Peter Knight
Active member
http://www.retrovox.com.au/proaudio.html
More stuff for everyone's archives. :thumb:
More stuff for everyone's archives. :thumb:
ABC (Australia) schematics
- Thread starter Peter Knight
- Start date
Help Support GroupDIY Audio Forum:
rlaury
Well-known member
Interesting stuff!
I wonder how the TriMax sounds? It looks a little like
the Fairchild topology.
RonL
I wonder how the TriMax sounds? It looks a little like
the Fairchild topology.
RonL
NewYorkDave
Well-known member
Great stuff! Thank you.
Many of these schematics include signal levels and transformer ratios, very useful data.
Many of these schematics include signal levels and transformer ratios, very useful data.
Bjorn Zetterlund
Well-known member
Nice! Thanks for that. Don't think that's a vari-mu/fairchild type of device though...uses some other kinda way which I haven't come across before. Interesting that it incorporates a delay network to help catch peaks; never seen that on a valve compressor.
Bjorn
Bjorn
Peter- Thanks for that link!
> I wonder how the TriMax sounds? It looks a little like the Fairchild topology.
Not at all. Fairchild is "one stage" (in the audio path). This A.30 cascoded out the wazzo.
The goal of very fast attack with "auto"-release is similar, true.
(And that is a lovely consise description of general limiter design issues.)
The gain-cell (V2) in the A.30 is quite different from any discussed here. It is not at all a vari-Mu, nor even an amplifier. It is actually a lot like a balanced diode attenuator, except with bottles, and the "diodes" self-buffer their control signal.
Look how vari-Mu and Gilbert cells work as limiters. When level rises, gain must drop. To reduce gain, these topologies reduce current. That means our loudest sounds have to ride very-low current. At some point it has to distort like mad.
Diode attenuators increase current to reduce gain. There is always a limit, but we are not "going the wrong way" like when pushing big signals through vari-Mu and Gilbert cells: starving them to reduce overloads. With common tubes, the output power with high GR is very weak, far below line level. Most limiters use a booster amp to bring signal out of the gain cell up to line level; Fairchild used a wide bank of tubes to boost power.
The A.30 feeds a balanced 600Ω signal through two 5KΩ resistors, to two cathodes. An amplifier extracts signal from these cathodes and boosts it up to line level. When not limiting, the current in the cathodes is nearly zero. To reduce signal, to "fight" over-loud inputs, we increase the current in the triode. That reduces dynamic cathode impedance from over 5KΩ to around 500Ω. With those 5KΩ resistors in series, this gives up to 20dB gain reduction. Also note that the loudest inputs force the highest currents in the triodes. They can fight large signals effectively.
I looked at this concept a few times and did not see a compelling advantage over a forward vari-Mu amplifier. Levels are higher, but noise is also higher. I may have to look at it again.
It will thump. Maybe it thumps less than a vari-Mu. But the booster stage has to handle up to maybe 25V of common mode thump and reject it. And it may thump most at high-GR, while a vari-Mu thumps worst as it just starts to GR. For sane engineers who don't live in high GR, this is better.
A tricky-bit: there is a 50uS delay line in the audio path. This cancels some of the attack-delay in the rectifier. I've seen that on recent gear, but not anything this old.
Another point: this bastard is feedforward. To do that, you need matching between detector and gain-cell. The classic vari-Mu control law is pretty non-linear, and when used with a linear rectifier won't give a flat limiting level. Feedforward mostly came of age when Drawmer and DBX developed exponential detectors and gain-cells that could drive each other with precision. But this guy is doing the same thing! With linear detector and semi-exponential tubes. I guess the 5KΩ cathode-drive resistors also linearize the control law, though I would not expect that to work well.
> I wonder how the TriMax sounds? It looks a little like the Fairchild topology.
Not at all. Fairchild is "one stage" (in the audio path). This A.30 cascoded out the wazzo.
The goal of very fast attack with "auto"-release is similar, true.
(And that is a lovely consise description of general limiter design issues.)
The gain-cell (V2) in the A.30 is quite different from any discussed here. It is not at all a vari-Mu, nor even an amplifier. It is actually a lot like a balanced diode attenuator, except with bottles, and the "diodes" self-buffer their control signal.
Look how vari-Mu and Gilbert cells work as limiters. When level rises, gain must drop. To reduce gain, these topologies reduce current. That means our loudest sounds have to ride very-low current. At some point it has to distort like mad.
