DC off-set versus no negative feedback...

Hello:

I had a discussion with Balanced Audio Technology's designer regarding DC off-set.
He stated that the companies power amps typically provide about 100 mv DC off-set at the amplifier outputs.
I was surprised, because I had heard many, many, many discussions stating that 10 or below is outstanding, 20 -50 is within the safe zone and anything approaching 100 and above is unacceptable - possibly causing voice-coil heating and displacement and a loss in speaker performance and some other issues as well.
When I asked Victor about that he said that 10 or especially 5 and below is due to high use of global negative feedback and the 100 mv in his design is nothing to worry about.
I would be interested in your comments please.
 

wakibaki

Banned
2008-01-08 11:51 pm
Yes, well, he would do, wouldn't he?

He also seems to have conveniently forgotten that the offset could be eliminated by AC (capacitor) coupling the output.

This is typical of the modern trend in amplifiers which is probably accurately described as 'design by dogma'.

The whole issue of amplifier design and specification has been muddied by the irresponsible behaviour of 'subjective' reviewers, or f*ckwits, as I prefer to call them.

The majority of cognoscenti would agree with your take on the issue, give or take a few mV, zero offset would be ideal.

What more can I say?

w
 
DC off-set continued...

Hello wakibaki:

Well, referring to your response... "what more can I say", probably a little more at least:eek:.
Since I am not part of the cognoscenti, by formal document, that means I need to know a lot more:
In your opinion, or fact:
At what amounts would DC off-set become unsafe?
Is it true that the DC displaces the voice coil such that it is no longer centered at rest?
If so, is this similar to DC off-set in applications such as ..."remove DC off-set" in my Digital Audio Workstation's audio memory (shown in waveform editor), where the zero-crossing is not centered 50% between the pos/neg excursions?
I assume the effect is similar in that you are not getting correct potential between pos/neg limits, as well, in the speaker drivers too?

What about the claim that normal and or high levels of global negative feedback tend to provide low DC off-set and that that lack of global neg.feedback produces higher levels of DC?
I'm taking a guess here, but having seen specifications listing DC off-set from those two design types, I haven't, as yet, noticed a direct correlation.
However, since B.A.T. products have been selling product for awhile, and if their offset is not safe or could be better, should I assume they are not designed properly and thus not a smart purchase regarding offset (if and only your response should recommend what are safe levels)?
Anything else interesting to know about this?
 
Last edited:
Any DC offset will cause voice coil displacement. Whether this matters or not depends on the amount, and whether the bass speaker is at its best at zero displacement. Lowest distortion need not necessarily coincide with zeo displacement, although it probably won't be far off. A few hundred mV should not do any harm, and won't cause much heating.

The real issue is that the standing current will displace the output stage (assuming normal PP). A few tens of mA offset will hardly be noticed in a Darlington output, but will push a complementary pair well away from the middle region. This may increase or reduce crossover distortion, depending on how well designed and accurately biassed the output is. For an unusual output stage it could be good or bad.
 
DC off-set...

If your voicoil has a dc resistance of 6ohm then 0.1V is about 1.7mW.

So any thought in my mind about needing 100mV or lower is nonsense.

Bigun:
Thanks for the math.
In your example, it would appear to be a small issue: 100 mv (0.1V), that is.
I'm not qualified to say that in most cases, this would be true as the DIY member, DF96, replied below, but I'll ask some more questions directly to him.

DF96 responded:
... "Any DC offset will cause voice coil displacement. Whether this matters or not depends on the amount, and whether the bass speaker is at its best at zero displacement. Lowest distortion need not necessarily coincide with zeo displacement, although it probably won't be far off. A few hundred mV should not do any harm, and won't cause much heating.

The real issue is that the standing current will displace the output stage (assuming normal PP). A few tens of mA offset will hardly be noticed in a Darlington output, but will push a complementary pair well away from the middle region. This may increase or reduce crossover distortion, depending on how well designed and accurately biassed the output is. For an unusual output stage it could be good or bad."
 

wakibaki

Banned
2008-01-08 11:51 pm
In your opinion, or fact:
At what amounts would DC off-set become unsafe?

The problem is really one of quantifying 'how much is too much.' A driver with small power handling will be more adversely affected than one with large power handling.

Is it true that the DC displaces the voice coil such that it is no longer centered at rest?
If so, is this similar to DC off-set in applications such as ..."remove DC off-set" in my Digital Audio Workstation's audio memory (shown in waveform editor), where the zero-crossing is not centered 50% between the pos/neg excursions?
I assume the effect is similar in that you are not getting correct potential between pos/neg limits, as well, in the speaker drivers too?

Yes. The voice coil is offset. Part of the driver excursion is lost. Part of the signal excursion is lost. The voice coil is heated while at rest. These effects are small, however.

What about the claim that normal and or high levels of global negative feedback tend to provide low DC off-set and that that lack of global neg.feedback produces higher levels of DC?

As I pointed out, an AC coupled amp with no global feedback could have no DC offset.

It really depends on topology, although it is true that negative feedback tends to mitigate against DC offset.

