Out of interest is there any reason for specific resitor values used as a voltage divider in global feedback circuits, eg 22k1 series and 1k to ground, or 10k series and 470r to ground. I was noticing in dselfs blamless he uses a series value of 10k and in the trimodal he uses 2k21 in series...2k21 being the lowest value i have ever seen used in most lin based topology amplifiers..
Thanks
Colin
Thanks
Colin
Basically you want to use the lowest values possible because of lowest noise. But there are limits. One is that the fb network loads the output (which is not a real issue in power amps of course; 1 k extra to the speaker is neglegible).
The other thing is that you want to DC value seen from boths amp inputs to be equal. Because the inputs need a small DC (base) current, this current through the DC equivalent resistance will cause DC offset. By keeping the DC equivalent values the same, the offsets from the two inputs cancel. (This is much less an issue with fet input amps)
The DC equivalent for the fb network (when DC coupled) is simply the parallel resistance of the two resistors. When the fb is AC coupled, the DC equivalence is the resistor going back to the output. It just is the resistance to ground seen from the input terminal without signal (so the output counts as 'ground' in this case). The other input DC equivalent resistance is normally the resistor from that input to gnd.
Jan Didden
The other thing is that you want to DC value seen from boths amp inputs to be equal. Because the inputs need a small DC (base) current, this current through the DC equivalent resistance will cause DC offset. By keeping the DC equivalent values the same, the offsets from the two inputs cancel. (This is much less an issue with fet input amps)
The DC equivalent for the fb network (when DC coupled) is simply the parallel resistance of the two resistors. When the fb is AC coupled, the DC equivalence is the resistor going back to the output. It just is the resistance to ground seen from the input terminal without signal (so the output counts as 'ground' in this case). The other input DC equivalent resistance is normally the resistor from that input to gnd.
Jan Didden
lumanauw said:
Hi David,
The difference in noise level is certainly easily measured. You know, we can now easily measure orders of magnitude below audibility threshold. Like 1 degree phase shift. Like the difference between 0.001% and 0.002% THD. And the difference between 1uV and 0.5uV wideband noise.
Audible? Depends on a lot of other factors, see posts above.
Jan Didden
In my experience and with fb resistor up to 47K, the noise issue with good quality MF resistors is almost inaudible.
It's certainly measureable, but scarcely audible.
OTOH, if you go to a very low fb network impedance, you can balance this with a small voltage generator to feed the input bias resistor. Using a diode works well here.
Then you can go down to 100R impedance for lowest possible noise!
Cheers,
Hugh
It's certainly measureable, but scarcely audible.
OTOH, if you go to a very low fb network impedance, you can balance this with a small voltage generator to feed the input bias resistor. Using a diode works well here.
Then you can go down to 100R impedance for lowest possible noise!
Cheers,
Hugh
AKSA said:
Hugh,
Good points. But I think the added complexity of the diode bias source is not worth the last drop of noise. Certainly with 1k - 22k fb networks which I normally use, I never had issues with audible noise. What I DID have was that the THD was so low that it disappeared into the noise
Jan Didden
It should be most sever with high power, high gain amps driving very sensitive speakers.
I've heard that the most common situation is professional gear (Hundreds of watts with horn type loudspeakers) set up to fill large spaces. Sometimes an audible hiss can be heard with no signal present.
Ive never heard anything like this home audio equipment -- 100-200W amp plus modest speaker sensitivity. There is a limit as to how low you can push these resistor values before before encountering impedance matching issues with respect top the preamp.
Hi,
another reason for keeping the FB resistors low is to minimize the effect of parasitic feedback: only 2pF overall capacitance from the output stage associated with a 220K resistor form a pole at 360KHz, requiring the amplifier to be unity gain stable (ie overcompensated), which is detrimental to slew-rate, etc.
LV
another reason for keeping the FB resistors low is to minimize the effect of parasitic feedback: only 2pF overall capacitance from the output stage associated with a 220K resistor form a pole at 360KHz, requiring the amplifier to be unity gain stable (ie overcompensated), which is detrimental to slew-rate, etc.
LV
lumanauw said:Hi, Elvee,
Good point
David,
You can only improve stability with changing the fb voltage divider if it actually changes the feedback ratio. Changing 20k to 10k+10k won't do that.
Jan Didden
SY said:
I used to know this stuff having working in HiV (2KV to 10KV)precision supplies.
I've just done a quick Google and cannot even find a single datasheet where this is specified anymore!
Used to be in uV/V applied, AFAIR.
And you are right, the effect can obviously be spread across a chain of resistors. Never heard of it being singnificant at <+_100V, though.
Hi guys,
Quite a nice array of answers, unfortunately I find absolutely no detailed information on this subject online and in the rare case can only find suggested values for op amps in relation to gaining the flattest possible bandwidth at a given voltage gain..I am using 22k in series with 1k shunt in my amplifier which was initially built of the blameless concept but with a few small changes. The input impedence is 22k and i am curious to whther or not i could use a series element as low as 2k2 or would 10k be better?. Also would i need to change the input resistance according also to something like 10k?. My pre uses the composite op amp by walt jung to drive the output so lower input impedence shouldnt pose a problem..
Thanks
Colin
Quite a nice array of answers, unfortunately I find absolutely no detailed information on this subject online and in the rare case can only find suggested values for op amps in relation to gaining the flattest possible bandwidth at a given voltage gain..I am using 22k in series with 1k shunt in my amplifier which was initially built of the blameless concept but with a few small changes. The input impedence is 22k and i am curious to whther or not i could use a series element as low as 2k2 or would 10k be better?. Also would i need to change the input resistance according also to something like 10k?. My pre uses the composite op amp by walt jung to drive the output so lower input impedence shouldnt pose a problem..
Thanks
Colin
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.