Diode attenuators increase current to reduce gain. There is always a limit, but we are not "going the wrong way" like when pushing big signals through vari-Mu and Gilbert cells: starving them to reduce overloads. With common tubes, the output power with high GR is very weak, far below line level. Most limiters use a booster amp to bring signal out of the gain cell up to line level; Fairchild used a wide bank of tubes to boost power.
The A.30 feeds a balanced 600Ω signal through two 5KΩ resistors, to two cathodes. An amplifier extracts signal from these cathodes and boosts it up to line level. When not limiting, the current in the cathodes is nearly zero. To reduce signal, to "fight" over-loud inputs, we increase the current in the triode. That reduces dynamic cathode impedance from over 5KΩ to around 500Ω. With those 5KΩ resistors in series, this gives up to 20dB gain reduction. Also note that the loudest inputs force the highest currents in the triodes. They can fight large signals effectively.
I looked at this concept a few times and did not see a compelling advantage over a forward vari-Mu amplifier. Levels are higher, but noise is also higher. I may have to look at it again.
It will thump. Maybe it thumps less than a vari-Mu. But the booster stage has to handle up to maybe 25V of common mode thump and reject it. And it may thump most at high-GR, while a vari-Mu thumps worst as it just starts to GR. For sane engineers who don't live in high GR, this is better.
A tricky-bit: there is a 50uS delay line in the audio path. This cancels some of the attack-delay in the rectifier. I've seen that on recent gear, but not anything this old.
Another point: this bastard is feedforward. To do that, you need matching between detector and gain-cell. The classic vari-Mu control law is pretty non-linear, and when used with a linear rectifier won't give a flat limiting level. Feedforward mostly came of age when Drawmer and DBX developed exponential detectors and gain-cells that could drive each other with precision. But this guy is doing the same thing! With linear detector and semi-exponential tubes. I guess the 5KΩ cathode-drive resistors also linearize the control law, though I would not expect that to work well.
Bjorn Zetterlund
Well-known member
Thanks PRR!
Top notch post as always. Soooo....looks like maybe Jakob isn't the first to do a feed-forward valve compressor with his G10 then eh?? Still the only feed-forward vari-mu I know of, though!
Bjorn
Top notch post as always. Soooo....looks like maybe Jakob isn't the first to do a feed-forward valve compressor with his G10 then eh?? Still the only feed-forward vari-mu I know of, though!
Bjorn
Well, I'll be snookered. You can feed-forward mis-matched rectifier-control laws and get "good" curves, at least over about 12dB of GR.
The problem is getting a best-fit between a straight line and a double-curved line. So it will never be "good" over 60dB of range, like the DBX circuits could be. But for many "sane" audio uses of limiting, 10-15dB is enough, and we don't need (may not want) ruler-flat action.
Simulation:
Red line is input signal, increasing from 0 to 10V.
Green lines are output signal, for five different initial biases (+18V to +26V) on the cathodes (trimmer R46).
The blue lines are the common-mode DC level at the cathodes, which must be rejected by the next stage. Note that I divided by 10 to make it fit on the graph; the CM voltage really shifts from about 20V to 56V.
(The lines kink at the top because the A.30 rectifier can't make more than about 25VDC, and I emulated that effect.)
Fiddling the bias shifts you from soft to over-compensated. The top green curve is +26V bias, gives 3dB over-compensation, and 0.4dB loss from the input signal when not in GR. The bottom curve, very soft action, is 4dB loss below limiting. Flattest action happens when the not-limiting gain is -2dB from the input signal. So to set-up: remove 6SN7 GR tube and control rectifier, feed a small signal (less than 1V at the cathodes), note the output. Now put the 6SN7 back in, and adjust trimmer R46 (or available 15V-25V source) for 2dB less output.
They say most but not all 6SN7s are balanced enough to work thumplessly. This also depends on the CMRR and common-mode clip-limits of the next stage(s). Their check is to simply put 3V AC (50Hz heater-power) on the 6SN7 grids. (This was a standard way to thump-test.) If balance is perfect, this signal will be rejected, no output. Apparently they will accept enough unbalance to show on the output meter. 3V is about 14dB higher than limiting level, so this suggests that CMRR including 6SN7 unbalance should be better than 20dB. If big hum on the control line is rejected at the output, then thumps should also be rejected.
I have not done detailed distortion checks.... my back needs a break from hunching over SPICE. My guess is that levels of 0.5V across two 6SN7 cathodes in push-pull will have "low" THD, well under 1%. But I need to check that around threshold where 6SN7 current is very low.
The problem is getting a best-fit between a straight line and a double-curved line. So it will never be "good" over 60dB of range, like the DBX circuits could be. But for many "sane" audio uses of limiting, 10-15dB is enough, and we don't need (may not want) ruler-flat action.
Simulation:
Red line is input signal, increasing from 0 to 10V.
Green lines are output signal, for five different initial biases (+18V to +26V) on the cathodes (trimmer R46).
The blue lines are the common-mode DC level at the cathodes, which must be rejected by the next stage. Note that I divided by 10 to make it fit on the graph; the CM voltage really shifts from about 20V to 56V.
(The lines kink at the top because the A.30 rectifier can't make more than about 25VDC, and I emulated that effect.)
Fiddling the bias shifts you from soft to over-compensated. The top green curve is +26V bias, gives 3dB over-compensation, and 0.4dB loss from the input signal when not in GR. The bottom curve, very soft action, is 4dB loss below limiting. Flattest action happens when the not-limiting gain is -2dB from the input signal. So to set-up: remove 6SN7 GR tube and control rectifier, feed a small signal (less than 1V at the cathodes), note the output. Now put the 6SN7 back in, and adjust trimmer R46 (or available 15V-25V source) for 2dB less output.
They say most but not all 6SN7s are balanced enough to work thumplessly. This also depends on the CMRR and common-mode clip-limits of the next stage(s). Their check is to simply put 3V AC (50Hz heater-power) on the 6SN7 grids. (This was a standard way to thump-test.) If balance is perfect, this signal will be rejected, no output. Apparently they will accept enough unbalance to show on the output meter. 3V is about 14dB higher than limiting level, so this suggests that CMRR including 6SN7 unbalance should be better than 20dB. If big hum on the control line is rejected at the output, then thumps should also be rejected.
I have not done detailed distortion checks.... my back needs a break from hunching over SPICE. My guess is that levels of 0.5V across two 6SN7 cathodes in push-pull will have "low" THD, well under 1%. But I need to check that around threshold where 6SN7 current is very low.
_in______out_____THD_
0.04 ___0.033 ___0.13% (noise?)
0.08 ___0.067 ___0.03%
0.16 ___0.132 ___0.06%
0.32 ___0.251 ___0.04%
0.64 ___0.444 ___0.07%
1.28 ___0.656 ___0.12%
2.56 ___0.756 ___0.19%
5.12 ___0.779 ___0.22%
Spectrum at 5.12V in:
fund=100%
2nd=0.0024%
3rd=0.22%
4th=0.0033%
5th=0.006%
0.04 ___0.033 ___0.13% (noise?)
0.08 ___0.067 ___0.03%
0.16 ___0.132 ___0.06%
0.32 ___0.251 ___0.04%
0.64 ___0.444 ___0.07%
1.28 ___0.656 ___0.12%
2.56 ___0.756 ___0.19%
5.12 ___0.779 ___0.22%
Spectrum at 5.12V in:
fund=100%
2nd=0.0024%
3rd=0.22%
4th=0.0033%
5th=0.006%
Kid Squid
Well-known member
Cheers Pete - some very interesting reading there ! thanks,
PRR - You are the Tech's Tech !!!! :grin: Respect to you :thumb:
Steve :guinness:
PRR - You are the Tech's Tech !!!! :grin: Respect to you :thumb:
Steve :guinness:
Distortion in this scheme is quite low, even at relatively high levels. 0.2% THD, mostly 3rd, is typical at 400mV, maybe 0.4% at 2V, and around 0.1% at 100mV. Clean below threshold, no kink of THD at threshold, and THD while limiting does not vary much before gain control is lost.
Around 20dB of good flat clean GR is possible.
The killer problem is: it needs LOTS of GR voltage, around 50V. That problem is slightly reduced by the fact you can use VERY high value grid (and release) resistors, because the DC bias is rock-solid (unlike typical vari-Mu circuits with small plate and cathode resistors). Even so, it tends to want more than the 20% of B+ peak swing you can easily get in resistance-coupled tube amps.
Even at these whopping voltages, rectifier drop is an issue. You note that the A.30 biases the tube rectifier a few tenths of a volt. Trying to use Silicon complicates things. (But I may have an elegant amp/rectifier idea....)
