Anyone built Sijosae's class AB amp ?

  1. FallenAngel Contributor
    After looking over Sijosae's CCS comparison, I'm not sure which CCS to use. It says the 2N5484 CCS isn't good but recommends the LM317 CCS as very good and I used it in a Sijosae Class-AB II amp before. Would it be a good idea to use the it in this application as well?

    This is what I'm thinking for the buffer using this modified LM317 CCS: Please take note, that I have absolutly no EE background and have totally no clue if I'm breaking any (or all) basic rules here, but I am fairly creative with mixing and matching different schematics in hopes of making something work


    I have read over this tread (Page 3 and 4 are the most important ones) to find some more information on the CCS to use in diamond buffers and now I'm more confused than ever.

    Unfortunatly I cannot get 2N5484 jfets, only 2N5486 which have a much higher current rating. After looking over at mono's design of a modified Sijosae Class-AB buffer, I've merged them into this (again, I'm hoping it makes sense [​IMG] ):


    Please let me know if you guys think either of these buffers would work well - as in better than the original design in the LM317 case and just as well in the 2N2586 case.

  2. steinchen Contributor
    Sijosae's ccs comparison only compares how the supply voltage affects the ccs current, imo there are several specs that are far more important: PSRR, impedance, noise and capacitance. As long as you don't need high currents (of more than 20mA or so) or run the ccs at high voltages cascoded jfets are superior. Even a lone jfet or a CRD is a good choice as a ccs.
  3. FallenAngel Contributor
    Thanks steinchen, I'll get my thoughts back to the 2N5486 version.

    cetoole: I just did a search on threads you started and came up with so much great info! [​IMG] After checking out these buffers, it seems I have very little if any originality in the above schematics so I'll just ask, how is the second from the bottom which is VERY similar to the 2N2586 I had. I added the inline input resistor of 4.7Ohms to match the output resistor.

    I'm actually very surprised at myself for even coming up with something similar (but of course somewhat worse) than the great looking design here. So I will (shamefully [​IMG] since I was thinking of coming up with something a little new) be turning my thoughts to breadboarding a slightly modified version of that design since I can't get any 2N5484's and I'll be probably using 2N5486's.

    Something interesting. From what I know, those jfets are used to run the buffers in Class-A. By adding them to the output of the amp, before the buffer, I'm thinking makes the opamp run in Class-A as well. So for using the 2N5486, would the 1K resistor be alright for current limiting, I saw mono having a 1K in his, so I'm hoping it'll be alright.

    After reading the schematic, I've taken guesses for a lot of the parts, but I can't figure all of them out. I'll post my guessed values tonight, but the main ones that I'm not sure about are:

    Caps around the OPA627: I'm guessing as per datasheet they're 0.1uF ceramics, but I'm not positive.

    R10 (R / L / G): As per Pimeta it's 1K, but since I'm using 2N5486 instead of 2N5484, will it be different and will I need to add another resistor for current limiting, but I'm guesing those are R19, R20 & R21. So, obviously, my next question is what's the recommended value for those 3?

  4. cetoole
    FallenAngel, there is more useful information for what you are looking for available on Headwize, so you may wanna check there. For example, in this thread on discrete buffers, AMB does a great job explaining what does what, and how it works, as well as how to calculate some of the data. I dont have any of these PCBs, due to the difficulty in getting small quantities manufactured for a reasonable cost, so unfortunatly, I have not been able to try it. The jfets seen in the buffer are not for class A biasing, they just set the quiescent current of the input stage of the buffer, which affects that of the output stage. It is actually the emitter resistors on the input stage which pushes it towards class a, and the ratio of input transistor emitter resistors to output transistor emitter resistors along with the idss of the jfets sets the quiescent current of the output stage (how deep into class A). AMB does a much better job explaining it in that thread I linked above, so check it out. 1k source resistors on the 2N5486 works, but you really should calculate the values to get the current you want.

    In that dDBed pimetaesque amp thread of mine, yes, the bypass caps default value would be 0.1uf ceramics. X7R is just fine here. Sounds like your amp will end up pretty close to the one I have done here.

