The experiments are for my pleasure only, I'm not producing equipment to sell - what I do is extremely untidy visually; I'm not entertaining the eyes, my ears are what has to be convinced.
How is that any different from a mandraulic switch though? Or was that your point?
An experiment is an orderly procedure carried out with the goal of verifying, refuting, or establishing the validity of a hypothesis. Experiments provide insight into cause-and-effect by demonstrating what outcome occurs when a particular factor is manipulated. (source wikipedia)
You are not experimenting. Not even sure what you do qualifies as goofing around.
The hypothesis is that if audio equipment is modified by identifying weaknesses and remedying them, one by one, then convincing sound can be achieved. The remedy doesn't has to be perfect, in some classic engineering sense - merely good enough to counter the weakness.
The experiments are to find what factors are the most important, that have the greatest impact; and what areas can be ignored, that don't inhibit success in reaching the goal. So far I reckon I've had pretty good success - I've spent quite a bit of time listening to ambitious systems of others, and they all show considerable, obvious flaws - I don't think I've been totally wasting my time ...
.
The experiments are to find what factors are the most important, that have the greatest impact; and what areas can be ignored, that don't inhibit success in reaching the goal. So far I reckon I've had pretty good success - I've spent quite a bit of time listening to ambitious systems of others, and they all show considerable, obvious flaws - I don't think I've been totally wasting my time ...
.
No, sorry, was a reaction to a quoted post that seemed to state that FET switches cause by definition spikes and clicks.
jan
Yes, due to capacitive feed-through.
SSM audio switches ramp the control voltage for this reason.
Dan.
Yes, and you can't get mechanical/electromechanical switches to make and break gradually.
There are also approaches that attempt to switch near zero-crossings. Of course with sufficient offset you will never switch, and if there is signal present just switching at a zero crossing is still a discontinuity and accordingly with energy. Around Harman some referred to this as Fourier noise.
Usually it's sufficient in a system to have one very good muting circuit at the end of whatever signal switching is used. It needs to be a gradual transition if there is any signal or offset present.
mini-Vendetta-style breadboard
With 2SK170BL (equivalent) and 2SJ74BL I did some rough matching for Idss and then adjusted the pullup resistors for a somewhat lower current (from 300 ohms to 453 ohms). I tied the input sources directly to common, and the output offset was tolerable. I installed a 101:1 voltage divider at the input so the noise of the generator could be attenuated. The arrangement is partially shielded and built on ground plane protoboard (Twin, plated-through pad-per-hole, one side plane). Gain came out at about 17.7 times, about 25dB. Noise density referred to the input and backing out the Ap noise, based on a 22kHz noise bandwidth was about 2.1nV per sq rt Hz. Some additional filtering was added for the bench supply and improved things slightly.
I added 1uF film bypassing to the uppermost and lowermost 1k gate bias resistors for the folded cascode parts. I also tied the middle 1k's to common. For the upper cap addition the circuit remained stable, but for the lower there was a few-hundred-MHz parasitic oscillation, which was quelled with 11 ohms in series with the cap, and I added a resistor to the other for symmetry and insurance. Now the noise went down a lot, with the input-referred noise density about 680pV per sq rt Hz. THD+N for the optimal input level of about 947uV is about 0.018% with the 22kHz filter and a 1kHz signal, again driven from a source impedance of about 10 ohms, which is not unrealistic for an MC cartridge. I also backed out the small noise contribution from the 10 ohms.
The likelihood of parasitics with these short-channel parts underscores the importance of having a decent 'scope to inspect things, and not to rely on the Audio Precision solely.
I got round to constructing a mini-Vendetta-style board yesterday, and have been working toward the best performance before exploring the floating I source addition.
With 2SK170BL (equivalent) and 2SJ74BL I did some rough matching for Idss and then adjusted the pullup resistors for a somewhat lower current (from 300 ohms to 453 ohms). I tied the input sources directly to common, and the output offset was tolerable. I installed a 101:1 voltage divider at the input so the noise of the generator could be attenuated. The arrangement is partially shielded and built on ground plane protoboard (Twin, plated-through pad-per-hole, one side plane). Gain came out at about 17.7 times, about 25dB. Noise density referred to the input and backing out the Ap noise, based on a 22kHz noise bandwidth was about 2.1nV per sq rt Hz. Some additional filtering was added for the bench supply and improved things slightly.
I added 1uF film bypassing to the uppermost and lowermost 1k gate bias resistors for the folded cascode parts. I also tied the middle 1k's to common. For the upper cap addition the circuit remained stable, but for the lower there was a few-hundred-MHz parasitic oscillation, which was quelled with 11 ohms in series with the cap, and I added a resistor to the other for symmetry and insurance. Now the noise went down a lot, with the input-referred noise density about 680pV per sq rt Hz. THD+N for the optimal input level of about 947uV is about 0.018% with the 22kHz filter and a 1kHz signal, again driven from a source impedance of about 10 ohms, which is not unrealistic for an MC cartridge. I also backed out the small noise contribution from the 10 ohms.
The likelihood of parasitics with these short-channel parts underscores the importance of having a decent 'scope to inspect things, and not to rely on the Audio Precision solely.
The integrated volume controllers all have 0-xing switching with (sometimes adjustable) timeout, like 50 ms. That way, you are sure to switch even in the unlikely event that there's little signal and lots of offset and you don't cross zero. Though, in many years of use of lots of these systems that has never happened to me.
They do sometimes exhibit what is called 'zipper noise' when you ramp the level up or down fast. Individual clicks too low to be audible can become audible if the ramp rate gets into the audio frequency range.
jan
> I added 1uF film bypassing to the uppermost and lowermost 1k gate bias resistors for the folded cascode parts ..... For the upper cap addition the circuit remained stable, but for the lower there was a few-hundred-MHz parasitic oscillation, which was quelled with 11 ohms in series with the cap, and I added a resistor to the other for symmetry and insurance.
I wonder whether a gate-stopper of say 220R between the cap and the actual gate pin would also do the job.
Or perhaps use a cap to make an RC of << 1Hz, which would then be likely to be an electrolytic.
> Now the noise went down a lot, with the input-referred noise density about 680pV per sq rt Hz.
Patrick
I wonder whether a gate-stopper of say 220R between the cap and the actual gate pin would also do the job.
Or perhaps use a cap to make an RC of << 1Hz, which would then be likely to be an electrolytic.
> Now the noise went down a lot, with the input-referred noise density about 680pV per sq rt Hz.
Patrick
Last edited:
A gate stopper would work as well I'm sure. And the 'lytics would have some ESR that might be enough to quell the parasitics, although one concern would be power cycling transients pulling momentary gate current---would have to look at that carefully.
I am slightly puzzled about the size of the change in noise for just the bypassing of the folded cascode gates, but it seems to be real. However, in the order I made changes, I added the PS filtering (11 ohms/2200uF each rail) after I added the gate bypassing, not before, so some of the excess noise may well have been from the power supply (a bog-standard Tektronix CPS250). The preamp needs a good supply.
The actual Vendetta schematic I saw, I believe has considerably larger and multiple input devices (two 2SK147, two 2SJ72), so again this is just a proof-of-concept exercise, not an attempt to outdo anything. But it is pretty promising.
When I add the floating current generator, in principle the importance of the cascode gate bypassing should be diminished. I should have those results soon. I am charging the NiMH batteries, and selected some 2SK381E devices (obsolete and not particularly recommended but which I have an ammo box of) and paralleled individual source-ballasted ones to get about the right total current formerly supplied by the pullup/pulldown R's.
If you have a long track between your switch (solid state or mechanical) and the fly back diode is not close to the relay coil, you will generate clicks from the relay coil inductive snap back. The trick is to locate the fly back diode right at the relay coil (I use a SMD device) and in parallel a 0.1 uF cap. I use a 100 ohm resistor in series with the relay coil.
This approach all but kills all relay switching noise.
It does not address of course clicks SMD oops due to DC on the signal.
Also be aware that cheap switches have excessive contact bounce, and this also induces clicks when switching. Use a snubber to solve this.
... and of course none of this (except DC) rears its ugly head with a ss integrated level control
Jan
... and of course none of this (except DC) rears its ugly head with a ss integrated level control
Jan
I get a pop whenever I switch from DAC to tubed phono stage but not
the other way. The DAC has OP AMP on the output and thus DC very close
to 0V. The phono stage is AC coupled with a pull down resistor. Hence 0V DC.
Both grounds are permanently connected together. So there should be no pop.
I should note that autotransformer volume control is after the switch.
One cause could be that when you open the tube circuit, the output cap has to (re)charge. What is the value of the resistor to ground on the output of the tube stage, and what is the DC on the other side of the cap?
Jan
- Status
- Not open for further replies.