Purifi class-D amplifiers

[theSven]

As I keep reading and trying to comprehend more about this class D amp, what power supply would you recommend if you wanted to make an amp that had 5 or 7 of these boards? I see that there have been two channels made, but I'm curious about making something with these that could be used to replace an older 5 channel amp.

[Evil Twin]

One might say it depends on the duty cycle and power delivery expected with real world loads and program material.

Adding some additional storage capacitance, it would not be unreasonable to be able to power 4 of the modules from a single SMPS1200A, say, for fronts and rears. Or three across for the L, R, and center with three channels and one SMPS1200A, and for a variety of other surrounds, four or more channels per power supply. Keep in mind the efficiency of the Class D amplifier itself is about 85-90%. So, for 150W delivered power, allow 200W of supply power.

For subs and centers, the power demands will be more substantial.

[fish fingers]

Is the 3e Audio supply suitable?

[Evil Twin]

No, that is single ended supply suitable for lower power IR and TI bridge type class D amplifier chips that use single supplies and bridge mode to control supply rail pumping, as this way the pumping with inductive loads always goes back into the single supply.

[ergo]

First buffer candidate working and did produce sine sines into 300 ohm load for test today.

My Purifi amp is still at friends place, so hopefully we get to listen to this one tomorrow and then at some point I'll measure it a bit more properly also. Did not want to create myself any measurement bias beyond just verifying that all is as expected and working fine.



****

BTW - Purifi folk have updated the "EVAL1 – User’s Guide". More info including a excerpt of opamp buffer and voltage stabilizer schematics.

[Evil Twin]

Thank you for the update, Ergo!

[Evil Twin]

I would not reasonably fault the design choices, such as the buffer amplifier, but it is interesting the trade offs involved. In my investigation of proposals for Class D amplifier front ends, I was favoring a solution developed with the LME49724. This part is not quite as quiet as the OPA1612, but the distortion floor is very similar, and it has a brutally capable output stage with nearly three times the current drive capability of the OPA1612. Oh, and did I mention it is a fully balanced amplifier with dual differential outputs by intrinsic design, not a combination of dual opamps operating independently?

I greatly appreciate the additional documentation, including the linear regulators and other schematics.

[dwk]

I would not reasonably fault the design choices, such as the buffer amplifier, but it is interesting the trade offs involved. In my investigation of proposals for Class D amplifier front ends, I was favoring a solution developed with the LME49724. This part is not quite as quiet as the OPA1612, but the distortion floor is very similar, and it has a brutally capable output stage with nearly three times the current drive capability of the OPA1612. Oh, and did I mention it is a fully balanced amplifier with dual differential outputs by intrinsic design, not a combination of dual opamps operating independently?

I greatly appreciate the additional documentation, including the linear regulators and other schematics.

In other words, something like this https://neurochrome.com/products/universal-buffer . Lots of folks over on diyaudio already scheming on the best way to use this as a front-end to the Purifi. Tom seems to know his stuff, so this seems like a rather promising option.

[Evil Twin]

In other words, something like this https://neurochrome.com/products/universal-buffer . Lots of folks over on diyaudio already scheming on the best way to use this as a front-end to the Purifi. Tom seems to know his stuff, so this seems like a rather promising option.

Exactly. He seems to have done his homework, and his testing seems reasonably rigorous. My investigations date from 2016, and were interrupted by a variety of personal and imperial issues... this included the development of a complete switching amplifier design using fully discrete circuitry, roughly based on Hypex concepts, with some of the newest MOSFETs available at that time from Seinar Industries.

I believe the LME49724 is a well engineered piece of silicon for many applications - it’s balanced outputs can drive a 600 ohm balanced load to up to 52V P-P.

[Evil Twin]

From 4 years ago...

This is the input buffer proposal I was verifying in 2016 for a possible build configuration for both Hypex modules and my own Class D module.

A specific concept of this input buffer is using a high performance high signal level capable Jensen line transformer to achieve very high CMRR between the inputs and outputs; not necessary to protect the Class D circuitry, but to provide isolation for the preamplifier from any switching noise contamination.


This is the simulation circuit in TINA Industrial; Texas Instruments provides an encrypted model of the LME4972 compatible with TINA Industrial for local simulation, which they use for online simulation work. The Jensen transformer is modeled in detail for key parameters.





