• 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

Armand

Member
Audio Company
Joined
Jun 25, 2018
Messages
93
Likes
321
@tomchr
When you measure the CMRR on your buffer, is it with equal source impedance or do you use the more rigid test with 10 Ohm imbalance?

Best Armand
 

tomchr

Addicted to Fun and Learning
Audio Company
Forum Donor
Joined
Nov 5, 2018
Messages
984
Likes
2,607
Location
Calgary, Canada
I measure with 10Ω imbalance and report the worst case result.

Tom
 

Armand

Member
Audio Company
Joined
Jun 25, 2018
Messages
93
Likes
321
I measure with 10Ω imbalance and report the worst case result.

Tom
Nice.
You should write something about that on your measurement page. I assume that without the imbalance the results are even better.

On the buffer I make for the Vera amplifiers the CMRR matches yours. Here is a plot with 10 Ohm worst case measured on the output of a complete amplifier.
1618518298575.png


Kind of magical to insert a 4V 50Hz signal on the inputs and measure 70uV on the speaker output :)
 

mocenigo

Major Contributor
Joined
Dec 8, 2018
Messages
1,274
Likes
1,034
Yeah. I doubt you'd gain anything by switching to the Purifi/Hypex Buffer. If you pluck C19, C20, C22, and C23 from the UB and replace R7, R9, R23, and R25 with 0 Ω resistors you end up with the same circuit as is used in the Purifi/Hypex Buffer Rev. 1.2 (production version). If SMD soldering is not your thing, it's sufficient to just pluck the capacitors. Leaving it alone works too. Ain't broken. Don't fix. :)

Tom

Oh no why should I “fix” something that is working incredibly well. I honestly think that the UB + 1ET400A is the best amp I have ever heard, and by a long shot. Again this is subjective but I know that it is also backed with good engineering.

I am in fact so happy with the UB that I would be tempted to build something around your Modulus amps (the only thing I dislike about them is that the names remind me of intel CPUs - and I am an Architect at ARM ;-) ). But this would probably induce a divorce…
 

boXem

Major Contributor
Audio Company
Joined
Jun 19, 2019
Messages
2,014
Likes
4,852
Location
Europe
The EVAL1 buffer is the front-end of an instrumentation amp. You can find a good treatment of it in just about any opamp text. Personally, I recommend Sergio Franco, "Design With Operational Amplifiers and Analog Integrated Circuits". There's a 4th edition out now, though I doubt it's significantly different from the 2nd or 3rd.

You can see the common-mode rejection simulation below. You'll notice the circuit has 0 dB CMRR, i.e. no CMRR.
View attachment 124071

Tom
Sorry Tom but I have to disagree here.
The buffer doesn't have 2 outputs, it has one output that is the difference between OUTP and OUTN. So displaying the common mode gain (and not the CMRR) is done like this:
1618525618246.png

I introduced errors in the resistors because they are the main reason for common mode errors. Otherwise what is displayed is mainly errors from the simulator.
 

boXem

Major Contributor
Audio Company
Joined
Jun 19, 2019
Messages
2,014
Likes
4,852
Location
Europe
I suppose I need to add some precision in my messaging regarding the single-ended vs differential. The output is differential, even with a single-ended input, but it's not balanced. If you drive the EVAL1 with a balanced source, the output becomes balanced.
View attachment 124079
View attachment 124080

In case of the Purifi/Hypex Buffer, I use a differential driver for the output, so the output is differential and balanced even with a single-ended input. That could explain why many have found the Purifi/Hypex Buffer to sound better, though I'm not aware of any actual study confirming this.

Hope this clears things up a bit.

Tom
Same comment here, we do not care about OUTP and OUTN individually, what is important is the difference between the 2.
1618526333915.png


Where your solution is superior is when the single ended input reaches values above 3.5 V RMS (10 Vpp). With EVAL1, the common mode voltage will reach values above the +/- 5 V recommended by Purifi in the 1ET400A datasheet, while staying close to 0 with your solution.
 

tomchr

Addicted to Fun and Learning
Audio Company
Forum Donor
Joined
Nov 5, 2018
Messages
984
Likes
2,607
Location
Calgary, Canada
Very intriguing. That doesn't change the fact, that with a single-ended source, the output of the EVAL1 will contain a common-mode component that the 1ET400A/NC500 will have to reject, whereas the output of the Neurochrome Purifi/Hypex Buffer will have zero common-mode component under the same conditions.
This is a clear and quantifiable difference between the circuits.

Tom
 
OP
J

JimB

Addicted to Fun and Learning
Forum Donor
Joined
Sep 19, 2019
Messages
731
Likes
493
Location
California
Very intriguing. That doesn't change the fact, that with a single-ended source, the output of the EVAL1 will contain a common-mode component that the 1ET400A/NC500 will have to reject, whereas the output of the Neurochrome Purifi/Hypex Buffer will have zero common-mode component under the same conditions.
This is a clear and quantifiable difference between the circuits.

Tom
Intriguing indeed. There certainly are real differences in the characteristics of the two approaches, as well as useful differences in features. A question of interest is how the differences in implementation effect the total amp performance, in particular cases (the later amp stages matter). Several claims of sonic superiority have been made for balanced inputs and the 1ET400A. Here we are focusing on the single-ended case (more challenging?). What is the significance of the CM voltage on the 1ET400A when that CMV is a only fraction of the audio signal itself?
 

tomchr

Addicted to Fun and Learning
Audio Company
Forum Donor
Joined
Nov 5, 2018
Messages
984
Likes
2,607
Location
Calgary, Canada
That's a great question. The common-mode component is a significant fraction of the signal. Without firing up the simulator and just looking at the plots I posted yesterday, I'd say the common-mode component is maybe ~15 dB (5x) lower than the differential signal. The CMRR of the 1ET400A is specified as ">60 dB", so the CM component will be attenuated ~75 dB. I don't know how audible that would be. If any of you would like to set up a scientifically controlled experiment with, say, 250 participants so we can tease that out I'm happy to help. :) But some claim to be able to hear -80 dB THD, so that's one "back of the envelope" figure that I keep in mind.
At least the CM component is of the same frequency and phase as the input signal, so I would think it would be hard to hear.

Another question is whether the Purifi 1ET400A suffers from common-mode distortion. That's an issue for many opamps, for example. That's why some prefer to use an inverting amp rather than a non-inverting one. The first half of my Super GainClone Review covers that in better detail (with measurements).

When I designed the Purifi 1ET400A/Hypex NC500 input buffer, I honestly didn't think that a differential and balanced output would make any difference compared to the instrumentation amplifier front-end used in the EVAL1. As pointed out earlier in this thread, both the EVAL1 buffer and my Purifi/Hypex Buffer measure well. And this is why @mocenigo's buffer does not have the differential and balanced output. Someone on DIY Audio pointed out that many claimed a differential and balanced output sounded better with the 1ET400A, and suggested that I offered that. I took that as an opportunity to offer something that is unique and different from the EVAL1. The output of the EVAL1 will contain a common-mode component when driven by a single-ended source (e.g., an RCA output), whereas the output of the Purifi/Hypex Buffer will not.
The differential and balanced output of the Purifi/Hypex Buffer increased the production cost by a few percent and there was a bit of room left on the board, so why not include it? Including it certainly won't harm anything and many say it sounds better. Sometimes you have to deliver what the market wants. There are many buffers out there that are basically copies of the EVAL1 with a few tweaks. I decided to be unique. I did no harm and I may have provided a benefit. What's not to like? I think we can all coexist.

