Hi,
I removed the original C1 and C2
I added:
R37 C1 C2 C14. I changed C6 from value and I replaced the CF output stage with an EF stage, because that one is more stable. The EF stage is also a standard Slone design.
Remco
I removed the original C1 and C2
I added:
R37 C1 C2 C14. I changed C6 from value and I replaced the CF output stage with an EF stage, because that one is more stable. The EF stage is also a standard Slone design.
Remco
Ok. I don't have the original schematic but I gether you are trying to reduce RF noise...is that right? Did you have some problem with this or are you taking precautions?
It is hard to say why you are seeing high THD. I just noticed C2 in particular and this rang an alarm bell for me. It may or may not be having an affect on distortion. I always use an input filter like R37/C1 so that should be fine. However, putting a cap to ground on the inverting input can cause instability. I just wanted to point this out in case you have not simulated the circuit. I'd also check the value of C6 (it is higher than normal with such a low R12). I cannot say the exact effect of C14 but again it is quite huge in value and will affect the open loop transfer function significantly.
Again, I don't know what is causing the THD for sure. Just pointing out some potential culprits that you have probably already considered and eliminated.
BAM
It is hard to say why you are seeing high THD. I just noticed C2 in particular and this rang an alarm bell for me. It may or may not be having an affect on distortion. I always use an input filter like R37/C1 so that should be fine. However, putting a cap to ground on the inverting input can cause instability. I just wanted to point this out in case you have not simulated the circuit. I'd also check the value of C6 (it is higher than normal with such a low R12). I cannot say the exact effect of C14 but again it is quite huge in value and will affect the open loop transfer function significantly.
Again, I don't know what is causing the THD for sure. Just pointing out some potential culprits that you have probably already considered and eliminated.
BAM
I made RF problems, because that is the assignment for my traineeship 🙂.
I didn't simulate this circuit, but I followed the guidlines presented by J. Goedbloed in 'Increasing RF immunity of audio amplifiers with negative feedback'
These guidlines give the presented values. They indeed modify the transfer function, because they severly reduce the bandwidth (50kHz in this case). C14 is needed to keep the circuit stable with R37/C1 added.
The problem is that the original circuit has similar THD specs (in my case), so the modification shouldn't be the main cause for it.
I didn't simulate this circuit, but I followed the guidlines presented by J. Goedbloed in 'Increasing RF immunity of audio amplifiers with negative feedback'
These guidlines give the presented values. They indeed modify the transfer function, because they severly reduce the bandwidth (50kHz in this case). C14 is needed to keep the circuit stable with R37/C1 added.
The problem is that the original circuit has similar THD specs (in my case), so the modification shouldn't be the main cause for it.
Hi Thoru, i follow traderbams remarks, C2 also rings alarm bells... C14 can help, but seems way too large to me with 2.2nf, it's likely this one causing the increase of 3rd harmonics with frequency.
But your main distortions are likely to come from the supplylines, i have following recommendations:
1, instead/parallel to C11/12 place a elyt cap 220uf-1000uf
2, place 47 to 150ohms into the supplylines between driverstage and vas, r13 will feed a lot of supply ripple into the vas.
3, i am missing an inputcap, never drive an asymetric input-ltp without.
4, greatly reduce r23/24, maybe 33ohms ?
5, C6 is way too big ?
6, skip r21 (short)
7, r4 should go to the other side of r37, this way you create DC-offset and unbalance.
Mike
But your main distortions are likely to come from the supplylines, i have following recommendations:
1, instead/parallel to C11/12 place a elyt cap 220uf-1000uf
2, place 47 to 150ohms into the supplylines between driverstage and vas, r13 will feed a lot of supply ripple into the vas.
3, i am missing an inputcap, never drive an asymetric input-ltp without.
4, greatly reduce r23/24, maybe 33ohms ?
5, C6 is way too big ?
6, skip r21 (short)
7, r4 should go to the other side of r37, this way you create DC-offset and unbalance.