The output filter displays the required bandwidth and bandwidth limiting, with excellent amplitude and phase response to 10Hz...

[ergo]

This looks cool and elegant indeed. Which Jensen did you have in mind? How big is it physically? It seems that most of them have a PCB mounted variant also. The coolest would be to redo the whole buffer board PCB of course instead of augmenting it with another board ... I might take that up at some point.

****
We did a brief listening of the WHAMMY circuit as the buffer yesterday. I only had OPA2134 as the dual opamp in parts bin for now, but it sounded rather good. On OLA speakers it seemed to add a bit of sparkle to top end and consequently more feel of depth. Bass balance and attack was also good.
I also made an option on a PCB to feed right into final MOSFET stage and skip the opamp. We did try that too. My friend preferred the version where opamp was in loop also. I was more unsure - both sounded slight different, but I was unable to tell with so short listening if I really prefer one of the other.



One thing is sure - Purifi power module is very revealing and it is very easy to hear differences upstream from it. So lets see where our searches lead.... it would be kinda cool to try out your above circuit too as it is probably technically closest to 'blameless'.

Other equipment up stream of WHAMMY buffer was Soekris multibit dac and BPBP - Bruno Putzeys Balanced Preamp for volume control.

[Evil Twin]

The transformer I selected for this application, and in fact sent a set to Steve Manning a few years ago for a Hypex NC400 build, is the Jensen JT-6110K-B. Unfortunately, it is not available in a PCB mount version, to my knowledge.
It is the highest level handling input transformer that Jensen makes, with a bandwidth of -3dB from 0.15 Hz to 120kHz. Yes, that is 0.15 Hz, as in 1/6 of a Hertz.

Low frequency fanatics need not fear the impact... It is designed for levels up to +26dBu at 20Hz.

In the recommended configuration, it presents an input impedance of 37.5kOhms to the driving source; this load configuration provides optimum damping of the step response of the transformer, and a relatively ideal input resistance to the active stage.

Common-mode rejection is 130 dB at 60Hz, though it obviously drops at higher frequencies due to winding coupling capacitance. It is a 4:1 input ratio, so using some gain after the transformer or having a high output capability preamp is preferred. The circuit above has the necessary gain, and due to the low "ideal" terminating impedance of the transformer, runs with a very low noise figure for the LME49724.


Projects like this were a key impetus for purchasing the APx555. Other issues have prevented finding the time to implement a PCB and testing, but now with this new opportunity with PuriFi I think steps must be taken once the immediate speaker project is completed. Well, possibly the next project, too... In my work for the Imperium I have further whetted my skills with Altium Designer, and having my own license for Altium Designer 20, I have no excuses with regards to personal productivity...

[ergo]

What would be a good place to order these in US (seems they are not available in big ones like Mouser and Digikey and no longer in PartsExpress). I'm coming over to US in end of February and it would be right time to stock up. I suspect these will be harder to get in Europe.

**
Altium is the 'real tool' for sure. I'm getting by with a 50EUR tool called SprintLayout for these small projects where a tie to circuit diagram is not a must. I did buy a perpetual license to Eagle also before they went all subscription, but I find the SprintLayout to be what the name says - so easy and fast that I typically have a board done in same time it would take to figure out the circuit and part match in Eagle.

[dwk]

What would be a good place to order these in US (seems they are not available in big ones like Mouser and Digikey and no longer in PartsExpress). I'm coming over to US in end of February and it would be right time to stock up. I suspect these will be harder to get in Europe.

quick hit: https://www.markertek.com/product/jt...r-4-1-stepdown

actually more reasonable than I was expecting.

[Evil Twin]

Markertek is the distributor I use, though one can order direct from Jensen, the pricing is higher. Markertek is a good distributor for many pro sound and studio oriented items, even items like the Mutec MC3+ digital data and clock reclocker.

Yes, I think this is good value component, all matters weighed and considered.

[Evil Twin]

What would be a good place to order these in US (seems they are not available in big ones like Mouser and Digikey and no longer in PartsExpress). I'm coming over to US in end of February and it would be right time to stock up. I suspect these will be harder to get in Europe.