Tom
 
Last edited:
OP
J

JimB

Addicted to Fun and Learning
Forum Donor
Joined
Sep 19, 2019
Messages
731
Likes
493
Location
California
That's a great question. The common-mode component is a significant fraction of the signal. Without firing up the simulator and just looking at the plots I posted yesterday, I'd say the common-mode component is maybe ~15 dB (5x) lower than the differential signal. The CMRR of the 1ET400A is specified as ">60 dB", so the CM component will be attenuated ~75 dB. I don't know how audible that would be. If any of you would like to set up a scientifically controlled experiment with, say, 250 participants so we can tease that out I'm happy to help. :) But some claim to be able to hear -80 dB THD, so that's one "back of the envelope" figure that I keep in mind.
At least the CM component is of the same frequency and phase as the input signal, so I would think it would be hard to hear.

Another question is whether the Purifi 1ET400A suffers from common-mode distortion. That's an issue for many opamps, for example. That's why some prefer to use an inverting amp rather than a non-inverting one. The first half of my Super GainClone Review covers that in better detail (with measurements).

When I designed the Purifi 1ET400A/Hypex NC500 input buffer, I honestly didn't think that a differential and balanced output would make any difference compared to the instrumentation amplifier front-end used in the EVAL1. As pointed out earlier in this thread, both the EVAL1 buffer and my Purifi/Hypex Buffer measure well. And this is why @mocenigo's buffer does not have the differential and balanced output. Someone on DIY Audio pointed out that many claimed a differential and balanced output sounded better with the 1ET400A, and suggested that I offered that. I took that as an opportunity to offer something that is unique and different from the EVAL1. The output of the EVAL1 will contain a common-mode component when driven by a single-ended source (e.g., an RCA output), whereas the output of the Purifi/Hypex Buffer will not.
The differential and balanced output of the Purifi/Hypex Buffer increased the production cost by a few percent and there was a bit of room left on the board, so why not include it? Including it certainly won't harm anything and many say it sounds better. Sometimes you have to deliver what the market wants. There are many buffers out there that are basically copies of the EVAL1 with a few tweaks. I decided to be unique. I did no harm and I may have provided a benefit. What's not to like? I think we can all coexist.

Tom
In this special case the CMV IS the audio signal. So, does that simply attenuate (or boost) the net signal in the 4th decimal place? Or, as you say, is it a challenge to the 1ET400A, producing some CM distortion? We don't see anything strange 70-80dB down in the results @amirm got when using the EVAL1-based Purifi demo. However, your implementation just seems a bit more fundamentally "sound" for marginal conditions. I don't think Purifi suggests that one cannot best the FE02 input buffer, just that it is adequate to demostrate/evaluate the performance of their amp.
 

tomchr

Addicted to Fun and Learning
Audio Company
Forum Donor
Joined
Nov 5, 2018
Messages
984
Likes
2,607
Location
Calgary, Canada
Can I ask what the purpose/benefit of the optional voltage regulator is?
As far as I can measure, there is no advantage of using the voltage regulator. I measure the same performance when I use the regulated AUX supply from the SMPS1200 as I do when I use the optional regulator and I'm happy to tell my customers that. Some choose to geek out over voltage regulators and I support that. DIY is a hobby.

We don't see anything strange 70-80dB down in the results @amirm got when using the EVAL1-based Purifi demo.
True. But I think he used the balanced input, so you would not see a difference. At least not one that could be attributed to the common-mode voltage. I could be wrong. I didn't read that specific review, but in general my observation is that Amir uses the balanced input if one is available, so I would assume that he did so on the Purifi+EVAL1 as well.

Tom
 
OP
J

JimB

Addicted to Fun and Learning
Forum Donor
Joined
Sep 19, 2019
Messages
731
Likes
493
Location
California
... I think he used the balanced input, so you would not see a difference. At least not one that could be attributed to the common-mode voltage. I could be wrong. I didn't read that specific review, but in general my observation is that Amir uses the balanced input if one is available, so I would assume that he did so on the Purifi+EVAL1 as well.

Tom
Good point. I think you're right. With many(?) using single ended sources, it would be interesting to know what the performance difference is, if any.

Thoughts, @amirm ?
 
OP
J

JimB

Addicted to Fun and Learning
Forum Donor
Joined
Sep 19, 2019
Messages
731
Likes
493
Location
California
Good point. I think you're right. With many(?) using single ended sources, it would be interesting to know what the performance difference is, if any.

Thoughts, @amirm ?
I also wonder whether the amp output being single ended (not balanced) matters. We lack the necessary details to assess that. A direct test is likely needed to show the differences.
 

ABall

Senior Member
Forum Donor
Joined
Jan 5, 2020
Messages
329
Likes
159
Location
Reading (UK)
Forgive me if its been mentioned but does anyone know if these amps produce enough power with the gain stage bypassed if you are using 94dB speakers? My pre has no gain, DCB1 buffer so source voltage only, room is 12x12. Also with the input impedance been low with no gain stage, I am wondering if the DCB1 would operate at its best? Oh it will be single ended, even less gain, maybe I should think about Toms input stage with my setup?
 
OP
J

JimB

Addicted to Fun and Learning
Forum Donor
Joined
Sep 19, 2019
Messages
731
Likes
493
Location
California
Forgive me if its been mentioned but does anyone know if these amps produce enough power with the gain stage bypassed if you are using 94dB speakers? My pre has no gain, DCB1 buffer so source voltage only, room is 12x12. Also with the input impedance been low with no gain stage, I am wondering if the DCB1 would operate at its best? Oh it will be single ended, even less gain, maybe I should think about Toms input stage with my setup?
The amp has the same power potential regardless of the gain of the input buffer gain stage. You need the SOURCE driving it to have the range to do that. I don't know what you have, so I can't comment on it's suitability to achieve full output. What is the nominal impedance of your speakers? With 94 dB sensitivity, I doubt you'd ever need to achieve the 200/400W these amps can produce.
 

tomchr

Addicted to Fun and Learning
Audio Company
Forum Donor
Joined
Nov 5, 2018
Messages
984
Likes
2,607
Location
Calgary, Canada
Forgive me if its been mentioned but does anyone know if these amps produce enough power with the gain stage bypassed if you are using 94dB speakers? My pre has no gain, DCB1 buffer so source voltage only, room is 12x12. Also with the input impedance been low with no gain stage, I am wondering if the DCB1 would operate at its best? Oh it will be single ended, even less gain, maybe I should think about Toms input stage with my setup?
There isn't enough information here to answer your question, but I'll try anyway.

I'm going to assume that you mean 12x12 feet (4x4 meter), so your listening position is probably ~2m from the speakers. I have no idea how loud you prefer your music, so I'm going to assume that you value your hearing and don't crank it past 85 dB SPL average, 100 dB peak at the listening position. I'm also going to assume that you have a stereo system (two speakers). I'm further going to assume that they're 4 Ω speakers.

The 2 meter distance results in 6 dB lower SPL than the 1 m reference level.
Because you're using two speakers instead of one you get 3 dB higher SPL, assuming the two speakers play completely uncorrelated signal. Naturally, the information in the left and right channels will be correlated to some degree, so 3 dB is a conservative estimate. If the two channels are 100% correlated (i.e. if you play a mono signal), you'd get 6 dB higher SPL.

So for a 1 W input, you'll get 94 - 6 + 3 = 91 dB SPL at the listening position. To get 100 dB peaks, you need an additional 100-91 = 9 dB. 9 = 3+3+3 and +3 dB is a doubling of power. So you'll need 1*2*2*2 = 8 W to produce 100 dB SPL at the listening position.

To dissipate 8 W into a 4 Ω speaker requires E = sqrt(P * R) = sqrt(8*4) = sqrt(32) = 5.65 V RMS.

