> the last thing you want on a VE bus is a big cap to ground!
It is often done. Even if in normal use you have enough mix resistors to be stable, it is nice to be able to bench-test the mixamp alone without jury-rigging an input termination. And this looks like a ribbon-cable board: when the ribbon connector falls off, it is nice if the mixamp will sit quietly instead of running off to MHz screams. Also a half-meter ribbon is very good at picking up 1 meter radio waves: if terminated in something like cable impedance (about 100Ω for most cable) then most radio signals give at most a few milliVolts, but if left open all that power rings-up to a high voltage that pushes the mixamp faster than it can fight back. The 100Ω+0.03uFd means that the bus WILL be loaded in 100Ω above 50KHz, so any ribbon-length and radio-wave that fits inside a building will be terminated.
Below 50KHz it is predominantly a high-NFB virtual-earth system. Bus impedance in the audio band is an Ohm or less. For a small number of high impedance mix resistors, above 50KHz it transitions to a loaded passive mixer. Worse-case, switching mix resistors in or out will cause ~1dB gain change above 50KHz for other inputs. Actually with this amp and a large number of 1K5 mix resistors, it will work virtual-earth with <0.1dB gain-change way past 100KHz.
> My way would be to use a 'middle of the road' value of bus resistor, say 6K8
I remember mixing on 3Meg resistors. 270K makes a lot of sense for expensive high-volt channel amps like 12AX7 (see all the Altec and Bogen low-price tube mixers). Gately used both 100K and 10K in various products. With six 100K inputs into a 301 chip, thermal and chip noise are similar, going lower-Z would not be a big improvement. But the noise-voltage equation promises lower resistor thermal noise voltage at lower impedance. This is bogus: what matters is POWER, not voltage. We were short-sighted because we were living in a +/-18V world. Anyhow, given a low-volt world, you get more power with lower impedance. ASSuming you can find an amp with very low noise voltage, and channels to drive low mix resistors, things get better though slowly.
1K5 seems like an extreme. Say one channel swings +15V into 1K5, we have 10mA current. The channel chip is likely a 5532. To make 10mA it has to go class AB. It has to take 10mA from the + rail. It always takes 3mA for idle current. So as it swings from 0V to +15V, it takes 3mA, 3mA, 3mA, 5mA, 7mA, 9mA.... the rail current is NOT proportional to signal current, it is signal+idle, and probably some glitch from booster tricks. If the rail impedance is one Ohm we have a very distorted 3mV to 10mV version of half the signal riding on the rail. And some of this will generally couple into the boards grounding system, and thus into the + input of the mixamp (and everywhere else).
I still think the real "mushy" here is the very low gain after the mixamp. The master faders should not be worked lower than -12dB, and preferably higher so the mix amp is working at lower level than the "+4dBm" outputs. +4 is hot enough for very complex mixes on ordinary chips. We want only one stage working that hard (actually that push-pull output works each half at -2dBu to get +4dBu output). We sure do not want the mixamp working at +10dBu nominal, +26dBu peaks, then pull-back the Master fader to give +4dBu at the output jack. While we could build a very fine mixamp to cruise +26dBu peaks sweetly (990 fer example), it ain't cheap and it is not necessary.
> My way would be to...
My way would be to: setup the recorder for unbalanced -10dBV input reference level, and take signal right from the mixamp, bypassing 3 chips and the $2 master pot. Now we do need to consider noise, and the BJTs may be some help in non-Rock work. Can't fade-out the end of a track with the Master, but I do that in post, long after the mixing stage. My way is not for everybody, true.
In random PA work, of course, I always want a Master. Feedback happens. Maybe with a long-running show, I would not need it, but all my work is ad-hoc and usually chaos.
It is often done. Even if in normal use you have enough mix resistors to be stable, it is nice to be able to bench-test the mixamp alone without jury-rigging an input termination. And this looks like a ribbon-cable board: when the ribbon connector falls off, it is nice if the mixamp will sit quietly instead of running off to MHz screams. Also a half-meter ribbon is very good at picking up 1 meter radio waves: if terminated in something like cable impedance (about 100Ω for most cable) then most radio signals give at most a few milliVolts, but if left open all that power rings-up to a high voltage that pushes the mixamp faster than it can fight back. The 100Ω+0.03uFd means that the bus WILL be loaded in 100Ω above 50KHz, so any ribbon-length and radio-wave that fits inside a building will be terminated.
Below 50KHz it is predominantly a high-NFB virtual-earth system. Bus impedance in the audio band is an Ohm or less. For a small number of high impedance mix resistors, above 50KHz it transitions to a loaded passive mixer. Worse-case, switching mix resistors in or out will cause ~1dB gain change above 50KHz for other inputs. Actually with this amp and a large number of 1K5 mix resistors, it will work virtual-earth with <0.1dB gain-change way past 100KHz.
> My way would be to use a 'middle of the road' value of bus resistor, say 6K8
I remember mixing on 3Meg resistors. 270K makes a lot of sense for expensive high-volt channel amps like 12AX7 (see all the Altec and Bogen low-price tube mixers). Gately used both 100K and 10K in various products. With six 100K inputs into a 301 chip, thermal and chip noise are similar, going lower-Z would not be a big improvement. But the noise-voltage equation promises lower resistor thermal noise voltage at lower impedance. This is bogus: what matters is POWER, not voltage. We were short-sighted because we were living in a +/-18V world. Anyhow, given a low-volt world, you get more power with lower impedance. ASSuming you can find an amp with very low noise voltage, and channels to drive low mix resistors, things get better though slowly.
1K5 seems like an extreme. Say one channel swings +15V into 1K5, we have 10mA current. The channel chip is likely a 5532. To make 10mA it has to go class AB. It has to take 10mA from the + rail. It always takes 3mA for idle current. So as it swings from 0V to +15V, it takes 3mA, 3mA, 3mA, 5mA, 7mA, 9mA.... the rail current is NOT proportional to signal current, it is signal+idle, and probably some glitch from booster tricks. If the rail impedance is one Ohm we have a very distorted 3mV to 10mV version of half the signal riding on the rail. And some of this will generally couple into the boards grounding system, and thus into the + input of the mixamp (and everywhere else).
I still think the real "mushy" here is the very low gain after the mixamp. The master faders should not be worked lower than -12dB, and preferably higher so the mix amp is working at lower level than the "+4dBm" outputs. +4 is hot enough for very complex mixes on ordinary chips. We want only one stage working that hard (actually that push-pull output works each half at -2dBu to get +4dBu output). We sure do not want the mixamp working at +10dBu nominal, +26dBu peaks, then pull-back the Master fader to give +4dBu at the output jack. While we could build a very fine mixamp to cruise +26dBu peaks sweetly (990 fer example), it ain't cheap and it is not necessary.
> My way would be to...
My way would be to: setup the recorder for unbalanced -10dBV input reference level, and take signal right from the mixamp, bypassing 3 chips and the $2 master pot. Now we do need to consider noise, and the BJTs may be some help in non-Rock work. Can't fade-out the end of a track with the Master, but I do that in post, long after the mixing stage. My way is not for everybody, true.
In random PA work, of course, I always want a Master. Feedback happens. Maybe with a long-running show, I would not need it, but all my work is ad-hoc and usually chaos.
