Well, an even easier and much better performance fix was already presented, see the 2nd simulation.
If you want a current output capacity beyond the opamp, buffer with an LME49600 (within the global feedback). This was done in the amazing The Wire headphone amp project here at DIYA.
The LME49600 buffers can be mounted on a perf board, then piggybacked on the "C-3850" board.
I'll upload the simulation files tonight.
That's perfectly true, in fact you can also buy boards from AliExpress which do just that - dual channel LME49720+LME49600 headphone amps straight out of the TI datasheet. I have some and they are excellent, but you are still restricted to opamp supply voltages, which restricts dynamic range. My mod is only a couple of resistors while your suggestion involves more expense and work. More importantly most of the contributors to this thread already have C3850 boards, or are going to have them made from Peter's excellent Gerbers, and are looking to get the best from them. I could easily take the C3850 theme and with fairly minimal redesign produce something which would give you the best of all worlds, but again I don't think that's what most contributors want? Incidentally there's another small mod ( no messing about with the pcb ) which is generally good, but particularly useful if you're powering the opamp separately as it makes the dc conditions of the rest of the ( higher voltage ) circuitry power supply independent. Otherwise you'd have to change resistor values to suit each power supply voltage. Details if anyone is interested.
I did a fair amount of poking around for Accuphase schematics on the web and found almost nothing on the C3850 - I doubt anybody bought one to just reverse engineer in order to make low cost pcbs for AliExpress. I'd love to be proven wrong. I did find a couple of instances where they'd used current feedback amplifiers to feed a diamond buffer+vas stage but these used darlington pairs, rather than single transistors, for the vas and a conventional emitter follower output stage. All of these were for much earlier products where service manuals are available. I'd love to see a C-3850 service manual, it must run to hundreds of pages.
I did a fair amount of poking around for Accuphase schematics on the web and found almost nothing on the C3850 - I doubt anybody bought one to just reverse engineer in order to make low cost pcbs for AliExpress. I'd love to be proven wrong. I did find a couple of instances where they'd used current feedback amplifiers to feed a diamond buffer+vas stage but these used darlington pairs, rather than single transistors, for the vas and a conventional emitter follower output stage. All of these were for much earlier products where service manuals are available. I'd love to see a C-3850 service manual, it must run to hundreds of pages.
tonight came sooner than usual.... here is the LtSpice files
Enjoy.
Enjoy.
Yes, @chalky . I would be very interested in your ideas for this board. A 5532 as preamp sounds very good, but an added buffer stage sounds even better. HPA as line preamp can sound really good.
Thanks @nattawa for info on "The Wire". This LME49600-based headphone amp mentioned near the end of the thread.
Thanks @nattawa for info on "The Wire". This LME49600-based headphone amp mentioned near the end of the thread.
Last edited:
Hi...
My guests had left and i had a little time to take another look at the pcb. I let the BD transistors move a little bit down to get more space to the big capacitors.
A few tracks moved fom f to b side. A few tracks, as voltage tracks are bigger. Some tracks i made shorter...
Small changes.
Most importend...my name is written on the pcb🙂😉
@chalky
(I dont know your real name...)
Do you you think it is possible to hear your changes ? I have a scope and function generator, but dont know tu use it...electronic newbie in some cases.😎
I have only my ears to verify, subjective
Peter
My guests had left and i had a little time to take another look at the pcb. I let the BD transistors move a little bit down to get more space to the big capacitors.
A few tracks moved fom f to b side. A few tracks, as voltage tracks are bigger. Some tracks i made shorter...
Small changes.
Most importend...my name is written on the pcb🙂😉
@chalky
(I dont know your real name...)
Do you you think it is possible to hear your changes ? I have a scope and function generator, but dont know tu use it...electronic newbie in some cases.😎
I have only my ears to verify, subjective
Peter
No, it was not for the gain, it changes the forward path frequency characteristics of the minor loop (consists of the opamp and the first buffer stage). I started out with 560R, had stability problem. Oscillation. Reduced to 22R and all was good. I didn't bother trying other resistance values.
Thank you guys for your effort. 👍
everybody is free to digging and modding....just do it. but lets keep this thread as it was .
chris
everybody is free to digging and modding....just do it. but lets keep this thread as it was .
chris
Last edited:
Hi nattawa
i put the symbol in the symbol folder under opamp. saved the lib file under the lib folder.
if i want to run the sim...--->
Why is LT spice asking of opening the lib file in /lib/sub...??
sorry for the noob question.
thanks chris
i put the symbol in the symbol folder under opamp. saved the lib file under the lib folder.
if i want to run the sim...--->
Why is LT spice asking of opening the lib file in /lib/sub...??
sorry for the noob question.
thanks chris
Perhaps OPA 1656 is missing in the libary. This way https://www.diyaudio.com/community/threads/opa1656-high-performance-cmos-audio-op-amp.335416/page-15
Post 291
I dont know the LT spice program but i know such problems in KiCad...
Post 291
I dont know the LT spice program but i know such problems in KiCad...
I did find service manuals for other models with schematics at audiocircuit . For example, the C-2410 . See page 22.
Last edited:
The .lib and the symbol file are in the same zip file I uploaded. You need to either copy them to the library/symbol folders in your LtSpice system, or add search path in your LtSpice setting to let it know where they are.
If copy files, . lib file should be in \sub, and the symbol file be in \sym. Did you try the other way by adding search path in the settings to point the way? From my not so reliable memory, the LTSpice may not allow altering its default library folders, depending version.
By the way, the error message shows the Spice was trying to find the . lib in my PC…
By the way, the error message shows the Spice was trying to find the . lib in my PC…
Ok, so here are my proposed mods, they don't involve hacking the pcb only changing parts. I'm sure the changes will be audible because the vas and output stages will now be operating in solid class A. So there's no doubt about the part numbers I'm using Peter's schematic below. Diamond buffers work best with high hfe transistors, which the BC550 and BC560 are. But a some of tha gain is being wasted because the hfe of the BC550/560 pair doesn't peak until a collector current of 5-10mA whereas they are all running at collector currents of 1.0-1.5mA here. So it would be advantageous to increase the collector currents of all the transistor currents here. Its common practice to set the current in diamond buffers with constant current sources and current mirrors and this gives a number of advantages including better PSSR. This would usually involve adding another six transistors to this circuit - so not really an option here. But wait, we can use constant current, or current regulating, diodes (CRDs). These look just like little 1n4148 small signal diodes but are in fact tiny ICs using a jfet and resistor to set the current. So the 15k resistors R12, R16, R19 and R21, which set the diamond quiescent current can be replaced with 3.5mA CRDs. This will up the hfe of all of the transistors and set the quiescent current of the vas transistors Q4 and Q10 to 7mA, solidly into class A with increased hfe. The emitter resistors of Q6 and Q12 can be reduced from 22R to 6.8R which will increase the quiescent current through these transistors to about 11mA, so they should run in class A even at the maximum output swing. You can get the CRDs ( they call them current limiting diodes ) from AliExpress at a reasonable price. Search for "current limiting diodes" and select the E-352. There is a bit of variation in the current between diodes and its best if the pair of CRDs for each diamond are matched - shouldn't be a problem because the best price for these parts on AliExpress are in a pack of 10 and that should be enough to get two matched pairs. If anyone wants to try this and is uncertain about how to use and match CRDs I can explain. Even if you dont want to try the CRDs reducing R12, R16, R19 and R21 from 15k to 5.1k will set the currents to the same level as the CRDs would. Don't forget to change the change emitter resistors of Q6 and Q12 from 22R to 6.8R. I'll try and find the time to populate one of my pcbs with all of these changes in the next two or three days and will post some photographs ( the orienation of the CRDs matters ). Hfiamps there are plenty of Accuphase schematics on the web just none that I could find that relate directly to this circuit, did you find any?
I forgot to say that the current through the CRDs is independent of supply voltage but you'd need to change the 15k ( or whatever value you'd set them too ) resistors if you changed the supply voltage by more than a little. Of course this only applies/matters if you're using separate supply voltages for the opamp and the rest of the circuitry.