advertisement


Ben Duncan's 'Supertuning CD' and 'Improved CD Electronics'

Well first measure the DC offset it has.

Does the input of your preamp have coupling capacitors?

The JVC 1010TN - Yes
The Croft/Tekni-fi/Aikido - sort of. The input signal would see an Elma switch then a TKD 100k pot and then a cap
The Type 76 - No
 
Pretty much any internally compensated monolithic op-amp is bad compromise here. The problem is the RF switching edges lie far outside the feedback bandwidth, and take the input stage in to a non-linear region. FET input devices have a larger linear region, so do slightly better.

Correct solution is to use an unconventional compensation approach, with the compensation cap directly on the virtual earth node. This needs a custom (i.e. probably discrete) gain stage to work. Compensation cap need to have very low ESL, as you care about its impedance all the way up to 10s of MHz - means probably a chip or feedthrough capacitor. Using hand made vintage PIO with straggly wires will not work!
 
There's a scheme for such an approach in the AD797 datasheet, iirc.
It's definitely not easy-homebrew territory, as PD makes clear.


ETA: found it:

797iv.jpg
 
Yes it is.

The part labelled 3K will set the output voltage (varies with Iout of the dac); so you'd pick C1 to best-suit the bandwidth required, and then you'd need to tune the RC combination (82pf+100R) for stability in-circuit.
And the AD797 need utmost-care in layout in the first place.

(and you can't just bung this in place of an existing I/V conversion ahead of an existing output stage, because the whole lot are conjugate in the sense of achieving a target output response. Lots of nasty filter maths to get to grips with /simulate. It's good fun, if you've nothing better to do...)
 
Yes it is.

The part labelled 3K will set the output voltage (varies with Iout of the dac); so you'd pick C1 to best-suit the bandwidth required, and then you'd need to tune the RC combination (82pf+100R) for stability in-circuit.
And the AD797 need utmost-care in layout in the first place.

(and you can't just bung this in place of an existing I/V conversion ahead of an existing output stage, because the whole lot are conjugate in the sense of achieving a target output response. Lots of nasty filter maths to get to grips with /simulate. It's good fun, if you've nothing better to do...)
I wonder if these points and those from @PigletsDad align with the way in which the infamous "ESS hump" in SABRE DAC intermodulation distortion has been traced to the I-V stage. On the surface it looks like tuning component values for reduced I-V stage gain and greater bandwidth have avoided triggering internal issues in the op-amp.
 
The AD797 datasheet is a good start, but its sensitivity is related to the design choices /architecture of the part.

The best single document - the most cogent and thorough single document I'd recommend reading in relation to using opamps is Linear Technologies' Application note AN47 - High Speed Amplifier Techniques, by Jim Williams.

There is a LOT of material there, but it's copiously illustrated and a brilliant all-round education - if taken in small bites! The most directly applicable are the detailed elements on how to use an oscilloscope, the effects and defects of bypass/decoupling; and breadboarding techniques. Bute there's very much more than that covered.
 
Purists gonna wince, but here's a stripboard layout based on a Busboard PR3U prototype board. This is about my fifth revision, and I'm happy enough with it to start looking at getting the relay logic and opamp regulation onto the bottom half of the board. PR100 is in reality off-board.

I've got the multiple ground lines to the edges, enough room around the opamps for power supply decoupling, and the relays also close to the bottom of the board to link to the logic circuitry. I've also got the offset compensation current injected right next to the IC101 input (bottom of R103).

A question for anyone following along. I've extended or added ground lines around the strip connected to the DAC output - the turquoise strip in the middle with the orange ground strips top and bottom. This was for screening purposes, but should I get rid of the extended ground lines because of potential capacitance issues?

There are a heck of a lot of components connected directly to the DAC's output, and I was wondering about the best way to handle that, because what little I know about layout is telling me that having most of them in a strip like this is probably suboptimal. Since this is DIY, I could go into the third dimension and come up with some sort of star arrangement, but I think that's probably attempting to guild a turd from the point of view of the effect of the rest of the layout?

y4moHbd7taSyT6JGqZLCeCsm-VIjikn6dBVM4g-pULFhbv86bSx3nJJlIfbZFwsEZjFUvfIE8nuHikOOQKH4ZikHl29DebWJmX20v-zOExjN-JZJCFy2khogQ3ACxdZmOqyNGuK3X4E2A-5X3eqRCFmCI6UsVILIT8gHJ-495aYhUN5U7w4VgAs29qB16VeQxuGBoHoWKBACij9Evhh8QIOPQ
 
If there is a way to directly couple the output from my 337 (or any Sony CDP) then please shout!

The TDA output current varies with temperature, so you would need a dc servo. Not worth the effort IMO.
BD added a precision reference to reduce the dc offset. This was an improvement over the emitter follower approach some manufacturers used, however I found it best to just get rid of the whole shebang and use capacitor coupling. I seem to remember you end up with 3.5V offset, but this will polarise any electrolytics nicely.

I also remember calling BD to tell him that his balanced phono stage had a pcb error which caused some dc offset (not at the output). I believe he fixed it after that.
 
Purists do wince at stripboard. To get a good low impedance I/V opamp circuit that is well behave to 1GHz is fairly easy using smd
 
Purists do wince at stripboard. To get a good low impedance I/V opamp circuit that is well behave to 1GHz is fairly easy using smd
Once I'd got the power supply on there, even with a different pre-fab stripboard with orthogonal rails at intervals, the layout was a complete rat's nest. If I go ahead with this project, I'll be teaching myself KiCAD in order to produce a well-laid-out double sided board.
 


advertisement


Back
Top