F5c (cascode) cviller boards 40 V rails build

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Power-up with a 100 watt light bulb in series went fine. Adjustments of P1 and P2 were able to increase bias and vary speaker offset. P1 and P2 set back to minimal resistance and I powered up without the light bulb. Had to use a 7 to 8 amp slow-blow fuse.

Rail voltages before bringing up bias current were 41.2 V on both sides (3rd photo).

I began with a 1 A bias, keeping offset close to zero, put the cover and front in place, and let the amplifier cook, intermittently readjusting to 1A (0.5 V across the 0.5 ohm source resistors). After a few hours of different biases, I found that 0.8 A kept the heat sinks tolerable to hand placement for 15 to 20 seconds or longer. I decided to see how various measurements, then, would look at a bias of 0.8 A, though a somewhat higher bias would be tolerable. I let the amp sit for a couple hours with the top and front in place at 0.8 A bias before taking the following measurements.

Measurements at 0.8 A bias, inputs shorted, no load on output:

Rail voltages 38.7 V (4th photo).
Ripple voltage on the rails using the scope was 38 mV.
At the wall, the amplifier drew 3.7A and 315 W.
Voltage to the base of Q101 and Q102 (cascoding transistors) was 15.1 V.


Temperatures were measured with meters (not with an IR thermometer, which can read all over the place).

Still at 0.8 A bias, inputs shorted, no load on output:

MOSFET temperature (measured on the device) averaged 69.4 C ( 157 F).
MOSFET source resistors were also 69.4 C (157 F).
Heatsink temperature in proximity to MOSFETs was 60.5 C (141 F).
Heatsink temperature on top surface was about 50 C (122 F).
JFET temperatures averaged about 57.2 C (135 F).

I’ll post distortion measurements and adjustment of P3 next.
 

Attachments

  • Cal#1.jpg
    Cal#1.jpg
    587 KB · Views: 533
  • Cal#2.jpg
    Cal#2.jpg
    354.2 KB · Views: 512
  • Cal#3.jpg
    Cal#3.jpg
    443.1 KB · Views: 497
  • Cal #4.jpg
    Cal #4.jpg
    465.3 KB · Views: 479
I would and do leave the output transistor end/edge of the PCB "floating".
The far edge of the PCB generally needs support. This prevents excessive flexing of the output transistor leads. Excessive flexing can fatigue the copper and eventually lead to cracking of the hardened copper.
Long term low level heating of the lead outs may anneal the copper, but I doubt the temperatures get high enough to anneal fully.
 
flg and Andrew. Thanks for your comments. I can easily remove the posts on the MOSFET sides of the boards as a precaution. As you know, it will take about 1 minute to do so.

At these temps, I don' know that this is of a concern, and I have additional temperature measurements taken at high amplifier outputs I will post later today that were taken last night, but I also certainly acknowledge that the the posts do not provide significant support at these locations given four MOSFETs bolted to the heatsinks that, each, are soldered in place on the board with 3 leads.

ZM or anyone else, do you have strong thoughts on this matter? Thanks, everyone, for their helpful advice and comments.
 
Distortion measurements and adjustments with amp biased at 0.8 A.

Initial measurements before connecting speakers:

1. 8 ohm, 100 W power resistor used to provide 8 ohm load on speaker terminals.

2. Signal generator for 1 kHz signal input.

3. Channel 1 of scope monitoring output at speaker terminals.

4. Output of signal generator (input to amp) adjusted for a given amplifier board output of about 2.83 V RMS (measured on scope). This would be 1 W RMS output (about 1.12 average W output I think).

The second photo shows a clean sine wave. I commonly forget to freeze the screen before grabbing a photo, so sometimes the waves look a bit saw-toothy. No ringing, though I didn’t really expect such. Square wave looked fine when I measured it later on channel 2. Both sides of the amplifier looked the same.

I plugged in the ipod and listened to music for about an hour. Sounded wonderful on Spendor S8e speakers.
 

Attachments

  • Cal#5.jpg
    Cal#5.jpg
    316.6 KB · Views: 442
  • Cal#6.jpg
    Cal#6.jpg
    342.6 KB · Views: 195
  • Cal#7.jpg
    Cal#7.jpg
    345.3 KB · Views: 193
Distortion measurements and adjustments with amp biased at 0.8 A.

Setup for each board:

1. HP339A distortion analyzer used for generate 1 kHz sine wave input into amplifier.

2. 8 ohm, 100 W power resistor at speaker terminals.

3. Input to the 339A from speaker terminals.

4. Channel 1 of scope monitoring amp output at speaker terminals.

5. Output of 339A (sine) adjusted to create amp output of about 2.83 V RMS for 1 W RMS (1.12 average W) output.

6. Channel 2 of scope monitored distortion residual output of the 339A.

7. Math channel of scope for FFT on the residual distortion waveform.

Channel 2 is the residual signal after the 339A removed the 1 kHz primary signal with a notch filter, and represents, of course, the residual distortion.

The FFT, again, is of the residual distortion, not of the entire amp output. I only used it to look at the relative strength (height) of each harmonic.

The first photo shows an overview of the setup. The 339A is on the bottom. On the active scope, channel 1 on top (yellow) is amp output. Channel two (light blue) on bottom is the distortion residual from the 339A. And the darker blue trace (which is purple in real life) in the center is the FFT.

