I use stainless cap head (hex) screws and a brass washer between the screw head and chip if using the insulated version. Also lightly sand the face of the chip and heat sink for smoother surface.
I think the non-insulated chips work better. Mica pads, thermal grease and shoulder washers. Plus a brass washer under the cap head.
I think the non-insulated chips work better. Mica pads, thermal grease and shoulder washers. Plus a brass washer under the cap head.
And I would be the second in line.
After hearing the FE, I can only praise the late Penasa's gift to the community.
I'll build the mod-686 in time, too. Measurements are mostly done (I'm sure also the last two I requested will be good), but only my ears will get the full picture.
Pass vfet part1 would be my first suggestion, with beautifully flat THD-freq measure, or perhaps a Hiraga if you fancy a faster amp.
ClaveFremen (Dario) sent me a couple of Rev. 1.6 MyRef FE boards a few years back so I could measure the performance of the MyRef FE. I built one of the boards and finally had time to measure it. I've attached a good summary graph of the MyRef FE vs Modulus-86.
The graph shows the THD+N vs output power for the MyRef FE and the Modulus-86 when powered by a pair of HP 6643A lab power supplies and when powered by a 2x25 V Antek AN-5225 power transformer.
You can find the rest of the measurements here: Review: MyRef Fremen Edition / MyRef FE – Neurochrome
Dario read through my review immediately after its publication and provided feedback, which I integrated. He also requested that I commented on the Build Guide, so I added that section.
Overall, the MyRef is a quite capable amp. It's a bit unfortunate that it has relatively high output noise (69 uV RMS, unweighted). Its high gain (30 dB) will also result in greater amplification of source noise. The MyRef is also a bit sensitive to the power supply as seen in the slight degradation in THD+N when used with the recommended power transformer. I doubt the degradation is audible, but it's a bit unfortunate that some performance was left on the table. I think those who wish to further improve on the MyRef FE can do so by following my recommendations for grounding provided here: Taming the LM3886 Chip Amplifier: Grounding – Neurochrome
Tom
The graph shows the THD+N vs output power for the MyRef FE and the Modulus-86 when powered by a pair of HP 6643A lab power supplies and when powered by a 2x25 V Antek AN-5225 power transformer.
You can find the rest of the measurements here: Review: MyRef Fremen Edition / MyRef FE – Neurochrome
Dario read through my review immediately after its publication and provided feedback, which I integrated. He also requested that I commented on the Build Guide, so I added that section.
Overall, the MyRef is a quite capable amp. It's a bit unfortunate that it has relatively high output noise (69 uV RMS, unweighted). Its high gain (30 dB) will also result in greater amplification of source noise. The MyRef is also a bit sensitive to the power supply as seen in the slight degradation in THD+N when used with the recommended power transformer. I doubt the degradation is audible, but it's a bit unfortunate that some performance was left on the table. I think those who wish to further improve on the MyRef FE can do so by following my recommendations for grounding provided here: Taming the LM3886 Chip Amplifier: Grounding – Neurochrome
Tom
Attachments
Last edited:
That's an interesting review, thanks 
Another possibility for the higher noise floor might be the big input cap ? This reminds me of this old thread : Big film caps and noise
Another possibility for the higher noise floor might be the big input cap ? This reminds me of this old thread : Big film caps and noise
It's a common misconception that film caps pick up noise. They can pick up mains hum and rectification hash, but that would show as an increase in the 60 Hz and harmonics, which does not seem to be the case. In fact, I tried shorting the capacitor by soldering a wire across it. That made no difference. That said, physically smaller is usually better for a wide range of things, EMI/RFI included.
Also, if the capacitor was picking up hum, it would do so even when the amp is at idle, so you would see a degradation in the noise floor when the amp is powered by a transformer. That's not the case for the MyRef FE. Rather the degradation occurs when the amp is producing a signal. That indicates to me that the source of the degradation should be found in the PCB layout.
Tom
Also, if the capacitor was picking up hum, it would do so even when the amp is at idle, so you would see a degradation in the noise floor when the amp is powered by a transformer. That's not the case for the MyRef FE. Rather the degradation occurs when the amp is producing a signal. That indicates to me that the source of the degradation should be found in the PCB layout.
Tom
Attachments
Last edited:
Tom's review
I want to thank Tom for the excellent work he did with measurements and for the suggestions he was so kind to give on how and where he would improve the design.
The review is fair, Tom higlights the flaws or things he consider to be improved (like noise figures and mounting limits) but at the same time appreciates the excellent (his words) performance of the amp in severeal aspects while it finds a pity that some of it is lost for noise and/or grounding.
