[DaveP]

Over the next few months I shall attempt to make a U73b limiter/compressor from scratch.  This project will (hopefully) be a tribute to Oliver Archut who we very sadly lost last year and who has been solely responsible for collecting the Telefunken Archive from which many of us have benefited.

This will be the first feed-forward limiter that I’ve made and Ruairi O Flaherty has kindly offered to test it for me (he is a mastering engineer).

I will be copying the original circuit closely but with a few exceptions:-
1. I will be using a matched pair of 6BJ6’s which I already have, they are the US equivalent of E99F’s
2. I will be using DC for the heaters.
3. I will use an ECF80 instead of an E80CF as I already have one.
4. I will use a resistor load instead of a choke in the make-up amp, this will reduce the output but I am using a 15k:600/ 5:1 OPT instead of the 10:1 in the original which should help with the loss.
5. I will be making a rack mount chassis to give me some more room to play with.
6.  I won't be including the balance test circuit, I don't want TX hum near the input grids.

I will cover some side issues in the course of the build which I hope will be useful:-
1  Splitting the primary of an Edcor XSM matching TX.
2. A series of tests on transformer screening.
3. An alternative to the expensive E88CC using TV tubes.

I’ve attached a redrawn schematic out of respect for Oliver’s estate.  I have used the original component numbering system as far as possible for reference.  If you want to check out his site you can find it here:-
http://www.tab-funkenwerk.com/id6.html

Best
DaveP

[DaveP]

The problem with many audio matching transformers sold at reasonable prices is that they have only a centre tap and not two separate windings/side.  The U73 needs two of this sort (see schematic).
I had several Edcor XMS15k/15k spare so I decided to risk one to see if I could split the centre tap wires, if it worked then we had a project.



I found that I could see two wires on the primary centre tap tag, but not on the secondary, but that did not matter for the U73.



 I used my wife’s kitchen knife (Shush) and heated the tag to melt the solder while gently scraping the two wires, after a bit one wire was visible so I very carefully unwound it while the solder was melted.  Do not do this unless you can afford to scrap a TX!



Next I cut a piece of Copper tube the same length as a tag and cut it in half and flattened it out.  I the cut a strip off the same width as a tag and drilled a suitable hole and “tinned it” with solder, a bit of snipping with decent scissors and we have a tag.



The Edcor has a spare slot so I super glued the new tag in place and soldered the unwound wire to it.  I had first checked the continuity with a DMM to check I had unwound the correct wire!!  Finally I secured the wire out of the way with super glue.
Don’t expect the DCR to be the same as one of the wires will have passed over the other and therefore be longer although the turns will be the same.  Flushed with success I attempted another Edcor XMS and was able to convert that one too,  so it was not a fluke and with these two split primary TX's I can now get on with the project.  I hope CJ is proud of his apprentice!

[ruairioflaherty]

That's great!  Really looking forward to this thread, and then ultimately hearing the unit.

Cheers,
Ruairi

[thomasdf]

Dave, I had the chance to hear and use a vintage pair of Siemens u73b and I have to tell... That is one of the sweetest compressors I ever heard in my life... So full bodied, and sweet but tight compression... Maybe a tad dark, but that matches pretty well our modern mixes and DAW environment. Very very musical!
And once again you are blowing my mind with one of your "from scratch" thread... Can't wait to read more!!
Keep up the great work!
Thomas

[gato]

'bout time for another dave p thread!

[DaveP]

Thanks for all the support guys, I'll do my best.

I have decided after looking at the pros and cons to leave out the 15kHz cut-off circuitry.

My reasoning is this,  it will not be used for FM transmission but mastering full frequency response recordings.  Studios that have used the highest quality mics and gear would not take kindly to having their mastering engineer wiping their top-end during mastering.  If that effect is required then someone can use their EQ.

Tonight I'm starting the transformer screening tests.

best
DaveP

[DaveP]

Transformer screen test

For this test I have used a small 6VA mains TX  as a source of hum radiation.  As I'm in France, the mains frequency is 50Hz so the hum comes up at 100Hz on the scope counter.   I have put the scope leads across one of the windings of an Edcor XSM15k/15k TX, the set up looks like this:-



This is what the trace looks like:-



The first test was to vary the distance,  I measured the induced voltage on the scope peak to peak while changing the distance from 10mm, 20mm, 50mm, 100mm through 200mm, this produced the exponential chart (attached)  because I can't upload pdfs to Photobucket and can't convert an excel chart to jpg, technology still has a long way to go!