**
Altium is the 'real tool' for sure. I'm getting by with a 50EUR tool called SprintLayout for these small projects where a tie to circuit diagram is not a must. I did buy a perpetual license to Eagle also before they went all subscription, but I find the SprintLayout to be what the name says - so easy and fast that I typically have a board done in same time it would take to figure out the circuit and part match in Eagle.

I, too, have Eagle PCB, on the Mac ("think differently" encompasses a wide range of values...) but never found it particularly pleasant to work with, and creating and managing component libraries is decidedly tedious and restrictive, and limiting in scope in Eagle. The Imperium has switched to a networked license model, and of course, never has enough licenses to check out, as I found when trying to train my Sith apprentice... we never could successfully check out a license for her on her work computer. With my own license, I can install on whatever systems are needed, as long as I only run on one at a time...

[5th element]

Not to point out the obvious but using the transformer on the input will destroy the immaculate linearity of the purifi amplifiers and all the work Bruno put in to better the ncore. Yes they help with regards to some issues but they should only be used if these issues are actually issues to begin with.

[Evil Twin]

So does hooking up the speakers to those amplifiers.

What is interesting is the difference in audibility of different effects, especially when they monotonically reduce in level.

This is a “design study” in progress that is taking far too long to complete. The potential benefits for the noise floor are also interesting...

To see real foolishness in action, I suggest reading the February issue of Stereophile and reviewing the measurements for Audi Research, D’Agostino, Pass, and Schitt.

[5th element]

Oh I know Jon I keep up to date with the reviews posted on Stereophile and find it saddening that so many expensive products are so poorly conceived.

Your inclusion of the transformer on the front end has actual engineering behind it and benefits that it can provide, so I'm certainly not bashing that aspect of it. I think it's a great idea should it be needed, but I would imagine that the instances where it is actually needed are few and far between. A transformer-less input stage can be designed with properties that will match and/or exceed the performance of the Purifi amplifier modules and going with a transformer would seem like a big step back. Then again if you like the sound that the transformer potentially introduces, that's an entirely different thing.

[kravi4ka]

@ergo - new impressions? Just silly, untrustworthy subjective impressions that would make you prefer one amp over the other?

[Evil Twin]

Ergo has impressions which he is cautious about sharing after testing out his current modifications- I will go so far as to say he is surprised, but also that I am not, based on experience derived from completely different circuits over 40 years ago.

I suggest a PM to him if you want more information before he has the confidence to post freely.

[ergo]

Yes, sorry indeed. Goes under Jon's saying that "Slow work takes time".... but there has been a progress behind the curtain.

Had some trouble getting the Pass's ACP+ going. Turned out to be my own mistake from translating a schematic into PCB, so one trace was a bit wrong. Now it works and input dif pair got tuned for lowest THD also over weekend.

I also got the BOZOS soldered up, but not power up yet as my lab power supply seems to need some attention of it's own and one polarity is not working...

Anyhow, I'm quite hopeful that we can put these 3 pre-stages side by side this week for listening and then I'll report back.


****

BTW there was an interesting offline tip from Purifi. They have found that magnetic speaker terminals can influence the performance of this amp quite a bit - "Just a single magnetic washer increased the 6kHz distortion by 15dB. removing it made wonders to the sound." We checked ours - My amp had Hypex branded clear speaker terminals - those turned out to be non magnetic. My friends amp had other terminals where both washers and nuts were magnetic. So that could have been a partial influence in our 'cap upgraded Purifi vs stock Purifi' listening test too.

****

[Evil Twin]

****

BTW there was an interesting offline tip from Purifi. They have found that magnetic speaker terminals can influence the performance of this amp quite a bit - "Just a single magnetic washer increased the 6kHz distortion by 15dB. removing it made wonders to the sound." We checked ours - My amp had Hypex branded clear speaker terminals - those turned out to be non magnetic. My friends amp had other terminals where both washers and nuts were magnetic. So that could have been a partial influence in our 'cap upgraded Purifi vs stock Purifi' listening test too.

****
[ATTACH=CONFIG]29936[/ATTACH]

This is no surprise either. In developing the Ayre Acoustics equipment, Charles Hansen found that any magnetic type materials in the signal path such as binding posts, resistors, etc, were audible and often measurable. Nickel especially seems a problem when it is asked to be a conductor... but in your friend's case, it's most likely that they are simple plated steel washers.


And yes, slow work takes time seems to be a governing point of existence, especially when the work is detailed and demanding...

[ergo]

Generating the +/-60V supplies for testing of Pass BOZOS takes some gear and effort

But it works... though from measurement result does not look too promising. But will be interesting to hear that type of distortion profile (dominated by 3rd harmonic as opposed to mostly 2nd for ACP+)

[Norm]

My dear ergo, I am with you in spirit as the leads from series connected bench supplies to device under test are checked, double checked, and checked once more before power is applied. With 140 or 150 mA showing on all the current readouts one is reminded the esteemed Mr Pass does like his bias, and why a line buffer has such big heat sinks.

[ergo]

I did indeed do just that - meaning verified all step by step with multimeter and used soldered connections instead of the small pin connectors for other circuits etc to make sure they don't come loose and I would not have to go too close to the 120V in end.

Yesterday we did some listening too. This this time the WHAMMY (opamp + complementary MOSFET power stage) and ACP+ (JFET / MOSFET based discrete opamp with very high bias). The impression after an hour or so listening was a small positive bias in WHAMMY direction. ACP+ seemed to add something very nice and pleasant to female vocals, but then listening through the to other instruments then the overall clarity and soundstage seemed better with WHAMMY. Due to BOZOS needing a lot of care with it's high supplies we did not get around to listening to that just yet.

Anyhow - it is becoming more clear every time we listen that the Purifi power stage is very revealing indeed. On the setup I had some years ago base on 'normal' mid class AB intergrated amps the two would probably have sounded pretty much same good. Being able to hear the small differences they have I think speaks about the quality of the amp.

It's very fun to do these experiments when the differences of this level can be heard, so that's why I guess I'm also not in super hurry to wrap it up and make the system static yet.

I might yet to go for 'Evil buffer' as well - just to get to hear what a stage that should measure super well can do and how it sound compared to these where one still can measure some distortion and put a for example '2nd harmonic dominated distortion' labels on them

[mocenigo]

Hello, new member here. I have been building my own Purifi+Neurochrome based amp for some time. Here's some pics.

I am using:
- The EVAL1 kit with the buffer bypassed;
- A Neurochrome Universal Buffer (also very, very revealing, configured for a 7.3 Db gain);
- Connex SMPS800re power supply;
- A +-18V superregulated linear power supply for the modulators; and
- A +-15V superregulated linear power supply for the Neurochrome Universal Buffer.

So far, it sounds just great, well above and beyond my previous NC500 based monos (with SI 994 opamps in the buffer).

[mocenigo]

I might yet to go for 'Evil buffer' as well - just to get to hear what a stage that should measure super well can do and how it sound compared to these where one still can measure some distortion and put a for example '2nd harmonic dominated distortion' labels on them

Hello, what is the "evil buffer"? Also, are you "the slow diyer"?

[ergo]

The 'evil buffer' refers to schematic Evil Twin posted a page ago here. So I guess my poor attempt to be funny.

"the slow diyer" - I don't have nothing against being called slow as long as it includes 'diyer' also in the title Goal is not so much to whip up something fast but rather to learn, gain experience and get things done eventually and as good as possible.

***
As for progress - there is some. Yesterday we had a listening session comparing my Soekris 1021 DAC to Denafrips Venus that my friend just got yesterday. Perhaps it will change a bit as it burns in, but initial impressions were good.
We tried with Soekris and Venus DAC driving the Purifi direct (EVK buffer bypassed). Here the Venus can clearly do a better job and Soekris struggles and does not sound as good as when WHAMMY is between the DAC-s and the Purifi. Overall it still seems that having the WHAMMY in path also does add a bit of ease to Venus presentation too. Overall these DAC-s sounded similar in nature. Neither of us felt that there is any big 'difference in sound character'. Venus did seem to have a bit better control of bass and bass presence.

We also tried a new opamp in WHAMMY - OPA1622 instead of the OPA2134. Similar feeling - very similar character overall, just a bit more air and freedom perhaps, but would take longer listening to really be able to tell one over the other.

Next up is soldering up two WHAMMY modules, so we could run full balanced signal chain. All listening this far has been by converting DAC out to single end and then feeding Purifi with single end signal (with cables built per Hypex documentation recommendations).

I also bought a pair of used Super TeddyReg regulators - so there will be a chance to compare those against the SilentSwitcher in use now.

[Evil Twin]

As always, your detailed comments are interesting and appreciated, Ergo.

I have a well known preference for R2R ladder DACs, and have a Soekris DAC1101 (for desktop and portable applications- bus powered, no usage issues whether in a Tie Fighter or Imperial Shuttle) and the "table top/system rack" DAC1541.

The 1021 board has an output impedance of about 640 ohms, which would lead me to expect it will do better with a high input impedance preamplifier, rather than driving a power amp direct. Only with the most recent TotalDAC have I seen a useful capability for direct drive, and I still find that I prefer preamp based operation- with the right preamp selection.

At some point I should do bench and listening comparisons, including against the Denfrips Terminator with the firmware upgrade, but for the immediate future (3-4 months) speaker development efforts will take precedence.

I have a planned bionic upgrade later this year which may give me more time in my personal laboratory for such explorations. Palpatine did not do a very thorough job for me... and that was many years ago.

[mocenigo]

The 'evil buffer' refers to schematic Evil Twin posted a page ago here. So I guess my poor attempt to be funny.

Ah ok, understood. Well, it seems that Tom Christiansen's new Universal Buffer goes already in that direction

"the slow diyer" - I don't have nothing against being called slow as long as it includes 'diyer' also in the title Goal is not so much to whip up something fast but rather to learn, gain experience and get things done eventually and as good as possible.

Well, there is a blog by "theslowdiyer" that was building the ACP+, WHAMMY, at the same time, and I thought it may have been you - but then why is this guy no reporting about the combinations of these

***
As for progress - there is some. Yesterday we had a listening session comparing my Soekris 1021 DAC to Denafrips Venus that my friend just got yesterday. Perhaps it will change a bit as it burns in, but initial impressions were good.
We tried with Soekris and Venus DAC driving the Purifi direct (EVK buffer bypassed). Here the Venus can clearly do a better job and Soekris struggles and does not sound as good as when WHAMMY is between the DAC-s and the Purifi. Overall it still seems that having the WHAMMY in path also does add a bit of ease to Venus presentation too. Overall these DAC-s sounded similar in nature. Neither of us felt that there is any big 'difference in sound character'. Venus did seem to have a bit better control of bass and bass presence.

We also tried a new opamp in WHAMMY - OPA1622 instead of the OPA2134. Similar feeling - very similar character overall, just a bit more air and freedom perhaps, but would take longer listening to really be able to tell one over the other.

Next up is soldering up two WHAMMY modules, so we could run full balanced signal chain. All listening this far has been by converting DAC out to single end and then feeding Purifi with single end signal (with cables built per Hypex documentation recommendations).

I also bought a pair of used Super TeddyReg regulators - so there will be a chance to compare those against the SilentSwitcher in use now.

I am very, very curious, keep us posted.

In the meantime, this is my build: Connex SMPS800RE regulated power supply at 54V (max power over 8ohm 155W), Neurochrome Universal Buffer, two superregulated power supplies. Sounds heavenly. It is progressing, now it has a case!

[ergo]

mocenigo - nice progress there and I see the amp has found a new home in a nice case too :T

I now have enough parts to solder up 3 more WHAMMY boards. That will allow us to test out the balanced drive throughout and then decide if we stick to WHAMMY as a buffer or experiment some more. My friend just got some OPA1656 for WHAMMY and says he likes these. He also listened to Burson V6 Classic, likes the bass a lot but says the high are not to his liking. I have not heard either of those yet. So I guess that is the plus in WHAMMY that it is easy to keep rolling the opamps to try find the 'best match' to taste and rest of the system. Measurement wise I suspect those would not look very different as newer opamps measure pretty good generally.

The Neurochrome Universal buffer looks pretty good and interesting indeed. I will be close enough to Canada week after next, so I was considering ordering one to try out as well.

[ergo]

I finally got around to building a cable for the purpose of measuring the input impedance of the Purifi module (EVK buffer bypassed)

I did build a dummy with 2k2 resistors in each leg and calibrated the series resistor with that... so result should be accurate. And it is 1.58kOhm over most of the audio range. It is close to what is expected. I just wanted to verify it does not change to lower impedance at some range as datasheet gives only one number and did not specific at what frequency. Datasheet does say 2.2k, but I did measure less. There might be a mistake in my method, but at least today I don't see it.