There is just one thing that I find weird in his review, where he states:
"The noise level is a bit on the high side at 69 µV RMS (unweighted, 20 Hz - 20 kHz bandwidth). That's high enough that I notice it with my ears within 30 cm of a relatively inefficient (85 dB @ 1 W, 1 m) bookshelf speaker."
In my several builds to be able to hear some noise I had to put my ears at 5-10 cm from the tweeter with 90dB speakers and other builders experience is similar.
How is this possible? Maybe Tom's place has lower ambient noise (mine is 33dB) or Tom has better ears.
To whom could misinterpret Tom's word, this will not limit, in any case, the enjoyment of the amp.
Another thing i would like to point out is about this:
"During this listening test I also noticed that my cellphone (40-50 cm away from the amp) would cause a burr/buzzing noise in the speaker when it synced with the cell tower."
This is absolutely true if boards are on free air or in an acrylic case but when the amp is mounted in a closed metal case no cellphone noise will be captured.
I want to thank Tom for the excellent work he did with measurements and for the suggestions he was so kind to give on how and where he would improve the design.
The review is fair, Tom higlights the flaws or things he consider to be improved (like noise figures and mounting limits) but at the same time appreciates the excellent (his words) performance of the amp in severeal aspects while it finds a pity that some of it is lost for noise and/or grounding.
There is just one thing that I find weird in his review, where he states:
"The noise level is a bit on the high side at 69 µV RMS (unweighted, 20 Hz - 20 kHz bandwidth). That's high enough that I notice it with my ears within 30 cm of a relatively inefficient (85 dB @ 1 W, 1 m) bookshelf speaker."
In my several builds to be able to hear some noise I had to put my ears at 5-10 cm from the tweeter with 90dB speakers and other builders experience is similar.
How is this possible? Maybe Tom's place has lower ambient noise (mine is 33dB) or Tom has better ears.
To whom could misinterpret Tom's word, this will not limit, in any case, the enjoyment of the amp.
Another thing i would like to point out is about this:
"During this listening test I also noticed that my cellphone (40-50 cm away from the amp) would cause a burr/buzzing noise in the speaker when it synced with the cell tower."
This is absolutely true if boards are on free air or in an acrylic case but when the amp is mounted in a closed metal case no cellphone noise will be captured.
I want to thank Tom for the excellent work he did with measurements and for the suggestions he was so kind to give on how and where he would improve the design.
Dario: in Tom’s review (build guide section) he states that "All [board] revisions up to Rev. 1.6 have the polarity on a capacitor marked incorrectly as well. "
I’m looking at a Rev1.05 board right now. What Cap position is he referencing?
Thanks, Pete.
According to page 21 in the current revision of the Build Guide, the polarity markings on C32 are incorrect up to Rev. 1.6. C32 is a film cap, which is not a polarized component, so I'm guessing the polarity markings on the board are intended to get the outer film in the cap connected to the lowest impedance node.
According to page 17, the direction markings on R11 are incorrect on revisions 1.5 and 1.6 as well. I missed that in my review. Resistors are not directional...
Tom
According to page 17, the direction markings on R11 are incorrect on revisions 1.5 and 1.6 as well. I missed that in my review. Resistors are not directional...
Tom
My hearing is normal according to my latest hearing test and I live within what's considered "inner city" in Calgary. My subjective experience is my subjective experience.
"33dB" is meaningless without a reference. 33 dB(A)? 33 dB(C)? Min lab sits at 29.7 dB(A).
On the other hand, 69 µV is 69 µV. You can compare that against other amps you've used to see if the MyRef FE is quiet enough for you. This is why I rely on measurements and not subjective experiences.
Tom
Last edited:
Tom already answered and I obviously confirm that cap and resistor orientation does not impact operation.
In case of the capacitors the orientation is about the outer foil as Tom correctly pointed out, for resitors is about my anedoctal experience that resistors have a direction where they sound a tiny bit better, don't expect Tom to support this last statement in any way
Hi all,
I would like to repeat here my notes about Tom's measurement, which posted in the other thread.
First of all, would like to thank Tom for his work. I think he made a very nice and correct assesment, as usual of him.
I do agree with most of the results - and ask to differ with many of the conclusions, but I hope that is permitted democratically..
There is only one thing which bugs me a bit, I would be happy if it could be discussed here, or somewhere else, doesn't matter.
