Page 2 of 2

Re: Distortion and source impedance in JFET-input op amps

Posted: Thu Jan 15, 2015 8:12 am
by Craig Buckingham
I'll have a stab at this from an analysis point of view. Mind you I don't believe I have any practical experience with issues that may necessitate the use of R8.

I read the previous analysis regarding breakthrough and I am not arguing that side when the pot wiper is at ground.

What I see is a DC compliance issue or spike potential when connecting or disconnecting equipment to the input when the pot wiper is at terminal 3, i.e maximum volume. Without R8 there is potential for large currents to flow, from stored charge on C46 , if the connecting equipment has a low output impedance. This could result in a spike on the +ve input that may cause damaging clipping of the IP to the power supply rails (depending on the IP stucture of the op-amp) and/or latch-up condition.

Of course if the preceding stage to the input is part of a composite circuit and therefore always connected then compliance issues would unlikely ever occur.

Re: Distortion and source impedance in JFET-input op amps

Posted: Thu Jan 15, 2015 10:16 am
by JR.
I don't think this clears up the R8 mystery but the HC6 had a known instability issue when racked in with long unterminated headphone cables plugged-in. The factory fix IIRC was to remove the 47pF feedback cap (perhaps a PCB layout problem.) . It was stable with output loaded down with cans, but didn't like cable capacitance alone.

Image

http://forums.prosoundweb.com/index.php ... 26.20.html

Here is a thread with some comments from Steve Macatee (an engineer from RANE).

JR

PS: The board image is not rendering correctly but not my image.

Re: Distortion and source impedance in JFET-input op amps

Posted: Fri Jan 16, 2015 9:23 am
by Craig Buckingham
That OP seems to be a little mmm well actually kind of sucks IMO. Class B with some deadband (crossover region) help from R50.

I can see why it would want to oscillate with capacitance on its output. R50 is between the class B stage and the feedback loop. That junction feeds the OP. So it seems to me that R50 is going to cause extra phase lag by forming a pole with the cable capacitance. That's my guess as to why C23 was included, to put some phase lead back to the -ve IP to try and keep it stable.

Taking the feedback from the op-amp OP would have made it stable, leaving the Class B OP stage outside the loop.

Is it a headphone amp??? Seams like it was designed to drive hi-impedance headphones. I mean look at R55 and R58 and tiny weeny 0.1uFd caps on the collectors of the OP drive transistors. Should have large electrolytics there and R55 and R58 be about 2Ω maximum.

Re: Distortion and source impedance in JFET-input op amps

Posted: Fri Jan 16, 2015 11:06 am
by JR.
Craig Buckingham wrote:That OP seems to be a little mmm well actually kind of sucks IMO. Class B with some deadband (crossover region) help from R50.
OP= original poster ?

Regarding the RANE HP amp that is a very common topology where the op amp carries the drive responsibility until the added transistors start conducting. I used almost that exact topology to drive some speakers in my first drum tuner (same but completely different set of parts and values.) 8-)

My initial thoughts is that the 15 ohm resistor is small, forcing the op amp to supply +/- 33 mA before the transistors come on, but looking inside the 4580 the op amp has a proper class A/B output stage that is specified for +/-50mA. I might second guess that down to transition at 25 mA but their design looks respectable in that regard.

I have designed several HP amps and it is not trivial to deal with an impedance range from as low as 3.2 ohms, to well above 600 ohms. or in their instability situation "no load just cable capacitance".
I can see why it would want to oscillate with capacitance on its output. R50 is between the class B stage and the feedback loop. That junction feeds the OP. So it seems to me that R50 is going to cause extra phase lag by forming a pole with the cable capacitance. That's my guess as to why C23 was included, to put some phase lead back to the -ve IP to try and keep it stable.
Oops, I thought C23 was the cap they pulled to make it stable, but nooooo, the running change to make it stable was to remove C22. Ignoring the buffer transistors I have seen the topology before used to decouple capacitive loads from simple line drivers. Without C22 there will be a rising HF response caused by the pole formed with output capacitance, C22 was added to knock down that HF bump and deliver flat response. Of course too much capacitance in this case forces it unstable (like a phase shift oscillator). The presence of resistive load to ground shifts that pole higher while not a ton with 600 ohms in parallel with 15 ohms.
Taking the feedback from the op-amp OP would have made it stable, leaving the Class B OP stage outside the loop.
but compromising accuracy, distortion and frequency response.
Is it a headphone amp??? Seams like it was designed to drive hi-impedance headphones. I mean look at R55 and R58 and tiny weeny 0.1uFd caps on the collectors of the OP drive transistors. Should have large electrolytics there and R55 and R58 be about 2Ω maximum.
In hindsight they could have stabilized the design and kept the C22 cap in place (to flatten response) by adding a simple RC load to ground across the output . It is easy to imagine not bench testing for a pure capacitive load (like a line output would), anticipating at least 600 ohm termination.

Probably some junior engineer given a very old design to copy (that may explain the R8). I have managed a engineering group before and management hates reinventing wheels when you have proved designs in house you can just cut and paste. But sh__ happens.

JR