LatFet Amp Based on Philips AH578

Member
Joined 2014
Paid Member
Here is a little, simple amp that turned out to be easy to build,
and with decent parameters.
It's based on Philips AH578 (which was a quasi amp), originally designed by Earl Rapp, but built with modern devices (LT1056 op-amp, Exicon dual-die FETs). Rails at 45V.
All screenshots taken with 8 Ohm load.
DC offset at the output: 2mV, and stable.
Idle current: 80mA per device.

PCB design: post #13
HexFet version of the amp: post #124
PCB design for HexFet version: post #130
Push Pull HexFet version (fast and stable): post #1
...
 

Attachments

  • philips22ah578.jpg
    philips22ah578.jpg
    232.4 KB · Views: 2,784
  • 1khz_thd.png
    1khz_thd.png
    23.5 KB · Views: 2,666
  • 1khz_fft.png
    1khz_fft.png
    19.6 KB · Views: 2,083
  • sinus_1khz.JPG
    sinus_1khz.JPG
    235.8 KB · Views: 1,983
  • sinus_34khz.JPG
    sinus_34khz.JPG
    225.4 KB · Views: 1,857
  • sinus_100khz.JPG
    sinus_100khz.JPG
    241.1 KB · Views: 453
  • square_1khz.JPG
    square_1khz.JPG
    241.2 KB · Views: 461
  • square_14khz.JPG
    square_14khz.JPG
    229.9 KB · Views: 495
  • square_34khz.JPG
    square_34khz.JPG
    232.6 KB · Views: 1,042
  • philips22ah578.asc
    10.4 KB · Views: 189
Last edited:
  • Like
Reactions: 1 user
Member
Joined 2014
Paid Member
Here is actual build.
Mounted on Alu chassis, with external PSU (with speaker protection and VU meters).
Heatsinks are adopted from Dell 4U server. ($6 each from Ebay).
PCB was originally designed for smaller devices (single die), but finally I went with dual-die Exicons, and they turned out little bit bigger..
 

Attachments

  • pcb_philips1.JPG
    pcb_philips1.JPG
    110.6 KB · Views: 930
  • test_bed.JPG
    test_bed.JPG
    264.6 KB · Views: 821
  • DSCN0526.JPG
    DSCN0526.JPG
    256.8 KB · Views: 610
  • DSCN0531.JPG
    DSCN0531.JPG
    214.5 KB · Views: 637
  • DSCN0537.JPG
    DSCN0537.JPG
    244.9 KB · Views: 641
  • DSCN0539.JPG
    DSCN0539.JPG
    142.8 KB · Views: 711
  • DSCN0540.JPG
    DSCN0540.JPG
    213.9 KB · Views: 785
Member
Joined 2014
Paid Member
Due to several inquires, I'm attaching here Sprint Layout design for the PCB.
PCB has 2 separate ground connections:
a) signal
b) dirty
These are NOT connected together on the PCB.
I run two separate wires from the PSU's ground to the PCB, one for a) and one for b).
Pinout for small transistors fits KSA992 type (BCE), however it's very easy to fit 2N5551 type pinout (CBE), they just need to be rotated slightly. Silkscreen clearly marks B,C and E holes.

R20/R22 need to be 1W
R18 : 2W
Q1/Q2 most likely can be replaced by KSA992/KSC1845
C5/C6 - small ceramic X7R 50V
C7 bipolar, 6V3 or 10V
C9/C12: 50pF - most likely this value can be lower (E.g. 33pF), I was just happy enough with oscilloscope results at 50pF, so didn't want to bother changing it to try lower values..
If using different latfets, gate stoppers R15/R16 may need to be re-adjusted depending on input capacitance of fets.
C1 in the original Philips amp was 100uF. 10uF is enough, but in my build I did use 100uF polar.
If C1 is polar, minus should be at the op-amp side.
When testing, set to Idle_Pot initially to minimum value to minimize idle current.
 

Attachments

  • philips578_drk.zip
    31.1 KB · Views: 267
Last edited:
Member
Joined 2014
Paid Member
Please explain your choice of R3, 4, 6 and 8, thanks.

All starting values for the input/vas stage were preserved from original Philips design.

Then tweaked for different rail voltage (original was running at 78V, and 4 output pairs of N bjt devices).

Then, tweaked for optimal Thd, FFT profile, square waves ringing and clipping behavior.

R4 (along with R6) defines gain - so it depends on your input voltage, and expected power you want to squeeze from this amp.
Original value was 220 Ohm, lowered to 150 to get around 100W of honest power from 1.2V input signal.

R3 was lowered from 2k to 1k to lower distortions, and C15 (missing in the original) was added to help handling square waves response.

R6, R8 value was preserved from original Philips design.
R8 values don't affect any measurements in any visible way..
 
Last edited: