How to lower source impedance for output transformer

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

PhilipMarlowe

Well-known member
Joined
Nov 28, 2019
Messages
96
Hello there.
I’m trying to balance an unbalanced monitor output which is 20ohms. The output trannies in question are the Lundahl 1539s which according the data sheet are best driven with a source impedance of negative 40ohms ( I didn’t know that’s a thing)

i wired them once as 1:2 and 1:1 with the latter giving me the best high frequencies response. The transformer sounds great and punchy however it totally kills any lower frequency and sounds like a low cut is applied.
Could this be because of the impedance mismatch? And if so: how would I go about lowering the source impedance of my monitor outs to -40 ohms without having to buy a new pair of transformers?

I’m a beginner so it might be that I mixed up some concepts of engineering just in case anyone reading this is confused.

thank you frens.
 
What is the load?

The negative source impedance refers to negative feedback drive (or perhaps more commonly referred to as negative impedance converter). It requires extra somewhat delicate circuitry. That circuit is actually listed in the datasheet. In particular you can see pin 4 going back to the + input of the driving amp. But you don't want to mess with that and that transformer will work just great without it. A negative impedance converter is special drive circuit arrangement that yields lower distortion at higher levels from smaller output transformers. It's a cool trick but not required.

A 20 ohm source should be able to drive that transformer provided that the downstream load isn't too low. If you're getting low frequency loss that makes me wonder what the load is. Is it a conventional 10K input? Is it low Z headphones or what?

Also, just because the source says it's 20 ohms, that doesn't mean it can provide a lot of current. If it's an old TEAC desk with a weak op amp it could be that it can only provide +- 10mA and before it starts to strain and loose level and distort and generally act badly. If that's the case, you could try wiring for step down 2:1.
 
As squarewave says, the load you are driving is an important factor. The other factor that would affect low frequency response is the value of output capacitor used in the monitor output. Telling exactly what is driving the transformer would help us to help you.

The other thing that affects low frequency response is the primary inductance of the transformer - which unfortunately Lundahl does not publish. It is possible that this is a typical modern low inductance transformer designed to be driven by a negative ipedance op amp which means it may perform poorly in a more conventional topology.

Cheers

ian
 
What is the load?

The negative source impedance refers to negative feedback drive (or perhaps more commonly referred to as negative impedance converter). It requires extra somewhat delicate circuitry. That circuit is actually listed in the datasheet. In particular you can see pin 4 going back to the + input of the driving amp. But you don't want to mess with that and that transformer will work just great without it. A negative impedance converter is special drive circuit arrangement that yields lower distortion at higher levels from smaller output transformers. It's a cool trick but not required.

A 20 ohm source should be able to drive that transformer provided that the downstream load isn't too low. If you're getting low frequency loss that makes me wonder what the load is. Is it a conventional 10K input? Is it low Z headphones or what?

Also, just because the source says it's 20 ohms, that doesn't mean it can provide a lot of current. If it's an old TEAC desk with a weak op amp it could be that it can only provide +- 10mA and before it starts to strain and loose level and distort and generally act badly. If that's the case, you could try wiring for step down 2
What exactly do you mean with load? The monitor outputs are 0dBu and 0 dBu ≣ 0.775 V with the output impedance is 20 ohms according to the manual. I*m not sure what this means though. Per Lundahl was saying 20ohms is quite low for monitor output- could that be a mistake in the manual and the actual impedance is quite alot higher? He also said it's strange how much low frequency seemed to be sucked out if it's just 20 ohms.
The mixer in question is a vintage austrian broadcast mixer Acousta P100 which they say is a clone of a Studer 961, or at least heavily inspired by it. I attached a pic of the monitor schematic floating around. The mixer has a master output which is transformer balanced, however in order to use 2 extra stereo auxs and tape inputs, I need to mix through the monitor output. Which has led me to the idea of transformer balancing the monitor output. I found a good deal on a pair of used 1539s and sought them out because they are in some of the best sounding gear I know (for instance SPL Mixdream). Little did I know that the impedances may not match up.

