RF filter for a transformer input line amp

shop deals on amazon
Advertise with us
Shop deals on ebay
GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

living sounds

Well-known member
Joined
Jul 26, 2006
Messages
3,934
Location
Cologne, Germany
I've modified the inputs on my console per the attached schematic.

AC 2 in this case denotes the output of a 4:1 input transformer.
R4 sets input impedance to 10k ohm.
R3 is the gain pot wired as a variable resistor.

C6 makes sure the circuit is stable, but also adds a bad sheen to the high end (that doesn't show up in THD measurements but is nonetheless unacceptable. Leaving C6 out I get oscillations in the following EQ stage (with gear connected to the balanced input). Increasing R1 to a higher value can again mitigate that, but doesn't sound good either (and affects frequency response). I also don't like the idea of RF getting to the op amp input in the first place.

My guess would be that C6 interacts with the transformer. I tried mitigating that by experimentally increasing R2 and lowering C6, but the result is pretty much the same as with the circuit in the schematic.

So how to you actually add an effective RF filter to a transformer input without it audibly affecting performance? I guess I could add an addtional balanced input stage, but that would be a massive overhaul on an already badly beaten board...


Thanks!
 

Attachments

  • MyConsoleInput.jpg
    MyConsoleInput.jpg
    93.9 KB · Views: 70
living sounds said:
I've modified the inputs on my console per the attached schematic.

AC 2 in this case denotes the output of a 4:1 input transformer.
R4 sets input impedance to 10k ohm.
R3 is the gain pot wired as a variable resistor.

C6 makes sure the circuit is stable, but also adds a bad sheen to the high end (that doesn't show up in THD measurements but is nonetheless unacceptable. Leaving C6 out I get oscillations in the following EQ stage (with gear connected to the balanced input). Increasing R1 to a higher value can again mitigate that, but doesn't sound good either (and affects frequency response). I also don't like the idea of RF getting to the op amp input in the first place.

My guess would be that C6 interacts with the transformer. I tried mitigating that by experimentally increasing R2 and lowering C6, but the result is pretty much the same as with the circuit in the schematic.

So how to you actually add an effective RF filter to a transformer input without it audibly affecting performance? I guess I could add an addtional balanced input stage, but that would be a massive overhaul on an already badly beaten board...


Thanks!
Click to expand...
Transformers are BP devices so already filter out RF.

I would figure out how to measure the "sheen" that bothers you.

If you can measure something you can control it.

JR
 
JohnRoberts said:
Transformers are BP devices so already filter out RF.

I would figure out how to measure the "sheen" that bothers you.

If you can measure something you can control it.

JR
Click to expand...

Thanks John. The "sheen" seems to be phase shift from the lowpass filter...

What capacitance value for C4 would you consider adequate?
 
living sounds said:
Thanks John. The "sheen" seems to be phase shift from the lowpass filter...
Click to expand...
I have never heard phase shift by itself...  Can you correlate increasing/decreasing "sheen" with increasing/decreasing phase shift. LPF are generally more audible for simple frequency response differences.

Probably hard to separate phase shift from frequency response in simple LPF.
What capacitance value for C4 would you consider adequate?
Click to expand...
I don't know...  As I recall Deane Jensen used to publish recommended terminations for his transformers, with response curves. Not a simple RxC or LxC...

JR

{edit- BTW I am not the transformer expert around here... /edit]
 
JohnRoberts said:
I don't know...  As I recall Deane Jensen used to publish recommended terminations for his transformers, with response curves. Not a simple RxC or LxC...
Click to expand...

I know you are the op amp expert - C4 is the compensation cap on the feedback path.
 
living sounds said:
I know you are the op amp expert - C4 is the compensation cap on the feedback path.
Click to expand...
OK you made me look...

5532 op amp is unity gain stable so that capacitor could be as large as you want...  About the only possible stability concern is providing a little lead in the feedback network to overcome lag from stray - input capacitance to ground (at most a few pF). 

I would generally set that feedback pole for half power at 200kHz or so... BUT your resistor values indicate a pretty modest gain (1+1.1/10). It's hard to have a -3dB point when the gain stage only has +1dB of gain... 

So C4 will cause response to asymptotically approach unity gain (from +1dB) not continue to roll off.

JR
 
As John stated, a transformer will knock out any RF entering the box so one shouldn't be needed here.

At first blush, if this is a 4:1 transformer with a design centre input impedance of 10K ohms, then your R4 value needs to be around half of the shown value, more like 600 ohms.  And whatever else you want or need in parallel to tame any overshoot.   
I would get the transformer operating nicely on its own. 
You shouldn't rely on op-amp compensation for taming transformer issues.
Are you using a 5534 and external compensation or is it half of a 5532? (edit - just looked again and see it's 5532) Regardless, use whatever the data sheet specifies for stability at a non-inverting gain of 2 is what I'd use as the C in // with your feedback R. 
I think the pot should be a 1K rev log, not a 10K btw. 

