Beefcake: a beefed-up BA-2

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've been working on something to drive my Maggies. It's essentially a BA-2 beefed up to dump 60 clean watts (and a couple hundred dirty ones) into 4 ohms.

The LTP's CCS uses BJTs to handle the higher rail voltage. They're a bit noisier than a JFET CCS, so the LTP is paralleled a la the J2 in an attempt to claw some of that back. The LTP is also cascoded for the higher rail voltage.

The VAS bootstrap CCS is also replaced with a BJT version. No particular reason, it just looks more symmetrical that way. ;)

The power supply is split with a 49V shunt reg for the front-end and a 46V CRC for the output stage.

Schematics in the PDFs; a somewhat simplified LTSpice one is also attached.
 

Attachments

  • beefcake frontend.pdf
    114 KB · Views: 117
  • beefcake output.pdf
    41.4 KB · Views: 97
  • beefcake.jpeg
    beefcake.jpeg
    544.8 KB · Views: 763
Draft of PCBs. Dissipation is on the order 350W/channel, so a monoblock design is used.

The first board contains a front end with + and - shunt regs; the second is one-half of an output stage. The OS board has output connections at each end so a single design can be used on both sides of the case.
 

Attachments

  • fe_front.jpg
    fe_front.jpg
    187.9 KB · Views: 770
  • os_front.jpg
    os_front.jpg
    77.7 KB · Views: 700
I found a more suitable opamp for the FE shunt regulator (LTC6090). It does require soldering a thermal pad on the bottom to the board. :cool:

I put a pair of large vias in the pad to flow some solder through. The datasheet recommends 7 smaller vias, but I think the larger ones will allow me to use the solder as a heat-wick and therefore (hopefully) avoid the whole toaster-oven-reflow thing.

In service the solder filling should more than compensate for the reduce via wall surface.
 

Attachments

  • opamp_back.jpeg
    opamp_back.jpeg
    22.6 KB · Views: 80
  • opamp_front.jpeg
    opamp_front.jpeg
    50.9 KB · Views: 561
Oh, and I also dropped to a lower voltage Zener for less noise. There's no bypass cap in the Spice sim (because Spice doesn't appear to model Zener noise anyway), but there is in the Kicad schematic and board file.

I've attached the Spice sim of the shunt reg. If anyone wants the Spice sims of the rest just holler.
 

Attachments

  • beefcake_fe_opamp.asc
    7.4 KB · Views: 39
Just started learning the basics of spice, got a long way to go. The numbers look good.

My bench tests with a dynamic load suggest cc; listening is another matter, sometimes I find with some circuits, a simple CRC best, you never know what is until you try all the combinations; thats where the bench and spice go out of the window.
 
Having completed my J2 (and determined that I do indeed like its sound signature), I've started to take a look at Beefcake again.

I matched a tube of Harris 240s and a tube of Harris 9240s, and it's easy enough to get close matches so I dropped the source resistors by 33% and marginally increased the bias.

I also swapped out the IRFP610 (in the VAS) for a 2SK2013, which is much more linear.

attachment.php



I've selected a pair of 120VA 2x46V transformers for the front ends and a pair of 1500VA 2x35V for the output stages. The output stage transformers have some pretty beefy wire coming out, so I've also done a small variation on the double-bridge board with larger holes and a bit more copper:

attachment.php


I did not update the PSU board because I'm planning on point-to-point wiring the cap bank.

Cheers,
Jeff.
 

Attachments

  • beefcake spice.jpg
    beefcake spice.jpg
    222.8 KB · Views: 800
  • beef bridge.jpg
    beef bridge.jpg
    92.3 KB · Views: 782
I discovered that the relatively large resistance in the LTP leg led to asymmetrical clipping. The easiest solution would be to raise the FE voltage... only I already have the transformers in hand.

I was able to lower the resistance and symmetricalize (new word, there) the clipping by going from a 1:4 LPT:VAS current ratio to a 1:3 (9mA:36mA -> 10mA:30mA).

The slight bump in LTP current also raised the gain slightly, allowing me to apply a bit more NFB to improve the HF response. (The simulated square-wave output is now, well, more square.)

Some of the annotations on the previous SPICE circuit were stale. They should have been:

FR-3dB: DC - 110kHz
Full-power bandwidth: 75kHz
Slew rate: 18V/uSec

AOL: 43dB
ACL: 22dB
Stability margins: 52º, 12dB

The new numbers are:

FR-3dB: DC - 130kHz
Full-power bandwidth: 85kHz
Slew rate: 20V/uSec

AOL: 44dB
ACL: 21dB
Stability margins: 50º, 13dB

Cheers,
Jeff.
 

PRR

Member
Joined 2003
Paid Member
> the relatively large resistance in the LTP leg led to asymmetrical clipping.

And higher even-order THD at all levels.

A current-mirror is an elegant and correct fix. Too expensive in 1969 but this plan dies not seem to be shy of parts; another 2 Q seems minor.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.