That's excellent Prof. Brown. You have apparently done your homework - and then some. I was aware the references I posted were not precisely matched to FE bypassing and needed exactly what you posted to get a clearer picture.
It now sounds like a daughter-board like someone suggested earlier would be useful for those who what to investigate. Dario will probably comment on the practicality and physical limitations involved - I hope 😉
Something that comes to mind is a spin-off of Siva's approach to LM318 modules. He packed a heck of a lot in a tiny package. That introduces a socket and/or pin format, but may not be out of the realm of possibilities with some cleverness in selecting the right connection element. Obviously above my pay grade. 😱
Thanks again - and long comprehensive posts like yours are necessary when the topic is complex and easily misunderstood.
It now sounds like a daughter-board like someone suggested earlier would be useful for those who what to investigate. Dario will probably comment on the practicality and physical limitations involved - I hope 😉
Something that comes to mind is a spin-off of Siva's approach to LM318 modules. He packed a heck of a lot in a tiny package. That introduces a socket and/or pin format, but may not be out of the realm of possibilities with some cleverness in selecting the right connection element. Obviously above my pay grade. 😱
Thanks again - and long comprehensive posts like yours are necessary when the topic is complex and easily misunderstood.

Attachments
Last edited:
In a My_Ref, the gain is set by R7 and R10, that is R7/R10 = 12k/390 ~ 30.8:1 voltage ratio. The problem is that the LM318 opamp has some stability issues as the frequency approaches DC. To make sure that the 318 remains stable, Mauro added C9 to R10, the combination making a high pass filter that acts like an open circuit at DC and causes the gain of the circuit to decrease as the frequency goes lower. That helps the low frequency stability.
The reason C9 is added is to reduce the the DC gain, as stated. However, the main intent is to reduce the output offset voltage, which would otherwise be the input offset voltage (10s of mV) multiplied by the DC gain (~30), thus going to the 100s of mV region, which tends to reduce the long-term reliability of speaker voice coils.
Some ways around having to use C9:
1) Use a DC servo, as in the Evolution.
2) Use an external discrete LTP with precisely Vbe-matched BJTs, instead of the internal darlington LTP of the LM318. I explored this in the Rev D, but didn't actually implement a prototype. This will probably work fine with BC849c/850c from the same reel, manually matched by Vbe within a few mV of each other at some known currents (say 0.1 to 1 mA).
Bob, that's actually the dual DIP8 LF03 discrete opamp in the pic, which doesn't directly substitute the LM318 single opamp. I did build a single-opamp version of the same, called the LF03s, but it was unstable in the MyRef Rev C (like a lot of other high-performance opamps including the OPA627, AD797B, LME49710, etc.). However, PreSapian has been able to stably use the LF03 in a variant of the MiniRef 1875 (with some mods).
Last edited:
I'm sure some of you have bypassed C9 in earlier versions of My_Ref. What kind of bypass value is typically used with C9? I would guess something in the range of 0.1 uF?
I've determined 'by ears', some years ago, that the best bypass value for C9 seems to be 22nF.
Dario - If you adjusted a new version of the FE to allow for bypassing, would that most likely be a surface mount item?
I don't even know if there will be another run... BTW in such case it will be probably a TH part (FKP2 or KP1830).
The reason C9 is added is to reduce the the DC gain, as stated. However, the main intent is to reduce the output offset voltage, which would otherwise be the input offset voltage (10s of mV) multiplied by the DC gain (~30), thus going to the 100s of mV region, which tends to reduce the long-term reliability of speaker voice coils.
Some ways around having to use C9:
1) Use a DC servo, as in the Evolution.
2) Use an external discrete LTP with precisely Vbe-matched BJTs, instead of the internal darlington LTP of the LM318. I explored this in the Rev D, but didn't actually implement a prototype. This will probably work fine with BC849c/850c from the same reel, manually matched by Vbe within a few mV of each other at some known currents (say 0.1 to 1 mA).
Siva,
Very interesting. I had not understood that reason for reducing the gain. I would love to learn more. Is there a place on the web where I can read about the DC servo from the Evolution or similar? Also, I'm not familiar with the term LTP. What does it stand for and where can I read more to understand?
Thanks in advance
Siva,
Never mind about the DC servo. I found some good links.
I'm still not clear about LTP.
Thanks in advance.
Jac
Never mind about the DC servo. I found some good links.
I'm still not clear about LTP.
Thanks in advance.