The datasheet also mentions balanced drive, so it may not be a good choice for unbalanced to balanced conversion.
I wrote Per Lundahl a message and he said for simple balancing it works with a 1:1 wiring, which I tried. High end and mids sound great! Only the low end has heavily diminished.


As squarewave says, the load you are driving is an important factor. The other factor that would affect low frequency response is the value of output capacitor used in the monitor output. Telling exactly what is driving the transformer would help us to help you.

The other thing that affects low frequency response is the primary inductance of the transformer - which unfortunately Lundahl does not publish. It is possible that this is a typical modern low inductance transformer designed to be driven by a negative ipedance op amp which means it may perform poorly in a more conventional topology.

Cheers

ian
I attached a picture of the schematic- does that help determine the value of the capacitor? Translated from german it says: "the monitor sum is made up of both Summing amps U2, whose negative feedback can be changed with two FET switches. There's an additional dim function of 16db accessible through a button. "
Unfortunately the schematic is a bit scrambled through a HTML bug.
 

Attachments

  • Screenshot 2021-05-08 at 02.42.13.png
    Screenshot 2021-05-08 at 02.42.13.png
    484.5 KB · Views: 94
By load we mean what the output is connected to. Meaning what the output is "loaded down" with. If the output is connected to a common 10K input, that is a 10K load and is considered to be a "light" load because it is easy to drive. But some older gear is 600 ohms which is harder to drive and is considered to be a "heavier" load. If you tried to connect an 8 ohm speaker to the output, a basic op amp circuit would not be able to drive it much at all.

Notice that there is a 100 ohm "buildout" resistor inside the feedback loop. That is also going to limit the current. So the output impedance might be 20 ohms (although the schematic shows 10 ohms) but if you connected the output to something that is 600 ohms that might actually be too "heavy" for the simple op amp output. Actually because of the buildout resistor, it will probably keep the op amp from distorting too much but it will limit the voltage swing. The op amp output probably cannot put out more than about +-12V. But with 100R and then 600R as the load, thats 12V/700R=17mA*600R=10.3V.

Having explained all of that, it should be able to drive just fine without low frequency loss. Especially if it's a 10K load. The output capacitor is 22uF which is more than good enough to get low frequencies. So honestly it's not crystal clear why you would be loosing lows.

How did you hook up the transformer. Draw a schematic, take a picture with your phone and post it. Make it look like the attached schematic and LL1539 datasheet wiring schematic but with your wiring bits added.

And of course now tell us what the load is. Are you going directly into a USB audio interface? And what software are you using to see the frequency response? Post a screenshot of the response ...
 
Hi
The 22uf might be 'adequate' when there is a 10K 'load' (input to your monitor amplifier/soundcard)
it is way too 'small', for a variety of different reasons if you wish to pass low frequencies with low phase shifts especially if you are introducing the LL1539.
I happen to have been testing (post repair) a Soundcraft B800 stereo master output module that has LL1539 transformers as an option and they use 500uF capacitors (well, 2 X 1000uf in series) and the transformer gets wired 1:1 using the two primaries in series and the two secondaries in series. the response is 'flat' down to approaching 10Hz although I forget exactly what it was now. It was also only minimally influenced by running into 600 Ohms. If the capacitor is too low a value there is always the possibility of 'ringing' (resonance/peaking) depending on the capacitance, the inductance of the transformer and how much load is on it. I actually tested this a fair while back with a Sowter output transformer which was being fed through 100uF capacitor. I think there was a nice 3db 'peak' at a few tens of Hz which then behaved when increasing the capacitor to 1000uF.
Depending what the op amp is in your circuit it could well run into internal current limiting and of course the 100 Ohm resistor will provide some 'safety' current limit.
Matt S
 