The gain of the Purifi 1ET400A is 12.8 dB. That's 10^(12.8/20) = 4.37 V/V, so the source must produce 5.65/4.37 = 1.30 V RMS. That's possible for just about any modern source except phones and tablets. So, yes. The Neurochrome Purifi/Hypex Buffer will provide sufficient gain when used at unity gain to produce sufficient SPL under the constraints given above. If your requirements are different, you'll need to work through the math.

Don't like math much? Here's a shortcut: SPL Calculator.

Note that this calculation does not account for the room. It also assumes that the two speakers produce completely uncorrelated signals. So in reality, you should be able to get a bit higher SPL - at least away from the room modes.

Now, if you build your amp with the Purifi/Hypex Buffer set to unity gain, you will probably find that it appears "less powerful", i.e., you have to turn the volume control past 9 o'clock to get sufficient SPL. That's a function of the low gain and says nothing about the amp's ability to provide power.

Tom
 
Last edited:

ABall

Senior Member
Forum Donor
Joined
Jan 5, 2020
Messages
329
Likes
159
Location
Reading (UK)
The amp has the same power potential regardless of the gain of the input buffer gain stage. You need the SOURCE driving it to have the range to do that. I don't know what you have, so I can't comment on it's suitability to achieve full output. What is the nominal impedance of your speakers? With 94 dB sensitivity, I doubt you'd ever need to achieve the 200/400W these amps can produce.
Thanks JimB, I realise the amps can still produce full power regardless, I guess I should of said that my sources are all standard line level devices, Gustard x16, Schiit phono, Onkyo AVR. Speakers are 8ohm. I doubt many people will need to drive these amps to full output so I dont see why it gets mentioned so much as if it MUST be factored for. I'm not aiming for full output, just loud enough. I think Tom has provided all the info I need, I just have to read it a few more times. :)
 

ABall

Senior Member
Forum Donor
Joined
Jan 5, 2020
Messages
329
Likes
159
Location
Reading (UK)
There isn't enough information here to answer your question, but I'll try anyway.

I'm going to assume that you mean 12x12 feet (4x4 meter), so your listening position is probably ~2m from the speakers. I have no idea how loud you prefer your music, so I'm going to assume that you value your hearing and don't crank it past 85 dB SPL average, 100 dB peak at the listening position. I'm also going to assume that you have a stereo system (two speakers). I'm further going to assume that they're 4 Ω speakers.

The 2 meter distance results in 6 dB lower SPL than the 1 m reference level.
Because you're using two speakers instead of one you get 3 dB higher SPL, assuming the two speakers play completely uncorrelated signal. Naturally, the information in the left and right channels will be correlated to some degree, so 3 dB is a conservative estimate. If the two channels are 100% correlated (i.e. if you play a mono signal), you'd get 6 dB higher SPL.

So for a 1 W input, you'll get 94 - 6 + 3 = 91 dB SPL at the listening position. To get 100 dB peaks, you need an additional 100-91 = 9 dB. 9 = 3+3+3 and +3 dB is a doubling of power. So you'll need 1*2*2*2 = 8 W to produce 100 dB SPL at the listening position.

To dissipate 8 W into a 4 Ω speaker requires E = sqrt(P * R) = sqrt(8*4) = sqrt(32) = 5.65 V RMS.

The gain of the Purifi 1ET400A is 12.8 dB. That's 10^(12.8/20) = 4.37 V/V, so the source must produce 5.65/4.37 = 1.30 V RMS. That's possible for just about any modern source except phones and tablets. So, yes. The Neurochrome Purifi/Hypex Buffer will provide sufficient gain when used at unity gain to produce sufficient SPL under the constraints given above. If your requirements are different, you'll need to work through the math.

Don't like math much? Here's a shortcut: SPL Calculator.

Note that this calculation does not account for the room. It also assumes that the two speakers produce completely uncorrelated signals. So in reality, you should be able to get a bit higher SPL - at least away from the room modes.

Now, if you build your amp with the Purifi/Hypex Buffer set to unity gain, you will probably find that it appears "less powerful", i.e., you have to turn the volume control past 9 o'clock to get sufficient SPL. That's a function of the low gain and says nothing about the amp's ability to provide power.

