refined, Mar 2, 2020
Last edited:
Thanks good information.
JimM
JimM
OP
- Thread Starter
- #490
Adding to the topic of power and cooling, Purifi recommends using a metal case with a base of at least 20 cm x 30 cm, at least 2-3 mm thick. For reference, the demo amps they loaned out (including the one to ASR) were 25 cm x 32 cm. We've seen quite a bit of discussion on ASR about amplifier power ratings. For these amp modules, this is mostly dependent on the heat-sinking we provide. If you are going DIY, and want relatively high, continuous output levels, you must provide a case with suitable heat transfer.
Last edited:
maty
Major Contributor
And / or silent fans noise < 19 dB (A).
Small boxes are usually problematic in amplifiers that are connected to current loudspeakers, many of them with true sensitivity of 83-85-89 dB and minimum impedance of about 4 Ohms (2-ways) or 3 Ohms (3-ways) is quite common. And if you listen to very good recording with high or very DR... as DR19 of 1812 Overture -> heat.
Small boxes are usually problematic in amplifiers that are connected to current loudspeakers, many of them with true sensitivity of 83-85-89 dB and minimum impedance of about 4 Ohms (2-ways) or 3 Ohms (3-ways) is quite common. And if you listen to very good recording with high or very DR... as DR19 of 1812 Overture -> heat.
Last edited:
OP
- Thread Starter
- #492
I take your points. I'm not a fan of fans in high performance audio gear, and one would hope that, surely, class D can be done without them. It requires a suitable enclosure - not too small, and thermally efficient.
Minimum impedance matters more if a relatively high percentage of the power is being delivered in those frequencies.
Regarding high DR recordings, those should help, no? The average power should be pretty low, UNlike more highly compressed material, that most of us would rather avoid, but for which we should make heat allowances. I'm interested to learn a bit more about how long the highest power levels are present with a range of high DR material. One second? 3? 10? 30?! That is, what is the average power level during the worst (highest) 3 seconds, 10 seconds, 30 seconds, not just the average for the whole piece, or recording. This pertains to the thermal capacity of a case - how long it can support higher heat levels than it can dissipate, continuously, at the target temperature. A larger case, and/or heavier heat conducting components, will ride through such conditions.
Last edited:
OP
- Thread Starter
- #493
I have heard from Ghent. They have been delayed in starting case design work for the Purifi modules, including the obvious reason, COVID-19 restrictions. They are not in Hubei province, but still. They appear to be planning a new case extrusion - a new case size. I am still hoping they will offer it with different back panel choices, similar to what they have for the Hypex Ncore modules. Latest guidance is 3-4 more months for shipments. So, a summer project. For those of us in the northern hemisphere, let's hope for good heat dissipation!
Last edited:
OP
- Thread Starter
- #494
For those who want to use binding posts or speakON output connectors with the Purifi EVAL1 set, BUT also want to use the existing XLR inputs directly on the back panel, you must either include five holes through the panel for the existing output sockets and standby switch, or remove them. If you want to use off-board sense feedback from your added output connectors, you must remove the on-board output sockets, as I have now done for some of my boards. If I'm removing the output sockets, I'm removing S1, since I have the required ampON signal coming into J3 via a 5-wire interface with Hypex SMPS1200. As is sometimes the case, after tearing into a couple, one gets ideas of how to do a thing, better.
The standby switch, S1, and the 4 mm output sockets can be unsoldered, but they are tough. They don't come, easy. Ya know, they don't come easy. For those doing it, note that the sense pin of the output socket (at the edge of the board) is the more delicate one (easiest to damage a pad), but also the easiest to heat and move. I recommend heating and then lifting that end of the socket to pull the pin out about flush with the circuit side of the board (not too much - you are stressing the joint at the other end of the socket). Then get a cutter in there and clip the exposed lead (you can remove it later, or not). Now heat the pin on the other end and pull the socket off. Be careful here - it is difficult, because there is little projecting to engage the iron, AND it connects through to a massive trace on the component side, AND the through hole is small. You don't want to damage the board - don't force it. The black ones are worse, because they connect to ground (even more copper heat-sinking). A good iron is required. At least the items are not heat sensitive, and are (likely) being discarded, anyway.
Removing S1 is difficult because there are 4 pins to unsolder, simultaneously! Fortunately, once the adjacent output socket is removed, you can clip two of the leads from behind, so there are only two to unsolder to free the switch, and one of those is 'mechanical' and can be separated, leaving only one to heat while pulling the switch body away. You can just clip off the remaining leads, or unsolder them, easily. I removed one of these while keeping the output sockets, and I don't recommend it (see my next post). NOTE! Removing S1 removes ampON, unless you provide it another way, such as via pin 1 of the J3 connector, as has been described elsewhere for 5-wire interfaces to the Hypex SMPS1200A400. If you do not implement another source of ampON signal, you should provide it here by jumping pins 2 and 3 together.