Tom reaches to a conclusion that the Myref amp is having a relatively high noise contribution, with respect to the Mod86.
My problem is, that I see it properly on the contrary, and would like to clarify it, if possible.
From Tom's measurements it results that the Myref produces at 20Hz - 20KHz range, unweighted: 69uV RMS noise on the output;
From his spec's for the mod86, it results that the mod86 produces at 20Hz-20kHz range, unweighted: 40uV RMS noise on the output.
Now Tom' s conclusion is that the Myref is noisy, and one should explore the reasons..
Let me put it differently: the total gain of the Myref is 31,6.
The input referred noise in 20Hz-20kHz, unweighted, RMS is 69uV/31,7 = 2,17uV
The total gain of mod86 is : 10.
The input referred noise in 20Hz-20kHz, unweighted, RMS is 40uV /10 = 4uV
That is ~double of the Myref's input stage contribution, and so the proper question here is: what is causing this in the Mod86?
Apart from this, I really would like to thank again for this effort of Tom.
All the best, George
I would like to repeat here my notes about Tom's measurement, which posted in the other thread.
First of all, would like to thank Tom for his work. I think he made a very nice and correct assesment, as usual of him.
I do agree with most of the results - and ask to differ with many of the conclusions, but I hope that is permitted democratically..
There is only one thing which bugs me a bit, I would be happy if it could be discussed here, or somewhere else, doesn't matter.
Tom reaches to a conclusion that the Myref amp is having a relatively high noise contribution, with respect to the Mod86.
My problem is, that I see it properly on the contrary, and would like to clarify it, if possible.
From Tom's measurements it results that the Myref produces at 20Hz - 20KHz range, unweighted: 69uV RMS noise on the output;
From his spec's for the mod86, it results that the mod86 produces at 20Hz-20kHz range, unweighted: 40uV RMS noise on the output.
Now Tom' s conclusion is that the Myref is noisy, and one should explore the reasons..
Let me put it differently: the total gain of the Myref is 31,6.
The input referred noise in 20Hz-20kHz, unweighted, RMS is 69uV/31,7 = 2,17uV
The total gain of mod86 is : 10.
The input referred noise in 20Hz-20kHz, unweighted, RMS is 40uV /10 = 4uV
That is ~double of the Myref's input stage contribution, and so the proper question here is: what is causing this in the Mod86?
Apart from this, I really would like to thank again for this effort of Tom.
All the best, George
Fine, we ruled out ambient or hearing differences
I'm not arguing about your measurement, nevertheless this is not my (and others) experience.
It would be nice to present 69uV of noise or even a simple sinewave from another source to your test speakers and see if you still can hear it from 30cm.
I'm sure it's not you case but other builders experienced noise/hum that disapperead simply rotating some degrees their toroidal transformer, just to say that what you heard could have been captured noise or something else.
So, I would like to repeat: it is not the myref FE that is noisy.. It is 2uV input referred noise, in the same bandwith and and test setup, where the Modulus86 measures 4uV.
Tom has noted that his more recent designs corrected that, and they are showing between 1.25uV - 1.4uV input referred noise, 20-20kHz, unweighted.
I would like to note that in my FE unit, and in my Evolution amplifier the total output noise, 20Hz-20kHz, unweighted, is 38uV. That is equivalent of 38uV/31,6 = 1,2uV input referred noise.
Furthermore, that result is greatly influenced by that 3,3kohm series input filter resistance, with it's thermalnoise contribution. Placing there a 100ohm resistance, (and eliminating temporarily the input RF filter) - the residual input referred noise of the total system is 650nV, producing 18uV total output noise on the amplifier with 31,6 times gain.
This is in response to Tom's note that he has lower noise choices in his product range.
Also do we, the Myref gang.
But.
Now I would like to ask.. 2uV is a problem, but 1,4uV is ok?
In other words: the difference between that MyRef 'classic' version and Mod86 is : 69uV output noise vs. 40uV output noise of the Mod86. (at 3 times less gain)
The power difference between 69uV RMS and 40uV RMS, on 8ohm speaker: is 400pW. 400*10^-12 watt..
The efficiency of a home-hifi system is ~1%.
So we are having a problem with a 4pW acoustic noise output in our room...
Really?
Last note: for me, it is a nice fresh news that 'Gain' in an amplifier - is a problem..
I would have thought that the main service from an amplifier module is... Hm.. Amplification of a signal..
Something that I am asked to pay for..
Now for seriously: I have a nice PCM63 based dac, amongst many other at home. Single ended valve output, cca 1Vrms.
