Safe level of DC offset?

generally, the lesser the better. For headphone amps, under 10mV is considered okay. My personal standard, however, is less than 1mV.

Offsets usually means there's something wrong with the amp, you should take care of it carefully.

I have wondered about this too, obviously we all want to get down to zero but things don't always work out that way and lobing a fat capacitor in the output path is not an option. I realise optimal design and component matching etc will reduce offset but there must be some point at which headroom is truly affected by the dc offset and not that it is merely a subconscious annoyance due to knowledge of its existence. I am assuming that we are restricting this discussion to hp amps and not expecting this signal itself to be amplified. The "golden rule" I have usually followed is to keep offset below 20mV but have never seen the reasoning for this ( nor questioned it ). Speaker amps are acceptable below 100mV ( again not sure of this derivative ) and have been aware of the discussion that high offset can reduce headroom in speaker amplifiers.

I guess my extension to fran's question would be this:

1. What is the "average" output swing of an amp into hp's say 100ohm at "normal" listening levels.

2. How would a 50mV offset affect the signal ( we are not speaking of several volts of offset which could damage headphone coils )

any thoughts? ..dB

i'm not capable of writing long wall of text just to explain all the ill effects of dc offset. However, apart from reduced voltage swing, headphone driver being pushed to one side (heating it), it actually made THD worse. I"ve did a simulation before and balancing the offset actually helps improve THD.

I can understand that the dc's heating effect could have ill effects on efficiency of a driver, but what element of dc is capable of 'pushing' it to one side? certainly not the electromagnetic field it doesn't produce?

I had thought that it had to do with the fact that both AC and the headphone driver are in constant "motion" whereas DC is steady and thus in conflict with the motion of the driver.

An electromagnet responds to current, whether it's AC or DC. A DC signal will cause a magnetic field of constant magnitude and direction to form around the driver's electromagnet, which will change the equilibrium position of the voice coil i.e. 'push it to one side'.

Hey Tze, no arguments that offset affects sound negatively and not even going to the abnormal voltages that push the coil out to the point where it overheats, I am more interested in getting a feel for how much does it take to impact audible sound. I understand how offset affects the driver and am generally critical of any amp I have built if it has an offset >5mV but I am not sure that I would be able to tell you that an amp has a 50,20,5 or 0 mV offset by listening to it. a quote from rock grotto " Basically, anything under 20mV is totally acceptable. 20mV to 50mV is borderline and anything greater than 50mV is a definite NO NO." - does this come from experience or is there hard data to back this up?
I would be very interested in your sim data if you have it online or handy.

The basis for my questioning is the following.
If we assume the hypothetical 100R headphones and drive them to a healthy 100mW , the voltage out for the amp is approximately 3.2V ( assuming my math is correct ) - will our theoretical 50mV offset impact significantly on this? ..dB

Doh. thanks for setting me staight

dBel84,

First and foremost i'm no expert. However my understanding of the possible ill effects of dc-offset are as follow:

1) Given a differential operational amplifier, offset usually means the great imbalance in the input bias and the chip might be oscillating or acting funny.
2) Generally, we consider offset as a "loss of voltage swing". An amplifier with the capabilities of swinging rail to rail (given +/- 15V and the offset is +5V), it means that we can only swing as much as +10V/-15V. Imbalanced clipping will occur.
3) DC offset is not just about the loss of dynamic range. Some other factors that greatly influence sound can also be the reason. For example: Chip oscillating at certain frequencies, lack of proper bypassing etc. These all affect sound. Also this is meant to answer your hypothetical question. In short your question is not broad enough to cover the causes and effect of dc offset.
4) Speaking from personal experience, simulations gave me higher THD which increases proportionally to the increase of dc offset. On the other hand, I had a cmoy amp (2132PA) which gave me >3mV of offset and once i swap the opamp with an OPA2227, offset immediately falls under 1mV. Guess what, the 2132PA has a lower input bias than the OPA2227. (EDIT: this is to show that offset is not only caused by the symmetrical imbalance, but other much more important factors that influence sound)

I guess based on my knowledge now i can only answer this much. I do hope someone comes in and correct me though lol.

Interesting thread!

I generally agree with TzeYang about trying to keep the offset below 5mV, but I think we're mixing up decimal points if we think that's going to offset the voltage swing enough for clipping. I would never even plug the phones in if the offset was truly a full 5V!

So, I also agree with dBel84. If you've built an amp and the offset is 30 - 50mV, you might consider using it anyway. Maybe a good precaution in that case is to never leave the headphones plugged in while the amp is on, but music is not being played?

your ( and everyone else's ) input is much appreciated..dB

Whether a certain amount of DC offset is excessive also depends on the headphones. Let's look at two extreme examples. Assuming two theoretical headphones, one is 16Ω and the other is 600Ω.

Since P = V² / R, a 50mV DC offset would dissipate 0.2mW on the 16Ω headphone but only 4.2µW on the 600Ω.

If both headphones are highly efficient at 98dB/mW, then 50mV of DC offset would be equivalent to "91dB" worth of offset pressure on the 16Ω headphone's diaphragm, whereas on the 600Ω it's 74dB. You can see from this that the DC offset would cause dynamic compression on both headphones, but much worse on the lower impedance model.

Of course not every headphone is 98dB/mW, so there is another variable in addition to the DC resistance.

Electrical characteristics aside, it would be useful to examine is just how much diaphragm displacement would 50mV of DC offset incur, and what the headphone's maximum diaphragm's excursion limit is. The difference would of course be the available dynamic range. But these are much more difficult to quantify.

so in conclusion...

it depends.