However, since B.A.T. products have been selling product for awhile, and if their offset is not safe or could be better, should I assume they are not designed properly and thus not a smart purchase regarding offset (if and only your response should recommend what are safe levels)?
Anything else interesting to know about this?

For myself, I wouldn't like to see more than a few mV of offset, not 10s of mV, and it is entirely possible that 100 or more will have an audible effect, although I have never knowingly run an amp with that much offset, that's based on what I've read and other people's observations.

100mV is probably not unsafe as such, given that it's ~2mW (DC), in typical modern domestic systems. A DCR of 6 ohms is frequently encountered in drivers. 2mW of AC is clearly audible, however, in many conditions.

Looking at Douglas Self's 'Audio power amplifier Design Handbook' we find: 'Most of my amplifier designs have assumed that a +/- 50mV offset is acceptable' ...'it could be argued that +/- 50mV is on the high side for a top-flight amplifier' ...'We might therefore hope to keep the DC output offset for the improved amplifier to within +/- 15mV without trimming or servos'. [My italics]

Obviously a design with trimmers or SOT components can see this improved upon, although a few mV of drift with time are unavoidable without a servomechanism. A servo can reduce offset to arbitrarily low levels.

As regards BAT, obviously the discussion of DC offset applies only to their SS offerings, the tube amps will be transformer coupled, thus being offset-free. For myself I would be happier to buy from a source that gives more prominence on their website to technical specifications, which are sketchy in the extreme, as opposed to reviews, of which there are many.

w
 

Elvee

Member
2006-09-08 2:04 pm
Speakers designed for certain types of acoustic enclosures have a very compliant suspension system as they rely on acoustic loading, and even a small DC voltage can cause a relatively large shift in the rest position of the voice coil.
100mV is probably safe, even for those speakers, but going much further doesn't seem very wise.
 
DC off-set...

Any DC offset will cause voice coil displacement. Whether this matters or not depends on the amount, and whether the bass speaker is at its best at zero displacement. Lowest distortion need not necessarily coincide with zeo displacement, although it probably won't be far off. A few hundred mV should not do any harm, and won't cause much heating.

The real issue is that the standing current will displace the output stage (assuming normal PP). A few tens of mA offset will hardly be noticed in a Darlington output, but will push a complementary pair well away from the middle region. This may increase or reduce crossover distortion, depending on how well designed and accurately biassed the output is. For an unusual output stage it could be good or bad.

Hi DF96:
Thanks for pointing-out several ideas here.
Some additional questions please:
Speakers:
Are some drivers not at their best when at zero displacement?
Apart from the amplifier Xover distortion issue you latter mention, are you referring to speaker distortion relative to DC and driver off-sets?
How much, generally, should one find as a safe level?

Amps:
PP refers to push/pull?
Issues in displacement with a PP design are?...
... displacement of the A.C. waveforms at the speaker outputs?
As for the rest of the implications for amplifiers, one would have to know about the design, then check for DC.
Could you elaborate a little more please?
Thank you for the reply.
 
Elvee makes a good point in post #8. Drive units for closed box cabinets can be very floppy without the trapped air to limit cone excursions. I had forgotten that (even though my own speakers are closed box types).

In theory a drive unit will be at its best at zero displacement, but who knows? Probably a small effect anyway.

PP is push-pull. The issue there is not voltage offset, but the consequent current offset. 100mV at 8ohms is 16mA. A Darlington output might have a quiescent current of 100-200mA so 16mA offset is nothing much. A complementary pair output might run at quiescent 8-12mA so 16mA pushes one side into switch-off. This will change the distortion spectrum for small signals - better or worse is difficult to say.

Having said all that, SS amps generally need plenty of negative feedback so DC offset need only be a few mV (from the input pair Vbe mismatch). Valve amps generally have zero offset from the transformer secondary. A design with large offset would have to justify it by being so much better in some other way. A designer's whim is not enough excuse, but I am sure your friend has a suitable explanation?
 
DC off-set

As I pointed out, an AC coupled amp with no global feedback could have no DC offset.

100mV is probably not unsafe as such, given that it's ~2mW (DC), in typical modern domestic systems. A DCR of 6 ohms is frequently encountered in drivers. 2mW of AC is clearly audible, however, in many conditions.

Obviously a design with trimmers or SOT components can see this improved upon, although a few mV of drift with time are unavoidable without a servomechanism. A servo can reduce offset to arbitrarily low levels.

Hi wakibaki:
Some additional questions based on your above...

AC coupled amps:
I assume this is the majority type as in: A system which is reported as fully DC coupled is the minority, so the AC coupled types are more common, correct?
An Ac coupled amp w/out global feedback: Can you name a few?

... "2mW of AC is clearly audible, however, in many conditions."
Do you mean audible as in "hum" or other detectable signal, or audible as in driver distortions/anomalies?

... "a design with trimmers or SOT components can see this improved"
I've heard that those devices do not help the over-all sonic quality. Further, that controlling off-set would be better accomplished by other means. Keep in mind that I have only heard these claims from designers/manufacturers that don't use them and advertise as such.

