Submixer/Monitor Controller - Comments & Criticisms Please

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

ej_whyte

Well-known member
Joined
Nov 12, 2010
Messages
263
Location
Cambridge, UK
So here is my first proper design, which I would like some feedback on from the great minds here. It is nothing groundbreaking though.

The idea for this was initially conjured up whilst in my 2nd year of my degree. We were asked to submit preliminary ideas for our final year major project, and this is basically a cut-down console master section, including a monitor controller. The key feature I wanted to have was the ability to route different groups to different types of summing make-up gain, and keep separate inserts for each group bus. The angle behind this was for small project/home studios without a console (I basically just thought up exactly what would suit my own situation :) )

The signal flow is as follows, showing 1 side of the hard-wired stereo only (16 inputs).

blockdiagram.png


Each gain stage can have be set between roughly 3 to 30 dB of gain, and also has an output fader (post insert).  This is enough that if all you want all 8 channels sent to one gain stage, there is still more than 10dB in hand.

The solid state gain stage is pretty much a standard 312 style circuit, with the output fader post-transformer.

The tube gain stage is a simple common cathode single stage which feeds into a 5532 voltage follower to keep the output impedance down for the 1646-based send. For tube purist this may be blasphemy, but I didn’t want to use an output transformer, mainly due to cost (already 6 transformers being used). Whilst I could have used a cathode follower, as most of the ‘tube’ sound comes from the common cathode stage, and the cathode follower being a pretty transparent topology, I chose to go for solid state for cost and board space. Besides, the audio has already been through many opamps so what is one more. Internal level is set at -2dBu nominal. The tube stage has a max input of about -4dBu, and when traced back to the main inputs that is about 19dBu. That seems quite low, however that is with tube gain set to full. At system unity it would increase to roughly 33dBu. The unit is designed to sit right on the converter DA outputs, so I was aiming for 18dBu at a minimum. With max input and max gain, the following opamp stage would be clipping before the tube anyway, so I do not see the tube stage as a bottleneck.

THAT 1646 & 1206 sends & returns throughout, except for the solid state gain stage, where the 2503 transformer is used as it is a key part of the API sound.

Bypass switch chooses between the summed outputs of the gain stages or just the first LR pair of inputs and then feeds this to the master insert.

From there it is sent to the main output fader and 1646-based outputs. The monitor controller section can then be fed either from the main outputs, or an aux in (front panel 3.5mm jack or rear panel RCAs). One interesting feature I have included is that when the monitor controller is fed from the aux input, the summing section can still be used independently (providing that it in not bypassed). The monitor controller is not overly complicated, including mute, dim, mono, main/alt outputs, a headphone output and obviously a volume pot.

The VUs are fed from whatever is hitting the monitor controller, and are fed by an adaptation of JLM’s meter driver.

I have used calibration trimmers throughout to maintain stereo balance.

I know that this is not the most pure or elegant design, for example, the whole summing section could be done with just one opamp and would be much lower noise, but that would then be missing the point of the main features I wanted to include.
Likewise, I could have used a circuit such as Ian’s mu follower for the tube stage, but I wasn’t looking for ultra-clean. I don't expect the noise performance to be amazing as there is quite a lot of attenuation at the start.

I am sure my inexperience will shine through here, but I have been through the design checking things over and over again and I am pretty happy with it now. It would have probably been a lot quicker and easier if I had just posted on here with what I wanted to build, but I wanted to learn how to design it myself not just follow instructions.

I’m sure there is loads I have forgotten to say or explain, but I have been working on this for a long time so there must be good reasons for most things I have done.

I am just looking for some feedback please and for all you more experienced people to point out my miserable mistakes lol.

Finally, massive thanks to Tom Waterman for helping me start out with this idea, as well as Abbey, JR, Ruffrecords Ian, PRR, NYDave, Ethan and everyone else who has posted extremely helpful information on this forum. I’ve only been in to this stuff about a year and a half, but after building my first SSL its all happened pretty quickly. I’m just about to finish my years placement at an Avid/Mackie/R.N.D service centre which has been an amazing year, but I would never have been able to get that opportunity without this place being around.

