GroupDIY Bridge Compressor.
- Thread starter DaveP
- Start date
Help Support GroupDIY Audio Forum:
I have made the CV part of this circuit and it works well.
This was designed by G.Q.Herrick for Electronics in Dec 1943. it comes from the Handbook of Industrial Electronic Circuits. I have used a 6BZ7 for the diode driver with silicon diodes, the tube is set up to have an internal plate Z of ~7k and in// with a 22k plate resistor, the output Z is ~5.3k.
With the 0.22uF cap fitted it gives a theoretical attack of 0.22 x 0.0053 = 1.1mS. According to my earlier maths this will need an instantaneous charging current of 6mA.
DaveP
This was designed by G.Q.Herrick for Electronics in Dec 1943. it comes from the Handbook of Industrial Electronic Circuits. I have used a 6BZ7 for the diode driver with silicon diodes, the tube is set up to have an internal plate Z of ~7k and in// with a 22k plate resistor, the output Z is ~5.3k.
With the 0.22uF cap fitted it gives a theoretical attack of 0.22 x 0.0053 = 1.1mS. According to my earlier maths this will need an instantaneous charging current of 6mA.
DaveP
0.015A/0.5uFd is 30,000 Volts per Second or 30V per milliSecond which is your apparent goal.
Being resistor-limited, the 15mA flows only at first, declining as voltage rises. However you do what nearly 15mA on tap.
This topology must thump; if EMRR says it is inaudible then Collins did something right. May be a good AC/DC ratio. May be good matching of the tube halfs.
> Why are resistor values in kilos marked with an M in the schematic? What is the historical reason for this?
Roman numerals. M=1000.
Transition to the French custom K=1000 happened around MCML in many US shops, but WE and Collins clung to M=1000 all through the 1950s.
Note that Collins spells-out "Meg" where used.
Being resistor-limited, the 15mA flows only at first, declining as voltage rises. However you do what nearly 15mA on tap.
This topology must thump; if EMRR says it is inaudible then Collins did something right. May be a good AC/DC ratio. May be good matching of the tube halfs.
> Why are resistor values in kilos marked with an M in the schematic? What is the historical reason for this?
Roman numerals. M=1000.
Transition to the French custom K=1000 happened around MCML in many US shops, but WE and Collins clung to M=1000 all through the 1950s.
Note that Collins spells-out "Meg" where used.
Andy Peters
Well-known member
Jonte Knif said:
For the same reason the financial people refer to thousands using a capital M. (A balance of a million dollars is written in their shorthand as $1MM.) It's a Roman numeral.
I suppose we can start arguing about whether kilo is a thousand or 1,024 ...
-a
Moving on to the output stage.
Looking at the 23C using 6K6G's or the 26W using 6F6's, then I'm guessing that both of these designs are working in AB2 as in the tube data; in other words the grids are being driven positive in triode mode, the plates have an rp of ~2.5K
There are more modern tubes available now which don't require the driver tube 6N7 or the TX of the 23C to do the same. Most obvious is the 6V6 in octal $14.00, or 6CM6 in B9A $5.50, or 6AQ5 $5.00 in B7G. these all have the same 6V6 characteristic, any preferences?
Any of these tubes will give less than 2.5k rp in triode mode, but the output power will probably be a lot less than 10W, is there any reason not to run them as pentodes?
Whatever we run then 300V and plenty of current will be available but the rp will be much higher in pentode mode, any thoughts on this?
best
DaveP
Looking at the 23C using 6K6G's or the 26W using 6F6's, then I'm guessing that both of these designs are working in AB2 as in the tube data; in other words the grids are being driven positive in triode mode, the plates have an rp of ~2.5K
There are more modern tubes available now which don't require the driver tube 6N7 or the TX of the 23C to do the same. Most obvious is the 6V6 in octal $14.00, or 6CM6 in B9A $5.50, or 6AQ5 $5.00 in B7G. these all have the same 6V6 characteristic, any preferences?
Any of these tubes will give less than 2.5k rp in triode mode, but the output power will probably be a lot less than 10W, is there any reason not to run them as pentodes?
Whatever we run then 300V and plenty of current will be available but the rp will be much higher in pentode mode, any thoughts on this?
best
DaveP
I can't speak factually toward AB2, other than to note I'd doubt anything would have been used that might cause large current fluctuations, and therefore possible output level changes, what with these being used as safety on transmitters, and FCC fines for over-modulation. But, the meter on a 26C only goes to -5 dB GR anyways. You weren't advised to push more than about 3 dB on limiters of the era, as general concept. The first RCA limiter used a separate regulated supply for the entire audio amp, specifically to avoid fluctuations that might cause over-modulation, and an unregulated supply with AB power tubes does not sound broadcast worthy to me. I may be alarmist there.