Mike
Hmmm, I'll re-check my calculations then, especially C2.
I've a few question about your remarks:
1. Aren't C3/C4 enough buffering?
2. This one is very difficult. I'm going to design a new PCB and I'll include possibilities for such a resistor then.
3. Why? Doesn't the input pair settle at GND level? There shouldn't be a problem then?
4. Ok, perhpas I should remove them altogether? They don't limit the VA driving current enough anyway, because my complete VA stage burned out on OPS failure. Why should this improve PSRR?
5.C6 and R10 make up an 50kHz LP filter. Should I change the R10/R12 ratio in order to keep C6 small, while maintaining the bandwidth and amplification?
6. Slone explains that R21 is needed to actually improve PSRR. Should I really short that one?
7. I'll fix that.
Remco
I've a few question about your remarks:
1. Aren't C3/C4 enough buffering?
2. This one is very difficult. I'm going to design a new PCB and I'll include possibilities for such a resistor then.
3. Why? Doesn't the input pair settle at GND level? There shouldn't be a problem then?
4. Ok, perhpas I should remove them altogether? They don't limit the VA driving current enough anyway, because my complete VA stage burned out on OPS failure. Why should this improve PSRR?
5.C6 and R10 make up an 50kHz LP filter. Should I change the R10/R12 ratio in order to keep C6 small, while maintaining the bandwidth and amplification?
6. Slone explains that R21 is needed to actually improve PSRR. Should I really short that one?
7. I'll fix that.
Remco
Okay, some answers:
1: No, because they get seperated by the resistor in the supply rails, you need caps for the frontend behind the resistor and in front of the resistor for the outputstage.
2: My experience with that change is that it dramatically improves sonics.
3: That's a big NO! These inputs only settle to GND with symetrical perfectly matched input or if using FETs for input. The basecurrent from the inputdevices (here ~7uA) gives through the 10k to gnd an inputoffset of ~70mv.
4: That's also a no, basestoppers help against local oscillation, but keep them small. That's not about PSRR, it's about 3rd harmonics.
5: You must be careful with filters inside feedbackloop, they might mess up stability because of wild phasehifts or increased feedback for higher freqs.
6: That's only a suggestion, r21 does not help with stability or distortions.
7: Also a suggestion, but balanced operation of input-LTP is quite important. It's hard to get balanced operation if you have different inputoffsets on both sides of the input-ltp. This only works if you have an input cap.
Mike
1: No, because they get seperated by the resistor in the supply rails, you need caps for the frontend behind the resistor and in front of the resistor for the outputstage.
2: My experience with that change is that it dramatically improves sonics.
3: That's a big NO! These inputs only settle to GND with symetrical perfectly matched input or if using FETs for input. The basecurrent from the inputdevices (here ~7uA) gives through the 10k to gnd an inputoffset of ~70mv.
4: That's also a no, basestoppers help against local oscillation, but keep them small. That's not about PSRR, it's about 3rd harmonics.
5: You must be careful with filters inside feedbackloop, they might mess up stability because of wild phasehifts or increased feedback for higher freqs.
6: That's only a suggestion, r21 does not help with stability or distortions.
7: Also a suggestion, but balanced operation of input-LTP is quite important. It's hard to get balanced operation if you have different inputoffsets on both sides of the input-ltp. This only works if you have an input cap.
Mike
I'd say Mike's advice is pretty good. There is a reason for R21 in these sort of circuits for reducing distortion, however, the way this has been implemented is a little unusual. It's not primarily to do with PSRR.
This won't be the main reason for your THD issues...it is more of a fine tuning thing.
If you want to fine tune at some stage then you'll want C8 connected across the CE of Q7 and R21 wants to be nearer to 30 ohms (if I read the schematic correctly and the quiescent current through Q8 is about 7mA).
BAM
This won't be the main reason for your THD issues...it is more of a fine tuning thing.
If you want to fine tune at some stage then you'll want C8 connected across the CE of Q7 and R21 wants to be nearer to 30 ohms (if I read the schematic correctly and the quiescent current through Q8 is about 7mA).
BAM
Made some modifications.
Replaced R4 and added input caps.
This doesn't help. Actually when going from 2kHz to 4kHz there is indeed a 6dB degradation in THD.
Since the extra filter capacitors are directly related to the extra resistor in de power supply lines, I'll add these later on the new PCB.
Since it seems that this has indeed a lot to do with the half-sine waves on the power lines, I'm going to re-layout my ppower supply.
After that I'll try improvements with R34/24 and R21. I think I won't be able to measure any imporvements there in the current situation.
BAM, could you explain why you think that this implementation is a bit unusual and how you come to the new value? Why is it better to have C8 over CE of Q7 instead of it's current position?
Remco
Replaced R4 and added input caps.
This doesn't help. Actually when going from 2kHz to 4kHz there is indeed a 6dB degradation in THD.
Since the extra filter capacitors are directly related to the extra resistor in de power supply lines, I'll add these later on the new PCB.
Since it seems that this has indeed a lot to do with the half-sine waves on the power lines, I'm going to re-layout my ppower supply.
After that I'll try improvements with R34/24 and R21. I think I won't be able to measure any imporvements there in the current situation.
BAM, could you explain why you think that this implementation is a bit unusual and how you come to the new value? Why is it better to have C8 over CE of Q7 instead of it's current position?
Remco
Slone typically set the input impedance at 10-12k ohms. There is nothing wrong with this but when doing some similar testing with a Turtle Beeach card I found that changing the bamplifiers inpur impedance to approx 25-30k ohms made a dramatic improvement. I could get the same effect just by inserting a fairly high value series resistor between the test signal and the AUT. I had a similar experience with an Audigy card.
I strongly suspect that much/some of the THD that is being seen in these cases originates in the sound card. They may not be able to drive a low (under 25k) input impedance with out significant distortion on their part.
I'm presently working on a 7-channel amp to be driven by an M-Audio card and have decided to set the input impedance at the traditional 47k just for that reason. No point in bulding a nice amp and then have your signal source go blah.
I strongly suspect that much/some of the THD that is being seen in these cases originates in the sound card. They may not be able to drive a low (under 25k) input impedance with out significant distortion on their part.
I'm presently working on a 7-channel amp to be driven by an M-Audio card and have decided to set the input impedance at the traditional 47k just for that reason. No point in bulding a nice amp and then have your signal source go blah.
Sure. BTW the alternative arrangement that I describe is the same one that Naim uses in its NAP250 and it's fiendishly clever.
The Q7 circuit is a feedback system in it's own right. By careful choice of R21 the sensitivity of Q7 Vce to Q8 Ic can be greatly reduced. And this is what one wants if the voltage between the output device inputs is to be held steady. The value of R21 depends upon the size of R16, R17 and the quiescent current of Q8 and the choice of transistor for Q7. To determine the optimal value you can either do a small-signal analysis on paper of the Q7 circuit or, more easily, simulate it.
Given that the "voltage output" of the system is across Q7 CE, this is a natural place to put C8, to reduce the output impedance at HF and to improve the stability of the Q7 system.
The arrangement in the schematic you posted is different from this and may be so for good reasons which i am unaware of. The difference is that because Q7 Vce is not decoupled by C8 the output devices "see" an output impedance of 15 ohms (R21). I estimate that value of R21 is not high enough to minimize the variation of Q7 Vce with Q8 Ic.
Finished the new power supply: no improvements on THD figure.
Measured the soundcard: Behaves very well under loading with 10kOhm resistor.
So it must be something in the amplifier itself. Design off the new PCB has started. I really hope that that modification is going to improve something, so that I at least have a starting point for further improvements....
Measured the soundcard: Behaves very well under loading with 10kOhm resistor.
So it must be something in the amplifier itself. Design off the new PCB has started. I really hope that that modification is going to improve something, so that I at least have a starting point for further improvements....
- Status
- Not open for further replies.