Pre amp therapy / therapy pre

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

I finally found what was causing the oscillation : it was the improper loading of the Lundahl mic trafo. Replaced the 220k with 9k+3nF series as recommended by Lundahl, and now the oscillations are gone.

Now, that buzz is annoying after all, I'd be happy if it was gone !

On the picture attached is a scope capture of the buzz at the output, scope settings being :
- Horizontal : 5ms/div
- Vertical : 200mV/div

The preamp gain is maxed, so that's about 83dB of gain (pentode, no feedback, gain and level maxed). Input is 150R across pin 2-3.

https://drive.google.com/file/d/0B6K-0QVHN7zUWlFVX1lBWDBXanM/view?usp=sharing

Any clue of where this comes from in the chain ? The buzz level is affected by both the gain and level control.

Thanx in advance.

Best regards.

Eric
 
ricothetroll said:
Any opinion on the routing ?
There is a major mistake in your routing. You made the ground entry on the input side. As a result, the current spikes that circulate in the ground trace create a noise voltage that is directly superimposed to the input signal.
 
abbey road d enfer said:
There is a major mistake in your routing. You made the ground entry on the input side. As a result, the current spikes that circulate in the ground trace create a noise voltage that is directly superimposed to the input signal.

I always thought grounding at the input was a good practice, as it's always done in guitar amps for example. Seems like I'm wrong ! Shouldn't be too much of a problem in such a "Class A" schematic, is it ? Anyway I joined a modified version.

BTW, is there already a thread here on this subject ? Couldn't find any. Might be a good idea to start this discussion.

For my buzz issue, I 'll try to increase B2-B3 series resistors, as I still have voltages much higher than Jonte's :
Post choke = 257V
B3 = 245V
B2 = 219V
B1 = 194V

I'll order them when back from holidays, in about 2 weeks.

To be continued

Best regards.

Eric
 

Attachments

  • v8.png
    v8.png
    64 KB · Views: 67
Hi !

I'm nearly getting to the end of the project. The buzz issue I had was actually caused by the mains transformer magnetic field. Solved it by wrapping it with soft iron lamination from the core of a garbage toroidal trafo. I also noticed that the buzz was quieter when the trafo was physicaly higher, so I built some kind of pedestal to take profit of this. Now the buzz is something like 15dB quieter, and is now quite as loud as the electronic noise on the 2nd channel, and barely audible on the first one.

I now have a question regarding the optimum load of the input trafo. I used Lundahl LL1528s, because I had them in stock fo quite a long time waiting for a project to use them. I know those are not optimised for high impedance tube input and so do not get as much gain as I could have, I'm OK with that.

Looking at Lundahl's datasheet, they claim that the optimal load is 9k in series with 3nF.  I implemented it (9k1 / 3n3) but I noticed that I was lacking high frequencies. So I took another LL1528 and started messing around with it and the network analyser of my Digilent Analog Discovery 2. I then noticed that with that load the frequency response isn't quite linear in the audio range. With only 9k1 the response is perfect though. See the Bode plots attached. I pushed the frequency up to 1MHz to to make sure there's no peak above the audio spectrum.

Is there an error on their datasheet ? Am I not understanding the concept of optimal load correctly ?

Thanx in advance.

Best regards.

Eric
 

Attachments

  • LL1528-Load-Comp.png
    LL1528-Load-Comp.png
    20.4 KB · Views: 22
I used 180R in series with the AD2 generator output, which has a low output impedance (opamp driven) :

https://reference.digilentinc.com/_media/reference/instrumentation/analog-discovery-2/ad2_rm.pdf -> page 26

I did some measurements to compare both loads with different source impedances, as I would like the preamp to behave well with different mics. The load without the 3n3 capacitor is still more linear, even with 1k source impedance (see attached charts)
 

Attachments

  • Capture.PNG
    Capture.PNG
    34 KB · Views: 27
ricothetroll said:
I used 180R in series with the AD2 generator output, which has a low output impedance (opamp driven) :

I did some measurements to compare both loads with different source impedances, as I would like the preamp to behave well with different mics. The load without the 3n3 capacitor is still more linear, even with 1k source impedance (see attached charts)
Your results are consistent with the basic simulation results. Really 9k is loading too much the secondary; it is not normal nor acceptable to clamp the signal by 4dB in order to optimize the transient response. Typically a good Zobel should not clamp by more than 1dB. That would suggest about 50kohms for the resistor.
I think you should interrogate LL about this issue. Absence of data rearding leakage inductance and stray capacitance does not allow more investigation.
 
Here are the results for different resistive loads. It's only at 12k (no near 9k1) that the peak at 80kHz becomes somehow flat. Looks like this load is necessary to keep the trafo from ringing (with 220k it did ring at the highest gain settings of the preamp)
 

Attachments

  • Capture2.PNG
    Capture2.PNG
    20.1 KB · Views: 29
Here are measures with 9k1 and different capacitor values. My guess is that the real optimal value is 330p !
 

Attachments

  • Capture3.PNG
    Capture3.PNG
    38.5 KB · Views: 24
Experimenting with different sources shows that it still doesn't stand high source impedance well.

BTW, what's the maximum microphone impedance one should expect ? I thought 1k was already high but I could find references to things as high as 50k....
 

Attachments

  • Capture4.PNG
    Capture4.PNG
    23.1 KB · Views: 22
ricothetroll said:
Experimenting with different sources shows that it still doesn't stand high source impedance well.