Edit: C3 could be raised somewhat for an easy try and see.  Make R4 620, and C3 1800pF or so.
 
Winston O'Boogie said:
As John stated, a transformer will knock out any RF entering the box so one shouldn't be needed here.

At first blush, if this is a 4:1 transformer with a design centre input impedance of 10K ohms, then your R4 value needs to be around half of the shown value, more like 600 ohms.  And whatever else you want or need in parallel to tame any overshoot.   
I would get the transformer operating nicely on its own. 
You shouldn't rely on op-amp compensation for taming transformer issues.
Are you using a 5534 and external compensation or is it half of a 5532? (edit - just looked again and see it's 5532) Regardless, use whatever the data sheet specifies for stability at a non-inverting gain of 2 is what I'd use as the C in // with your feedback R. 
I think the pot should be a 1K rev log, not a 10K btw. 

Edit: C3 could be raised somewhat for an easy try and see.  Make R4 620, and C3 1800pF or so.
Click to expand...

Perhaps TMI but 5534 was only stable down to closed loop gain of 3x w/o over compensation. The 5532 as already shared is unity gain stable.

JR
 
JohnRoberts said:
Perhaps TMI but 5534 was only stable down to closed loop gain of 3x w/o over compensation. The 5532 as already shared is unity gain stable.

JR
Click to expand...

Yes exactly.  I looked at the scheme quickly and mistakenly read 5534 but no external comp.  which is why I asked while writing.  I looked again and saw it was 5532 which, as you say, is unity gain stable.  I then recommended that he use whatever it says to use on the data sheet which as you said, could almost be anything. 

I think it's possibly an issue with a bit of ringing from the transformer getting things upstream a bit excited.
Again, I would look at taming any ring with a proper load R rather than what is shown (I say 620R or so), then a zobel if needed, or a shunt C of bigger value than currently shown.
Cheers
 
You want the RF filter on the other side of the transformer,  where the XLR or TRS connector is.  Two capacitors,  one on the hot input and one on the cold input,  both going to chassis ground. There are more elaborate schemes,  but this is simple and often good enough.  100p is typically fine.

Get rid of C6, either nothing or make it an appropriate zobel network.

Do you even need U1 at all? How much gain / attenuation do you need? Also remember that U1 needs to partially drive R1, it will do 1.1k, but won't be that happy about it.  Something like 4.7k would be much better for a NE5532 / 34 line stage.
 
john12ax7 said:
Get rid of C6, either nothing or make it an appropriate zobel network.
Click to expand...

Maybe.  Maybe not.  Neve did it that way on all the desk line inputs and no one complained.  It'll work, but so will the right secondary R or a zobel.  Might depend on the transformer and your mood that day I guess.

But with the value as is, if you Thevenise a typical low impedance of 50ish ohms connected to the input, it reflects down to practically nothing on the secondary so the series resistance and impedance is a mix of that plus DCR etc and ends up not a whole lot of series R to feed a 680p or whatever cap he has now.  Put a higher value C in like Neve and it may just do the trick.
 
john12ax7 said:
Do you even need U1 at all? How much gain / attenuation do you need? Also remember that U1 needs to partially drive R1, it will do 1.1k, but won't be that happy about it.  Something like 4.7k would be much better for a NE5532 / 34 line stage.
Click to expand...

Depends on what the next stage load is, but a 5532 is quite happy driving down to 820 ohms (they say down to 600 ohms but 820 is where the  limit for best design practice is I think). and a bigger value feedback R would have a noise penalty.
 
Imo it's best to shunt the RF to chassis as soon as possible, but as they say ymmv.

For ICs, a lot tend to degrade performance when you drop below a few k load, Samuel Groner did some very nice work measuring a bunch of them. The noise / distortion trade off would be worth it for line level imo.

This is definitely worth a read / download (opamp distortion file)
http://www.nanovolt.ch/resources/ic_opamps/

In looking at it again now I also notice that NE5532 common mode performance is rather poor, so probably not the best choice for a non-inverting stage.
 
john12ax7 said:
Imo it's best to shunt the RF to chassis as soon as possible, but as they say ymmv.
Click to expand...

Yes of course, I agree.  :)  I was just doubting that any RF is getting through the transformer whatever brand or type it is.
My bigger shunt C was something I suggested as something which would bring down the very upper end and help tame any possible overshoot.
I do think a zobel would be more elegant  if you have the means to get the right values.
Me, I put a R on that works the best these days and generally don't bother with zobels as I can't hear high end ringing anyway. 