Here is "what lies beneath" (board trace details - you can see the output and sense traces). Note that with the output sockets removed, you can, and you MUST, provide sense feedback connections (as described, elsewhere)!
For those who cannot, or don't want to unsolder any of these parts, there is another way to eliminate the need for panel holes for these 5 items, in my next post.
The standby switch, S1, and the 4 mm output sockets can be unsoldered, but they are tough. They don't come, easy. Ya know, they don't come easy. For those doing it, note that the sense pin of the output socket (at the edge of the board) is the more delicate one (easiest to damage a pad), but also the easiest to heat and move. I recommend heating and then lifting that end of the socket to pull the pin out about flush with the circuit side of the board (not too much - you are stressing the joint at the other end of the socket). Then get a cutter in there and clip the exposed lead (you can remove it later, or not). Now heat the pin on the other end and pull the socket off. Be careful here - it is difficult, because there is little projecting to engage the iron, AND it connects through to a massive trace on the component side, AND the through hole is small. You don't want to damage the board - don't force it. The black ones are worse, because they connect to ground (even more copper heat-sinking). A good iron is required. At least the items are not heat sensitive, and are (likely) being discarded, anyway.
Removing S1 is difficult because there are 4 pins to unsolder, simultaneously! Fortunately, once the adjacent output socket is removed, you can clip two of the leads from behind, so there are only two to unsolder to free the switch, and one of those is 'mechanical' and can be separated, leaving only one to heat while pulling the switch body away. You can just clip off the remaining leads, or unsolder them, easily. I removed one of these while keeping the output sockets, and I don't recommend it (see my next post). NOTE! Removing S1 removes ampON, unless you provide it another way, such as via pin 1 of the J3 connector, as has been described elsewhere for 5-wire interfaces to the Hypex SMPS1200A400. If you do not implement another source of ampON signal, you should provide it here by jumping pins 2 and 3 together.
Here is "what lies beneath" (board trace details - you can see the output and sense traces). Note that with the output sockets removed, you can, and you MUST, provide sense feedback connections (as described, elsewhere)!
For those who cannot, or don't want to unsolder any of these parts, there is another way to eliminate the need for panel holes for these 5 items, in my next post.
Last edited:
Hi Jim - based on your description and the pictures I have to wonder if you used anything to actually desolder the contacts or just went for the heat and lift?
You should look into desoldering wick/braid. Fwiw I find that works way better than any of the spring loaded vacuums. Now if you can afford a proper desoldering station that's a whole other thing...
Tom
You should look into desoldering wick/braid. Fwiw I find that works way better than any of the spring loaded vacuums. Now if you can afford a proper desoldering station that's a whole other thing...
Tom
OP
- Thread Starter
- #496
So you want to use binding posts or speakONs, with the on-board XLRs, but NOT remove the five offending components described above, without having to put holes in your back panel for them? You can do that. And leaving the output sockets in place means you do not need to provide a sense feedback path - it is still there in the existing output connectors. First, you can remove the 'face' of the output sockets. You can clip them, cut them, file them - even melt them. No dis'ing my tools - they are over 50 years old.
I chose to clip them off (nibble around the base of the lip), then touch up the face by scraping with a knife, but filing and melting work just as well. You just want the faces of the sockets to be flush with the edge of the board so there is no need to make a hole for them through the back panel.
Fortuitously, the contact inside does NOT project past the face of the remaining body - there is no need to provide insulation to prevent them from shorting against a back panel.
That remaining S1 switch? If you are good with a jewelers screwdriver, it is clean and easy. From my previous post, you can see how the face and handle of the switch mate with the body and latch together. Slip the blade of screwdriver between the metal frame and the switch body and twist to bend it out, releasing the side catches and permitting the front to bend out and down. If it is too tight against the body of the black output socket, you can use a pen knife to cut down the little ridge on the side of the socket, by the switch (though I didn't).
Pry the sides away as far as reasonably possible, then pull the front out and down (maybe pry in the gap at the top to get started).
This metal frame is easily fatigued. Bending out 90 degrees (as shown), then back up and down again broke it off, short and clean for me. It does not protrude, but even if it did, you could file it back or let it touch the back panel - it is grounded. Finally, you'll want to pick out the greasy, little gold rocker plate inside - you don't want it to fall out later and short something.
The end result is perfectly for flush mounting. As I described in my previous post, IF you are not providing ampON by another path, you should provide it here, by bridging pins 2 and 3 together, under the board.