    Steinchen got it dead on with regards to what is important in the CCS for this purpose. What Sijosae was measuring has little relevance to this, as our supplies will remain almost constant. This is why buffers like those used on the PPAv2 work great, even with the inaccuracy of the standard current mirror, and the simple jfet+source (adjustable) resistor. The bias current may not be dead on, but it tracks between the positive and negative supplies evenly. If I was to use a different CCS besides jfets or crds, I would definitly go for the LED based one, which, IIRC, is what Mr. Walt Jung used in his origional diamond buffer that started all of this. Also, for the LM317 CCS, which I would not recommend for use here, you should have the wire from Adj to the base of the input transistors go instead to the junction between the 620ohm and >4.7ohm resistors. The 2N5486 CCSed diamond buffer looks good though, and I would expect very nice results using this. I say this however without running the numbers to determine quiescent current. It will definitly be easier to build than my drop in buffer for the pimeta, due to your addition of source resistors on the ccs jfets.

    Keep up the good work though, I cant wait to see this built.
  5. FallenAngel Contributor
    Oh great, after reading the latest schematic on this amp, I realized my amp will be almost EXACTLY like it [​IMG], call me a little theif, but I totally love that DB design *which someone should have told me already existed [​IMG]* with a jfet cascode.

    So that more or less seals my fate at this buffer design:


    Also, since I think i'll be moving much more in the direction of cetoole's design, further discussion on this should be moved to his tread.
  6. sphinxllama

    this amp looks great and I plan on building it. most of the parts I got at my local electronic dealer. but I have difficulty to get one of this opamps I listed below:
    AD823 / OPA2134PA / OPA132 / OPA134 / OPA2107AP / OPA2227PA / OPA2132PA

    does anybody have experience with other opamps besides the above or knows any good one I can use with this amp and that are easy to get?
    I have two opamps from other projects:
    can I use one of these?

    I have to more questions about the layout. I drawed the battery supply in the shematic is it right like this:

    and @thedoctor why do you use this parts i marked at the picture:
    cant find it one sijosaes shematic.

    thanks for your help hopefully [​IMG]
  7. heatmizer
    You can use the ne5532 but it is a bipolar opamp you need to be aware of your resistors they have to match on the input and feedback. Maybe add a 1K between pot and input of opamp and a 100K to ground. Use 6K and 1.2 K in feedback loop for a gain of 6. No input cap so watch your offset. As far as the railsplitter goes good luck getting it balanced. Look at The Doctor's schematic on his ground channel with a opamp in the ground channel the the railsplitter will balance itself. The 2 things you circled are small box caps bypassing the electrolytics
  8. Nordic
    You still got to love that guy.

    I just finished a minor mod now to get his Poor Man's Zen working.... I was naughty and didn't follow his layout, and paid the price in hours wasted.... copying him more closely this headphone amp certainly is amongst the best I built.... top 2.
  9. sphinxllama

    thanks for your quick answers! i do not understand a thing you wrote. mybe some of it but i am a newbe in diy and my english is not that good eather [​IMG] just finished my first cmoy. it worked great and now i want to step up.

    so the easiest way would be to buy the opa2134 and just build as shown on sijosaes shematic. should i use the electrolytics 100nF (right???) and put both between the +4.5 and GND and the other one between -4.5 and GND and connect both GNDs?
    is the connection from the battery correct? sorry but i am confused because on the pictures from thedoctor the battery power goes straight to the opamp. or do i look wrong [​IMG]

  10. heatmizer
    build the amplifier part as shown but you will have problems with the power section. sijosae's amp sound really nice but the Railsplitter (read power section is not very accurate i could never get it within 1/2 volt from positive to negative. if you really want to build the power section as is protoboard the thing 1st and switch out bc327 and bc337 until you have an even split 1/10 of a volt is good so say 1, 9 volt battery +4 and -5 volts would be bad +4.5 and -4.6 volts would be good. my solution was a buf634 for my railsplitter/power section.

    the yellow things are small film capacitors bypassing the electrolytics which are on their side
  11. TzeYang
    ^alternatively, just match the transistors using a HFE capable DMM. (which is cheap)
  12. .:tuxianer:.
    Hi,is there anywhere a part list? An i want to build this amp for my Grado sr325. Can I use the standard parts or should i change they to optimize them for my Grado?Thanks
  13. FallenAngel Contributor

    Originally Posted by .:tuxianer:. /img/forum/go_quote.gif
    Hi,is there anywhere a part list? An i want to build this amp for my Grado sr325. Can I use the standard parts or should i change they to optimize them for my Grado?Thanks

    All the parts are in the schematics.

    Customize all you want [​IMG], you can lower the gain if it'll only be for Grados, but I didn't see any real reason to with mine.
  14. .:tuxianer:.
    Ok. Ithink I will take the standard parts. Is ther any final schematic? And wich version sounds better? I or II?

    And what's this: [​IMG]?
  15. TzeYang
    it's a small intergrated TREAD regulator.

Share This Page