My_Ref Fremen Edition - Collaborative DC-Servo design - Page 2 - diyAudio
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 6th June 2014, 12:49 AM   #11
Reality first!
diyAudio Member
 
ClaveFremen's Avatar
 
Join Date: May 2005
Location: Milan, Italy
Quote:
Originally Posted by KSTR View Post
Why not try the easy things first? LM318 has offset balancing pins (pin 1&5) that have not been used in the design so far. Part count is minmal, one 100k trimpot to positive supply.
Hi Klaus,

interesting and easy, never though about it!

With my limited knowledge I've had the impression it was a somewhat dynamic behaviour of the input transistors biasing...So it's possible to remove C9 from the circuit simply trimming the DC offset?

That sort of correction will work well also with an applied signal?

Will be stable with temperature?
__________________
Dario
ClaveFremen "Bailando Salsa en el Sietch"
  Reply With Quote
Old 6th June 2014, 12:59 AM   #12
diyAudio Member
 
lehmanhill's Avatar
 
Join Date: Jan 2008
Location: Chelsea, Michigan
Klaus,

This is why we love to hear from you. You keep us thinking. And the first thought is to read the LM318 datasheet again, which I should have done already. As you point out, there is offset balancing designed into pins 1 and 5 and they even show a circuit for balancing offset. Assuming that DC offset is fairly constant, it looks like it could work.

If an automatic offset balancing is desired or necessary, it seems like using pins 1 and 5 would be worth considering instead of feedback.

Thanks again.

Jac
  Reply With Quote
Old 6th June 2014, 01:20 AM   #13
diyAudio Member
 
lehmanhill's Avatar
 
Join Date: Jan 2008
Location: Chelsea, Michigan
Quote:
Originally Posted by ClaveFremen View Post
Hi Klaus,

interesting and easy, never though about it!

With my limited knowledge I've had the impression it was a somewhat dynamic behaviour of the input transistors biasing...So it's possible to remove C9 from the circuit simply trimming the DC offset?

That sort of correction will work well also with an applied signal?

Will be stable with temperature?
Of course, I hope Klaus answers with his usual expertise, but in the meantime, it seems like something that should be easy to test. At one point, I had the impression that output DC offset was due to a combination of internal offset from the ICs and mis-matching of external components that caused an imbalance. Maybe I was thinking that was why one build had 0.5 mV offset with the input shorted and another build had 5 mV offset. If that were the case, then it seems like trimming might work.

Would it be possible to jumper C9, measure the DC across a dummy load with shorted input, a single frequency, say 1 kHz sine, and white noise to see if the DC measurement changes? For temperature, a heat lamp or hot air gun could make 10 deg C change and help understand the temperature sensitivity of the amp. Note the datasheet shows some offset slope with temperature, but it is very small. Of course, that is only the LM318 and we need to consider the whole amp.
  Reply With Quote
Old 6th June 2014, 09:31 AM   #14
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
I will join in shortly
__________________
regards Andrew T.
  Reply With Quote
Old 6th June 2014, 09:51 AM   #15
Reality first!
diyAudio Member
 
ClaveFremen's Avatar
 
Join Date: May 2005
Location: Milan, Italy
Quote:
Originally Posted by AndrewT View Post
I will join in shortly
__________________
Dario
ClaveFremen "Bailando Salsa en el Sietch"
  Reply With Quote
Old 6th June 2014, 11:36 AM   #16
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
Quote:
The RSO1 resistor, from what I've read, is usually set 10x the feedback resistor of the amp so to lower noise and AC injection to the main amp.
the feed in resistor can be from ~1times R10 value to 20times R10 value.

You referenced R7 instead.
Linux agrees
Quote:
2) As a quick back-of-the envelope calculation, +/- 1 mA gives a pull range of 1 mA times 390 R, or ~390 mV at the feedback node. This seems adequate to correct about +/- 12V at the output. To get +/- 1 mA injection current at +/- 10V swings at the output of the DC servo, we need a series injection resistor (RS01) of about 10k.
If you want to save resources and space then you must use the inverting DC servo topology.

Do not "add" an inverting opamp to either the non-inverting or the inverting servo. Just feed into the appropriate feed in node.