Schematics:

https://www.dropbox.com/sh/3vueak9kkvm911k/WTf4aLYfcN/Complete%20Edited.pdf

Cheers
 
Well that was a disappointing response lol  :-\ I guess it is bigger than the average mic pre schematic and looking through it is quite a bit of effort.

I'm going to start the board anyway.
 
Cool project.

I didn't go through everything, just eyeballed around. But one stage in particular raises a question. That's the tube and the question is "why?".

12AX7 is just about the most linear and boring tube you could choose. With 300V B+ it's  like having no tube at all, pointless. You won't be able to saturate it the way it's set up, and even if you could, it would sound harsh. Save the tube experimentation for some future project, and never ever do anything with a 12AX7. You will only be disappointed.

I suggest you stick to solid state for the whole project. It will simplify the PSU greatly as well. That DOA stage with a transformer alone looks like a good idea, and will sound much more interesting. Maybe duplicate that instead, and use some other exotic DOA?

In general the routing of the project looks good.
 
Sorry, only just notcied your post.


Just so I understand this right, you passive balanced mix four inputs into a 1:10 mic transformer which is then buffered. I reckon your bus loss with the values shown is close to 30.5dB so your transformer output is around -4.5dB relative to the input. It then goes through some no gain routing and on the the semi/tube gain stages.

You say the max input level to the 12AX7 is -4dBu. By my calculator, -4dBu is about 1.15 volts peak which is plenty to drive the average 12AX7 into grid current producing severe nasty sounding distortion. To avoid grid current an obtain the regular euphonic tube distortion you probably want, I would recommend this level be reduced by 6dB. Alternatively use a lower mu tube which would require greater bias and could hence take a larger signal without grid current.

Cheers

Ian
 
ruffrecords said:
Just so I understand this right, you passive balanced mix four inputs into a 1:10 mic transformer which is then buffered. I reckon your bus loss with the values shown is close to 30.5dB so your transformer output is around -4.5dB relative to the input.

The transformer is wired for 1:5, so I calculated the levels as follows

Loss from summing four channels: -12.04dB
Loss from 680R shunt: around -18.5dB (so yes around 30.5dB bus loss)
Gain from transformer: +13.98dB
Overall stage gain relative to input: -16.92dB

I'm not sure how you got -4.5dB though unless i'm being stupid. With the transformer wired for 1:10 it would give 20dB of gain, making the output around -10.5dB relative to input no?

Regarding the tube section, I guess my inexperience shows! I have never actually built any tube based equipment, but I did quite a lot of reading up on it before attempting to design that stage.

By my calcs nominal level to the tube would be -12.56dBu (0.52Vp-p), whilst the absolute peak would be -4dBu as previously mentioned. However, this is the level at the input of the tube stage, not the tube itself. As the 'gain' is set by a stepped input attenuator and is designed with 10dB in hand, with the stage set for system unity this would actually be reduced down to -22.56dBu nominal (0.16Vp-p), -14dBu peak (0.44Vp-p).

I have biased the tube for around -1.8 volts, so thought that would give enough headroom for 1.8Vp-p before grid current at 0.9V (1.8 volt bias to 0.9v for grid current makes 0.9v for half the wave form, 1.8Vp-p?).

ruffrecords said:
To avoid grid current an obtain the regular euphonic tube distortion you probably want, I would recommend this level be reduced by 6dB.

If you're hitting the inputs with 19dBu and keeping the tube gain at full system +10dB then you must be deaf is my thinking. Keep the gain at system unity and you should be alright hitting it with 19dBu (i think). So in 'normal operation' I think it will be alright, but yes if you absolutely smash the inputs and have full gain then it can distort.

Kingston said:
12AX7 is just about the most linear and boring tube you could choose. With 300V B+ it's  like having no tube at all, pointless. You won't be able to saturate it the way it's set up, and even if you could, it would sound harsh. Save the tube experimentation for some future project, and never ever do anything with a 12AX7. You will only be disappointed.

I had read that the 12AX7 is linear, but I thought that using such a simple stage, as opposed to something more linear like a mu follower would counteract that a bit. Do you have any suggestions of a better way to do this?

I am going to make the board pretty modular anyway, so I can just bolt in a different tube stage if it all sounds pear shaped.

