advertisement


Dolby B/C decoding open source software

The patent contains example circuits. Does anybody know if these are actual values used? It would be a very poor patent that could be bypassed by changing a few part values

Looking at Jims scans above, the original circuit must have been a calibration nightmare, with 10% components and no attempt at temperature compensation.
I know from my own experience that jfets are very inconsistent variable resistors, even from a single brand thanks to fab variations. I used to buy house numbered selected jfets from Motorola long ago

Section 6.6 is the key to avoiding pumping, the slow and fast attack and decay have to be just right
 
Last edited:
You could convert one of the analogue schematics to LTSpice and use the audio file as a stimulus, output another file.

edit
What is missing is the selection parameters for the 2SK30D jfet, a very broad spread chopper device. The VCA behaviour is very dependent on Vgs vs Rds
 
Last edited:
I can tell you from personal experience that the JFETs were a particular single-sourced manufacturer type in all the Dolby discrete NR systems then usually further selected and/or matched for specific parameters. Each compressor stage then had "adjust-on-test" (AOT) resistors selected to trim the performance so that all encoders/decoder modules behaved the same to a matrix of amplitude and frequency points across the operating bandwidth and dynamic range.

In the case of the final Dolby SR professional encoder there were almost 100 AOTs to select per module, many of which were interdependent and required sequential adjustment in a specific order - which eventually we got a computer to do once we hit mass production.

The IC versions that were in most consumer tape decks for B/C/S were much simpler usually just 1-2 AOTs per channel.

The early versions of C in consumer implementations were more prone to "pumping" because they were essentially two B systems in series and not quite as well matched between different systems. Only when a full single chip C encoder was produced did consumer decks start to behave more like the professional versions
 
You need a very specific Vgs/Rds curve for this circuit to work

Another variable in the design is the use of OA91 germanium point contact diodes. These have very low forward voltage at tiny currents, but Vf is quite high and very variable at 10mA, while the reverse leakage can be poor. Nowhere near as good or predictable as a decent low voltage Schottky
 
davidsrsb said:
You need a very specific Vgs/Rds curve for this circuit to work

Another variable in the design is the use of OA91 germanium point contact diodes. These have very low forward voltage at tiny currents, but Vf is quite high and very variable at 10mA, while the reverse leakage can be poor. Nowhere near as good or predictable as a decent low voltage Schottky
Click to expand...

yes later circuit versions used BAT-41
 
I am surprised that the Nakamichi schematic is almost identical to the Wireless World. Small component differences would have caused different behaviour. I suspect Nakamichi spent some extra effort on selection and adjustment to get their quality
 
wd40addict said:
Well apart from the one I linked to up thread...
Click to expand...
The source for that project doesn't seem to include that for all the small libraries that do the interesting work. From their names it appears likely that it more or less directly emulates the electronics (eg 'diode.lib', or 'FETres.lib'.)

So, in the usual meaning of the term, it's not an open source project. Which is a shame. Pascal isn't a problem per se.
 
Glancing at the code, it is a circuit simulation. This is doing the same job as my LTSpice model suggestion.
As the code is using simple ideal diode and jfet models, LTSpice is likely to give a more realistic and accurate result
 
JensenHealey said:
This month's HiFi News has landed with a letter and replies about this very subject.
Click to expand...

Yes. Prompted by Barry raising this in an earlier issue in his HFN column. This in turn reminded me it was on my 'to do list' from ages past but I've lacked the round tuits to chase it down. I then talked to Barry about it and decided to see if I could find a solution rather than try to DIY one. Hence asking here.

Doing it via Spice is an interesting idea. However I've not used it for many years. 'C' would be quicker from that POV for me, but Spice might be better. So if no other solution/volunteer pops up I can put that on my to-do list.

In ye olde days I tended to find it quicker to build an actual circuit and measure it than use Spice. But I can't 'embroider' tinned copper wire onto a board as quickly as I used to. 8-/
 
You must log in or register to reply here.


advertisement


Back
Top