I much prefer the non-inverting.
THREE reasons.
It has an input filter. this gives the servo amplifier less work to do.
It has a separate integrating capacitor and this dedicated cap gets on and does it's sole job without interference
It can be fitted with an output filter. I consider this last mandatory (see next comment)

These three factors result in less noise and less distortion and less audio, being injected into the feed in node.

Once you add anything to that node it cannot then be removed. It comes out of the speaker output.
__________________
regards Andrew T.

Last edited by AndrewT; 6th June 2014 at 11:41 AM.
  Reply With Quote
Old 6th June 2014, 11:36 AM   #17
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
What about the comp pins of the 318?

Can these be used for injecting a servo signal?
__________________
regards Andrew T.
  Reply With Quote
Old 6th June 2014, 11:39 AM   #18
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
Add on a window comparator, that identifies excessive DC servo correction signal.
This is equivalent to DC protection and could activate the protection relay BEFORE the output rail goes to supply rail and welds the relay contacts together.

The protection relay should output a signal to drive a "warning LED".
__________________
regards Andrew T.
  Reply With Quote
Old 6th June 2014, 06:07 PM   #19
KSTR is offline KSTR  Germany
diyAudio Member
 
KSTR's Avatar
 
Join Date: Jul 2007
Location: Central Berlin, Germany
Let's run through the current version to check how a manual offset trim would be feasible.

First, it's important to note that only the LM318 is contributing, the LM3886 is inside it's feedback loop.

LM318 max offset spec is 10mV (voltage) and 200nA (current). Bias current max spec is 500nA. LM218 would be significantly better both in max and typical specs (but only available in metal can or cerdip, probably harder to get and at a price).

DC gain is unity, so the voltage offset comes in as is 10mV. A non-issue.

Problem is the offset current, which means bias currents of inputs are matched only to this point. If the matching were perfect (zero offset current) *AND* DC input resistances were equal, no offset would result.

But we don't have matched input resistances, one leg is 103k (R12+R13) while the other is 12k (R7). That alone could give a lot of current offset.
Input voltage at the 12k leg is 12k*500nA=6mV and input voltage at the 103k leg is 103k*500nA=51.5mV(!!) which results in an additional voltage offset of 45.5mV. And that is with zero offset current spec.
Actually bias current can be as much as 200nA apart, assuming worst-case that would be in the 12k leg, reducing its offset amount to 3.6mV. The difference would still be dominated by 51.5mV from the higher resistance leg, though.
So, in the current design we could expect a worst-case offset (at 25degC) of about 60mV -- typically it might be only 1/5th of it, say 10mV. Thankfully the gain at DC is 1, so that value doesn't get multiplied.

Now even 60mV wouldn't do any bad to most speakers, only some very high efficiency types with very short excursion might be off-centered by an amount that would be seen in a slightly increased distortion measurement (but not necessarily audible).

To put numbers into perspective, a LM3886 used in datasheet standard config does have similar, or even worse offset, depending again on source resistance mismatch on the input pins, and it will be drifty

If we still want to have the lowest offset that can be had from the LM318, we'd need to a) balance the input resistance and b) make them as low as possible. The chip's offset trims are not intended to be used to null extra offsets from input resistance mismatch and/or too high source resistance. BTW I looked up the wrong figure (sorry), for the offset trim parts and connection see upper left figure on page 7 of the datasheet.

We could reduce R13 to 10k and that would match and lower source resistances as seen by the chip, any remaining offset would be convenient to trim out then. Any preamp (except exotic "passive" pre's with larger than 10k pots) should be happy to drive a 10k load.

We'd need to increase input blocking capacitor (C13) to 10uF to get the same LF input roll-off. Actually 3.3uF of 4.7uF would be enough, still below 1Hz corner frequency. Andrew would certainly make a point with the notion that the input roll-off shouldn't be any lower anyway than the roll-off of the feeback (C9), which currently is at about 2Hz.

Further, matching source impedances also reduced distortion. If we want to get rid of C9 (and C13) as a possible contributor to distortion and use a servo instead, low and matching source impedance would quite wise to have as well for that reason. And making a good servo that doesn't add significant noise and distortion itself isn't trivial either so the net benefit to this circuit is questionable im my view.