So, no unsoldering. Just solder your jumper wires from the available output holes near the back of the on-board output sockets to your new sockets on the back panel. (Sorry for any confusion in this last photo - I had already removed S1, which was a royal pain with the output socket in place, and totally unnecessary with the new technique I just described here.)
I chose to clip them off (nibble around the base of the lip), then touch up the face by scraping with a knife, but filing and melting work just as well. You just want the faces of the sockets to be flush with the edge of the board so there is no need to make a hole for them through the back panel.
Fortuitously, the contact inside does NOT project past the face of the remaining body - there is no need to provide insulation to prevent them from shorting against a back panel.
That remaining S1 switch? If you are good with a jewelers screwdriver, it is clean and easy. From my previous post, you can see how the face and handle of the switch mate with the body and latch together. Slip the blade of screwdriver between the metal frame and the switch body and twist to bend it out, releasing the side catches and permitting the front to bend out and down. If it is too tight against the body of the black output socket, you can use a pen knife to cut down the little ridge on the side of the socket, by the switch (though I didn't).
Pry the sides away as far as reasonably possible, then pull the front out and down (maybe pry in the gap at the top to get started).
This metal frame is easily fatigued. Bending out 90 degrees (as shown), then back up and down again broke it off, short and clean for me. It does not protrude, but even if it did, you could file it back or let it touch the back panel - it is grounded. Finally, you'll want to pick out the greasy, little gold rocker plate inside - you don't want it to fall out later and short something.
The end result is perfectly for flush mounting. As I described in my previous post, IF you are not providing ampON by another path, you should provide it here, by bridging pins 2 and 3 together, under the board.
So, no unsoldering. Just solder your jumper wires from the available output holes near the back of the on-board output sockets to your new sockets on the back panel. (Sorry for any confusion in this last photo - I had already removed S1, which was a royal pain with the output socket in place, and totally unnecessary with the new technique I just described here.)
Last edited:
OP
- Thread Starter
- #497
WELL! No desoldering station for me. Sigh. I did use both a handheld soldervac, and braid. They just weren't good enough to free the inner pins of the output sockets, which becomes clear once you try those buggers. In the end, I needed to wiggle them out, gently, while heating. This is where it is important to work one end at a time, in stages. In fact, leaving solder on them was required, since I couldn't get enough heat into the pins when they were dry (after using solder vac and wick). Everything was freehand for me - not even a vise to hold the board (which really would have helped). I have more of them to do, so I might dig out a vise (I moved house, recently).
But, for anyone not wanting to use off-board sensing feedback, my post just above is really the way to go.
Last edited:
Just a follow up, I chose to leave the original speaker connectors along with the additional 5 way speaker connectors which allows a bi wire connection if I ever want to go that way (see post #320) and the switch is just another hole. The chassis I used is a DIYAudio store chassis from Italy a nice chassis for the price but it is a steel chassis and hard to work, would have been nice if the back panel had been aluminum. I mounted all the components on separate pieces of aluminum so they could easily be removed to work on. The EVAL1 and amps are mounted on aluminum angle and bar brackets for two reasons 1. to raise the modules up from the floor of the chassis and 2. Provide additional heat sink for the amp. They are mounted with silicon heat paste for good heat transfer. I made a template from the Purifi EVAL1 manual showing the connectors with dimensions by copying and printing the page until I got the right dimensions, I think it ended up at 120% of the PDF page. Then measured the template with my steel rule. If you have a drill press then drilling the holes is possible with sufficient precision, I wouldn’t try it with a hand drill but It could be done. I suppose you could also just do a rectangular cut out for the connector area. I’m glad I went ahead and used the chassis, as of right now I consider it a sturdy finished trouble-free amp with all the features (post #428) I wanted to incorporate in the amp (at least at this time) hey it’s DIY after all.
MMorrill88
Member
- Joined
- Feb 4, 2020
- Messages
- 13
- Likes
- 4
Hey guys I just want to give a real quick opinion on my experience. So for the last month I had the 1et400's in my Nord monos with sonic imagery 990s. They did sound real good to me, wasn't sold on them bettering the Nc500 though. Put the nc500s back in and yes, to me the hypex were in fact better to my ears. BUT then I reassembled the Eval1... And to me this beat out either of the other 2 amp setups. This is of course just my experience. Now I just need a permanent home for the Eval1, hope ghent comes through. Just another note, you can definitely feel the extra head room of the Nc500 but the purifi eval1 is just better sounding to me, like its more refined? I don't know, I'm still new to the audiophile world so I don't have the ears of the more experienced.
Last edited:
What speaker were you using?
Similar threads
