Panson -- did you ever talk to Mark about the caps? When they did the first application note on the LM4702 he had spent a lot of time on values and placement.
The mylar or polypropylene bypass caps -- 3 of them -- one across the rails where the power comes in, the others are as immediately adjacent to the chip as possible. I used 0.47u and 0.68uF pp. They are labeled as Cs1, Cs2 and Cs5 in AN1490.
data for the new universal power board
I tested the new universal power board with two-pair MJL21194/93. The supply is +/- 39 V, no load. Bias generation/tracking is my preferred four-diode stack. Each diode is attached to the top of each output device. LME49811 with 22 pF comp. cap is employed.
THD at 20 kHz is below 0.01 % for both 8 Ohm and 4 Ohm. Very promising result!
I tested the new universal power board with two-pair MJL21194/93. The supply is +/- 39 V, no load. Bias generation/tracking is my preferred four-diode stack. Each diode is attached to the top of each output device. LME49811 with 22 pF comp. cap is employed.
THD at 20 kHz is below 0.01 % for both 8 Ohm and 4 Ohm. Very promising result!
Attachments
So...how do the Sankens outputs perform?
Better than....21194/93? I used to work or ON Semi and have a few ON parts but like the Sanken 2 screw mounts. Plots look good for the 21194/93.
Better than....21194/93? I used to work or ON Semi and have a few ON parts but like the Sanken 2 screw mounts. Plots look good for the 21194/93.
Yes, I think two screws making better thermal contact.
I am going to measure the board with Sanken. Here is clipping performance for the ON parts. We compare the clipping point for power supply using one and two transformers (two in parallel). We get more from the amp by using a stronger power supply.
Attachments
Last edited:
Sanken output devices Super!!!
I tested the board with two-pair Sanken 2SA1295 and 2SC3264. They exhibit much better THD performance in high freq region than ON Semi 21194/93 (same bias level for them). The later pair performs better in the freq region THD. The Sanken pair got much higher fT compared to the ON Semi pair. Will it be the reason for lower high-freq THD? They also got better hfe vs Ic characteristic. Distortion rise for heavier load, 4 Ohm, is less than that of 21194/93.
From the FFT plots, we can see that THD content is largely 2rd harmonic. Nice feature?!
I tested the board with two-pair Sanken 2SA1295 and 2SC3264. They exhibit much better THD performance in high freq region than ON Semi 21194/93 (same bias level for them). The later pair performs better in the freq region THD. The Sanken pair got much higher fT compared to the ON Semi pair. Will it be the reason for lower high-freq THD? They also got better hfe vs Ic characteristic. Distortion rise for heavier load, 4 Ohm, is less than that of 21194/93.
From the FFT plots, we can see that THD content is largely 2rd harmonic. Nice feature?!
Attachments
Last edited:
Panson,
Nice work! That's a very interesting result for the Sanken vs On Semi will you also be trying the mjl1302/3281 on semi devices?
Ken
More data for ON Semi 21194/93
Compared to the Sanken pair, 21194/93 has better low-freq THD performance. The plot below is 1 kHz THD vs level for 2.7, 4 and 8 Ohm. The clipping levels for 2.7 Ohm and 4 Ohm are more less the same which is not for the Sanken pair. The curves for Sanken in previous post are placed here for comparison.
Compared to the Sanken pair, 21194/93 has better low-freq THD performance. The plot below is 1 kHz THD vs level for 2.7, 4 and 8 Ohm. The clipping levels for 2.7 Ohm and 4 Ohm are more less the same which is not for the Sanken pair. The curves for Sanken in previous post are placed here for comparison.
Attachments
Further results are placed here http://www.diyaudio.com/forums/solid-state/170618-comparing-power-bjts-sanken-semi.html#post2250274
Last edited:
Impact of layout on performance
We all knew that layout pays vital role in amp performance. Here is an evidence. The amp under test is LME49811 with two-pair NJL3281D/1302D.
From left: 1kHz for previous version power board, 1 kHz for new version power board, 10 kHz for previous version power board, 10 kHz for new version power board
The performance for 8 Ohm is nearly the same for both boards. However, it is very clear that the new board achieves better performance (lower THD and higher clipping level) for heavier loads.
I just told you that I did not make a very good board 😱 or I am improving. 🙄 Nevertheless, this is a good example to share with you.
I am going to take trace resistance measurement. See whether I can get some useful data.
We all knew that layout pays vital role in amp performance. Here is an evidence. The amp under test is LME49811 with two-pair NJL3281D/1302D.
From left: 1kHz for previous version power board, 1 kHz for new version power board, 10 kHz for previous version power board, 10 kHz for new version power board
The performance for 8 Ohm is nearly the same for both boards. However, it is very clear that the new board achieves better performance (lower THD and higher clipping level) for heavier loads.
I just told you that I did not make a very good board 😱 or I am improving. 🙄 Nevertheless, this is a good example to share with you.
I am going to take trace resistance measurement. See whether I can get some useful data.
Attachments
Last edited:
Impact of layout on performance (cont.)
Let's take a look the FFT of THD residual of the old and new board.
Left: new board
Right: previous board
The new board has no 50 Hz related harmonics. It has lower signal related harmonics as well. On the other hand, there are 50 Hz harmonics measured from the old board. The old board exhibits higher signal related harmonics where the third harmonic dominates. Since trace resistance measurement does not favor the new board, I think that component placement makes the difference.
Let's take a look the FFT of THD residual of the old and new board.
Left: new board
Right: previous board
The new board has no 50 Hz related harmonics. It has lower signal related harmonics as well. On the other hand, there are 50 Hz harmonics measured from the old board. The old board exhibits higher signal related harmonics where the third harmonic dominates. Since trace resistance measurement does not favor the new board, I think that component placement makes the difference.