[5th element]

So 1.58k up until the low pass on the input kicks in it seems.

[JonMarsh]

As usual, tells the story about why low driving impedance is desirable... and tells it in a very systematic way.

[5th element]

I wouldn't call 1.8k low impedance but now you've mentioned it some driver stages would have trouble with it.

The Purifi modules have something like 12dB of gain on their own. To reach maximum amplifier output you'd need a buffer that can swing considerable voltage with, potentially, decent capacitive load drive for the best linearity. Assuming you're only playing with a single ended 2VRMS source.

Most modern, high performance, op amps can do this, but it's certainly not a universal trait. Nor could you opamp roll your favourite of the month into the output of the buffer and assume it would be happy.

Of course the best solution would be driving the inputs of the Purifi module directly with a capable balanced source. This would give you the lowest noise too.

[ergo]

I have two WHAMMY modules up and running now since weekend. Plan is to try to run one as a balanced buffer for one channel and the other for the other channel. I will power them from MEAN WELL AC/DC 24V/900mA supplies -> Super TeddyReg regulators. Each WHAMMY MOSFET stage runs at 37mA class-A at moment, but I can increase that once I move to TeddyRegs instead of the SilentSwitchers I used now that have a total load of 150mA. Will be seen if they are 'symmetrical enough' to get some more benefit out of that than we get with SE drive already.

Why I did measure the impedance in the first place was that I did build the WHAMMY boards so that one can extract the signal also after the opamp. So we did try to drive the Purifi also from opamps direct versus the MOSFET class-A stage in between. Surprisingly even the opamps that have ample current output capability of 50+mA seem to do better with the extra power stage in between. And it's not only the bass control, but also the depth of the sound stage for example. So pretty interesting and not easy for me to grasp either as I would also like to know what it is that makes up that difference and how could I measure it out or put numbers to it

[theSven]

What super regulated linear power supplies are you using? Where did you order your case from?

[ergo]

The case is from Modushop in Italy

https://modushop.biz/site/

I ordered also an extra 4mm front panel and then cut it to dimensions on a small bandsaw to be used as the separator internal wall between PS and the amp.


Meanwhile there was also progress and I got 2xWHAMMY modules ready for balanced buffer operation. I was running a single module with Jan Diddens SilentSwitcher regulator, but with 150mA load capability it was not enough for 2 modules. So upgrade to SuperTeddyReg-s was done feed by two 24V AC to DC modules by MeanWell.

I finished those just before my trip to US and my friend was listening to them meanwhile. Now I'm back but he is on trip, so I still have not yet heard these myself. According to him though there was another small incremental improvement in driving the Purifi with true balanced signal (from Denafrips Venus DAC)

I also replaced the heatsinks on the modules for less tall, so that it would be ready to fit the 80mm height I have inside the enclosure.

[Evil Twin]

Slow work always takes time, and some processes cannot be rushed...

[theSven]

Thank you Ergo. As the days get closer to my Calliope being ready to pick up from Steve I want to start looking into building a 2 channel amp for these. Longer term I want to build the Ardent D’s, but it gets me closer to my goals for the basement project.

[Supernova]

Hey all. Tom has made a a buffer specifically for the 1ET400A and NC500. Plug and play with the SMPS1200.

https://www.diyaudio.com/forums/vendor-s-bazaar/351387-purifi-1et400a-hypex-nc500-input-buffer.html#post6127335

I thought y'all would find this interesting.

Cheers

Peter

[JonMarsh]

Hey all. Tom has made a a buffer specifically for the 1ET400A and NC500. Plug and play with the SMPS1200.

https://www.diyaudio.com/forums/vendor-s-bazaar/351387-purifi-1et400a-hypex-nc500-input-buffer.html#post6127335

I thought y'all would find this interesting.

Cheers

Peter

Very interesting thread, thanks for the link!

[mocenigo]

And now the amp has new connection boards by Neurochrome! In the following thread at DIYaudio I share my experiences.

https://www.diyaudio.com/forums/vendor-s-bazaar/351387-purifi-1et400a-hypex-nc500-input-buffer-2.html#post6133929

High resolution pics:

[theSven]

Ergo, What do you think about that new board that neurochrome is offering for the card instead of needing the evaluation board? Would you still use the buffer with that other card setup?