Go Back   Home > Forums > Amplifiers > Chip Amps

Chip Amps Amplifiers based on integrated circuits

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 17th November 2012, 02:09 AM   #61
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Quote:
Originally Posted by kmibb View Post
Any news?
I've been able to audition the LM3886TF/OPA2134 version with a Marantz CD63 source and 2-way, 8-ohm folded-horn towers designed by Sreekanth (the best 2-way speakers I've heard) in Bangalore. Interconnects to the speakers were also Sreekanth's braided design. Trafo was 160 VA, EI core.

The good news: It sounded very transparent, clean and dark with no trace of hum or interference. It's comparable to the MyRef on sonics, probably even more neutral.

The bad news: One channel kept going into thermal runaway, which indicates that it's probably at the edge of stability and requires changes to the compensation to increase phase margin.

So it went back to the drawing board for compensation tweaks. I made some mods (on the existing board, without any track cuts but one wire jumper) but haven't tested/auditioned it yet. I was busy with simulations of various discrete opamps over the last month - my existing LF03 discrete opamp, though stable, was current-starved and distorting in the MiniRef. There's a simple mod to the LF03 to bias it so that it draws lower current (say 8-10 mA per channel) and stays in Class-A in the MiniRef. I plan to test all these changes when I get around to the next series of tests this month.
  Reply With Quote
Old 12th December 2012, 03:43 PM   #62
kmibb is offline kmibb  United States
diyAudio Member
 
Join Date: Mar 2010
Keep us updated, I'm very interested in this amp.
  Reply With Quote
Old 15th December 2012, 02:07 AM   #63
gootee is offline gootee  United States
diyAudio Member
 
Join Date: Nov 2006
Location: Indiana
Blog Entries: 1
Quote:
Originally Posted by linuxguru View Post
I've been able to audition the LM3886TF/OPA2134 version with a Marantz CD63 source and 2-way, 8-ohm folded-horn towers designed by Sreekanth (the best 2-way speakers I've heard) in Bangalore. Interconnects to the speakers were also Sreekanth's braided design. Trafo was 160 VA, EI core.

The good news: It sounded very transparent, clean and dark with no trace of hum or interference. It's comparable to the MyRef on sonics, probably even more neutral.

The bad news: One channel kept going into thermal runaway, which indicates that it's probably at the edge of stability and requires changes to the compensation to increase phase margin.

So it went back to the drawing board for compensation tweaks. I made some mods (on the existing board, without any track cuts but one wire jumper) but haven't tested/auditioned it yet. I was busy with simulations of various discrete opamps over the last month - my existing LF03 discrete opamp, though stable, was current-starved and distorting in the MiniRef. There's a simple mod to the LF03 to bias it so that it draws lower current (say 8-10 mA per channel) and stays in Class-A in the MiniRef. I plan to test all these changes when I get around to the next series of tests this month.
It could be as simple as not having small bypass capacitors right at each power pin. Also, the decoupling/local reservoir capacitors seem too far from the power pins.
  Reply With Quote
Old 15th December 2012, 06:35 AM   #64
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Quote:
Originally Posted by gootee View Post
It could be as simple as not having small bypass capacitors right at each power pin. Also, the decoupling/local reservoir capacitors seem too far from the power pins.
You're probably correct on both counts. The distances are even more on the MyRef, which doesn't use rail-to-ground small bypasses near the LM3886 power pins, but it's likely that they're mandatory here. I just got a few Panasonic ECHU 22nF/50V SMD film caps in 1210 size, which should be perfect for bypassing in this application.
  Reply With Quote
Old 15th December 2012, 09:23 PM   #65
gootee is offline gootee  United States
diyAudio Member
 
Join Date: Nov 2006
Location: Indiana
Blog Entries: 1
I'm very surprised that Mauro didn't use HF bypass caps. I guess I never checked. Most transistor-based amplifiers have a hidden positive HF feedback path through the power rail. Without bypass caps, it's like asking for trouble. The bypass caps need to be within a millimeter or two of each power pin.

Decoupling is another matter. The decoupling caps can be thought of as a small local power supply, to provide fast-changing transient currents on demand. So they should be as close as possible to the chipamp power pins, to minimize the impedance presented to the pins.

Maybe I should first say that the sound from the speakers is made by current that is almost all coming directly from the power supply and decoupling capacitors, which is modulated by the amplifier. Quite-convincing proof of that is shown in this plot: http://www.diyaudio.com/forums/attac...om3a_33kuf.jpg , which is part of a post at Power Supply Resevoir Size .

"The rails currents ARE the signal and they come from the capacitors!"

With so many inches of trace between the smaller electros and the chipamp power pins, there is significant inductance (and resistance). There are several unwanted effects from that, all inter-related:

Any demand for a fast transient current change will have to pull the current through that inductance. That can create a significant voltage glitch on the rail, proportional to the rate of change of the current (V = L x di/dt). At the same time, the current's rise time is slowed, and the current is basically delayed, i.e. there is distortion of the commanded transient's amplitude vs time. That will smear the edges of output signal waveforms, and their Fourier components, in the time domain, degrading the soundstage image. Also, the feedback loop needs a very fast response, since it is trying to cancel things like high-frequency harmonic distortions. So that is affected, too. (Most chipamps need to have very low impedance, as seen by the power pins, out to somewhere between 80 kHz and 400 kHz.) Also, the transient currents will now have to travel in a much larger loop. So you also have a much larger antenna, there, than necessary.