I’m now going to try to cut and past photos together in single jpegs to illustrate measurements. Jim (6L6) previously posted a nice YouTube link of watching the scope in real time as adjustments are made, though he wasn’t able to show the THD measurements on the meter at the same time. But his video helps put this all together and worth watching if you are interested.

https://www.youtube.com/watch?v=CQG2j9fkr9E

Back to this amp.

First, the LEFT channel.

Baseline THD was 0.009% (note the 0.03% scale). While the 2nd harmonic was dominant, there was a relatively strong 3rd harmonic as well. I then adjusted P3 (circled in red in the photo) to lower THD to its minimal value. Turning P3 after reaching minimum, of course, caused THD to begin to rise again, and an adjustment to minimum value was quite easy and quick. Sorry, again, for not freezing the scope tracing before photography.
 

Attachments

  • Cal#8.jpg
    Cal#8.jpg
    550.7 KB · Views: 536
  • Cal#9.jpg
    Cal#9.jpg
    236.7 KB · Views: 515
  • Cal#10.jpg
    Cal#10.jpg
    91 KB · Views: 517
  • Cal#11.jpg
    Cal#11.jpg
    255.3 KB · Views: 524
Distortion measurements and adjustments with amp biased at 0.8 A.

Now the RIGHT channel.

Baseline THD was 0.004%, which is near the lower limits of accurate measurement by the 339A. And on this side, the 3rd harmonic was quite dominant. After adjustment of P3, the meter moved to read about 0.001%, which is below what the 339A can read accurately, as best my understanding. This is probably best interpreted as being < about 0.003%. Interestingly, while adjusting P3, one could watch the 3rd harmonic almost disappear while the 2nd harmonic become dominant. In the second photo, taken after adjustment of P3, the scale for the FFT is one half that in the first photo, and the scale of the 339A is set on 0.01% (was 0.03% in previous photos).

So….then I listened to the same music as before P3 adjustments. I was extremely skeptical I would be able to detect a difference and wasn’t expecting to notice anything at all. I was quite taken back. I definitely thought the sound was better. This is completely unblinded, but I firmly believe there was a noticeable difference, as commented on by NP. Minimizing the 3rd harmonic and whatever else happened during P3 adjustments made a definite improvement.

In this build, efforts to minimize THD also resulted in the greatest dominance of the 2nd over the 3rd harmonic. That is, both results occurred at the same setting of P3. In 6L6’s case on YouTube, he found that he had to accept slightly higher than minimal THD to minimize the 3rd harmonic, but this wasn’t the case for this amp.

I am skeptical of being able to tell differences listening to an amplifier built with 4-pole “audio” caps in the power supply (which I have used) compared to 2-pole caps of the same capacitance, ripple current and ESR. I don’t believe that brands of resistors or types of resistors make a difference in most locations of an audio amp, despite testimonials found across the internet. But I was actually impressed by the improvement with adjustments to P3. I really wasn’t expecting this.

I’ll continue to post measurements tomorrow, but must work on an unrelated paper tonight.
 

Attachments

  • Cal#12.jpg
    Cal#12.jpg
    243.9 KB · Views: 509
  • Cal#13.jpg
    Cal#13.jpg
    320.6 KB · Views: 160
Distortion measurements continued

Continuing with the same setup as above.

I incrementally increased input signal into the amplifier and measured THD at different amplifier outputs. I took measurements on the left side of the amp, which had the higher THD after adjusting P3 (.006%). As a reminder, all measurements with 8 ohm load.

I placed the results in the first figure, as I could not get a table to survive formatting, though I'm sure there is a way to do it.

With a bias a 0.8 A and 4 devices per channel, one would anticipate the amp would remain in class A until 3.2A. 3.2A into 8 ohms would be a peak of 81.92 W, which is about 41 average W per channel. Please correct me if my calculations are incorrect.

Watching wave forms on the scope, clipping was first barely evident on the top of the sine wave at around 23 to 24 V RMS. I grabbed a photo (again, forgetting to freeze the scope) well into clipping at 24.8 V RMS (10X probe) to demonstrate the sudden appearance of almost countless harmonics around this threshold. 24 V RMS would be about 72 W RMS, and 23 V RMS would be about 66 W RMS.
 

Attachments

  • Cal#15.jpg
    Cal#15.jpg
    77.1 KB · Views: 211
  • Cal#14.jpg
    Cal#14.jpg
    527.6 KB · Views: 220
3.2Apk into 8r0 is the same as 40W into 8r0.
that bit is correct.

But I don't follow how you arrived at the maximum ClassA current.

4 pairs of devices biased to 0.8A results in 1.6Apk of ClassA current for a Push-Pull output stage.
To get 3.2Apk of ClassA current, the output stage would need to be biased to 1.6A. That would would be 400mA of output bias current through each pair of output devices.
 
Last edited:
With a bias a 0.8 A and 4 devices per channel,
4 pairs of devices biased to 0.8A results in 1.6Apk of ClassA current
where did I get that from?
Andrew. it's 2 pairs. at 0.8A pr device.
is this me having another bad hair day?

4devices = 2pair
2pr @ 800mA per pr gives output bias = 1.6A
Maximum ClassA current for a push-pull amplifier is double the bias current and equals 3.2A.
Thus the power calculation is correct.
one would anticipate the amp would remain in class A until 3.2A. 3.2A into 8 ohms .......... which is about 41 average W per channel. Please correct me if my calculations are incorrect.
P = I²R = Ipk²R/2 for an unclipped sinewave signal.

Do not get into using peak power. Reserve that for the very few occasions where you need to check peak dissipations in a junction/component.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.