22 µF is incredibly inadequate to drive any transformer, even a high inductance one like a Jensen. Putting it in series with 10 Ω almost seems like a joke - the reactance of the capacitor will be the same as the resistor at about 700 Hz (and there's where your bass is going!). Make that capacitor at least 10 or 20 times larger ... and then hope the op-amp has the current capability to drive the transformer to high output levels.
Incidentally, you may be disappointed with the result if you're trying to create a symmetrical (equal magnitude, opposite polarity) output signal pair using an output transformer. Into typical high-impedance balanced inputs, the inter-winding capacitance of any output-type (no Faraday shield) transformer will have the same symmetry as on the primary. But symmetry is not the same as balance ... which I've tried to get the world to understand for decades.
Bill Whitlock
AES Life Fellow
 
22 µF is incredibly inadequate to drive any transformer, even a high inductance one like a Jensen. Putting it in series with 10 Ω almost seems like a joke - the reactance of the capacitor will be the same as the resistor at about 700 Hz (and there's where your bass is going!). Make that capacitor at least 10 or 20 times larger ... and then hope the op-amp has the current capability to drive the transformer to high output levels.
Incidentally, you may be disappointed with the result if you're trying to create a symmetrical (equal magnitude, opposite polarity) output signal pair using an output transformer. Into typical high-impedance balanced inputs, the inter-winding capacitance of any output-type (no Faraday shield) transformer will have the same symmetry as on the primary. But symmetry is not the same as balance ... which I've tried to get the world to understand for decades.
Bill Whitlock
AES Life Fellow

I must confess that I don't understand magnetics as well as I would like. But why does the higher cap impedance cause LF loss? I could see how the smaller cap value would cause LF resonance but if I just do a simple RCL calc with 10R, 22uF and 1H, the corner is 34Hz. What am I missing? Leakage inductance?

Now the inter-winding capacitance symmetry vs balance I understand perfectly. (just kidding - I have no idea what that is)
 
I must confess that I don't understand magnetics as well as I would like. But why does the higher cap impedance cause LF loss? I could see how the smaller cap value would cause LF resonance but if I just do a simple RCL calc with 10R, 22uF and 1H, the corner is 34Hz. What am I missing? Leakage inductance?

Now the inter-winding capacitance symmetry vs balance I understand perfectly. (just kidding - I have no idea what that is)
When 22uF and 1H resonate at 34Hz the inductive and reactive impedances are equal and opposite so they cancel. The load is therefore just the 10ohms.

In practice you want the resonance to be well below the the bottom of the audio band. For it to be one tenth of 34Hx the capacitance needs to be 10 times bigger and the inductance also needs to be 10 times bigger. So with a 220uF output cap you want the transformer inductance at very low frequencies to be at least 10H.

Cheers

Ian
 
By load we mean what the output is connected to. Meaning what the output is "loaded down" with. If the output is connected to a common 10K input, that is a 10K load and is considered to be a "light" load because it is easy to drive. But some older gear is 600 ohms which is harder to drive and is considered to be a "heavier" load. If you tried to connect an 8 ohm speaker to the output, a basic op amp circuit would not be able to drive it much at all.

Notice that there is a 100 ohm "buildout" resistor inside the feedback loop. That is also going to limit the current. So the output impedance might be 20 ohms (although the schematic shows 10 ohms) but if you connected the output to something that is 600 ohms that might actually be too "heavy" for the simple op amp output. Actually because of the buildout resistor, it will probably keep the op amp from distorting too much but it will limit the voltage swing. The op amp output probably cannot put out more than about +-12V. But with 100R and then 600R as the load, thats 12V/700R=17mA*600R=10.3V.

Having explained all of that, it should be able to drive just fine without low frequency loss. Especially if it's a 10K load. The output capacitor is 22uF which is more than good enough to get low frequencies. So honestly it's not crystal clear why you would be loosing lows.

How did you hook up the transformer. Draw a schematic, take a picture with your phone and post it. Make it look like the attached schematic and LL1539 datasheet wiring schematic but with your wiring bits added.

And of course now tell us what the load is. Are you going directly into a USB audio interface? And what software are you using to see the frequency response? Post a screenshot of the response ...
Ah gotcha. According to the manual the line input impedance is 10k, just like you said.