Does any of that matter, in this case? I can't say, offhand, of course. But there are simple calculations that can be done, to find out. You can easily calculate the minimum required decoupling capacitance, and, the maximum acceptable inductance and resistance, which will give you the maximum acceptable round-trip distance through the decoupling capacitor connections to the chipamp power and ground decoupling points. Check out the links to my posts with example decoupling capacitor calculations, collected at Power Supply Resevoir Size .

Sorry that the decoupling calculations are so disorganized. I still haven't had time to go back and actually finish them.

In the chipamp cases I have looked at, the connection length could not even be made small-enough to get the inductance low-enough to be able to have a low-enough impedance, from the point of view of the power pins. The solution was to use at least three parallel caps, extremely close to each power pin. I guess maybe you could connect three 47 uF electros underneath the board, at each power pin. (The "47 uF" part is just a guess. And I would actually use the largest capacitance that comes in whatever size case that will fit.)

The links in the post at the following link are also quite important, and are fairly-closely related:

LM3886 component selection

Cheers,

Tom

Last edited by gootee; 15th December 2012 at 09:51 PM.
  Reply With Quote
Old 16th December 2012, 10:41 AM   #66
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
I came very late to the MyRef Threads.
But I recall that some Members who had acceess to Mauro's thoughts/experiments posting
that Mauro had tried local decoupling and had rejected it as sounding bad.

Sorry if I am misquoting.

That brings in (at least for me) the question of "how did Mauro implement the local decoupling"?
__________________
regards Andrew T.
  Reply With Quote
Old 16th January 2013, 02:02 PM   #67
gootee is offline gootee  United States
diyAudio Member
 
Join Date: Nov 2006
Location: Indiana
Blog Entries: 1
Good question, Andrew. Maybe he didn't, in any specific way. I think that the myref is essentially a Howland current source topology. Those involve a positive feedback loop and can be touchy in terms of stability, especially when pushed close to maximum performance. This is merely a guess but perhaps improving the high-frequency current response of the supply rails caused some type of stability problem that negatively affected the sound quality.

I seem to have a vague memory of having to limit the decoupling cap values for a small (1.5 Amp) Howland current source that I designed for a test instrument, about ten years ago. It was the output amplifier for a staircase generator and I think that maybe when the decoupling caps were too large, the waveform was either slightly distorted or the maximum output level couldn't quite be reached. I just can't quite remember. But that circuit was pushed to several extremes to meet its specs and almost anything I changed ruined something else.

Last edited by gootee; 16th January 2013 at 02:09 PM.
  Reply With Quote
Old 9th February 2013, 03:32 AM   #68
prezden is offline prezden  Canada
diyAudio Member
 
Join Date: Nov 2008
Did the MiniRef ever get to the board stage?
  Reply With Quote
Old 9th February 2013, 05:35 PM   #69
diyAudio Member
 
Join Date: Dec 2009
Send a message via Yahoo to Lil Knight
Very interested in this mini revision. I am going to order a large batch of PCBs soon, if there is a proven layout for this design, I could ask for some boards made.
__________________

  Reply With Quote
Old 13th February 2013, 06:25 AM   #70
diyAudio Member
 
Join Date: Sep 2007
Send a message via Yahoo to linuxguru
Update: The instability/thermal runaway problem in the first MiniRef 3886 prototype has finally been solved.

There are probably many ways to tweak the compensation, typically with a large Cdom, to achieve this. I simply went with an RC network between the inputs of the LM3886, similar to the approach used by Mauro in the MyRef Rev C - 820 pF in series with 3.3k, between the non-inverting and inverting inputs seems to work just fine. Higher C values up to about 1.5 nF should be Ok.

This compensation mod is preferable because it involves no track cuts on the existing boards. I'll do the same mod on the second prototype with a pair of LM3876TFs, and eventually respin the PCB if all works fine (the same board layout is used with both the 3886 and 3876). Meanwhile, the existing boards are usable with the mod.

The test setup was a Linux notebook running Rhythmbox -> HiFiMeDIY USB 2.0 24/96 ES9023 Sabre miniDAC -> MiniRef 3886 with LME49720 as the outer-loop opamp. Chipamp rail voltages were / - 28 VDC, a bit lower than usual, but excellent for LM3886 sonics.

I listened to it only for about 45 mins, and it sounded airy, well-balanced, with plenty of micro-detail, thanks to the LME49720 in the outer loop.
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Eagle vs. Sprint-Layout for PCB design/layout hollowman Parts 11 12th January 2014 09:01 PM
Can someone check My PCB layout against the Schematic for mistakes please?? Minion Solid State 9 6th June 2007 03:26 PM
need a signal generator pcb layout or schematic hizz_eye Solid State 8 3rd March 2005 06:09 PM
Denon PMA 770 - Schematic & PCB layout - Anyone got one? GeorgeBoles Solid State 1 29th September 2004 11:50 PM
Does anyone have a schematic or pcb layout for eq punchpeanut Parts 3 22nd September 2003 08:20 AM


New To Site? Need Help?

All times are GMT. The time now is 06:53 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2