But take all that with a pinch of salt, as I said, I have no tube design experience except for what the books say.

Thanks for the comments.
 
Ahh I have also just realised that I forgot to mention the 6dB loss from the routing stage. As there is the usual 6dB loss from summing the two group buses when the blend mode is on, I have put in a 6dB for when the blend is not engaged also, so that the level doesn't jump. User-friendliness over absolute performance in this case.

That means that nominal input to the gain stage would actually be around -18.5dBu at full gain, -28.5dBu with gain set for system unity. Peak input would still be around -4dBu full, -14dBu unity as explained in my last post.

I will draw up a system levels diagram tomorrow to clear this up.
 
ej_whyte said:
I had read that the 12AX7 is linear, but I thought that using such a simple stage, as opposed to something more linear like a mu follower would counteract that a bit. Do you have any suggestions of a better way to do this?

Simple yes, but still awfully linear. Less than 0.1THD%. The saturation point is very abrupt - almost transistor like. It doesn't have any "euphonic tube distortion". Cathode is unbypassed, and you have negative feedback happening. THD should be very low indeed. This is a workhorse high impedance output very high gain stage, no good for color.

The fix is somewhat simple. First pick some other tube. Optimise cathode bias and plate resistor for the new tube and finally bypass the cathode resistor (100-220uF electrolytic). ECC88, ECC99, perhaps 6SN7, maybe EF86 in a triode configuration. Or any of their russian equivalents. And most importantly, lower the B+ voltage significantly. For a simple stage intended for color 200VDC is a better starting point, but B+/cathode bias/saturation is a good place for experiments.

It's still somewhat pedestrian for color, but already much more interesting than the original invisible or harshly distorting stage. You should consider a transformer output form the tube, just like you already did for the DOA. What you will end up with is a line driver you can easily saturate.

Happy hunting!
 
Kingston said:
ej_whyte said:
I had read that the 12AX7 is linear, but I thought that using such a simple stage, as opposed to something more linear like a mu follower would counteract that a bit. Do you have any suggestions of a better way to do this?

Simple yes, but still awfully linear. Less than 0.1THD%. The saturation point is very abrupt - almost transistor like. It doesn't have any "euphonic tube distortion". Cathode is unbypassed, and you have negative feedback happening. THD should be very low indeed. This is a workhorse high impedance output very high gain stage, no good for color.

The fix is somewhat simple. First pick some other tube. Optimise cathode bias and plate resistor for the new tube and finally bypass the cathode resistor (100-220uF electrolytic). ECC88, ECC99, perhaps 6SN7, maybe EF86 in a triode configuration. Or any of their russian equivalents. And most importantly, lower the B+ voltage significantly. For a simple stage intended for color 200VDC is a better starting point, but B+/cathode bias/saturation is a good place for experiments.

It's still somewhat pedestrian for color, but already much more interesting than the original invisible or harshly distorting stage. You should consider a transformer output form the tube, just like you already did for the DOA. What you will end up with is a line driver you can easily saturate.

Happy hunting!

OK thanks for the info, I was under the impression that EF86 in triode was pretty linear though? I'll look in to it more later.

System level diagram attached for clarity.

Cheers
 

Attachments

  • Levels2.png
    Levels2.png
    85.3 KB · Views: 22
ej_whyte said:
OK thanks for the info, I was under the impression that EF86 in triode was pretty linear though?

They all can be very linear, especially at high B+. EF86 is indeed somewhat a different sound compared to the others I mentioned and needs much higher step-down transformer, if you want to steer that way. (more third order distortion, the others like to distort in invisible 2nd order).
 
The thing about triodes is their distortion is directly proportional to output level (all other things remaining equal). From you system diagram I see the nominal level into the tube stage is around -22dBu with a corresponding output of around +8dBu. Almost any triode will give fairly low distortion at this sort of output level. Assuming your maximum signal is 18dB above these levels, you do have the option to drive the tube harder and attenuate the output correspondingly. Distortion in this sort of circuit is also load dependent so you can increase the distortion by increasing the load (lowering its resistance). For example, if we take your circuit as it stands and just decouple the cathode with 220UF, a simulation indicates that a 0.5V peak 1KHz input gives +26.8dBV output and 1.4% distortion. If you remove the 1Meg load the distortion drops to 1.2% and the output goes up 0.5dB. Increasing the load to 160K increases the distortion to 2.5% and the output drops to 24.8dBu. Of course you should always take tube simulation results with a pinch of salt but the basic principles are right. One way to exploit this is to implement a sort of drive/gain control that operates on both the input and the output of the tube stage so as to vary the operating level of the tube and hence its distortion

