Straight from the horse's mouth, OPA2604 problems are thermal. The silicon is running hot and at +/-24V supply it is out of spec, mostly DC.
Since there is no immediate technical solution (short of a full redesign, not economically feasible), TI choosed to pull the chips from the usual distribution channels. You can still order it, but only if you sign a waiver/disclaimer, stating that the data sheet parameters are not guaranteed over the whole temperature and DC bias conditions. So the OPA2604 is currently kind of Schroedinger cat.
Nein, that would be a major change, breaking the backwards compatibility. Imagine such chips sold as spare parts for an existing application that doesn't use a solder mask (pretty common for tightly controlled impedance traces boards).
Am I old fashioned with my LF412 servo?
The servo in picture below is keeping my amp below 1mv DC offset, measured with fluke 8808.
This one has -60dB gain at 10Hz, it takes about 1 minute after powering up to lower DC offset from 0.4v to 0.
Would a better performing opamp make a significant change in noise?
The servo in picture below is keeping my amp below 1mv DC offset, measured with fluke 8808.
This one has -60dB gain at 10Hz, it takes about 1 minute after powering up to lower DC offset from 0.4v to 0.
Would a better performing opamp make a significant change in noise?
Attachments
I would have to do some research on when exactly I came up with the idea to use servo control for off-set and drift correction. Cant go by publications exactly because they can take months to get written up and put into the publishing pipe-line. But, before I had done direct-coupled amp attempts in my own preamp in order too hear an amp without coupling caps long before and this is what I talked about.... better cap development lead to no cap coupling (line stage). A no-cap couple line stage used in a Crown IC150 was reviewed in IAR. I told Walt Jung that it sounded better without a cap. In the conversation, I suggested a dc servo. I am sure Walt Jung knew about chopper stabilized amps. But, he came up with what we all use today. First used in his published RIAA preamp article in TAA. Then, I went straight to direct coupled amp designs..... as has most of the planet.
Since then, I have seen some other clever ways to do it in CMA circuits. All good. Thats further progress in the same direction of refinement.
THx-RNMarsh
Last edited:
The way to view this servo thing is that in a decent amp with feedback, the output offset and drift are 2nd or 3rd order effects i.e. the main amplifier loop for the most part has the offset under control to within a few tens of mV. You therefore only need a little bit of assistance from the servo to bring the offset under control.
You can see this in practice if you dial the amplifier offset out manually so that the servo controller output is exactly 0. The servo output remains at or close to 0 (within a few mV) right down to a few Hz (and over temp - a good LTP or CFA DB stage is quite DC stable over temp in a practical audio amp). Its only when you get to 1-2 Hz and large rail excursions that you see any appreciable change in the servo output. A lot implementations actually attenuate the servo output substantially by using a series resistor feeding into the servo summing node on the amplifier - so they really are quite 'hands off' wrt to the main loop above a few Hz.
Separately, if you are hearing a big sonic difference with a servo, you have other issues. either its a sub-optimal servo design, or the amplifier is suboptimal and the servo happens to be improving it . . .
You can see this in practice if you dial the amplifier offset out manually so that the servo controller output is exactly 0. The servo output remains at or close to 0 (within a few mV) right down to a few Hz (and over temp - a good LTP or CFA DB stage is quite DC stable over temp in a practical audio amp). Its only when you get to 1-2 Hz and large rail excursions that you see any appreciable change in the servo output. A lot implementations actually attenuate the servo output substantially by using a series resistor feeding into the servo summing node on the amplifier - so they really are quite 'hands off' wrt to the main loop above a few Hz.
Separately, if you are hearing a big sonic difference with a servo, you have other issues. either its a sub-optimal servo design, or the amplifier is suboptimal and the servo happens to be improving it . . .
Last edited:
in a line stage or power amp with low gain, it is not so hard to deal with DC off-set drifts. But for the high gain, direct-coupled phono stage which WJ had in mind, the drift in dc would/was more severe issue. Even my little low gain headphone amp I made has no DC servo. Isn't needed.
THx-RNMarsh
THx-RNMarsh
Last edited:
Servos aren't toys, and IF they can be safely eliminated, so much the better. BUT if you do, you are liable to 'throw the baby out with the bathwater' and you will pay more, and get less. It is relatively easy to make a lousy servo. You can make the bandwidth too high so that it dances along with the music. Not good. Or you can make it from 'junkbox' parts and add some sort of nonlinearity to the gain stage. Or you can feedback to an input position that creates some measurable offset, because you are too lazy to make the amp, etc, virtually DC accurate in the first place and too much error correction causes added input stage distortion. That is why the Blowtorch has second stage servo correction. The first stage 4Q jfets are fully floating, so they are not much effected. And finially, NOISE or INSTABILITY from the servo ckt. This is a real engineering set of tradeoffs, amateurs could make a mistake and then for evermore tell everyone how servos don't work. '-)
6 years ago, I was working on a balanced to SE converter circuit with Joachim & Brad.
Back then I could not get the DC offset to stabilise enough.
The XCEN -- Balanced to Single Ended Converter
After leaving it for a few years, I finally got round to a revised version last year which is stable to +/-3mV or so.
This is done by changes in the hardware to make them track thermally.
XCEN Balanced to Single-Ended Converter Revisited
So even without global NFB, one can still do without a servo.
Patrick
Back then I could not get the DC offset to stabilise enough.
The XCEN -- Balanced to Single Ended Converter
After leaving it for a few years, I finally got round to a revised version last year which is stable to +/-3mV or so.
This is done by changes in the hardware to make them track thermally.
XCEN Balanced to Single-Ended Converter Revisited
So even without global NFB, one can still do without a servo.
Patrick
- Status
- Not open for further replies.