I've read Sam Groner's paper (and his other in reply to Self on Amplifier Design,  both are good and very useful).  I would also recommend Doug Self's 'Small Signal Design'  in regard to noise and distortion in mixer design. 
 
It is usually best to stop RFI before it reaches the transformer. Often you don't need it as the transformer does a very good job unless you are in a studio right next to a 1KW
transmitter.

The circuit seems odd to me. The 4:1 transformer drops the level by 12dB. Then you fo into an inverting op amp with a 1K feed back and 10K input resistor which drops the signal another 20dB. I know you are altering gain by changing the 10K at the input but this is a really bad way to control gain..

Cheers

Ian
 
ruffrecords said:
The circuit seems odd to me. The 4:1 transformer drops the level by 12dB. Then you fo into an inverting op amp with a 1K feed back and 10K input resistor which drops the signal another 20dB. I know you are altering gain by changing the 10K at the input but this is a really bad way to control gain..

Cheers

Ian
Click to expand...


Hey Ian,
it's just drawn a little strangely, upside downey if you will.
Unless I completely got the wrong end of it, it's a non-inverting amp stage, the non earthy side of the secondary goes to +tve in of the 5532. 
Gain is less than 2 as shown, although I think that for best control on the pot, the value should be 1K rather than 10K.  As is, too much of the pot resistance will be getting the gain up to 2 or 6dB.
Anyroad...  Whatever works  :)
 
Thanks guys!

The termination value of 1.1k is correct after all, since the transformer actually has a 3:1 turns ratio.

When I originally did the mod I found the non-inverting circuit to work better, but I can't remember why. I use Signetics 5532, which have quite a bit lower common mode distortion than the TI counterpart, but yes, it's not ideal.

The preamp is followed by a voltage follower/active lowpass filter, so no significant additional load. I could change the feedback resistor to 4.7k without any ill effects. I'd like to keep the 10k pot, since that's what the console came with originally.

I could change the circuit to inverting simply by moving the ground connection from the - to the + input - would you recommend that?

I'll also add the 100p caps on the input of the transformer (which is mounted off the channel on the input board).
 
Then yes, 1.1K would be about right for a 3:1.  Still not much that your 680pF is doing at the moment given the reflected and secondary DCR etc.
Adding in a bit of series R where you show an R of 0 ohm would be OK and not much of a noise penalty really. 

What is the desk, is there a schematic anywhere?  What was there before for line in?  That sort of stuff etc.
If the line input was happy before, there's no reason it can't be now with a transformer there instead. 

How much butchery do ya wanna do?  ;D

Edit:  If the transformers are mounted off board a ways, that's also a round-the-houses way for the 5532 to get a ground reference for the positive in.  Ideally, the feedback and ground return are the same value but since you already have that 1.1K off-board transformer load, see what happens if you put a 12K or so R at the 5532 +tve input.  Maybe a bit of series R as well and the 680p will do it.
 
Apologies, I misread the schematic. Confused by ground not being at the bottom but I see now  it is non-inverting. I would leave it that way, it will be quieter.

Cheers

Ian
 
Winston O'Boogie said:
Then yes, 1.1K would be about right for a 3:1.  Still not much that your 680pF is doing at the moment given the reflected and secondary DCR etc.
Adding in a bit of series R where you show an R of 0 ohm would be OK and not much of a noise penalty really. 

What is the desk, is there a schematic anywhere?  What was there before for line in?  That sort of stuff etc.
If the line input was happy before, there's no reason it can't be now with a transformer there instead. 

How much butchery do ya wanna do?  ;D

Edit:  If the transformers are mounted off board a ways, that's also a round-the-houses way for the 5532 to get a ground reference for the positive in.  Ideally, the feedback and ground return are the same value but since you already have that 1.1K off-board transformer load, see what happens if you put a 12K or so R at the 5532 +tve input.  Maybe a bit of series R as well and the 680p will do it.
Click to expand...

The transformers are mounted inside the console chassis, but at the input PCBs. The 1.1k termination resistors are on the channel PCBs close by to the input op amps. I have removed the direct ground path from the input PCB to the channel PCB in favour of a single ground path for every channel to a star ground point at the console center to reduce hum (although there is a second connection from each channel's star ground point to each channels chassis).
 

Similar threads

z11111
Driving transformers to play nice with audio interface line outs & line ins
2
Replies
24
Views
3K
Cqwet Dbdfte
Cqwet Dbdfte
K
Adding RF filtering to 'too-simple' input.
Replies
4
Views
402
k brown
K
S
6V6 Output Transformer NFB resistor mod / value
Replies
14
Views
660
ruffrecords
ruffrecords
M
Qs re Helios 2182 —> 22113 Microphone Preamp
Replies
11
Views
6K
dogears
D
V
Design an analog K-weighting filter
Replies
3
Views
2K
mbn
mbn

Latest posts

Back
Top