big high frequency rise when preamp is at low gain

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

plumsolly

Well-known member
Joined
Nov 3, 2007
Messages
969
Location
Denver, Colorado
I was running some frequency tests on my Auditronics 110 console today and I realized that all the channels have a disturbing high frequency rise when they are at their lowest two gain settings. See the graph here:
lowgain1.jpg

When the gain is turned up beyond the first two points, it evens out to a normal looking response. I changed the mic amp chips in most of these to 5534s and some of them I put DOAs in - the symptoms are the same for both. The first thing I did when I realized what was happening was to add a 150/120pf cap across the feedback network - (this is recommended in the Auditronics literature for upgrading to 5534's to limit the bandwidth of the amp stage to prevent high frequency oscillation) This didn't help. Then I added a 22pf comp cap between pins 5 and 8 of a 5534 channel to see if that would help. It did not.

Any thoughts would be really appreciated - thanks, Ben
 
Hi!

It would be useful to know exactly how you're measuring this curve. It might help with troubleshooting.
 
Hi - Thanks for responding. I am sending a sweep into the line in on the channel and measuring the signal off of the direct out - it shows up the same coming off the 2-mix buss as well, though. I am generating the sweep and doing the measuring with this software: http://www.audiotester.de/ - I get the same result with other software. Here is a schematic of the channel:

Thanks a million, Ben
 
Decipher SW2-2. Your last two settings insert 1K in series with the transformer primary. Whatever resonance-control they use to tame this works fine on the other 9 settings, but less perfectly on the last two.

 
Thanks for the insight PRR. I thought it wouldn't make sense that it was an oscillation because the feedback r is the same at positions 1 and 3 and it doesn't exhibit the same symptoms at these positions - so your diagnosis makes sense. I have another revision of this channel (the 110b) with a different pad network at the first two positions and it does roughly the same thing **Edit - No I don't - I have 110a's with 110b faceplates and 110a circuits, so never mind this **Edit - here is that schematic: When you say "resonance control", are you talking about the Zobel network? Do you suppose there is any way to address this problem? - 7db up at 20k is pretty drastic. On a more mysterious note, I have 2 channels that drop off at roughly the same slope in the first two positions?? Thanks a whole heap, Ben
 
For full assessment of the problem, you should check the frequency response in all positions with a 200 ohms source impedance generator. I suspect you have driven the input directly from the too low impedance of your soundcard. You may find that the ringing happens at all settings with a standard 200 ohms load.
The application of a Zobel network is valid only if the source impedance is correct. According to the schematic, the 3 & 4th position should be worse, because the series resistance is increased from 1k to 1k + 220R.
On the 110B, you could experiment putting individual Zobel networks on the parallel elements of the L pad (R9, 10 &11).
Unfortunately, on the 110 there is no such possibility. I suspect you'll have to optimise  certain positions and accept the fact that some others are flawed...
 
Hi Abbey  - thanks for responding. Most of the measurements were done directly from my converter which does not list an output impedance. I did, however try running the output of my converter into one channel on my console with the gain set at the middle/flat position and took the direct out of that channel into another channel set at a problem gain setting, in order to see if the response was different when driven from a transformer-balanced 600-ohm source. I got the same result with that setup. I will look into a hardware signal generator. The line inputs will mostly be connected to the output of my converters for mixing, so I guess I may want to optimize for that source impedance?
Thanks and Best, Ben
 
It is very likely that the actual output impedance of the line output is < 60 ohms. This mixer adopts the bridging principle, not matching. So, in both cases the input has not been loaded properly.
You must determine what is the optimum source impedance either by varying externally the source impedance of your generator, or internally by using the line input and varying R7.
You may find that the Zobel network needs be modified. The only proper and convenient way is to use a square wave gen at ca. 5kHz and an o'scope.
This is a quite lengthy procedure and can be puzzling if you're not familiar enough with the intricacies of leakage inductance and spreaded capacitance. A Spice simulator helps a lot too.
But in fact, I suspect Auditronics have targetted a correct line input frequency response in their design brief... ?
 
Ok - I did some experimenting. First, I found that the mic input does not have this problem - so it is arising from the combination of the pad on the line input and the pad inserted by the gain switch reacting with the transformer. The module I experimented with was a 110a (the first schematic) - I found that raising r10 can start to flatten the response out - which makes sense because when the gain switch puts 220r in series with r10 it flattens out completely. Obviously, if i did this, I would have to raise r11 to get the same amount of attenuation, and then raise r9 because of that - I didn't experiment with this. I did install trimmers for r6, 7, and 8 to see I could rework the pad to make the response flatten out. Changing r6 and r8 didn't seem to have much of an effect besides level. Raising r7 high enough got a flat response at the lower gain settings, but it was ruined elsewhere. I imagine Auditronics went through all this, and settled on what's there. I also tried wiring the input transformer primaries in parallel to see if that would help (figuring i could switch that with the line/mic switch) - no luck really. So now I am looking at bypassing the transformer for the line input. I tried tying + and - together and going straight into the opamp - flat as a pancake. So that is an option if I do not need balanced line inputs - but it may not work as well for sources other than my converters. Another option would be something like this:
Picture1.png

I could also use That Corp balancing chips, or install line in transformers - this is probably price prohibitive, however. Do any of these sound like a good idea? Let me know what you think, Thanks, Ben
 
Have you thought of isolating teh stage that is causing the abberation in frequency response? Try repeating the test with the input at the same point, but probing the signal at the various stages of the circuit (IC1 output, IC2 output etc). You should be able to work out exactly which part of the circuit is causing the problem, rather than making assumptions.
 
Hi Gswan  - I think I have it pretty well narrowed down - it doesn't do it when you bypass the transformer and it doesn't do it when you take out either or both of the pad stages. I have not tried it yet with both pads in and the transformer bypassed, and I will, but I don't anticipate this to be a problem. Best, ben
 
It was not clear in your original post that it was happening on the line input, but in the end, I think the solution may be much easier. It just looks like the impedance of the pad is incorrect (I can't read the values of R6, 7 & 8). You just have to find what source impedance gives the flattest response and use this value for R7, then adjust R6 & 8 for desired attenuation.
 
HI Abbey -
abbey road d enfer said:
It was not clear in your original post that it was happening on the line input
Yes, I'm sorry. I should have made that clear.
abbey road d enfer said:
I think the solution may be much easier. It just looks like the impedance of the pad is incorrect (I can't read the values of R6, 7 & 8). You just have to find what source impedance gives the flattest response and use this value for R7, then adjust R6 & 8 for desired attenuation.
r6 and r8 are 5.1k and r7 is 62r. I tried installing trimmers for r6, 7, and 8 and by raising r7 to 1k (the highest setting on the trimmer) I was able to get the response to start to flatten out at the lowest 2 gain settings, but it adversely affected the response at other gain settings. I think you were right when you said "I suspect you'll have to optimise  certain positions and accept the fact that some others are flawed..." I think Auditronics probably went through this process and decided that a rise after 15k on the first two gain settings was the best compromise. That is not good enough for me however, so if I can't fix it and still go through the mic transformer, I will find another solution: Either dedicated line in transformers, electronic balancing, or unbalanced line inputs - a note on that; some of the Sphere Eclipse consoles (my dream console) had unbalanced line inputs - so maybe its not such a bad/crazy idea to just go unbalanced in right to the opamp.
Thanks, Ben
 
My auditronics 501 has all unbalanced line inputs/returns etc. No problems here.

Or, edcor wsm series transformers are about $11 each, wouldn't cost too much to add them, Im considering it for my board. 
 
So I figured out that I do not, in fact, have any 110b modules. I have some modules with faceplates that say 110b and that have a 110b control layout, but they have 110a pcbs and the circuit is almost exactly the same as my modules with 110a faceplates. So it seems probable that they may have fixed my high frequency rise problem with the 110b revision. Unfortunately, I cannot just change my 110a's to a 110b circuit, because the U-pad structure in the 110b calls for a 3-pole gain switch, and all mine are 2-pole. What I can do is change everything but the pad; mine will be an L-pad version of the pad in the 110b. This will have three benefits: 1) Hopefully I will be able to avoid the pops from the gain switch because I will be switching the feedback>ground resistor, rather than the feedback resistor itself 2) Hopefully I will have better gain staging and less noise (if you look at how the gain setting and pad integrate with one another on the 110a it is just wacky. For instance, at -30db there is actually more gain than at -20) and, 3) The lowest pad setting will be -30 db and that may bring r10 up high enough to avoid the high frequency bump. If am still getting a high frequency bump, I will look into bypassing the transformer for the line input. On a related note, why is it ok to use an L-pad instead of a U-pad even though this is a balanced signal? Is it because the cmrr of the transformer is not affected by the different resistances on + and -? (it may or may not be clear that I have no idea what I am talking about here :))
Thanks, Ben
 
> because the cmrr of the transformer is not affected by the different resistances on + and -?

Not much. Often not enough to fool with an extra switch pole.
 
So I changed the gain/pad scheme in a couple of modules. It's worked very well so far; I have less clicks and pops and the frequency response has leveled out considerably at the lowest gain/pad setting - instead of 7db up at 20khz, I have 2db down - a marked improvement (though still a little irritating, mostly because the rest of the settings are pretty flat) I ran into something funny with the pad, though: I chose the resistors using this formula: Rseries = Rshunt * (k - 1) I left Rseries at 1k and calculated the values for Rshunt for -30db, -20db, and -10db. I got 33r, 111.7r, and 466r (closest resistor values) respectively. When I installed the pad, I found that my -20db and -10db setting were off. I installed a trimmer and found 121.4r got me -20db and 713r got me -10db. Looking at the 110b schematic, it looks like the values for these resistors don't adhere to the formula either. I know what resistors to use now, I'm just curious why the calculated values weren't right. Thanks, Ben
 

Latest posts

Back
Top