Take a look at the phase shift at 20K or however high a frequency you can hear. I have been able to detect 0.8 degree of phase shift with certain circuits and electrostatic headphones and would aim for over 200KHz as the corner of a single pole filter. This may indeed not provide sufficient filtering and additional steps may be called for. I offer this only as a potential explanation of an audible effect of a filter that on first inspection would seem that it should be inaudible.
I hear
ya on the phase shift as far as possible audible effects go, But the plots he posted show a low pass filter with a corner frequency well above 500 kHz reducing distortion components in the audio band.
I would like to hear possible explanations why such a high frequency filter (>500 kHz), which should not effect frequencies anywhere near the audible bandwidth appears to reduce distortion components from 10 kHz to 25 kHz? I must be missing something, because this makes zero sense to me.
ya on the phase shift as far as possible audible effects go, But the plots he posted show a low pass filter with a corner frequency well above 500 kHz reducing distortion components in the audio band.
I would like to hear possible explanations why such a high frequency filter (>500 kHz), which should not effect frequencies anywhere near the audible bandwidth appears to reduce distortion components from 10 kHz to 25 kHz? I must be missing something, because this makes zero sense to me.
Interestingly modelling the Legato 2 circuit in SPICE with and without the 15nF caps also showed an effect on reduction of harmonics at much lower frequencies than the filter knee. Without the caps there was a much increased presence of a first order ringing at twice the frequency being pumped in. Off the top of my head the reduction was in the order of 60dB. I do prefer the Legato without the caps though, even if I too detect a very slight increase in 'glassiness'.
Had a thought about my measurements in the traffic on my way to work - I'm measuring on single ended output whilst the filtering is before the balanced to SE opamp. Could some high freq rubbish that slips through when not filtering upset the opamp and cause oscillation/distortion?
Anyway, I'll do some more measurements using my 60MHz DSO and FFT and correlate my results.
Cornelius
Anyway, I'll do some more measurements using my 60MHz DSO and FFT and correlate my results.
Cornelius
What you are seeing is the limitation of the test setup.
What I am extremely pleased with is the extremely low measured harmonics! All of the stuff I see above what would be harmonics looks like noise.
It also could be a the FFT is not measuring the balanced signal? You would get much better noise immunity then.
Last edited:
Yes those caps are important to getting the best THD.
Notice that both the 2 and 3rd harmonics are basically nil even when using a low value cap. Even 1nf will work fine.
If you look at the noise bands(they are not related to the fundamental at all) they are consistent in all three shots starting especially at about 6.8Khz and there are several harmonics of that. This is pretty normal looking noise in sound card measurements. You can usually tell the FFT to ignore them. As you said this could also be re-sampling hash. Likely a combination.
As for the Bal/SE opamp if you must run without the filter caps you will definitely want to add something like 20pf compensation on both nodes. This will make the opamp a bit happier
But really to get the best picture of the I/V stage itself you should measure the balanced output. If your test setup can sum two signals it should be pretty easy.
I am very please looking at your FFT in regards to actual distortion on the fundamental. That is pretty amazingly low.
Cheers!
Russ
Notice that both the 2 and 3rd harmonics are basically nil even when using a low value cap. Even 1nf will work fine.
If you look at the noise bands(they are not related to the fundamental at all) they are consistent in all three shots starting especially at about 6.8Khz and there are several harmonics of that. This is pretty normal looking noise in sound card measurements. You can usually tell the FFT to ignore them. As you said this could also be re-sampling hash. Likely a combination.
As for the Bal/SE opamp if you must run without the filter caps you will definitely want to add something like 20pf compensation on both nodes. This will make the opamp a bit happier
But really to get the best picture of the I/V stage itself you should measure the balanced output. If your test setup can sum two signals it should be pretty easy.
I am very please looking at your FFT in regards to actual distortion on the fundamental. That is pretty amazingly low.
Cheers!
Russ
Last edited:
I already said that the filter is first order. It very easy to tell that from the circuit too.
65khz is not low at all for a low pass. In fact it is higher than the filter ESS uses on their demo board.
The short answer to your question is that the noise that needs filtering is well above 200khz. The fact that it is so high freq makes it hard to detect unless you have a very very fast FFT to use.
You could actually knock a lot of it out with just a 1nf cap, the trade off is how far down it will be. You could also add another RC and increase the slope at the expense of higher output impedance. That would be fine too.
You have so many options it must be hard to choose one.
Last edited:
I am always happy to post with a morning cup of coffee.
Brian is likely busy putting in his much needed boiler.
If you are running balanced only. Then yes you have a lot of freedom to experiment. the 25V caps on the legato can be replaced or even omitted.
What I have found is that with the CFP there is a point of diminishing returns and that happens to be about 15V.
That said, if you want you can actually short the CFP resistor and omit the P channel devices and use a TO-220 N channel FET. If you do that you will have a classic "D1" type common gate stage. Then increasing the positive rail will help a lot.
You can always leave the negative rail relatively low. Its up to you. It's purpose in the cct is just to provide current for the uppper part of the circuit.
Here is how I am running my Legato (2.0.0 PCB with 3 FETs per side per half):
I/V resistor 105R(I don't need 2VRMs, 1.2 or so is fine).
Current source resistors are 475R (2 x in parallel)
Basically you can use the negative rail to tweak the amount of current and thus the bias point for the top half.
I aim to bias the output at about (VCC/2)-1V So in this case I was aiming for about 6.5V. This in my experience is a sweet spot.
Positive rail 15V.
Negative rail 13V.
It sounds delightful.
Now if you are going to experiment, be sure you understand the implications. Calculate the power dissipated and the supply requirements carefully. Your on your own here, but it is certainly manageable. I will help out if I can.
Cheers!
Russ
Hi Russ,
Are you saying that it's okay to use Legato without the 100uf/25v caps? Any caveats/possible benefits we should be aware of?
Thanks,
Alex
Omit the caps in Legato if and only if you know exactly what you are doing.
Basically the decision will have to be based on what is coming after the Legato.
Hi Russ,
can I omit BP caps, if output is going into transformers? If I understood correctly, those caps needed to eliminate the DC offset voltage, and the trafos will take care of that?
Thanks,
Fedor
No, you generally don't want DC into a transformer. If you get it down to a couple mv then it will probably be ok, otherwise, no.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.