The next test was to keep the distance at 100mm but put various metals in the way,   The reading with nothing was 3.6 but with 1.5mm thick steel it fell to 2.4



with a 1mm thick Aluminium sheet it was 3.4



with two Aluminium 1mm sheets it fell to 2.8 but with the steel as well it went down to 1.8.  This proves that a sandwich arrangement is better at disrupting the magnetic lines of force than a single metal sheet alone.

I shall bear all this in mind when designing the layout.  One final point, it did not help to turn the TX through 90 degrees which surprised me.
best
DaveP

[DaveP]

OK, Lets put this data into context.

The input to a U73 is nominally +6dB, so if we say that we need a minimum S:N ratio of 70dB it tells us the  level of screening required.

+6dB is 1.55Vrms and 70dB is 3162.

The maximum hum we can tolerate is therefore 1.55V/3162 = 0.00049Vrms

To relate to the peak to peak voltage we have been recording from the scope we need to multiply by 2.828 which gives us 1.39mVp2p.

Here's the distance v mV chart (copied and pasted Doh)



We can see that we need our input TX at least 200mm from our PTX (distance is the best insulator) and that it will need screening as well.  The voltage at 200mm was 2.1mVp2p.

The steel plate managed to reduce the voltage to 67% so it would reduce our example to 2.1 x 0.67 = 1.4mVp2p, this would almost give us our target but we have to go further than this.  The U73 has an input pad which reduces the voltage from 1.55V to 155mV so we have to find another 20dB of screening from somewhere or lose the pad!

best
DaveP

[DaveP]

Now for some weird stuff!

Just to remind you, this arrangement produces 3.9 @ 2mV/Div=7.8mVp2p



Turning through 90 degrees on axis makes no difference!



But turning the axis through 90 degrees drops the hum to 1.2 @ 2mV/Div=2.4mVp2p, and reduced the hum to 30.7%



Now comes the weird stuff.   When I put the steel plate next to either the source TX or the XSM TX it made no difference!
But moving it near the middle finds a null point and the hum drops dramatically to [email protected]/div =0.5mVp2p, that's a hum reduction to 6.4%



The croc clips are to earth all the metal work.

To go back to our previous example,  we now have 0.5mVp2p which is 177uVrms.

With 1.55V that is a S:N ratio of  78.8dB and with 155mV after the pad its 58.8dB, but that is with a distance of only100mm.

If we take the distance back out to 200mm,  then the hum falls to 0.7mV with no screen, but after finding the best placement for the screen the hum has disappeared and only hiss is visible on the scope.



These measurements are not absolute but they are all completely relative to each other and we can learn the following from our tests.

1. The distance between the PT and the IPT should be as long as possible.
2. Screens work better when sandwiched with dissimilar metals.
3. The axis of the IPT should be 90 degrees to the axis of the PT.
4. Screens should be placed after trial positioning for lowest hum.

This is the end of the transformer tests, I'll now get on with the layout.
Best
DaveP

[EmRR]

Now for some weird stuff!

 When I put the steel plate next to either the source TX or the XSM TX it made no difference!
But moving it near the middle finds a null point and the hum drops dramatically to [email protected]/div =0.5mVp2p, that's a hum reduction to 6.4%

I saw similar results when tweaking the GDIY26C.  It looks even weirder on a spectrum analyzer, watching the harmonics swim around.   In some cases, you'd improve one thing while making something else worse, and have to make unclear choices. 

[DaveP]

Hi Doug,

Yes, its because of our experience with the GDIY26C that I'm doing these tests now, I needed to sort out some fundamental principles for this new layout.
best
DaveP

[alexc]

I really love the 6BJ6 tube - the smoothness of the grab and release characteristic is the best I have heard and used.
Such an interesting tube; a bit like the 'mouse who roared'.

I'm currently using it again in a guitar pre+limiter box which is a mashup of a :

- 6BJ6 preamp + discrete line amp + utc insterstage -> 6BJ6 vari-gm pair  + utc interstage -> 6V6 SE makeup/line amp
- 'power assisted' AU7 sidechain with tube rectification and glow tube regulator/threshhold

Inspired by the the 'AWA' limiter. Quite a bit like the u73b but with SE primary-fed make up stage.

It is my foreground project for the now winter DIY season next to the fireplace 

Tally-HO    Dave!  Looking forward to reading on your new limiter project   

AlexC

[DaveP]

I've been testing some  alternative tubes for the E88CC/ECC88/6DJ8.
I’ve found some nice TV tubes with the right spec to replace the  E88CC and the choke load.

The first part of the amp section I replaced with an RCA 6BN4A, which is a B7G format with the same spec as half a 6BZ7.  At the 0.6mA current used in the schematic, the mu/gm/rp is the same as the ECC88.



As I was not planning to use the choke load of the original in the other half of an  E88CC, I needed something beefier for just a simple resistor load.  I found that the 6GZ5 was a B7G format TV output tube that fitted the bill perfectly.  It’s spec’ed for a 15k load, so I used it triode wired with said resistor parallel fed to a 15k:600 OPT terminated with a 600 ohm resistor.
For such humble parts, I was staggered by the frequency response, it was flat from 10Hz to 70kHz.
I left out the 500pF by-pass cap as it needed no help in the treble department.
An ECC88/6DJ8 would have cost me $12.50, an E88CC would have been $25, but the 6BN4A was $3.50 and the 6GZ5 was $4.75, do the maths!



You can put in a +6dB signal for limiting and a +16dB signal for compression, either way the output is still +6dB, the by-pass cap  C18 increases the gain by 10dB when in circuit.

I have tabulated the gain structure as its quite complicated, especially as the gain changes between limiting and compression.
The dB figures are from the original schematic.  The corresponding multiplication or division factors are in the row below  and the actual Voltages are the last row.  It’s important to note that the way the  audio transformers are wired compensates for the push-pull split and the return to unbalanced amplification.

best
DaveP

[DaveP]

I've made a rough birds nest of the 6BJ6 section using the Edcor 15k:15k as an interstage.



All the voltages are fine but the frequency response is not as expected.  I expected to lose bass but instead I found that the treble drops at 6dB/octave from around 13kHz, depending on whether I alter some resistance values or not.

The resistor R40 acts like a feedback resistor between the pos and neg halves of the cycle, (Remember early Vox tone control?).
If I change the value it alters the frequency response as in this table below:-



You can see that the output level varies as expected, but with  R40 at 10k the level starts to fall before 10kHz,
With 6.8k it stays flat to 13kHz, but going lower as with 4.7k or 5.6k, the bass starts to roll off.  I'm guessing this may be due to interaction with the Edcors capacitance?
Any ideas??

I have also found that the frequency response is improved (and the square wave performance) with a 15k resistor across the secondary of the interstage on the far right of the schematic above.

best
DaveP

[MatthisD]

Hello Dave,

I mentioned in my message that I had high frequency loss when testing the circuit with various transformers for the input and I found that by connecting their primaries in parallel (disconnecting the side-chain amp) gave an even response so the problem was with using the two primary windings individually. A Marinair/Neve input transformer gave much better response than anything else but I think the input transformer is the trickiest as it handles four high-impedance signals.

Regarding the low frequency, I believe the interstage transformer's primary windings are tuned to reduce frequencies below 40Hz which masks a bump down there, and forms a HPF to add to the LPF to give the overall frequency response of 40Hz - 15Khz. I found this, again by comparing two different transformers in that position.

[DaveP]

Thanks for that MatthisD

I'm on the case
DaveP

[dmp]

Wow - I just saw this thread - this is great. The transformer tests you showed could be it's own thread. Really interesting to see that.
I remember reading that transformers should be aligned at right angles, off axis.  Your confirmation of the reduction in hum when you turned the transformer to have the axis point perpendicular to each other (instead of parallel) was nice evidence of that. I wonder how much more it would reduce to move the axis to be not pointing at each other?
Also, hum vs distance should follow the inverse square law as the distance increases...  like other things in nature (gravity, pressure). I was going to see how closely your plot followed that, but I'm running out of time right now.
Something radiating outwards from a point decreases proportionally to the increase in the surface area of the sphere. 

[DaveP]

There's good news and bad news!

I have been doing dozens of tests on various ways to connect the input transformer but this 15k:15k has a deep dip in response at 20kHz that I can't correct.  Even without the extra winding being used there is a problem with it in this circuit.  This confirms what MatthisD wrote earlier.  However.....

What does work very well is the humble Edcor WSM600:15k, this means that I can't connect the side chain supply in the same way, but I can get around that problem by using another pad and another transformer like this.......



With this arrangement the frequency response goes out to 35kHz.

Other points:  The interstage needs a terminating/ grid resistor of 15k and the value of R40 plays a very big part in the level and frequency response of the amp.  Finally, the dropping resistor to the 5.1V zener needs to be changed from the original 80k to 33k as the BZX79 draws more current than the old OAZ 201

best
DaveP.

[DaveP]

This is the situation to date.

I have made a rough mock up of this with old parts in birds nest format, see my other projects for what that looks like!

It does work and all the voltages and audio levels are on the money.  I have two Edcors winging their way to sunny France as we speak and I'm working on the layout for all the nice new parts.

I will post some pics when I start cutting the metal.

best
DaveP

[thomasdf]

Yes!! Keep these posts coming Dave!!
Can't wait to see more!