BTW, what's the maximum microphone impedance one should expect ? I thought 1k was already high but I could find references to things as high as 50k....
Microphones are typically designed for 5-10x nominal Z, so designing mic pres for 2kohms is a safe bet. Note that some users appreciate the possibility to explore higher values, since some mics seem to "open up". I tend to agree with that. Some mics will react more than others to seeing a high impedance. The effects are well understood; they can be easily EQ'd. The electrical damping in microphones is so low that microphone loading has no electromechanical effects, only the electrical interaction betewwen the mic impedance and the preamp impedance.
 
Thanx for your answer !

I wrote to Lundahl and already got and answer from Per Lundahl :

"Hello Eric,

The purpose of the RC load is to give high frequency attenuation, as this is considered to results in a better-looking square wave response, in particular when the source impedance is low.
Personally I have become skeptical to exaggerated RC loading. Even if you are outside the 20-20k frequency range, in my experience RC loading makes the sound more dull.

Do you actually need any RC load?

Beware that audio analyzer inputs sometimes have a fairly high capacitance to ground which will make your HF response look even worse when you use step-up microphone transformers...


- Per Lundahl"

I then sent him my measures. To be continued...
 
ricothetroll said:
The purpose of the RC load is to give high frequency attenuation, as this is considered to results in a better-looking square wave response, in particular when the source impedance is low.
I would tend to somewhat disagree with the formulation. The Zobel network is an image impedance of the reflected impedance at the secondary, i.e. the capacitive effect compensates the inductive nature of the xfmr. The result should a purely resistive impedance. It will certainly reduce HF, compared to the original resonant (peaky) response. But the resulting response should be maximally flat.

Personally I have become skeptical to exaggerated RC loading.
"Exaggerated" is sure bad, but there should be a nice compromise. I'd really like to know the values of nominal and leakage inductance as well as parallel capacitance. A graph of the response without any load would help understanding.

  Even if you are outside the 20-20k frequency range, in my experience RC loading makes the sound more dull.
That is not the experience I had with other mic xfmrs; proper Zobel optimization flattens the resonant peak, but the asymptote after cut-off is pretty much the same.
 
Hi,

Here is the graph without load, input drectly connected to the generator and output directly connected to the scope. Still done with AD2.

I tried to measure the primary inductance but couldn't obtain consistent results with the resonant circuit technique described here :

http://sound.whsites.net/articles/audio-xfmrs.htm

For example (input signal is 10mV peak) :

- Series cap = 10nF-> peak at 900Hz-> 3.13H
- Series cap = 100nF -> peak at 160Hz -> 9.65H
- Series cap = 1uF -> peak at 30Hz -> 28.14 H

How should I interpret those results ?
 

Attachments

  • ZeroSource-NoLoad.PNG
    ZeroSource-NoLoad.PNG
    16.5 KB · Views: 36
Hi,

After a lot of measures and simulations I came to the following conclusions :

- I've used the "Jensen" procedure to determine the optimal load, and finally the best load is 9k1 + 470pF

- That trafo (LL1528) can't with one load show no ringing for low impedance sources (20R typical condenser mic) AND show a full bandwidth for high impedance ones : 300R seems to be the limit for an acceptable bandwidth, 200R is fine.

- The mic/line switch, that adds 4k7 on Jonte's schematic, cannot be adapted for the LL1528 : to have 20dB attenuation, you have to add too much series resistance, or to add another resistance in parallel to the primary to lower the source impedance and get back low/highs. I think I'll remove that switch and use the preamp with the PAD instead, or even better : through the unbalanced DI input. Input impedance is around 770R with the PAD engaged for all audio frequencies, considering the reactance of the Zobel network.

- Even without any load, the LL1528 at the input stage shows some high frequency lossed (1-2dB notch at 5kHz, starting at around 1kHz). When the gain is at the maximum (no NFB at all in the first two stages) this can be seen at the output of the preamp, but for other values of the gain the NFB flattens the response.

Final conclusion : you can use a LL1528 for this project, but it's not the ideal choice.

BTW I measures the leakage inductance by shorting the secondary and measuring the primary's inductance with an LCR. I also measured the inter-winding capacitance, and found the following values :
- L(leak) = 1mH
- Cw = 120pF
I calculated the primary inductance from the datasheet low frequency response and found 12H, so the secondary is 300H. That model seems reliable after some simulation/measures comparisons.

Best regards.

Eric
 
Hey, I'm really interested in building this preamp.
And wanted to ask, if the cinemag s217d would be a suitable output transformer for this preamp?
Original spec sheet says it can take 25ma DC. Ratio is around 5:1 (or 22:1) I think with primary inductance of 38H instead of 35H of the Lundahl. It Can be used in either push pull or single ended. And it has "a tertiary winding for feedback" if that is of any use... or electrostatic shielding.
Would be glad to hear some thoughts about using this transformer.
 
ricothetroll said:
Hi,

Here is the graph without load, input drectly connected to the generator and output directly connected to the scope. Still done with AD2.

I tried to measure the primary inductance but couldn't obtain consistent results with the resonant circuit technique described here :

http://sound.whsites.net/articles/audio-xfmrs.htm

For example (input signal is 10mV peak) :

- Series cap = 10nF-> peak at 900Hz-> 3.13H
- Series cap = 100nF -> peak at 160Hz -> 9.65H
- Series cap = 1uF -> peak at 30Hz -> 28.14 H

How should I interpret those results ?

This is at least in part because the inductance of iron based transformers is frequency dependent and in general it does increase as the frequency decreases.

Cheers

Ian
 
Back
Top