As has been mentioned, the 12AX7 is a fairly linear tube, in some cases even more so than my favourites the 6SN7 and 6CG7. For your purposes it is probably best to select a tube with a greater intrinsic distortion. Unfortunately, the tube most renowned for its distortion, the 12AU7, has a mu less than 20 so it cannot alone produce  the gain you need. An alternative strategy might be to look at topologies that produce plenty of gain and distortion. I would expect a 12AU7 cascode for example to produce relatively high levels of distortion.

And I have absolutely no idea how I came up with -4dBu. Arithmetic is not my strong point.

Cheers

Ian
 
From you system diagram I see the nominal level into the tube stage is around -22dBu with a corresponding output of around +8dBu. Almost any triode will give fairly low distortion at this sort of output level.

The levels diagram is just relative to whatever input is supplied. With a +4dBu input the level to the tube stage would be around -18dBu, +12dBu output.

I think part of the problem is that I don't actually know how much distortion is the right amount. I want enough that you can differentiate that it is a tube stage, but not too much as it is on the master bus, but I don't really have a clue what that relates to in %.

What simulator are you using Ian?

I'm not sure what to do now, there are too many options!

Thanks
 
That's the 64K dollar question - how much distortion is enough? I guess the answer is as much as sounds good to you. The golden eared brigade will claim to be able to detect values well below 1% but there is plenty of research to indicate that most people find it hard to hear 5% provided it is mostly 2nd and 3rd harmonic and decreasing the higher the harmonic. What kind of distortion sounds 'euphonic' is even harder to define but my view is that the only harmonic distortion that is likely to be so is 2nd because it is musically an octave so it changes the timbre of a sound but not its musicality. 3rd harmonic distortion is effectively a musical 5th which clearly alters the chord being played as do all higher odd harmonics. This probably explains why triodes are popular because they generate mainly 2nd and 3rd harmonic distortion and generally the 3rd harmonic is about 20dB below the second until you hit them really when when the 3rd and higher harmonics start to rise closely followed by the awful grid current distortion.

The simulator I am using is LTspice. It is free, runs on Windows and Linux (with Wine), and is well supported with tube models. I said before that you should take tube simulation with a pinch of salt, especially when it comes to distortion measurements. The models I use are generated direct from the manufacturers data sheets so you would expect them to be accurate. Despite that, there is a significant difference between modelled and measured distortion under the same conditions. This is not, as you might suspect, due to differences in tubes made today and yesteryear - most NOS and current production tubes of the same type measure very much the same but they all differ from the models generated from data sheets. One can only conclude that the data sheets are not that accurate. Anyway, the bottom line is there is no substitute for measurement and listening.

What to do now? Build, test, listen, rebuild, retest, listen....  you get the idea.

Cheers

Ian
 
OK, I have done quite a major redesign. Decided to drop the tube section for now due to lack of experience, i'll save that for another time. Dropping the tube has simplified it quite a lot, as I can now use virtual earth for both gain stages. This also meant that I was able to change it so that every input can be either routed to group 1 or group 2, and the two groups can then be blended together. The downside of using 'active summing' is that I loose the input transformers, so every input is now debalanced by THAT 1200 before any routing. Each gain stage will be the same on the PC board, with the option of either TX output or THAT 1646. I intend to stuff one side with 2520 + 2503, and the other with 990C + 1646.

Maybe it is less interesting with the tube stage (and input transformers), but oh well, it should save a lot of money in the PSU and input transformers, but it has allowed me to enhance the routing functionality and I hope there will still be quite a sonic difference between the 2520 and 990C stages. I will also make the board modular, so that I can add a tube stage at a later stage if I really want to.

Updated schematic linked below. I will be starting the PC board soon.

http://dl.dropbox.com/u/7095191/Complete%20v2.pdf
 
Back
Top