The 26C manual mentions using 6F6 is 6K6 doesn't provide enough output power.
6AQ5 at $5 sounds just fine. We have enough desirable antiques causing 6V6 prices to rise steadily already.
I would question whether this is aimed at general purpose compression, or brick wall limiting, or some of both. Plenty of lower power compression units have been sold, many with 1W max output stages, and in practice they don't sound a lot different from the compressors using 10W output stages. If we are chasing high ratio limiting, it's another matter and we want max reasonable power for the side chain.
The 26C manual mentions using 6F6 is 6K6 doesn't provide enough output power.
6AQ5 at $5 sounds just fine. We have enough desirable antiques causing 6V6 prices to rise steadily already.
I would question whether this is aimed at general purpose compression, or brick wall limiting, or some of both. Plenty of lower power compression units have been sold, many with 1W max output stages, and in practice they don't sound a lot different from the compressors using 10W output stages. If we are chasing high ratio limiting, it's another matter and we want max reasonable power for the side chain.
Thanks Doug,
I'll see how the rest feel about the output tubes as well, but I'm glad we're on the same wavelength here.
Regarding the attack and release times: that was going to be my next question. The 26C/W were broadcast limiters as you say, but this compressor is only going to be used for recording. So the next question is "give me the attack and release times you all want." I can easily get some 6 way Lorlin switches, so let's come up with some ideal times starting with 0.1mS attack? The stock release times seem too long for recording at 5 sec.
By the way, I've discovered a typo error on the drawing, C-109 is 0.5uF on the later drawing making all the figures wrong. But on the parts list its 0.25uF (same as 23C), this makes all the figures correct, especially putting back the original resistors removed for the revision.
best
DaveP
I'll see how the rest feel about the output tubes as well, but I'm glad we're on the same wavelength here.
Regarding the attack and release times: that was going to be my next question. The 26C/W were broadcast limiters as you say, but this compressor is only going to be used for recording. So the next question is "give me the attack and release times you all want." I can easily get some 6 way Lorlin switches, so let's come up with some ideal times starting with 0.1mS attack? The stock release times seem too long for recording at 5 sec.
By the way, I've discovered a typo error on the drawing, C-109 is 0.5uF on the later drawing making all the figures wrong. But on the parts list its 0.25uF (same as 23C), this makes all the figures correct, especially putting back the original resistors removed for the revision.
best
DaveP
On the choice of output tubes, although 6V6s are somewhat more expensive,
the russian eqivalent 6P6S is around 1/3 of the price of a current production 6V6
Still, as this is planned to be a point-to-point build, one could make suitable mods and
use which ever finals/traffos one prefers.
My personal pref right now are TV tubes like tbe 6S4A
Up to 8.5W dissipation for a single triode envelope for around $4.
High plate resistance tho - need a higher input impedance traffo.
Again, Edcor 10W 600ohm output traffos to the rescue for a reasonable price.
For release times, 1s or so is the max I would ever use - maybe 6 steps
Attack would be 0.1ms to around 1ms - again 6 steps or so
Cheers
the russian eqivalent 6P6S is around 1/3 of the price of a current production 6V6
Still, as this is planned to be a point-to-point build, one could make suitable mods and
use which ever finals/traffos one prefers.
My personal pref right now are TV tubes like tbe 6S4A
Up to 8.5W dissipation for a single triode envelope for around $4.
High plate resistance tho - need a higher input impedance traffo.
Again, Edcor 10W 600ohm output traffos to the rescue for a reasonable price.
For release times, 1s or so is the max I would ever use - maybe 6 steps
Attack would be 0.1ms to around 1ms - again 6 steps or so
Cheers
alexc said:
Attack: 0.1 to 1ms is a very small grouping for 6 steps to me. I like slower attacks a lot of times so I'd push for 1 or 2 longer time constants, 10ms and 20ms??
Release: again I'd push for one that's still a little longer than 1 sec. Some days you just can't beat a slow release.
For contrast ... these are the listed times from the Manley Vari
RECOVERY 5 steps: 0.2s, 0.4s, 0.6s, 4sec., 8sec.
Variable ATTACK: 25msec-70msec
> I'm guessing that both of these designs are working in AB2 as in the tube data; in other words the grids are being driven positive
No. Can't "drive positive" with C-R coupling to grids.
Look at the single-tube triode rating: 0.85W. Figure 1.7W for a pair. This may be conservative: I can imagine >5W/pair. But the big thing is that Collins would have MATCHed to 600 ohms. OT may be 5KCT:600. Not a max-power condition, but the output follows simple resistor math.
> The stock release times seem too long for recording at 5 sec.
Some of us don't do pop. I have used 20 sec decay.
No. Can't "drive positive" with C-R coupling to grids.
Look at the single-tube triode rating: 0.85W. Figure 1.7W for a pair. This may be conservative: I can imagine >5W/pair. But the big thing is that Collins would have MATCHed to 600 ohms. OT may be 5KCT:600. Not a max-power condition, but the output follows simple resistor math.
> The stock release times seem too long for recording at 5 sec.
Some of us don't do pop. I have used 20 sec decay.
benlindell
Well-known member
Wow, I can't say I've ever ventured over a 1-1.5 sec release, but I do pop almost all the timePRR said:
Thanks PRR for clearing up the AB2 question.
So the output stage is far weaker than first thought, not a 10~15W output at all, but it still manages very fast attack times.
The published max output is 28dBm which is 19.5V/600 ohms, this is only 0.63 W!
With the 5K:600 OPT it has a voltage ratio of 2.887:1 so the voltage on the primary is only 56.3Vrms.
I shall be doing some calculations later today but it looks like the output stage is only a few watts as you say. This does of course make the output stage easier to design.
Re the attack and release times, it looks like it will need a wide range to cater for all tastes, or maybe people can modify to suit themselves. I'll put in some values that hopefully most will agree with and if possible include the first two position timings of a Fairchild as well.
best
DaveP
So the output stage is far weaker than first thought, not a 10~15W output at all, but it still manages very fast attack times.
The published max output is 28dBm which is 19.5V/600 ohms, this is only 0.63 W!
With the 5K:600 OPT it has a voltage ratio of 2.887:1 so the voltage on the primary is only 56.3Vrms.
I shall be doing some calculations later today but it looks like the output stage is only a few watts as you say. This does of course make the output stage easier to design.
Re the attack and release times, it looks like it will need a wide range to cater for all tastes, or maybe people can modify to suit themselves. I'll put in some values that hopefully most will agree with and if possible include the first two position timings of a Fairchild as well.
best
DaveP
OK I've now done the calculations:
5K opt/4 = class B load of 1250 ohms
B+ is 320V giving 103Vpeak/tube from 1.25k loadline.
2x103x0.707=145.64Vrms
(145.64V)^ /5000=4.24W classAB
Class A load gives 2.2W
You can always rely on PRR to get you in the right ballpark!
best
Davep
5K opt/4 = class B load of 1250 ohms
B+ is 320V giving 103Vpeak/tube from 1.25k loadline.
2x103x0.707=145.64Vrms
(145.64V)^ /5000=4.24W classAB
Class A load gives 2.2W
You can always rely on PRR to get you in the right ballpark!
best
Davep
> The published max output is 28dBm which is 19.5V/600 ohms, this is only 0.63 W!
At what THD?
It was customary to run a large stage and rate it 1/4th or 1/10th of max to get the THD way down.
Whatever the THD number, the actual unit probably does better; some headroom for tube-wear or a near-reject tube.
It was already known that while 5% THD in a single stage (radio output) was tolerable, a long chain (such as any non-minimal broadcast system) of 1% stages gave a haze over the signal. NFB was new and posed new design problems (such as impedance matching), so stages were over-built and under-pushed. (There's also the haze of IM when small NFB is applied, though this was not understood at the time.)
What is the difference between total re-engineering versus just plagiarism? The tube cost seems negligible. The power supply and interstage transformer are not cheap, but neither is your design and breadboard time. If you were making a thousand units for profit you sure would design-out cost, but the likely market is dozens and many of those "love iron". Iron haters already have THAT, Pico, RNC, etc.
At what THD?
It was customary to run a large stage and rate it 1/4th or 1/10th of max to get the THD way down.
Whatever the THD number, the actual unit probably does better; some headroom for tube-wear or a near-reject tube.
It was already known that while 5% THD in a single stage (radio output) was tolerable, a long chain (such as any non-minimal broadcast system) of 1% stages gave a haze over the signal. NFB was new and posed new design problems (such as impedance matching), so stages were over-built and under-pushed. (There's also the haze of IM when small NFB is applied, though this was not understood at the time.)
What is the difference between total re-engineering versus just plagiarism? The tube cost seems negligible. The power supply and interstage transformer are not cheap, but neither is your design and breadboard time. If you were making a thousand units for profit you sure would design-out cost, but the likely market is dozens and many of those "love iron". Iron haters already have THAT, Pico, RNC, etc.
I'm not enough of a design engineer to be able to address the inductance requirements of the transformers associated with the bridge portion, but I do notice the published L values are significantly higher than anything that can be purchased off the shelf today. EDIT: the Hammond 850Q and 835 are suitable, tons of inductance. Is it a simple redesign, or a stumbling block? I'd have to hack around on the bench to know. Can you deliver the related sound without a transformer involved? I'm doubting it on one level, as the shifting loads interaction with the inductance has to be a large sonic contributor. Does this matter? If not, then the other angle would be to develop the most economical bridge compression circuit that could be used with a variety of different make up amps. There are so many weirdo variations out there, what's the hook, and is it a big enough hook? I may be wrong, but recall the Neve 2254 being a bridge circuit variation. I've never looked at it. If so, a study of the parallels and tangents between the various eras could be illuminating.
Hmm, unless we try new approaches then we are condemned to serve out our time as clone slaves; not my definition of DIY.
I have 3 years before I retire and go to finish the renovation of my house in France, I don't know when I will ever get the time to do electronics again, so I'm making the most of it now. PRR you did a great job with the design of your varimu and got a lot of people fired up, I would like to continue the tradition with a bridge based comp. If people like it then great, if they don't then I won't lose any sleep, I will have just had fun doing it and my design time is free at the moment.
I think you are right about the over engineering to achieve low distn. and the haze angle.
Doug,
For a transformer the f-3dB = Tube output Z/6.28xL
When the 12AU7 is fully on its output Z is 15k// with 10K (6K) so the inductance required for 20Hz would be 48H, when the tube is best part off then 15k is more the mark so it needs more like 120H. The original tube was another 6N7 which has a higher rp than a 12AU7, so changing the tube improved the low end and reduced the heater requirement as well. The 12AU7 was also capable of being semi remote, which is why PRR used it too.
I don't think it would be popular coming up with a design that needed $300 interstages. If guys want to clone a 26W or C then they are free to do it, always have been.
best
DaveP
I have 3 years before I retire and go to finish the renovation of my house in France, I don't know when I will ever get the time to do electronics again, so I'm making the most of it now. PRR you did a great job with the design of your varimu and got a lot of people fired up, I would like to continue the tradition with a bridge based comp. If people like it then great, if they don't then I won't lose any sleep, I will have just had fun doing it and my design time is free at the moment.
I think you are right about the over engineering to achieve low distn. and the haze angle.
Doug,
For a transformer the f-3dB = Tube output Z/6.28xL
When the 12AU7 is fully on its output Z is 15k// with 10K (6K) so the inductance required for 20Hz would be 48H, when the tube is best part off then 15k is more the mark so it needs more like 120H. The original tube was another 6N7 which has a higher rp than a 12AU7, so changing the tube improved the low end and reduced the heater requirement as well. The 12AU7 was also capable of being semi remote, which is why PRR used it too.
I don't think it would be popular coming up with a design that needed $300 interstages. If guys want to clone a 26W or C then they are free to do it, always have been.
best
DaveP
Quoting UTC design principles, one wants good response down to 7 Hz to assure low distortion at 20-30 Hz. I have found many transformers that are -3 dB/20 Hz to sound lacking in sub bass, even though the charted losses appear small. FWIW. So, agreed, a $300 interstage would limit popularity greatly, and it may well take one to get response as good as many 50-70 year old units. If there's another path that sounds good, it should be explored.
That UTC principle sounds good to me Doug,
Back to work, I've just done the calculation on a pair of 6AQ5A's triode wired on a 10K load and it works out at 4.68W, so they are a reasonable replacement for the 6F6's. I'm ordering a pair today.
The only downside of these tubes is that the output Z will be slightly higher with a 10K TX at 1250 ohms, the 5K TX on the 26W gives the 6F6's an output Z of ~800 ohms. So the attack time might be slightly longer on the first setting.
best
DaveP
Back to work, I've just done the calculation on a pair of 6AQ5A's triode wired on a 10K load and it works out at 4.68W, so they are a reasonable replacement for the 6F6's. I'm ordering a pair today.
The only downside of these tubes is that the output Z will be slightly higher with a 10K TX at 1250 ohms, the 5K TX on the 26W gives the 6F6's an output Z of ~800 ohms. So the attack time might be slightly longer on the first setting.
best
DaveP