I tried 1:2, 2:1 and what worked best is the wiring with 1:1 recommended by Per Lundahl. The 1:1 wiring loses the least low end. It's most sub energy that is missing. It's decent but still lacking and I doubt I'd wanna sum low end through this unit like this.
I attached a pic of the wiring. Forgive the sloppiness, as it's just for testing purposes. I will put this in a donor case as soon as I figured it out.
"Monitor hot LL1539 pins 1+6
Monitor cold/gnd LL1539 pins 2+5
connect LL1539 pins3+4

LL1539 pin 12 XLR pin 2
LL1539 pin 8 XLR pin 3
LL1539 pin 10 local ground and XLR pin 1
LL1539 pin 11 + 7 connect"

I'm feeding the line inputs of my mixer straight from my interface. The output impedance of the interface line outs is 51 ohms, and the line ins of the mixer are all transformer balanced. I'm monitoring directly in ableton. It's not my interface as I compared the frequency response to the main outs of my mixer and the monitor outs without the lundahls in the signal path. Both have more low end information. Mid and high end does sound way better and punchier through the lundahls though. And there's a nice spatial effect happening which was the reason I wanted the 1539s in the first place.
Hi
The 22uf might be 'adequate' when there is a 10K 'load' (input to your monitor amplifier/soundcard)
it is way too 'small', for a variety of different reasons if you wish to pass low frequencies with low phase shifts especially if you are introducing the LL1539.
I happen to have been testing (post repair) a Soundcraft B800 stereo master output module that has LL1539 transformers as an option and they use 500uF capacitors (well, 2 X 1000uf in series) and the transformer gets wired 1:1 using the two primaries in series and the two secondaries in series. the response is 'flat' down to approaching 10Hz although I forget exactly what it was now. It was also only minimally influenced by running into 600 Ohms. If the capacitor is too low a value there is always the possibility of 'ringing' (resonance/peaking) depending on the capacitance, the inductance of the transformer and how much load is on it. I actually tested this a fair while back with a Sowter output transformer which was being fed through 100uF capacitor. I think there was a nice 3db 'peak' at a few tens of Hz which then behaved when increasing the capacitor to 1000uF.
Depending what the op amp is in your circuit it could well run into internal current limiting and of course the 100 Ohm resistor will provide some 'safety' current limit.
Matt S


22 µF is incredibly inadequate to drive any transformer, even a high inductance one like a Jensen. Putting it in series with 10 Ω almost seems like a joke - the reactance of the capacitor will be the same as the resistor at about 700 Hz (and there's where your bass is going!). Make that capacitor at least 10 or 20 times larger ... and then hope the op-amp has the current capability to drive the transformer to high output levels.
Incidentally, you may be disappointed with the result if you're trying to create a symmetrical (equal magnitude, opposite polarity) output signal pair using an output transformer. Into typical high-impedance balanced inputs, the inter-winding capacitance of any output-type (no Faraday shield) transformer will have the same symmetry as on the primary. But symmetry is not the same as balance ... which I've tried to get the world to understand for decades.
Bill Whitlock
AES Life Fellow

Interesting. So what you're saying the culprits are the capacitors which are way too low for the task of reproducing low end? Would I just be able to screw open the mixer and replace the capacitors with a larger value and that would fix the problem? I'm having trouble understand all the technical specifics to be frank. But as far as I understand the higher the voltage in the mixer, the higher the headroom, the larger the frequency response and the output levels can be?

I found the schematics of the master output section. As I understand it the monitor out mirrors the master output but differs in that it doesn't go through the output trannies of the master out but is a direct out connection. I attached a schematic of the main outputs, if that's any help.
 

Attachments

  • Screenshot 2021-05-08 at 23.08.20.png
    Screenshot 2021-05-08 at 23.08.20.png
    1.2 MB · Views: 73
  • Screenshot 2021-05-08 at 23.16.56.png
    Screenshot 2021-05-08 at 23.16.56.png
    461.3 KB · Views: 81
I found the schematics of the master output section. As I understand it the monitor out mirrors the master output but differs in that it doesn't go through the output trannies of the master out but is a direct out connection. I attached a schematic of the main outputs, if that's any help.
You can see that the output capacitor on this one is 10x that on the monitor outputs.

