Audio Asylum Thread Printer
Get a view of an entire thread on one page
|For Sale Ads|
We had a customer's Jolida power amp in for light repair and the customer was interested in upgrade. We replaced the Audyn coupling capacitors with Russian K40Y paper in oil caps.
There are three coupling caps per channel, two .22uF and one .47uF. We missed the fact that one of them was .47uF at first and replaced it with a K40Y .22uF. We did one channel at a time so the customer and we could hear the difference between the stock Audyns and the K40ys.
We played the same cut in mono on the same speaker with both sides of the amp in turn.
With the inadvertently too-small coupling cap the sound was worse than original, with constricted dynamics and lack of detail. This is the opposite of what I've come to expect with K40Ys. That's when we went looking and found the undersized capacitor.
We then added a second .22uF in parallel for .47uF (close enough to the original .47uF). The sound then opened up with more detail and impact than the other channel, exactly what we expected initially.
It seems odd that a too-small coupling capacitor would affect dynamics and detail rather than bass, but that's exactly what happened. And using the right values of the K40Ys did what I've come to expect of them-detail, impact, macro and micro dynamics.
The 0.47uf is not a coupling cap. It is involved in the phase inverter and you probably introduced a phase problem into the audio band with the lower value cap.
Yes, that's a decouplimg capacitor and it ties the grid to AC ground. It is a bypass cap, albeit a small one. The smaller the value, the less effective its bypass function becomes allowing a bit of positive feedback into the grid of the non-inverting tube.
Anyway, that's my theory.
The problem with this circuit isn't so much undesired signal amplitude at the grid. Rather, it's the phase shift imparted by the RC. Any decoupling value of 1uF or less will create phase shift at the grid that is roughly 90 degrees throughout much of the vocal range. Whatever the amplitude, the lower section of the splitter will be modulated by this out-of-phase signal. The result is output that's smeared and lacking in definition. This type of splitter is a good topology to avoid in any high-end amplifier.
Buy Chinese. Bury freedom.
Isn't this simply a SRPP driving a differential splitter? Granted the 15K in the tail doesn't make it the best one but that is the designers fault not the topologies fault.
Unless... Are you speaking of the 1M / 47µ combo being bad practice rather than simply grounding the undriven grid.
Being a SE guy, I never really looked into what that actually does.
Yes, that's the problem. The grid must be at AC ground. I suppose the reason this doesn't immediately grab everyone's attention is that the same RC network creates significant attenuation of the signal at the grid. After all, that's its purpose (and to bias the grid to the same DC level as the upper section). The problem is, the phase shift is nearly constant at 90 degrees across the audio band, whereas attenuation varies as a function of frequency. Use of this circuit demands that we assume the shift to be inaudible across the entire range of attenuation, whatever it might be, and I don't think it's reasonable to do that. I also think this is the cause of the phenomenon the OP is describing.
Buy Chinese. Bury freedom.
...the section of the PI under discussion is cathode drive so under ideal conditions the output at the anode is in phase with the input at the cathode. Under ideal conditions, g1 of this section should be at AC ground at all frequencies and phase shift would not be an issue. In the practical implementation with a finite value grid bias resistor and a finite value decoupling cap, there is phase shift approaching -90 deg at g1 of the cathode coupled section. But at all but the very lowest frequencies, say <10 hz, the amplitude of the phase shifted signal applied to g1 is minuscule such that the output at the anode is shifted by just a few degrees and not the -90 deg at g1. IOW, the signal applied to the cathode dominates the output at the anode.
The "long tailed pair" PI certainly isn't perfect but gross phase shift at the non-inverting output is typically not an issue. As you know, this style PI was used successfully in many classic designs including those from Marantz and McIntosh.
The problem is, the phase shift is nearly constant at 90 degrees across the audio band, whereas attenuation varies as a function of frequency. Use of this circuit demands that we assume the shift to be inaudible across the entire range of attenuation, whatever it might be, and I don't think it's reasonable to do that.
I will agree with you on this. I wonder what the result would be if the 1M resistor was replaced with a grid choke. The resultant LC filter would be a 2nd order filter with a 180 instead of a 90-degree phase shift across the frequency band. Given a sufficiently large decoupling cap to ground, the same signal would appear (same relative phase angle) at the grid as it would at the cathode input. Of course this would be a feed-forward signal, but it might not be too detrimental given that it'd be very small.
Could it be a way to attempt to compensate for LF phase lag caused by finite inductance or is it simply to deal with the direct coupling as CB says below?
I get all turned around and my head starts to hurt when I start attempting to think of what is going on. You are feeding a frequency attenuated 90º shifted signal that is 180º out of phase with the cathode to the grid. I do see that cutting the cap value in half will give a 6dB increase of whatever effect this does.
I've been bothered by that structure when I last saw it (in some Dynaco monoblock driver boards from somewhere) and I wondered why you'd bother with the direct coupling only to have a coupling cap stuck on the other grid.
In these fixed bias amps with negative rails, why in the world wouldn't they put a CCS between the negative rail and the LTP, then ground one grid and have a single coupling cap to the other? You'd get the added bonus of being able to drastically raise the impedance of the plate load resistors on the LTP which doesn't hurt anything either.
I totally agree with your premise and I've been thinking the same thing for a long time now, at least on the few occasions that I've contemplated building an amp with an LTP splitter - haven't built one yet. Here's the basic idea of the circuit taken from tubecad.com with the only difference being that in the case of the cathode coupled inverter, we would need a plate load at the first tube.
That's not a phase inverter, but could be turned into one with some parts added.
Phase shift was my initial thought. I still don't see it as a very plausible cause. The cutoff frequency's -3db point is at 0.34 Hz with the 1 Meg and 0.47 uF components. The -1db point is at 3.4 Hz and the FR is essentially flat by 34 Hz. With the 0.22 uF cap in place the -3db point is at 0.72 Hz
I could be wrong though and the few degrees of phase rotation at 20 Hz could have a larger effect than I'm allowing for, hence the qualifier about this being a theory.
The -3dB point isn't the issue. I was referring to phase shift in the midrange. Note that this is only occurring in the lower section, so it represents an undesired phase differential between the two outputs of the splitter.
Buy Chinese. Bury freedom.
I wasn't aware that such a situation could occur within the passband from a first order cut-off filter. Can't argue with the sim results though. Having said that, the phase shift at 10 KHz is only 2 deg which is probably less than what is imparted to the signal by the OPT at that frequency. I guess the effect could be cumulative leading to distortion and higher order harmonics at a range where the ear is sensitive.
If you don't mind, could you please post a sim with a 1 uF or 1.5 uF capacitor? Just curious to see the difference.
"the phase shift at 10 KHz is only 2 deg"
As shown in the graph, phase shift at 10 kHz is nearly 90 degrees. It won't change much with a 1u cap.
Buy Chinese. Bury freedom.
I stand corrected, I saw the deviation from 1 KHz to 10 KHz as a 2 deg difference - which it is, not realizing that it's centered at 90 Degrees across the entire passband.
Was the wrong cap installed on the un-driven side of the phase inverter? Don't know the amp topology but just a guess.
Post a Message!
This post is made possible by the generous support of people like you and our sponsors: