Hifisonix kx-Amplifier

Putting my fingers on D6-D7 quiets the oscillation, so next I tried a 15pF cap parallel with C3, and that also damped it. Possibly because I was holding the cap.

Transistors:
BC546BTA
BC556BTA
KSA1881ESTU
KSC3503DSTU
KSA1220AYS
KSC2690AYSTU
NJW3281G
NJW1302G
 
Last edited:
Have you got the Zobel connected?

It you scope on tne supply rails (at the fuses) and check them as well

Zobel=RC? No. But I tried a 0,3u cap in series with the load resistor, and that did the same as just load resistor without cap.

There is a hint of the oscillation on the supply voltage, but very weak. I can see it on the scope FFT. I think that is a consequence, not the cause. I used the same supply with many amps, and never had a problem. I actually tried it with a switched supply too, and saw the same on the output of the amp.
 
However, with the load resistor (short cables) connected it should act as a HF impedance for the amp too. Ok, maybe the load resistor has some tiny inductance..

It concerns me that I have not had any problems testing several DIY amps like this before, seems like stability is on the limit if the RC on output and twisted power cables are needed to keep it quiet even with shorted input?

Will give it another try tomorrow!
 
From stability simulation I got some idea to try and play with increasing R42 too.
I tried 8k in sim which gave some peaking, but better phase margin. Oscillation seems to be in the area with less than 20dB OLG and not great phase margin. However my sim differs from yours, I get higher gain, and lower bandwith in my sim. Anyway, it gives me some idea on what I can tweak.

Maybe I'll even try removing R42 completely if nothing else helps, -going single pole, that seems to give great stability margins (with other trade-offs of course).
 
Last edited:
Thank you for the support!

I tried the supply cables first (easiest), and that did nothing, but after adding the RC on the output it was quiet.
I made some measurements, and things are looking better I think. I'm surprised to see the high emitter current spikes on the flanks of the square wave. I never measured that before, so I have no reference. There is some asymmetry on the emitter currents as seen on the sine.

Yellow trace is Vout, and blue trace is diff probe on emitter resistor(s). If you read the file name Vetoe is voltage over two emitter resistors(emitter to emitter), and VetoVout is over one emitter resistor.
 

Attachments

  • 20k sine Vout Ve1toVout .jpg
    20k sine Vout Ve1toVout .jpg
    167.3 KB · Views: 150
  • 20k sine Vout Ve2toVout .jpg
    20k sine Vout Ve2toVout .jpg
    166.1 KB · Views: 166
  • 20k square Vout VetoOut .jpg
    20k square Vout VetoOut .jpg
    150.7 KB · Views: 151
  • 20k sine Vout Vetoe .jpg
    20k sine Vout Vetoe .jpg
    144.6 KB · Views: 152
  • 20k square Vout Vetoe .jpg
    20k square Vout Vetoe .jpg
    137.5 KB · Views: 154
www.hifisonix.com
Joined 2003
Paid Member
Just so I understand

Pic 1 and pic 2 are sine wave into resistive load (they look the same to me)?
Pic 3 - I expect some current overshoot and fast rise rise times because you are charging and then discharging the Zobel which at HF is 2x the LF current. You will not ever see this with a music signal since the rise/fall times are 10us or thereabouts. I assume your Zobel is 10 ohms and 0.1 uF.
Pic 3 ?
Pic 5 and pic 3. You are seeing HF ringing on the waveform tops, again unlikely you will ever see this with a music signal, but it is a good test of the quality of your PSU, cabling and local decoupling. You have quite a bit of inductance in your power cabling and the 1000uF decoupling caps have ESR and EL (same inside your lab PSU as well BTW) that will be contributing to this. You could try soldering a 0.1 to 1uF SMD ceramic cap across the 1000uF caps under the board. With the actual PSU this problem should not be present - driving <1us ruse times at high power is a brutal test for the amp+Lab PSU combo.
 
1&2 7ohm load but voltages over different emitter resistors (upper & lower)
3 Ok, I never lookaed at this before, so I had no reference, was only a bit surprised to see such high current peaks at the transitions (I'm learning here :) )
5 I would say is the shoot-through current seen on the small blips on the blue trace.

Just some basic tests to see that it looks ok. Lab supply actually seems very stable in these tests until current limit is reached. If I lower the frequency, the current limit kicks in on the current peaks.

I'm just about to post some more measurements on the current sharing I simulated before.. just give me a few minutes.
 
www.hifisonix.com
Joined 2003
Paid Member
You should try to do these tests with the actual amp supply. The amp is much faster the lab PSU (almost always the case with any kind of amp btw).