"as far as I understand the higher the voltage in the mixer, the higher the headroom, the larger the frequency response and the output levels can be?"
Most of it is true, except the frequency response, that is unaffected by the voltage.
 
You can see that the output capacitor on this one is 10x that on the monitor outputs.

"as far as I understand the higher the voltage in the mixer, the higher the headroom, the larger the frequency response and the output levels can be?"
Most of it is true, except the frequency response, that is unaffected by the voltage.
Alright. That’s good to know. So would I just be able to simply replace the capacitors with 20x the value since the signal path is otherwise the same?

edit: it also says I can use an external power supply instead of the internal one with a dc voltage range of 12v to 30v max. Would that in any way affect the headroom?


thanks
 
Last edited:
Alright. That’s good to know. So would I just be able to simply replace the capacitors with 20x the value since the signal path is otherwise the same?
Yes; beware, though, that the opamp may struggle do deliver, particularly at low frequencies.
edit: it also says I can use an external power supply instead of the internal one with a dc voltage range of 12v to 30v max. Would that in any way affect the headroom?
The answer is it depends. If there are voltage regulators in the unit, it won't change, if not, the higher the voltage, the higher the headroom.
What type is U1?
 
The question of the supply voltage has many possible answers of which in a sense the voltage that arrives at the op amp will determine the maximum voltage headroom so, assuming the op amp (indeed all of the circuitry) can withstand a typical 'full' plus and minus 18 Volts (36 Volts total) that most op amps can tolerate safely that in a simplistic way is the 'best' for highest voltage headroom.
The fact (reported above) that the mixer will accept 12 to 30 Volts external supply input suggests it uses a DC-DC convertor to obtain the plus and minus rails INTERNALLY so as JR hints at, there are many things potentially being changed here.
Higher Voltage headroom may allow a degree of 'improvement but may then be constrained because the internal current limiting of the OP amps involved will shift the point where they start to current limit. Remembering that the current limit is there in an attempt to prevent chip failure and 'chopping chunks out of your signal' to achieve this 'safety' is not 'musical'. The op amp makers are more interested in ensuring reliability against customers 'abuse'.
I believe you can add nitrous oxide to gasoline engines for great performance, until you burn holes in the piston crowns.
 
Yes; beware, though, that the opamp may struggle do deliver, particularly at low frequencies.

The answer is it depends. If there are voltage regulators in the unit, it won't change, if not, the higher the voltage, the higher the headroom.
What type is U1?
Pardon my late response. With U1 you mean the Opamp? I attached a picture. It's the LF442 type OPAMP.

The question of the supply voltage has many possible answers of which in a sense the voltage that arrives at the op amp will determine the maximum voltage headroom so, assuming the op amp (indeed all of the circuitry) can withstand a typical 'full' plus and minus 18 Volts (36 Volts total) that most op amps can tolerate safely that in a simplistic way is the 'best' for highest voltage headroom.
The fact (reported above) that the mixer will accept 12 to 30 Volts external supply input suggests it uses a DC-DC convertor to obtain the plus and minus rails INTERNALLY so as JR hints at, there are many things potentially being changed here.
Higher Voltage headroom may allow a degree of 'improvement but may then be constrained because the internal current limiting of the OP amps involved will shift the point where they start to current limit. Remembering that the current limit is there in an attempt to prevent chip failure and 'chopping chunks out of your signal' to achieve this 'safety' is not 'musical'. The op amp makers are more interested in ensuring reliability against customers 'abuse'.
I believe you can add nitrous oxide to gasoline engines for great performance, until you burn holes in the piston crowns.
Interesting. I guess that's that then? From what I gathered in this thread mods seem very hard to come by on this mini mixer. Upping the voltage through the use of an external power supply would come with a risk of damaging the board I'm interpreting?

It does sound very good as is, but the urge to mod it to make it better is strong haha
 

Attachments

  • tempImageGveo7g.png
    tempImageGveo7g.png
    1.7 MB · Views: 35
Back
Top