Member
Joined 2009
Paid Member
Attached is the file Nait1.asc that relates to the image posted. I am including a .txt file with some .models to play with. You should replace these with the models for the devices you are planning to use.
Also attached is a file Naim2SC3281.asc where I changed the output drivers from BD139/140 to MJE243/253 (higher voltage) to align with the NAP250.
You can open circuit the phase correction networks to judge whether or not BD139/140 might have been a better choice.
I changed the RC input filter capacitor in this asc file from 330p to 180 p to open the bandwidth. I think that will make an improvement in the dynamics including the bass. This has no unsettling effects.
Very interesting thread, Bigun! Here is the measurements from my Naim versions at 20kHz squares:
First is the standard circuit with 200pF input cap and the R/C circuits in the drivers.
Second one is the same circuit, but without the R/C circuit and Lat. MOSFET output with about 300ohms in the gates.
I know - different signal levels - but still the LATFETs look definitely better. Soundwise I don't think so, unfortunately. I suspect bandwidth itself is important, but it's not just that that makes the good sound. My original circuit always sounded "dark" and I thought the MOSFET one being wider bandwidth will open up the sound, but after I got used to it I found it darker and somewhat dull. I ended up with tube amp few years ago and haven't built a single thing since then. I'm ashamed to upload the 20kHz behaviour of my tube project
, but I love the sound. There is music.
Wish you luck and hope to hear about great sound from your version!
First is the standard circuit with 200pF input cap and the R/C circuits in the drivers.
Second one is the same circuit, but without the R/C circuit and Lat. MOSFET output with about 300ohms in the gates.
I know - different signal levels - but still the LATFETs look definitely better. Soundwise I don't think so, unfortunately. I suspect bandwidth itself is important, but it's not just that that makes the good sound. My original circuit always sounded "dark" and I thought the MOSFET one being wider bandwidth will open up the sound, but after I got used to it I found it darker and somewhat dull. I ended up with tube amp few years ago and haven't built a single thing since then. I'm ashamed to upload the 20kHz behaviour of my tube project
, but I love the sound. There is music.Wish you luck and hope to hear about great sound from your version!
Attachments
Last edited:
Very interesting thread, Bigun! Here is the measurements from my Naim versions at 20kHz squares:
First is the standard circuit with 200pF input cap and the R/C circuits in the drivers.
Second one is the same circuit, but without the R/C circuit and Lat. MOSFET output with about 300ohms in the gates.
I know - different signal levels - but still the LATFETs look definitely better. Soundwise I don't think so, unfortunately. I suspect bandwidth itself is important, but it's not just that that makes the good sound. My original circuit always sounded "dark" and I thought the MOSFET one being wider bandwidth will open up the sound, but after I got used to it I found it darker and somewhat dull. I ended up with tube amp few years ago and haven't built a single thing since then. I'm ashamed to upload the 20kHz behaviour of my tube project
Wish you luck and hope to hear about great sound from your version!
Apart from changing the RC input capacitor value did you alter any compensation capacitor values or any phase correction capacitor values.
It is still possible with some changes to my simulation - square wave input 10kHz output into 8R//2uf, to get some small improvement in the shape of the square wave output, to make more out of MJL3281 devices if one dares.
BD139/140 drivers perform better than MJE243/253 in experimental trial adjustments to my simulation. Bigun has chosen a different transistor line-up which should have the potential to do a better job. For that we will have to wait and see.
In the meantime it would be interesting to know the line-up in your original Naim project and if anything changed when you adopted LATFET outputs and the identity of the latter parts.
Your disappointment with "dark"sound has me curious. I does take a while to adjust to the sound of Naim equipment.
I had an amplifier to try out on approval and was dissatisfied to the point I was always going to return it. Several weeks later it got a reprieve. I have a few other systems including DIY which is not allowed in the lounge.
Last edited:
Member
Joined 2009
Paid Member
Michael,
That’s very useful, you found what I was seeking and I think you have shed some light on this topic. I’ve embedded the device models into the spice file you provided and uploaded it here so that others can use it more easily. I adjusted the Vbe to give 30mA idle current through the outputs and reduced the feedback cap value to reduce the simulation time - you need to run it for awhile to ensure the feedback cap has time to charge up so that the output is symmetrical about 0V. I reduced the Zobel cap to 100nF (it seemed better behaved with this value).
There’s a lot going on at h.f. with this amplifier and rather than think about the math of poles and zeroes I find it less demanding to use the simulation to explore different components and component values.
I’ve uploaded a .pdf file that captures the results of my fiddling around with it this morning. I didn't aim for lots of prose so you may have to read it twice to get what I'm rambling about.
It has shown me the value of the phase correction networks. The issue is the stability with capacitive loads (of course) but not the 2uF load of an electrostatic speaker which doesn’t seem to be any issue, but rather the parasitic capacitance of ‘bad cables’. I used a 2nF load across the 8R speaker and it worked nicely to tip the amplifier into instability. The phase correction networks addressed this nicely, although with different resistor values than the original NAP 110 design.
Also interesting to discover was a clear benefit of reducing the emitter resistor on the low-side driver (the one in parallel with the diode). I don’t think this has been explored much before and it may be worth further investigation.
Last edited:
Member
Joined 2009
Paid Member
this was a red herring, ignore it (I forgot to reset the idle current).
I found some device models for my choice of drivers and updated the spice file again.
A couple of plots to share. First one shows the benefit of my alteration to the Vbe multiplier to better control the temp-comp. The idle current is shown for my improved approach (described earlier) in blue I(R56) and without my improvement in green I(R42). The temperature was set at 40C and it produced the two lines at the top of around 32mA. The temperature was increased to 60C and it produced the two lower lines. My improved circuit shows a 10% fall in the idle current compared with a drastic fall for the original. I could tune it for 0% change but I'd prefer to have a slightly -Ve temp. co for the build.
The second plot shows the distortion profile at 500Hz into an 8R load (dissipating 46W av.) with 2nF in parallel.
Attachments
Hi, These are the schematics. Please ignore the dual diff. input. That's not the version that I measured. I was obsessed with the Naim circuit from 2007 to about 2012. I have tried all possible tweaks and never heard it sound as good as the very first one that I built. And I always found it "dark". I thought it was the attenuation of the high freq. response in the original. But everything that I tried didn't improve the sound only the measurements. I've tried several combinations of transistors, various capacitors etc, output inductor vs. resitor etc... In the end, maybe it just my years...
I'd like to see a simulation. Unfortunately, the ones that I've done rarely correlate with the real built.
I have a "memory picture" of Naim demos in stores - this dry, bold, super clear sound... Even too clear sometimes, like in a jazz club with good acoustics. Never heard the same at home with the clone. It could be that all theses demos were with Naim speakers, and I know that they sound really bright with other brands. Maybe to compensate the "dark" amps
... I don't know....Piece of my experience...
Last edited:
Hi Gareth,
Thanks for these explanations.
There is something for everyone here - simulating is stimulating of a lot of purposeful thought and examination - as you have shown in your attachments.
I am taking from this your advice to let the simulations run a little longer to allow charging of the feedback capacitor and will apply that to revisit some single ended circuits I looked into recently.
Cheers,
Michael
Member
Joined 2009
Paid Member
I've found simulations very handy for learning a bit more about how things work as my background is not electronics (it's physics). I've built amplifiers based on simulations that have worked first time. I've also been thrown off by silly things, like real-life LEDs having different voltage drops and affecting the current level of a CCS. But they don't tell you how an amplifier will sound. I remember learning from Hugh the benefit of trying different values of the compensation capacitor - you can change an amp from dark to bright and find a 'magic' in between.
Agree. Especially if many people collaborate and correct models are built, simulation is no different from the real stuff. For DIY hobby though, I use it mostly for parts of circuits to understand their operation and to experiment, or for passive circuits, like loudspeaker filters where the devices are easy to define. Its great time saving tool and fun to use. I'm always happy to try somebody else's work using simulation models, but if I don't build it, I know nothing, except that it seems to work.
It's not likely for me to return and build again a Naim amp, but I'll be interested to know how you'll find the sound and what would affect it the most. I experimented for many hours and couldn't squeeze more performance.
We are independently thinking along the same lines about capacitor values - I think you are spot on with changing the value of the zobel capacitor from 220n to 100n.
Member
Joined 2009
Paid Member
let's wait and see if you like the next one, TGM11, which if I go forward with it, will be inspired by NAIM S1 on a very small scale (singe pair outputs)
Hi, These are the schematics. Please ignore the dual diff. input. That's not the version that I measured. I was obsessed with the Naim circuit from 2007 to about 2012. I have tried all possible tweaks and never heard it sound as good as the very first one that I built. And I always found it "dark". I thought it was the attenuation of the high freq. response in the original. But everything that I tried didn't improve the sound only the measurements. I've tried several combinations of transistors, various capacitors etc, output inductor vs. resitor etc... In the end, maybe it just my years...
I'd like to see a simulation. Unfortunately, the ones that I've done rarely correlate with the real built.
I have a "memory picture" of Naim demos in stores - this dry, bold, super clear sound... Even too clear sometimes, like in a jazz club with good acoustics. Never heard the same at home with the clone. It could be that all theses demos were with Naim speakers, and I know that they sound really bright with other brands. Maybe to compensate the "dark" amps
...Piece of my experience...
You have not mentioned the earthing arrangements. Do you still have your 140 Clone?
No, all were dismantled long time ago. I keep only some pictures and schematics.
I experimented with grounding too, but because I couldn't hear or measure a difference with my home equipment, I ended up with grounding arrangements by Self's diagram, from his book. The path to ground though was the old Naim-like - power amp grounded in (or better said through) the preamp, except for the safety chassis ground of course.
BTW, Self's book basically describes how to make an amp with the same classic topology, but with all stages properly designed - The Blameless amp. I don't know how Blameless sounds, but I suspect it's the typical solid state type. Huge amounts of feedback anywhere.
Last edited:
Hey, just give me a good reason and I'll warm-up the soldering iron in no time ...let's wait and see if you like the next one, TGM11, which if I go forward with it, will be inspired by NAIM S1 on a very small scale (singe pair outputs)
Hi,
sadly I don't see any overload protection. The Naim originals were equipped with these, though.
Best regards!
"Blameless Amp" is a misnomer. At least Self got this right: "THREE NON-EXISTENT DISTORTIONS. Audio engineering suffers from misinformation, disinformation, and downright lying more than most fields of endeavour."
Not that they are "non-existent" at all.
Last edited:
Member
Joined 2009
Paid Member
The single slope SOAR protection of the original has been widely used in other simple amplifier designs and from what I've read it has been widely condemned for it's impact on the sound. I also read that Naim needed it to prevent their electrostatic speakers from blowing up their own amplifiers due to the occasional short circuit inside the speaker. It looks difficult to avoid sonic penalties with this simple approach because there's very little, if any, sweet spot between degrading the performance under normal conditions or failing to protect the output devices under poor conditions.
A good IV protection system will allow all valid audio signals to pass to all valid audio loads.
None of the wanted signal should be limited/interupted/corrupted.
This requires quite prodigious output current to be delivered to speakers that can demand transient currents upto and beyond three times what an equivalent rated resistor load would demand.
i.e for a 100W into 8ohms amplifier connected to an 8r0 dummy load you expect the current demand to never exceed 5Apk
A reactive 8ohms speaker presented with fast changing transients can demand transient currents exceeding 15Apk.
The 8ohms amplifier's IV must allow those 15Apk transients to pass without modification.
None of the wanted signal should be limited/interupted/corrupted.
This requires quite prodigious output current to be delivered to speakers that can demand transient currents upto and beyond three times what an equivalent rated resistor load would demand.
i.e for a 100W into 8ohms amplifier connected to an 8r0 dummy load you expect the current demand to never exceed 5Apk
A reactive 8ohms speaker presented with fast changing transients can demand transient currents exceeding 15Apk.
The 8ohms amplifier's IV must allow those 15Apk transients to pass without modification.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.