One of the Top Solid-State CFA amp design

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Right. So what your seeing at the output of your amplifier might not be the amplifier. You have to isolate from what you see on the scope.

Is your amplifier in a case or just trapped to a heat sink sitting on your bench.

Try to find the source of the RF and shut it down.

I know of one DIYer who spent two days trying to find an oscillation in his amplifier only to find it was ingress. His amp is actually well behaved.

I have no case yet, just amp pcb trapped to a heat sink.

The most probable sources of RF ingress is PC and wi-fi router. I'll try to measure when they are off.
 
double layer pcb

Well, in the good old days we easily got away with this. Now days with modern components you need take care of the return path much more diligently than in the past. It is advisable to consider a solid ground plane between your routing layers. Also it is advisable to take great care of the signals need a dedicated ground trace to the common reference point.

To sum it up as previously stated, more sophisticated layout following basic high speed guide lines is needed.

Also, if you have to put more than 10pF capacitors everywhere to tame things down you have major problems with the circuit.
 
I thought the best solution is star grounding technic, not a ground plane.
There is a dedicated ground trace for input part of the amp pcb which than goes to common power GND refference by wire (not shown on layout).
I redesinged the layout a bit. Will this variant be better than previous? From decoupling caps there will be separate wires to a common power GND plane
 

Attachments

  • SSA.jpg
    SSA.jpg
    649.2 KB · Views: 902
I thought the best solution is star grounding technic, not a ground plane.
There is a dedicated ground trace for input part of the amp pcb which than goes to common power GND refference by wire (not shown on layout).
I redesinged the layout a bit. Will this variant be better than previous? From decoupling caps there will be separate wires to a common power GND plane

You do want a ground plane, you do want to connect this ground plane to your common reference point. You do want to route ground signals you care for as dedicated traces on your routing layer to the common reference point.

I am afraid your variant layout without a plane layer between the routing layers will exhibit the same issue you are facing now.

You should familiarize with the concept of ground planes and high speed routing techniques. This will help you understand that what I am trying to tell you.
You need to develop a habit of thinking about the return path of your signals, just connecting something to 'ground' and then forget about it is not helping. This is not DC but AC and at slightly elevated AC frequency a new set of layout routing rules apply than for plain DC stuff.
 
If it's RF interference, it will be "talking". Set the time base to audio/video frequency range and see if there is any modulation present. Zoomed in to see the carrier frequency this is often missed.

Once that is ruled out, you need to determine whether it's local or global (loop). The fixes are different.
 
RF ingress has absolutely nothing to do with your PCB design or layout.
And adding large planes would only make it worse for RF ingress problems because large planes act as antennas and do nothing for the problem.

Your problem, assuming it is RF and not local oscillation is solely because the circuit is not in an grounded enclosure. If the source is local then turning it off will help but if it is broadband then there is no hope.

It's a simple solution that doesn't require complexity.
 
I tried with all home appliances off in the room, but the same result. Perhaps the scope itself produces RF noise or just FM radio signal in an air.
RF ingress has absolutely nothing to do with your PCB design or layout.
And adding large planes would only make it worse for RF ingress problems because large planes act as antennas and do nothing for the problem.

Your problem, assuming it is RF and not local oscillation is solely because the circuit is not in an grounded enclosure. If the source is local then turning it off will help but if it is broadband then there is no hope.

It's a simple solution that doesn't require complexity.
Will try to find some steel or alu case tomorrow
 

Attachments

  • 20171126_225852[1].jpg
    20171126_225852[1].jpg
    248.9 KB · Views: 734
Last edited:
Earth ground is for safety only. I'm also not convinced the enclosure needs to be connected to the circuit, a faraday cage for instance

With the ultra low distortion oscillator I designed and build it made a big difference grounding the enclosure to the circuit ground. It wasn't enough just to have a solid ground plan on the PCB.

These days audio frequency is my hobby. My day job is RF.

I agree with the Faraday cage. I'm really just trying to make a strong point.

A deep turkey pan would be good enough to conduct a test to see if this is ingress or an unstable circuit.
 
I have no Earth GND in my wall socket (have to rebuilt a mains wiring in my house). I can only connect to circuit GND currently.
This is fine. It will do more than an earth ground.

We are just testing a theory. You want to see if the apparent oscillation amplitude decreases with shielding. If it does then it is likely ingress. But if there is no difference then you likely have a local oscillation in your circuit.

Who knows you may have both.
 
Last edited:
When tried to check the other ch. out (with different output bias method - diodes string) the scope showed little less oscillation if I'm not wrong. However this test is done with different position of slope level on a time base unit of the scope. On the first photo the triggering signal isn't going through time base unit, on second photo with setting a slope level on some single point, the signal goes through time base to the scope (it's seen by a light of small green led in the midle).

Found an alu case. Will I have to put in there the amp on heatsink and probe with only wires going out from a cover?


The last two photos are inside a case.
 

Attachments

  • 20171126_234922[1].jpg
    20171126_234922[1].jpg
    713.4 KB · Views: 700
  • 20171126_235006[1].jpg
    20171126_235006[1].jpg
    707.3 KB · Views: 669
  • 20171127_002424[1].jpg
    20171127_002424[1].jpg
    556.3 KB · Views: 628
  • 20171127_002506[1].jpg
    20171127_002506[1].jpg
    542.3 KB · Views: 613
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.