If you use the feedforward topology, you MUST avoid phase-shifts between signal and control paths. That may be why they have all those small R-C networks across plate resistors: null the transformer's phase shift in the bass.
Around 20dB of good flat clean GR is possible.
The killer problem is: it needs LOTS of GR voltage, around 50V. That problem is slightly reduced by the fact you can use VERY high value grid (and release) resistors, because the DC bias is rock-solid (unlike typical vari-Mu circuits with small plate and cathode resistors). Even so, it tends to want more than the 20% of B+ peak swing you can easily get in resistance-coupled tube amps.
Even at these whopping voltages, rectifier drop is an issue. You note that the A.30 biases the tube rectifier a few tenths of a volt. Trying to use Silicon complicates things. (But I may have an elegant amp/rectifier idea....)
If you use the feedforward topology, you MUST avoid phase-shifts between signal and control paths. That may be why they have all those small R-C networks across plate resistors: null the transformer's phase shift in the bass.
Peter Knight
Active member
More Trimax stuff...
http://www.retrovox.com.au/Images/type1spec.gif
http://www.retrovox.com.au/Images/type1cct.gif
http://www.retrovox.com.au/Images/type1cct_1.gif
Not sure what else this guy has got on his site.
http://www.retrovox.com.au/Images/type1spec.gif
http://www.retrovox.com.au/Images/type1cct.gif
http://www.retrovox.com.au/Images/type1cct_1.gif
Not sure what else this guy has got on his site.
If I am not mis-taken, Fairchild had a line of stuff made in or for Australia.
Peter Knight
Active member
Uh, I think the story was that Fairchild had their Vari-mu tubes made in Australia - or that the vari-mu tube was invented in Australia, I'm not sure. Someone correct me if I'm wrong.
Got a Trimax transformer here, which has gone into my LA2 clone - so I didn't hack it apart. :razz:
Got a Trimax transformer here, which has gone into my LA2 clone - so I didn't hack it apart. :razz:
ggoerss
Member
As i understand it AWA contracted Fairchild to produce limiters for them under the Amalgamated Wireless Australia name.These were not the fabled 670 ones though.
I seem to recall an earlier thread on the AWA limiters.
Rick O`neil at Turtlerock mastering in Sydney seems quite knowledgeable about the evolution.
www.turtlerockmastering.com
I seem to recall an earlier thread on the AWA limiters.
Rick O`neil at Turtlerock mastering in Sydney seems quite knowledgeable about the evolution.
www.turtlerockmastering.com
ggoerss
Member
Has anyone got any pinouts or specs for Trimax transformers as i have a couple that i`d like to use if i knew what they were.
ciminosound
Well-known member
Yeah, I have some Trimax trannys too.....I need to dig those out....
mylesgm
Well-known member
I've some schematics, lots of transformers and some trimax equipment if anybody is interested... great aussie built gear from the 50-60s
M
M
In the PRINCIPLES OF OPERATION, they clearly mention:PRR said:
" Time delay circuit:
Since all limiter devices depend on the charging of a condenser (sic), some time must elapse before limiting can take place.
...
...by inserting a time delay network..., the signal may be sufficiently delayed so that the limiting takes place simultaneously with the arrival of the signal."
OTOH, this APF is very much the same as the "time-smearing" filter found in all broadcast processors.
You sure are excavating the past.
> same as the "time-smearing" filter
Disagree. There is a modern fashion to smear the *whole* overtone zone, say 100Hz-5KHz. This tends to prevent peak addition and high peak/RMS ratio. I do not believe this was practical before transistor active filters. (Though, there were SSB phase-shifters for the 300-3KHz speech band potted in large soup-cans.)
The A.30 APF is a "small" 50 microSecond job. It just anticipates the ~~100 microSecond rise time of the control rectifier for "zero" attack, no transient overmodulation.
> same as the "time-smearing" filter
Disagree. There is a modern fashion to smear the *whole* overtone zone, say 100Hz-5KHz. This tends to prevent peak addition and high peak/RMS ratio. I do not believe this was practical before transistor active filters. (Though, there were SSB phase-shifters for the 300-3KHz speech band potted in large soup-cans.)
The A.30 APF is a "small" 50 microSecond job. It just anticipates the ~~100 microSecond rise time of the control rectifier for "zero" attack, no transient overmodulation.