How could I listen to such a device, supposing I had voted for a Mod86 at the heart of my system?
Or a valve based RIAA phono pre, with a similar output, of my friend. Or an FM radio tuner, for those friends in the UK & BBC...
Just asking..
All the best, George
Tom has noted that his more recent designs corrected that, and they are showing between 1.25uV - 1.4uV input referred noise, 20-20kHz, unweighted.
I would like to note that in my FE unit, and in my Evolution amplifier the total output noise, 20Hz-20kHz, unweighted, is 38uV. That is equivalent of 38uV/31,6 = 1,2uV input referred noise.
Furthermore, that result is greatly influenced by that 3,3kohm series input filter resistance, with it's thermalnoise contribution. Placing there a 100ohm resistance, (and eliminating temporarily the input RF filter) - the residual input referred noise of the total system is 650nV, producing 18uV total output noise on the amplifier with 31,6 times gain.
This is in response to Tom's note that he has lower noise choices in his product range.
Also do we, the Myref gang.
But.
Now I would like to ask.. 2uV is a problem, but 1,4uV is ok?
In other words: the difference between that MyRef 'classic' version and Mod86 is : 69uV output noise vs. 40uV output noise of the Mod86. (at 3 times less gain)
The power difference between 69uV RMS and 40uV RMS, on 8ohm speaker: is 400pW. 400*10^-12 watt..
The efficiency of a home-hifi system is ~1%.
So we are having a problem with a 4pW acoustic noise output in our room...
Really?
Last note: for me, it is a nice fresh news that 'Gain' in an amplifier - is a problem..
I would have thought that the main service from an amplifier module is... Hm.. Amplification of a signal..
Something that I am asked to pay for..
Now for seriously: I have a nice PCM63 based dac, amongst many other at home. Single ended valve output, cca 1Vrms.
How could I listen to such a device, supposing I had voted for a Mod86 at the heart of my system?
Or a valve based RIAA phono pre, with a similar output, of my friend. Or an FM radio tuner, for those friends in the UK & BBC...
Just asking..
All the best, George
The MyRef FE produces 69 µV of white noise at its output. I can hear this with my relatively inefficient bookshelf speakers as hiss from the tweeters. I hear this when I have my head within 30 cm of the speakers.
Unless you believe that 69 µV of white noise from the MyRef FE is somehow different from 69 µV of white noise from another noise source, my speakers will produce the same SPL with the two sources. Implying otherwise would mean that the speakers had some sort of mind of their own or memory that allowed them to reproduce higher SPL with one 69 µV source than with another. That's absurd. Last I checked speakers were linear (relatively anyway), time-invariant, and causal systems.
As we discussed via email, I believe the LM318 is the dominant noise contributor in the MyRef FE. You can find the voltage noise graph for the National Semiconductor LM318 attached here. You can find the full data sheet here: http://www.mit.edu/~6.301/LM118.pdf. The TI data sheet does not have any noise data.
You're looking at around 11 nV/rtHz at 10 kHz and higher at lower frequencies. The calculation gets a bit involved with the relatively high 1/f component, but I think a reasonable estimate is:
En_out = Gain * En_in * sqrt(BW) = 31.62*(11e-9 * sqrt(19000) + 20e-9 * sqrt(1000)) = 67.95 µV RMS
Basically I'm estimating the total noise by approximating the noise voltage as outlined with the red line in the second plot. It's probably a bit pessimistic, but not overly so.
Tom
Unless you believe that 69 µV of white noise from the MyRef FE is somehow different from 69 µV of white noise from another noise source, my speakers will produce the same SPL with the two sources. Implying otherwise would mean that the speakers had some sort of mind of their own or memory that allowed them to reproduce higher SPL with one 69 µV source than with another. That's absurd. Last I checked speakers were linear (relatively anyway), time-invariant, and causal systems.
As we discussed via email, I believe the LM318 is the dominant noise contributor in the MyRef FE. You can find the voltage noise graph for the National Semiconductor LM318 attached here. You can find the full data sheet here: http://www.mit.edu/~6.301/LM118.pdf. The TI data sheet does not have any noise data.
You're looking at around 11 nV/rtHz at 10 kHz and higher at lower frequencies. The calculation gets a bit involved with the relatively high 1/f component, but I think a reasonable estimate is:
En_out = Gain * En_in * sqrt(BW) = 31.62*(11e-9 * sqrt(19000) + 20e-9 * sqrt(1000)) = 67.95 µV RMS
Basically I'm estimating the total noise by approximating the noise voltage as outlined with the red line in the second plot. It's probably a bit pessimistic, but not overly so.
Tom
Attachments
That's fine and dandy, but it's the output noise that matters for the sound quality as that's what you hear as hiss in the speakers.
The output noise of the Modulus-86 is 40 µV RMS (unweighted). The output noise of the Modulus-186 and -286 is 28.5 µV RMS (unweighted). The output noise of the Modulus-686 is 25 µV RMS (unweighted).
The MyRef FE has very high gain (30 dB). This means that any noise of the source will be gained up by a factor of 31.62. Take something like the MiniDSP 4x10HD, which many use for active crossovers. The unbalanced output of the 4x10HD has a noise voltage of 13.3 µV RMS (unweighted). This means your speakers will see a noise voltage of:
- Modulus-86: En_total = sqrt(En_amp^2 + (En_source*Gain)^2) = sqrt(40e-6^2 + (10*13.3e-6)^2) = 138.9 µV
- MyRef FE: En_total = sqrt(En_amp^2 + (En_source*Gain)^2) = sqrt(69e-6^2 + (31.62*13.3e-6)^2) = 426.2 µV
I know which I prefer...
Tom
Good that You brought it up again, almost forgot about..
If that is a problem, there are some choices: if the MiniDSP is that noisy - use something better..
If the MiniDSP is not really noisy, but rather and simply 'hot' in output - hm, it's called a 'potentiometer', which would also provide that extra service of not sending in clipping the amplifier..
If that is a problem, there are some choices: if the MiniDSP is that noisy - use something better..
If the MiniDSP is not really noisy, but rather and simply 'hot' in output - hm, it's called a 'potentiometer', which would also provide that extra service of not sending in clipping the amplifier..
I have done a short jump to ASR, Stereophile tests.
XTZ EDGE A2-300 Review - set to 29dB by Amir (suggested by Thx)
Cambridge Edge A - 31,5dB and 37,5dB (balanced)
Mytek Brooklyn Amp - switchable, 23dB and 29dB
Musical Fidelity M2si - set to 29dB by Amir, (so probably even higher default gain)
Again : the Myref family : 30dB.
XTZ EDGE A2-300 Review - set to 29dB by Amir (suggested by Thx)
Cambridge Edge A - 31,5dB and 37,5dB (balanced)
Mytek Brooklyn Amp - switchable, 23dB and 29dB
Musical Fidelity M2si - set to 29dB by Amir, (so probably even higher default gain)
Again : the Myref family : 30dB.
The MiniDSP 4x10HD is unfortunately that noisy: Review: MiniDSP 4x10HD – Neurochrome
The SHD is much better, but also more expensive.
You're right. You can use a potentiometer. Then you have the noise of that to deal with. It'll will be amplified by the gain of the amp as well.
You could use the 0.9 V output of the MiniDSP. That would get you to a noise voltage of 272.9 µV RMS at the speaker with the MyRef.
Given the prevalence of 2-4 V sources, I really don't see why a power amp needs 30 dB of gain. But that's me. I seem to recall that the THX standard is 26 dB.
I suspect some amp designers want higher gain because it results in deafening SPLs at very low settings on the volume control, which gives the perception of a very powerful amp.
Tom
The SHD is much better, but also more expensive.
You're right. You can use a potentiometer. Then you have the noise of that to deal with. It'll will be amplified by the gain of the amp as well.
You could use the 0.9 V output of the MiniDSP. That would get you to a noise voltage of 272.9 µV RMS at the speaker with the MyRef.
Given the prevalence of 2-4 V sources, I really don't see why a power amp needs 30 dB of gain. But that's me. I seem to recall that the THX standard is 26 dB.
I suspect some amp designers want higher gain because it results in deafening SPLs at very low settings on the volume control, which gives the perception of a very powerful amp.
Tom
Last edited:
Just did a quick survey of some name brands. Parasound, Cambridge, Jeff Rowland, Krell. I'm seeing gains ranging from 22 dB (Cambridge) to 29 dB (Parasound). Pascal M-PRO2, S-PRO2, T-PRO2: 26 dB. The Pascal amps reach 53 µV RMS at the output. Purifi and Hypex set their amps up for 26-27 dB gain as well (when used with their buffers). The Hypex NC500 reaches 10 µV RMS (unweighted) on the output, by the way.
Regardless, the concept of gain structure has been known for a while now.
Tom
Regardless, the concept of gain structure has been known for a while now.
Tom
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.