DIY preamp design - gain topology, IC vs. discrete, OPA627 vs. Zen... ?

hi everyone,

ok i know people have already asked all the questions i'm about to ask, but i'd like to pick everyone's brain and have a nice open discussion about some high-level as well as finer points of designing a DIY line stage...

so i'm thinking of building a simple preamp for my system... first thing that comes to mind is to use IC opamps. i know a lot of audiophiles hold their nose when you mention "IC opamp" but i think if you lay out your circuit properly and use the proper devices you can get pretty good sound out of ICs. in particular, the OPA627 seems to be a popular pick. however, i'm also interested in how the Pass Labs projects sound (e.g. DIY discrete opamps, Zen line stage), and how they would compare with a properly-executed IC design.

if i use a OPA627, i am going to go with the surface mount components and get the very shortest signal path possible - i intend to have my entire feedback path less than 1cm long and with very short leads to signal and power supply bypasses. i think if the layout is super-tight, the performance of the device can be maximized. furthermore i will be using a choke-regulated power supply and LT1085/1033 regulators at the very least.

that said, i have a couple questions:
<ul>
<li>should i use two active stages, one before the volume pot (for input buffering) and one after? or just one after the volume pot? i prefer the latter (less is more) but there are performance considerations...
<li>if i use two stages, how should i split the gain between the two? maybe 6dB at the input buffer and 12dB at the output stage? or just buffer the input? i guess i am looking for the best balance between noise performance (more gain up front) and distortion (more gain at the output). since i am limited to using +/-15V rails w/most ICs, most of the gain probably needs to be in the output.
<li>Walt Jung, the opamp guru, seems to favor always buffering the output of gain-stage opamps due to current-drive, thermal, and distortion considerations. like i said though, i hate having more circuitry than is absolutely necessary. do the benefits of a buffer outweigh the drawbacks?
<li>if i pull it off the way i said i would, do you think my suface-mount IC design (using probably OPA627) will sound ok? :)
<li>will the Pass Zen design, if properly built, still kick my opamp's ***?
</ul>

ok those are my thoughts for now... if anyone cares to add additional questions or comments, feel free! i'd really appreciate any input whatsoever, not only for my own benefit but also because i'll probably be writing a technical column for Bound for Sound magazine in the near future so this would be good research...

thanks,
dorkus (marc)
 
As far as Line Stages Go I made a Headphone amp consisting of a Buffer/Opamp combo. One of the opamp's i evaluated was Both the OPA-627 & OPA-637. The Buffer is (3) BUF_634's in Parr. Both of the above required a Constant Curremnt source on the Output of the Opamp, or the Sound was Quite Gritty. The Current source Cleened things up alot. The Opamp I ended up with and since you are going SMT is the AD-825 in my view the Best Sounding Opamp i have Heared. The AD-825 is alot more anilitical than thr 627 or 637. also a less foward madrange. The Voices on the AD-825 & Buf-634 are What i would Consider Amoung the Finest sounding Line level stages i have heared. I did make a All Casscoded Pure Complimentry Headpone amp that did sound slightly Better. Cost Less in Semiconductors. Opamps and Buffers are Not Cheep Cost wise. I like the Multiloop Feedback Technique as discribed by Walt Jung.
 
hmm interesting, did you try your designs as line stages only, instead of driving headphones? i'm surprised the OPA627 sounded gritty, i've tried it before in different applications and thought it sounded pretty smooth. did you keep the BUF634 inside the opamp's feedback loop, or run it open loop? do you also recommend using the buffer even if running just a line stage? i will look into the AD825...
 
The OPA-627 is smooth sounding untill you hear it with a Current load on the out put. Then you Can Notice the Lack of Grain.Then Drop in an AD-825 and Wow you will never go back to a OPA-627. I recomend a Buffer in line stages also because the Cable capacitence of most High End cables and most surly any long lenghts of any cable.Some people like use the Buffer open-loop so as to remove the Possibility of Load intereaction with the Feedback Loop. However the Distortion of just the Buffer stage will not be reduced by Feedback. So my Ideal solution is to have the Best of both Words and use a Multiloop feedback Topology. This way the Open-loop band width of the Gain stage Can Be increced well over 100KHz. rather than the the 100Hz. or so if reling just upon the Gain slope of the Opamp.See My email attachment.
 