As for Andrew's suggestion of using the offset pins of the LM318 to inject a servo signal, yes this can be done (has been done eg to NE5534's) and might be better approach but it's again not exactly trivial.

For general improvement possibilities, one could put the feedback around the relay (see Cordell's book).... and use a better spec'd opamp like OPA627 (no current offset issues), though that's a different amp design, more like the MiniRefs...

Last edited by KSTR; 6th June 2014 at 06:15 PM.
  Reply With Quote
Old 6th June 2014, 06:28 PM   #20
Reality first!
diyAudio Member
 
ClaveFremen's Avatar
 
Join Date: May 2005
Location: Milan, Italy
Hi Andrew,

first of all, thanks for contributing

I appreciate it.

I'll ask you some deepening on your notes, it's to understand your points, nothing else.

Quote:
Originally Posted by AndrewT View Post
the feed in resistor can be from ~1times R10 value to 20times R10 value.

You referenced R7 instead.
Why should R10 be used instead of R7 (which is the feedback resistor of the composite amp, right?)?

I've simply followed AN-1192:

Click the image to open in full size.
(excerpt from TI AN-1192)

Can you, please, elaborate?

Quote:
Originally Posted by AndrewT View Post
Do not "add" an inverting opamp to either the non-inverting or the inverting servo. Just feed into the appropriate feed in node.
Not my work, too.

It's Cordell's preferred topology for the DC-Servo:

Figure 8.6 shows a noninverting integrator that requires only one op amp. It is like a single op-amp differential amplifier but with the feedback and shunt resistors replaced with capacitors to make it into an integrator. It requires two capacitors, and this is a disadvantage.
I prefer DC servos that employ a dual op amp and only require a single integrating capacitor, perhaps on the order of 1 mF. It is quite economical to employ a high-quality 1-mF film capacitor. Dual op amps that are of high quality are also relatively inexpensive.

(excerpt from Bob Cordell's Designing Audio Power Amplifiers Book)

Quote:
Originally Posted by AndrewT View Post
I much prefer the non-inverting.
THREE reasons.
It has an input filter. this gives the servo amplifier less work to do.
It has a separate integrating capacitor and this dedicated cap gets on and does it's sole job without interference
It can be fitted with an output filter. I consider this last mandatory (see next comment)
I agree but the third point has some problems too.

According Cordell the capacitor in the post filter would be critical for sound quality and the goal is to remove as much as possible this problem.

Quote:
Originally Posted by AndrewT View Post
These three factors result in less noise and less distortion and less audio, being injected into the feed in node.

Once you add anything to that node it cannot then be removed. It comes out of the speaker output.
Absolutely, but with a so low Fc (0.16Hz) how much AC would pass past the DC Servo? Isn't filtered enough?

Quote:
Originally Posted by AndrewT View Post
What about the comp pins of the 318?

Can these be used for injecting a servo signal?
Sincerely I don't know but I suspect it would be a not trivial task...

Quote:
Originally Posted by AndrewT View Post
Add on a window comparator, that identifies excessive DC servo correction signal.
This is equivalent to DC protection and could activate the protection relay BEFORE the output rail goes to supply rail and welds the relay contacts together.
It seems pretty interesting but way out of my design capabilities...sadly.
Attached Images
File Type: jpg Figure 14.jpg (148.1 KB, 377 views)
__________________
Dario
ClaveFremen "Bailando Salsa en el Sietch"
  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
My_Ref Fremen Edition - Build thread and tutorial ClaveFremen Chip Amps 2000 Yesterday 10:58 PM
My_Ref Fremen Edition RC - Build thread ClaveFremen Chip Amps 1250 19th September 2014 11:56 AM
My_Ref Fremen Edition GB (Second run) ClaveFremen Group Buys 52 22nd January 2014 09:13 PM
My_Ref Fremen Edition GB (interest for a third run) ClaveFremen Group Buys 107 13th January 2014 12:12 AM
My_Ref Fremen Edition - need help on PCB evaluation ClaveFremen Chip Amps 450 23rd October 2013 04:38 PM


New To Site? Need Help?

All times are GMT. The time now is 03:53 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