Tom
Thank you for such a detailed answer Tom, its much appreciated. It looks like I will be a bit short of power then as my speakers are 8ohm, I was really trying to take all the different buffer offerings off the table. I am currently running TPA3116 which I modified for lower gain and better linear response as per the original spec and it has more than enough power. The Purifi is just one of those itches I feel I have to scratch, if your buffer does no more than enable full output for single ended sources then I would probably stick to the eval board, I certainly dont need full output. It's very refreshing to here you are open about the performance of your designs, I was on the cusp of trying your 686 after reading about it on AOS before the purifi hit the scene and I purchased a new pair of speakers. The chap on AOS who builds and sells valve amps shot himself in the foot when he purchased your 686 then said it was an "end game amplifier" I doubt he sells many valve amps now. :D
 
Last edited:

tomchr

Addicted to Fun and Learning
Audio Company
Forum Donor
Joined
Nov 5, 2018
Messages
984
Likes
2,607
Location
Calgary, Canada
It looks like I will be a bit short of power then as my speakers are 8ohm, I was really trying to take all the different buffer offerings off the table.
You could add a resistor (SMD or through-hole) to the Purifi/Hypex Buffer and up the gain a bit. It would be pretty easy to configure it for, say, +6 dB which could be the optimum you're looking for.

The Purifi is just one of those itches I feel I have to scratch, if your buffer does no more than enable full output for single ended sources then I would probably stick to the eval board, I certainly dont need full output.
The Purifi/Hypex Buffer can be used in place of the EVAL1 or EVAL2 for those who prefer a bit more flexibility. It's more of a "tweaker's delight" as it allows for experimenting with voltage regulators, remote sensing, gain, etc. Its main advantage over the EVAL1/2 is the differential and balanced output, which remains balanced even with a single-ended source. Many say that type of drive sounds better with the Purifi/Hypex amps and better than the EVAL1/2 buffers. Both buffers measure well, so I can't point at a difference in the measurements that would be responsible for the perceived difference, but at the same time I also can't deny that these people perceive a difference.

The chap on AOS who builds and sells valve amps shot himself in the foot when he purchased your 686 then said it was an "end game amplifier" I doubt he sells many valve amps now. :D
My bad... :)

I sold my Modulus-686 which left me without an amp for a while. I wanted to build a demo amp with my Purifi/Hypex Buffer and the 1ET400A, so that's what I'm currently using. I miss the Modulus-686. I need to build more. That's another case of "both amps measure well, but...". I definitely prefer the Modulus-686. I find the Purifi amp to be a bit harsher and to have a narrower sound stage. But if you want a Class D amp, I will recommend the Purifi without hesitation. I power two channels with a Hypex SMPS1200A400.

If you don't quite need the power of the Modulus-686, the Modulus-286 and Modulus-186 is where it's at. And Modulus-86 if you prefer to solder everything yourself. All four have the same sonic signature (and stellar measurements) when operated below clipping. The MOD86 is a bit noisier than the rest as it uses a different front-end. For more information about the tradeoffs between the different amps, have a look here: Choosing Your Neurochrome Amplifier.
The Modulus-86 and -186 provide 40 W (8 Ω), 65 W (4 Ω). The -286 provides 65 W (8 Ω), 125 W (4 Ω). The Modulus-686 is the most powerful and provides 240 W (8 Ω), 360 W (4 Ω) when powered by a ±36 V supply. I prefer to use slightly lower voltage (±27 V) as that allows me to use an SMPS and a smaller chassis. I call that the Safe-n-Sane build. It provides 130 W (8 Ω), 200 W (4 Ω) under those conditions.

Note that I rate my amps conservatively. I.e. the power specs are for 0.1 % THD+N. The Purifi amps are rated at 1 % THD+N.

Tom
 
Last edited:
Top Bottom