Sorry i have the Schematics in a MS Word doc. and havent found a good place to host them. am still looking. If you want the same amount of Dynamics as the Zen. then you will need to parralell about 4 BUF-634's to get a Low output impedance or you can use My fav. Buffer the EL-2008CT or EL-2009CT Both are going to be Discontinued the First of the Year. But the EL-2008CT is available from insight Electronics $10.52 Ea. The data sheet is at http://www.elantec.com This Buffer is The Best sounding Buffer i have Heared. The BUF-634 is almost as good if you use 4 of them. But The BUF-634 is like the Burr Brown Opamps Soft sounding. The Elantec Buffers are Verry Anilitical sounding.
 
Line Driver

Michael,

Great circuit for obtaining a balanced output, but I think it still needs a buffer to get the best out of it, like ppl recomends, as the transformer has a pretty low input impedance.

Jam

P.S. I noticed that you had built a Lang Class A amplifier how does it sound and what have you done to modify it?

[Edited by jam on 06-30-2001 at 01:07 PM]
 

mlloyd1

Member
Paid Member
2001-02-25 7:10 pm
Northern Iliinois
Things I did to the Lang:

Added Servo for no DC offset

Changed cascode to some funky design from a Malcolm Hawksford paper I was curious about (supposed to be much lower distortion than conventional cascode). I'll dig it up for a reference if you're interested.

Probably a few other things, too. That's what I remember. It's been a while.

I like my Borbely amp better. My speakers are 4 ohm (Kef 104/2). I spent A LOT of time and attention to the Borbely to make it sound as good as I could at the time: lots of output devices, good grounding, monstrous power supply, matched devices, etc.

Some day :) I'll get around to tuning the Lang; there's a lot of potential there. I spent my time on the Borbely because I wanted to play with a JFET front end.
 
Fet input modifications

ppl,

I would first set the voltage on the base of the cascode transistors (of the differential pairs) to about 10volts. Replace the differential transistors with the jfets you mentioned. Fets like to see about 10 volts on their drains for lowest noise and distortion.
Reduce the value of the emitter resistors of the differential (if there are any there )because fets have lower transconductance than bipolars.
Hopefully the differentials are current sourced so you won't have to adjust the value of the drain resistors.
Fire the amp up with a variac and check for stability and offset. If everything is ok short out the large electrolytic in the shunt part of the feedback circuit.
You can then play with the values of the source resistors on the differentials or adjust the current sources for zero offset.
I have modified several Aragon amplifiers (which use a similar topology to the Lang and Leach amplifiers) replacing the diffefentials with jfets or mosfets with great results.

Removing the large feedbeck cap has a great impact on the sonics of the amplifier (for the better).
Another area to experiment would be to reduce the amount of feedback used, by reducing the open loop gain of the amp.

Jam

[Edited by jam on 07-01-2001 at 11:36 AM]
 
Jam: Thanks for your solid advice. My topology is similar not exackly like the Leach. The amp is already Direct Coupled thank god i hate Coup. caps. Some of the other input details are the voltage ref for the input Diff amp cascodes are set at 15 volts (Base of cascodes to Emmiter of Diff amp) A current source supplies the 15 volt Ref. The input stage is a complimentry version of what was in the Mark levonson #23 (I thought that was a cleen amp and felt comfortable with their aproach. My amp is i gess you could say made up of Ideas from manny Designs. Like Passes Cascoded output stage ect. i am also thinking of replacing the BC-107's and BC177's used as the active device on the Second voltage gain stages with Complimentry jfets also like maby 2N5458 & 2N5460? other recomendations would be helpfull. I think the BUT used on the Non active Half of the second stage cascode breing a High voltage Types can stay. One again thanks for you input.