Jac
long tailed pair, type of amp configuration
Thanks Tom. There is so much to learn. Actually, a long tailed pair sounds more like a bird or maybe a Russ White creation. I think I'll put this one near the bottom of the things to learn about list.
I've been reading about the DC servo circuit around the internet and it seems fairly easy to implement. Anyone care to comment on the sound tradeoff between a DC servo and our current C9 configuration?
Jac
Have you read the LONG and very informative DC servo Thread that Gootee contributed to?
Thanks for that connection. I haven't, but will. I had found Gootee's own web page and read the summarized material on DC servo circuit's there.
Thanks again.
DC servo
I've started a new thread for a collaborative effort on designing a DC-Servo for the FE:
http://www.diyaudio.com/forums/chip...en-edition-collaborative-dc-servo-design.html
I've started a new thread for a collaborative effort on designing a DC-Servo for the FE:
http://www.diyaudio.com/forums/chip...en-edition-collaborative-dc-servo-design.html
AC and DC coupling
While waiting for the chassis for MyRef FE I have a couple of questions. I am thinking of implementing AC and DC coupled RCA inputs for MyRef FE. One would feed the signal through the DC blocking capacitor and the other would bypass the capacitor if the source already has a one at the output (or speaker has a capacitor in signal pass). Would it make a problem to have one RCA connector open while the other is in use since they are effectively connected in parallel? Would the source output impedance be sufficient to disregard any noise pick up via the open RCA connector?
Also, should I consider swapping the speaker cable polarity if I feed the signal to the amp bypassing the DC blocking capacitor? It seems that swapping the polarity of the input RCA wires is not an option due to the ground connections. Or am I wrong here?😕
Oleg
While waiting for the chassis for MyRef FE I have a couple of questions. I am thinking of implementing AC and DC coupled RCA inputs for MyRef FE. One would feed the signal through the DC blocking capacitor and the other would bypass the capacitor if the source already has a one at the output (or speaker has a capacitor in signal pass). Would it make a problem to have one RCA connector open while the other is in use since they are effectively connected in parallel? Would the source output impedance be sufficient to disregard any noise pick up via the open RCA connector?
Also, should I consider swapping the speaker cable polarity if I feed the signal to the amp bypassing the DC blocking capacitor? It seems that swapping the polarity of the input RCA wires is not an option due to the ground connections. Or am I wrong here?😕
Oleg
Two RCA inputs for each channel is an excellent way to set up every Power Amplifier.
The DC input feeds the Power Amp PCB directly.
The AC input passes through the blocking capacitor to the same PCB input pin.
However, add RF attenuation to both input RCA sockets. Add a 1M0 to 2M2 resistor across the AC input RCA socket.
DO NOT connect the Ground/Return wire of the interconnect to the "hot" input of either the DC nor to the AC input.
That Return wire is connected to a thousand "aerials" that WILL INJECT massive interference into the Power Amplifier.
The DC input feeds the Power Amp PCB directly.
The AC input passes through the blocking capacitor to the same PCB input pin.
However, add RF attenuation to both input RCA sockets. Add a 1M0 to 2M2 resistor across the AC input RCA socket.
DO NOT connect the Ground/Return wire of the interconnect to the "hot" input of either the DC nor to the AC input.
That Return wire is connected to a thousand "aerials" that WILL INJECT massive interference into the Power Amplifier.
Last edited:
Too many distractions of late so I missed out on Dario's group buy but I think I'm finally getting back on track.
I'm wondering if these Elna Silmic II - 220uF 50V caps would work well for C1/C2.
Elna Cerafine and Elna Silmic Capacitors homepage
I'm wondering if these Elna Silmic II - 220uF 50V caps would work well for C1/C2.
Elna Cerafine and Elna Silmic Capacitors homepage
Too many distractions of late so I missed out on Dario's group buy but I think I'm finally getting back on track.
Welcome back 😉
I'm wondering if these Elna Silmic II - 220uF 50V caps would work well for C1/C2.http://www.hificollective.co.uk/components/elna_capacitors.html
They actually do but Elna RJHs are a better alternative.
You would want low ESR caps there.
Thanks Dario. Life gets in the way of our hobby sometimes 😉
Not sure where to get RJH's, there's none on ebay that I can find, maybe mouser has them.
Not sure where to get RJH's, there's none on ebay that I can find, maybe mouser has them.
- Home
- Amplifiers
- Chip Amps
- My_Ref Fremen Edition - Build thread and tutorial