Well I should have known by now (from my observations here at DIY), that there must be a robust and mature understanding of audio design, such that, a complete understanding of a particular system is in order to make most comments valid.
Very interesting responses as usual, but feel free to keep them coming!
 
Bottom line: 100mV offset=pis-poor design.
Yes , B_$ design !! :sad: Even with a halfarse match in the ltp , I get 5-10mv , if I bother to hfe match ... under 2mV. 1% degeneration resistors help as well. :) Then there is a little thing called an "offset adjust" trimmer. I get .1mV or better as the final result. My last amps (luxman clones) were .1-.2mV 4 months after being built.

OS
 
I use an LTP current mirror with a preset in one leg.
The offset then can be adjusted very accurately.

You want to get the pair to have as little offset as possible BEFORE any trimming (2-5mv) , as if you trim too far you will affect the gain (degeneration) of one transistor too much. Most of the time , what little "trimmin' " I have to do is to compensate for the 1% tolerance of the fixed Re's.

I use this (below). 1/4 to 1/2 turn (25 turn trimmer) is all that is usually needed .. more , and I know that there is a crappy match.
PS .. it is usually wise to match the current mirror as well in a topology like below , the cascode is less critical.
OS
 

Attachments

  • trimmer.gif
    trimmer.gif
    4.3 KB · Views: 82
Last edited:

wakibaki

Banned
2008-01-08 11:51 pm
Some additional questions based on your above...

AC coupled amps:
I assume this is the majority type as in: A system which is reported as fully DC coupled is the minority, so the AC coupled types are more common, correct?
An Ac coupled amp w/out global feedback: Can you name a few?

Any amp can be AC coupled. Simply connect a capacitor between the output and the speaker. Single supply amps are typically AC coupled, since they are biassed to put the output halfway between ground and the supply e.g. 6V for a 12V supply. The cap is necessary to stop the speaker seeing the 6V.

I couldn't name an amp without GFNFB.

An amp with a DC offset will, in the majority of cases, be a dual-supply DC coupled amplifier. Dual supply MEANS that the output quiescent (silent) voltage can be at ground potential (halfway between V+ and V-). This MEANS that DC coupling (no capacitor) becomes possible. If DC coupling is employed throughout this MEANS that a very small offset at the input can be amplified to a comparatively large offset at the output unless steps are taken to nullify it, such as close matching, trimmers or (servo) feedback.

... "2mW of AC is clearly audible, however, in many conditions."
Do you mean audible as in "hum" or other detectable signal, or audible as in driver distortions/anomalies?

OK, I've just been looking at a headphone amplifier, the guy built it to accept an input from a Sony MZ-E40 minidisk, it has ~80mV RMS output and we're talking about 100mV. It was quoted as being 'quiet' but it was obviously far from being inaudible, giving 74dB (SPL) in his headphones, i.e. 74dB above the threshold of hearing. His headphones have a higher impedance than the 6 ohms I used to calculate the ~2mW, there's probably < 0.2mW going to produce the sound in the phones. 2mW of 1kHz sinewave (or music) will be clearly audible at a metre from just about any conventional stereo speaker of modern manufacture regardless of efficiency. Quiet but audible, and not so quiet in some cases.

... "a design with trimmers or SOT components can see this improved"
I've heard that those devices do not help the over-all sonic quality.

Well, I'm just talking about selecting resistors (Select On Test) with a DMM as opposed to just taking them off a bandolier, so this is not going to worsen the sonics, the resistors are going to be there anyway, they just won't be as accurately matched in the amp with greater DC offset. Manufacturers prefer to just trust the nominal value because this is cheaper than having a guy on the production line making sure they match, or having a trimmer that someone has to adjust.

Further, that controlling off-set would be better accomplished by other means.

Depending on the design the offset can also be improved by (hand)(or machine) matching the individual active devices (transistors), however the total offset is likely to be a result of slight mismatches between both the active devices and the passive devices (resistors). Eliminating the mismatches all adds to the cost.

w
 

MJL21193

Disabled Account
2007-03-10 1:20 am
I have some small experience with 2 NGFB designs and I can understand why there is offset. It can be tricky to get it down but it is doable. Normally, in this type of amp, feedback is picked from the VAS and to keep offset low, you need to pick from the high side and the low side using 2 resistors instead of one:

111.PNG

R13 and R21 in the above pic. These also need to be very closely matched in value.

With that said, there is absolutely no apparent benefit to this type of feedback scheme, other than the bragging rights to say no global negative feedback. Possibly, the less technically minded among us see that and assume no feedback period, and think that is good - they've read how bad feedback if is some well respected audio review magazine.
 

Bonsai

Member
Paid Member
2003-07-25 10:44 pm
Europe
www.hifisonix.com
"Yes , B_$ design !! Even with a halfarse match in the ltp , I get 5-10mv , if I bother to hfe match ... under 2mV. 1% degeneration resistors help as well. Then there is a little thing called an "offset adjust" trimmer. I get .1mV or better as the final result. My last amps (luxman clones) were .1-.2mV 4 months after being built."

My experience is similar. 100mV is not acceptable in my view because its just sloppy engineering.