The probable reason the lab supply goes into current limit when you lower the F is that fast high current spikes are no longer being supplied from the 1000uF decoupling caps on the amp module or the PSU output caps.
 
So, I wanted to test the simulated current sharing I posted before. Here the amp is driven harder into class B, still in low bias mode. I changed to a switched 26V supply that can deliver enough current. Load is 7ohm resistor. Input is 120Hz sine. Yellow trace is Vout, blue is the diff probe placed in different locations (see file name for picture for description).
Blue trace:
1 voltage over Q7
2 voltage over one emitter resistor (cant remember if it's upper or lower)
3 voltage over opposite emitter resistor
4 voltage over both emitter resistors (one upper and one lower)

I did not keep track of the polarity of the diff probe.. sorry..

This seems a bit similar to my simulation. I hope you can have a look at this. I played around with a 10-100u cap from D1 to Q7 collector in sim, and that seems to even it up, but I have not tried it on the bench.
There definitely seems to be something going on with the bias circuit with current in R12, and possibly due to some voltage drop over R21 too?
 

Attachments

  • 120Hz Vout Ve1to2.jpg
    120Hz Vout Ve1to2.jpg
    166.6 KB · Views: 171
  • 120Hz Vout Ve2out.jpg
    120Hz Vout Ve2out.jpg
    162 KB · Views: 177
  • 120Hz Vout Ve1out.jpg
    120Hz Vout Ve1out.jpg
    162.4 KB · Views: 178
  • 120Hz Vout Vbias.jpg
    120Hz Vout Vbias.jpg
    170.4 KB · Views: 178
Last edited:
You should try to do these tests with the actual amp supply. The amp is much faster the lab PSU (almost always the case with any kind of amp btw).

The probable reason the lab supply goes into current limit when you lower the F is that fast high current spikes are no longer being supplied from the 1000uF decoupling caps on the amp module or the PSU output caps.

Yes, thought so too, but it was ok on the high frequencies. I checked the supply voltage during the previous tests, and it was rock steady.

EDIT: Just checked the ripple under load @120Hz (as above) with the switched PSU, and it was 200mVpp
 
Last edited:
www.hifisonix.com
Joined 2003
Paid Member
I've just checked my sims. If I look at the current through just one of the 0.33 Ohms resistors, I get a clean sine wave.

If I look at the current across two of the resistors as in eg across R14 and R15, I get exactly the same current waveform you are showing.

Your amplifier is working correctly. I should have plotted the total current across two resistors in my write-up, but never thought about this so thanks for raising it. I will come back with a full explanation after I have studied it a bit.

[re your post #334 above -that's my point - you wont get current limiting at HF but with the same input level it will probably trigger at LF on your PSU. The current limit circuit in the PSU is probably also filtered so this is another factor why it does not limit at HF]
 
Last edited:
I cant check at the moment (I have all the stuff at work), I can only say total consumption was around 0,8A and fairly evenly distributed between the outputs when I checked, just can't remember the readings.

Are you running the sim at abt full power (far into class B)? Try to look at the bias spreader voltage in this situation, like pic1. From what I know it should ideally be constant (only vary slightly to regulate DC current), but here it seems to be modulated by the output current. Only slightly modulated when working in class A, but heavily modulated when going to class B. I think this means the total current from rail to tail will be 'pumping' with the varying bias voltage.

Putting the cap as I mentioned, I figured it could be 'dc coupled' to sense the DC current from the emitter resistor as I think you intended, but not modulated by AC current in the resistor (more like a normal bias spreader at AC). For some reason my sim sometimes got into slow motion depending on the cap value I entered, so I lost my patience with it. I think it got worse at LF too, so I was running the sim at 10Hz.
 

Attachments

  • cap.jpg
    cap.jpg
    171.3 KB · Views: 171
Last edited:
I'm sorry, I think I was looking at it the wrong way, the 'Vbe-loop' for the bias circuit is the lower half, so both upper and lower Re voltage drops are in the loop, right?
I'm no expert, it only behaves a bit odd. All that extra 'work' done by the outputs and odd currents circulating would probably cause some extra distortion at high output?
 
I went back to my simulation on the added cap (C17), and it did not look too good after all. The currents look good, but when I send a burst of 10 cycles at 20Hz to the amp, the cap slowly discharges during the burst, and when the burst stops, there is a dc offset at the output until the cap is charged again. :(
EDIT: I tried the same without the cap, and get a DC offset after the burst too.. maybe input cap or something doing this.


I wait for the pro to come up with an elegant solution :)
 
Last edited: