Yes, you're right. D3 isn't doing any good. I guess trying to get the same dissipation from the folded cascode transistors has a more intelectual than practical appeal. Regarding RB, I just did what "all" the otheres are doing. Using a 470k - 1M resistor in that position seem to be the standard procedure for JFET input amps. Could you please explain how to choose the right resistor. I've heard that it should be at least 10 times the pot resistance not to affect the input impedance too much.Lumba Ogir said:
By adding source resistors, the OLG is reduced and because of that the THD is increased by a couple of dBs. I just thought it was appealing to have the amp stable without a lag comp cap.The increased noise can be heard when listening through IEMs, but I find it's still low enough not to be annoying.aparatusonitus said:
In sim I can see the amp is peaking when the volume is lowered. This can be cured by using a pot with less resistance, which I don't have. Is there another way to fix this problem. Could this be why the amp is a little "edgy"?
Also, I would try something like 4k99/1~2k for feedback devider to avoid HF peaking.
I'm glad I have you people
. I twisted the amp in LTSice to get rid of the peaking both at max and at 50% volume. As a result I added a 22 pF lag comp cap as per Jonathan's suggestion and I increased the input capacitance to 66 pF. I had to keep the source resistors, and I could keep the feedback resistors unaltered. The amp sounds smoother and more neutral. The tonality's the same as in the amp without the input cascode, which I consider neutral. I'm not sure yet if cascoding the input makes it better. I have to give it some time. The amp is slower. Shown in the picture is a simulated 100 kHz sqaure.
Now I think I know what peaking does to the sound.
Next step is to try Lumba's suggestion, to lower the RB resistance.
. I twisted the amp in LTSice to get rid of the peaking both at max and at 50% volume. As a result I added a 22 pF lag comp cap as per Jonathan's suggestion and I increased the input capacitance to 66 pF. I had to keep the source resistors, and I could keep the feedback resistors unaltered. The amp sounds smoother and more neutral. The tonality's the same as in the amp without the input cascode, which I consider neutral. I'm not sure yet if cascoding the input makes it better. I have to give it some time. The amp is slower. Shown in the picture is a simulated 100 kHz sqaure.Now I think I know what peaking does to the sound.
Next step is to try Lumba's suggestion, to lower the RB resistance.
Attachments
Nelson,
your steadfastness has result in a nice circuit.
I`m sorry, again. I had a blackout on RB, disregarding the pot, of course its value is not critical. Very confusing...
After a closer look:
10k is a proper value to me, don´t see the need for going lower.
You have a number of capacitors around, which might be required for RF filtering and stability but are sonically deteriorating. If some of these are used to alter tonality, then that, in my orthodox view, should be considered as a failure.
I would keep the position of C5, but check the need for C10, C6, C9, C13.
What kind of capacitor is Csource (22uF)?
My impression is that 2SK170 can operate at very low Vds, the problem is that capacitances increase with decreasing voltages (and with increasing gain). Varying parameters are especially undesirable, therefore the goal should be to hold as many as possible constant. You would be the right man to explore the proper biasing conditions for 2SK170, which has probably not been done yet on a sonic bases.
Andrew,
I have noticed many times that you are a good observer, nothing escapes your eyes.
your steadfastness has result in a nice circuit.
Right, higher THD due to the GNF mechanism but a more favorable harmonic spectrum.
I`m sorry, again. I had a blackout on RB, disregarding the pot, of course its value is not critical. Very confusing...
After a closer look:
10k is a proper value to me, don´t see the need for going lower.
You have a number of capacitors around, which might be required for RF filtering and stability but are sonically deteriorating. If some of these are used to alter tonality, then that, in my orthodox view, should be considered as a failure.
I would keep the position of C5, but check the need for C10, C6, C9, C13.
What kind of capacitor is Csource (22uF)?
My impression is that 2SK170 can operate at very low Vds, the problem is that capacitances increase with decreasing voltages (and with increasing gain). Varying parameters are especially undesirable, therefore the goal should be to hold as many as possible constant. You would be the right man to explore the proper biasing conditions for 2SK170, which has probably not been done yet on a sonic bases.
Andrew,
I have noticed many times that you are a good observer, nothing escapes your eyes.
Well, actually it seems I can get rid of all capacitors except C5. I just took C10 for granted since most folded cascode/DB amps use it. I thought it was needed for stability. If I remove it in LTSpice, I can remove C6/9 if I increase the feedback resistors to 1k/4k7. I'm too tired to do it tonight. Hopefully tomorrow.Lumba Ogir said:
Csource is just to simulate the cap in my portable player (it's actually 47u). I would never use an input cap unless absolutely necessary.
In the beginning of this project I was advised to use small caps in C6 and C13 positions and resistor RZ to keep RF out. Do you think they're pointless and should be removed?
Rz & C13 have only reached -3dB at 160MHz.
You could easily lower this to 1.6MHz or even 300kHz. But this is an operational fix for operational interference.
Omit it for sims that assume ideal conditions.
I don't like all the add ons between the input gate and the 100r anti-oscillation resistor.
Move them outboard or remove completely.
You could easily lower this to 1.6MHz or even 300kHz. But this is an operational fix for operational interference.
Omit it for sims that assume ideal conditions.
I don't like all the add ons between the input gate and the 100r anti-oscillation resistor.
Move them outboard or remove completely.
Nelson,
the benefit of the push-pull topology is higher efficiency at the price of sonic impairment. The symmetric transfer characteristic of the push-pull mode creates only odd order harmonic and intermodulation distortion products. IM is much worse as it is scarcely related to the fundamental tones and is three times the THD level. GNF will reduce, or rather push up all that in frequency out of sight (for measurement, but not quite for the ear). The cancellation of even-order harmonics implies a tangible loss, a tonal impoverishment.
Single-ended amps produce both even and odd harmonics, nevertheless, don`t measure nice, don`t look nice in LTSpice and sound cards (if that would bother you), just sound nice.
Besides, the push-pull action gives rise to displacements on the time axis, which is by no means less regrettable, even if some of its impact cannot be explicitly pointed at, it`s there in a latent manner to the ear, whose fundamental intention is to localize the sounds. Such a sound field appears chaotic, taking up much brain resources for processing and leads to exhaustion, shortening the listening sessions and saving battery power.
the benefit of the push-pull topology is higher efficiency at the price of sonic impairment. The symmetric transfer characteristic of the push-pull mode creates only odd order harmonic and intermodulation distortion products. IM is much worse as it is scarcely related to the fundamental tones and is three times the THD level. GNF will reduce, or rather push up all that in frequency out of sight (for measurement, but not quite for the ear). The cancellation of even-order harmonics implies a tangible loss, a tonal impoverishment.
Single-ended amps produce both even and odd harmonics, nevertheless, don`t measure nice, don`t look nice in LTSpice and sound cards (if that would bother you), just sound nice.
Besides, the push-pull action gives rise to displacements on the time axis, which is by no means less regrettable, even if some of its impact cannot be explicitly pointed at, it`s there in a latent manner to the ear, whose fundamental intention is to localize the sounds. Such a sound field appears chaotic, taking up much brain resources for processing and leads to exhaustion, shortening the listening sessions and saving battery power.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.