It's definitely cheaper purchasing extrusions than manufacturing your own from billet unless you have access to cheap metal. I inherited a couple tons of aluminum when a customer of ours shut down an airplane wing manufacturing facility and asked us to dispose of some of their moulds. It's very monotonous machining fins as well unless you have a CNC machine. It is cost effective and fast making heat sinks out of stacked sheet metal though. Fly cutting does a wonderful job of finishing them quickly too. I've done a little research into the anodizing. It looks pretty painless. That's my next experiment.
http://sound.westhost.com/articles/diy-heatsink.htm
Attachments
Could you make fins in an aluminum block quickly with a drill press and a cross slide vice using the right bit?
Could you make fins in an aluminum block quickly with a drill press and a cross slide vice using the right bit?
You would destroy the drill press quickly. The bearings aren't designed to be side loaded. It's even tricky to do on a milling machine depending on the depth of the fin. The fin will vibrate as you are cutting leaving a poor finish.
Updated Kypton-V2 "sansui" ....
Thimios responded with a 😱 ,when he heard it in it's "classic form".
Bimo had the same reaction ... and kept asking me to design it. 😕
Removed - The super-pair and the bad sound.
Added -
- C13 - lead comp.
-The anti-parallel diodes in the first stage.
- R34 for servo "settling".
- BJT OR FET at input. Perfect with every Jfet I tried !
Fet's are different ....
.007%- 20hz/ .001%- 200hz/ 3ppm 500-10Khz/ 14ppm-20K
Package below.
OS
Thimios responded with a 😱 ,when he heard it in it's "classic form".
Bimo had the same reaction ... and kept asking me to design it. 😕
Removed - The super-pair and the bad sound.
Added -
- C13 - lead comp.
-The anti-parallel diodes in the first stage.
- R34 for servo "settling".
- BJT OR FET at input. Perfect with every Jfet I tried !
Fet's are different ....
.007%- 20hz/ .001%- 200hz/ 3ppm 500-10Khz/ 14ppm-20K
Package below.
OS
Attachments
Last edited:
Infidel v1.0
Now it's alive!
Now it's alive!
Attachments
Last edited:
That's a nice square response! Similar to my MF80 project, which uses MOSFETs as OPS 🙂 (See here..)
Last edited:
VAS current is high and it's not controlled by the trimmer...
What is the control chanching R17?
Attachments
R28/R29 "steal" VAS
current from the Cascodes , lowering those values (trimmer) will
lower the cascode current with more (I) for the LED's.
You WON'T be able to see this on Q13/Q16 emitter resistors. To
calculate the cascode current , subtract the LED forward current
from the current you measure on R24/R31.
Another way I told still4given is to place a 10R between IPS/OPS PD+
and measure VAS current directly.
The full range on R28 should be 10K=3.8ma to 100k 5.8ma. Your CCS's
are running "hotter" 1.8ma .... you should have 4.5 - 6.5ma range.
Edit - the RED LED Vf must be 1.8v (newer red's).
R17 - "rbridge" .... you can reduce this to 5 to 10R and get 300khz
square-waves that look like your 50khz ones !! 😱
If you want to "max" this one out - 10-12pf for C7-8 and 5R for R17.
This will give 4.2mhz CLG and 1000V/uS slew with a little overshoot.
PS - this will "flat out" beat most forum CFA's and most definitely the VFA's.
You can even degenerate the diamond (and bridge) Re's - 27-33R and get
>>1Kv/us (1.2Kv/us is the record).
OS
Last edited:
OS, I wonder how Kypton-V sound with Hawksford cascode VAS instead single transistor VAS....
Oscillates even better than with the Baxandall. 🙁 Too many (poles & zero's).
You need some Ccb for such a high Z 2 stage IPS.
Baxandall and the Hawksford both work VERY well for the Infidel. I
chose the Hawksford because it's faster. 😀
OS
Should it reduce gain of high frequency at VAS like vzaichenko did with TMC compensation?
Hi, Val. I shipped the returns out monday@noon. You should see credit
when it reaches Minnesota.
What I meant was any 35mm snap-in. Most 6800's are 35 X 50mm and
8200's are 35mm X 60mm. I have over 100mm (between Wilhelm's aluminum
plates). I can fit 12 of these = 98400uf .... all for <10$ each. 6800's = 81Kuf
6-7$ each.
Even as this is less than the 4 - 33k's , 3 pair paralleled banks have lower ESR. Ripple would be reduced. 🙂 Much easier to source and afford.
OS
OS,
Are you still thinking of developing a power supply for all these different power amps? I imagine a universal unit won't work with the variations in numbers of output devices but perhaps that isn't true?
Are you still thinking of developing a power supply for all these different power amps? I imagine a universal unit won't work with the variations in numbers of output devices but perhaps that isn't true?
I don't know about Val's version , but I was overcompensating to keep
the baxandall local FB loop from misbehaving. My probe shows the
superpair to have a 15mhz peak. Thimios made a 19mhz oscillator out of it. 😀
Without the superpair , I could "unleash" the high gain of this IPS.
A smaller C3/C6 (last kypton V2 schema) gave me 2mhz CL unity gain and
excellent gain margin at 20K. THD dropped to single digit ... better than
with the superpair. I could even push UG further with the new C13 (lead
comp.) to gain some stability margin.
PS - this amp can't be fully evaluated for stability by probing
the main FB loop ! It's a "freak"
In the real world listening tests ... dropping the superpair caught thimios's
attention instantly. 2 post saying "please" 😱😱 fix this amp !! 😀
Baxandall = good for LIN's and CFA's (apex's amps - kypton C). NOT for
this one.
Why fix what is not broken ? I ask ....
OS
Just get a cheap multi cap PCB for 35mm snap-ins from ebay. 4 caps per
channel for 2 pair , 6-8 caps per channel for the monster.
You can adjust total capacitance - 4700u , 6800u , even 10KuF(per cap).
All are 35mm w ..10mm leads and will fit.
I intend to etch my own cap PCB's - real simple with quick connects.
OS
Attachments
Although it may look strange, my prototypes with Baxandall + Hawksford (with the tube at the input) feel very good. Soldered the second channel, hooked up OPS - just works. Situation with Terry's build is not so easy, but my two channels don't show any problem. What's going on? 😕 🙂 It behaves very close to simulation... 🙄
Would you try BJT version?
I still do not understand clearly how tube behave 😀
It look like tube have high capacitance "base - collector" (if BJT, what it called for tube?).
Attachments
For the tube it would be "grid - plate" 🙂
It's somewhat closer to jfet than to bjt... high impedance, high linearity, low gain and - you're right - rather high input capacitance.
I will try bjt version a bit later - prototyping the "clean CFA" now 😉