U73b from scratch

[DaveP]

I've managed to make some progress today as some parts have finally arrived.



This shows the insulating pad on the heat sink for the  schottky diodes for the heater supply.  I've been waiting more than a week for that.

The heater wiring is now in, please note that the colour of the wiring does not infer the sexual orientation of the DC supply :



I also had to order a metal clad power resistor for the heater supply, luckily that came out as exactly 2 ohms.



That tucks away nicely in the power supply box.

Finally, the OPT is in place.



This was one of five I had made for me by Danbury Electronics in the UK,  15k:600, goes by the number DB1607.  I needed some with screens which is something Edcor does not offer, pity that.
The components are now starting to go in, more on that next post.
best
DaveP

 

[ruairioflaherty]

Looks superb Dave.

 

[DaveP]

Thanks Ruairi,
I've been working on the mock up I've made to check that the parts I'm using actually work.  I've not shown a pic of this because its lethal and a bad example to set.

I have set it up using the instructions I translated and it does work ok.  However I found that the Edcor TX I'm using for the negative CV output did not work at its best when wired with the primaries in parallel like the original circuit.  What happened is that the load was too low and the CF flunked out past a certain point.  This meant that it limited for a while but could not maintain the reduction at higher inputs and the output went up again.  By wiring it in series (15k:15k) it worked fine and could drive the 6BJ6's down much further.  I guess that the original TX is something like 40k:40k or thereabouts

I have found from experience that its better to make a mock-up first to iron out these types of problems.  If you don't do this, then you have to make alterations to the finished amp and it starts to get spoiled and tatty looking.
best
DaveP

 

[Dr Gris]

I've been working on the mock up I've made to check that the parts I'm using actually work.  I've not shown a pic of this because its lethal and a bad example to set.

DaveP

I still remember that fantastic first pic of your SA-39B  ;D
As usual, I'm very impressed. You're a true diy hero!!

Cheers //M

 

[DaveP]

Oh dear what a lot to live up to :-\

Well the tags strips and a couple of transformers are now in and it looks like this:-



The rest of the transformers are still in the mock-up where I'm doing my final tests before tearing it down.

In order to keep a careful check of what's done and what ain't, I trace it out on the schematic, so the red lines are representing the pic above.  This avoids expensive smoky mistakes and helps when you have to go back to it after a break.



That's all for the moment, I have more heroic work to do. :
best
DaveP

 

[DaveP]

I have just finished my work with the mock-up compressor.  It all hinged on problems with the way the side-chain was configured.

This is the original circuit configuration.



I was able to try this with one of my modified Edcor 15k:15k transformers mentioned at the beginning of this thread.  Notice that there is a step-up of 1:2 in the voltage and that the 100k resistor is in parallel with the driving cap, putting a small amount of DC through the windings.  It also means that the impedance of the primary is 3.75k which turns out to be too low. 3.75k:15k.

The symptoms of this are as follows,  you can set the compressor up for small inputs and it follows the correct output reduction characteristic until the cathode follower can no longer drive the load.  The output then shoots up in an uncontrolled manner.  After I sussed this out I tried it wired as 15k:15k and it was a little better but again the same thing happened.  This time it was not because the load was too low but because there was not enough current and or voltage being transferred.

The 6BJ6 tubes need to be able to run to cut-off to reduce the output properly, so if there is insufficient voltage or current then there is insufficient gain reduction.  The next thing I tried was to put a diode bridge in place of the CT diodes in the original schematic, this worked too well and the gain reduction started way to early and put it out of spec.  Finally I adopted the following configuration and it worked perfectly.



This seems to work because the primary load is high enough (15K) for the CF to drive it through the full range of inputs.  But the voltage is not too high as there is a 2:1 reduction in voltage before the bridge multiplies it by 1.414.  As the voltage on the secondaries is halved, the current is doubled, so this again helps to drive the diodes to produce sufficient CV to drive the tubes to cut-off.

You will notice that the original CF used 30k//100k which is 23k (hence 22k).  I am not sure what the purpose of the 100k resistor is, but it may be to do with interference reduction or frequency response smoothing, if anyone out there has an inkling then please let me know.  The TX on the original U73b was probably nearer 40k:40k which would have made it 10k:40k in the original configuration which the CF could have driven successfully.  Unfortunately we find it difficult to source such transformers nowadays, whereas the Edcors are both plentiful and affordable.

best
DaveP

 

[MatthisD]

You will notice that the original CF used 30k//100k which is 23k (hence 22k).  I am not sure what the purpose of the 100k resistor is, but it may be to do with interference reduction or frequency response smoothing, if anyone out there has an inkling then please let me know.  The TX on the original U73b was probably nearer 40k:40k which would have made it 10k:40k in the original configuration which the CF could have driven successfully.  Unfortunately we find it difficult to source such transformers nowadays, whereas the Edcors are both plentiful and affordable.

best
DaveP

I imagine its to minimize the resistive load. I had no problems with the original side-chain circuit. My transformer primary inductance is about 50H and I measured the frequency response with the secondary disconnected from the rest of the circuit to make sure its reasonably flat with no resonance.

 

[DaveP]

Some interesting info on testing Edcor matching transformers.

There are two ranges,  WSM which are 0.5W and XSM which are 2 .5W, I am assuming this refers to the ac signal power that can be dissipated in the TX.  They are also rated in terms of ac voltage which is 10V for WSM and 50V volts for XSM.

I started testing with the XSM 15k:15k as the interstage as it seemed the most robust.  Unfortunately Edcor do not publish the frequency response for this TX .  It works fine until you test the frequency response, but in this U73 application it dives at 20kHz and even treble boost can't rescue it.  I guess there is an intrinsic resonance in there somewhere.

I then realised that it was safe to use a WSM15k:15k because in the U73 there is only 1-2V max  on it and its rated for 10V.  The published frequency response for this TX is totally flat (attached) and so it proved to be in practice.  There is 160V DC on it but this makes no difference as it is totally balanced in push-pull.  I am still using one XSM15k:15k for the  CV -ve voltage as the output from the CF drive can exceed 10V.

It just goes to show that you don't always have to spend megabucks on a TX, my little $10 WSM is flat from 20-40kHz!
best
DaveP

 

[DaveP]

I made some more progress tonight but I have one last WSM TX in the post from Edcor as we speak.
You can see where it will go on the right in this pic.



This is the state of play on the schematic.
I am now over halfway with the build, old style point to point ain't quick



The power supply is almost complete, I have the main HT dropper to fit after I check it's value when the amp is fired up.


All the TX wiring is now set for 115V

best
DaveP

 

[ruairioflaherty]

Dave,

When this finally gets here for a listen I won't know whether to hang it on the wall as art or plug it in for a listen!

Superb.

Cheers,
Ruairi

 

[DaveP]

Ruairi,

When I tested the rough hash-up I was expecting some problems as there usually are when making vari-mu types.  But this compressor, because it's feed-forward, doesn't have any feedback issues at all, I was pleasantly surprised.  It also flattens the output without any distortion that I can see on the scope.

It does take some time to calibrate, but once that is done its good to go.  The calibration is essential as there is no feedback, so the side chain output has to be set relative to the input.

I shall be very interested to see what you make of it.

all the best
DaveP

 

[DaveP]

Had a disaster with the original meter and managed to break off the pointer! :

My hands are a bit too large for finicky work like this, but I have successfully made a new meter scale:-



I had to scan in the original scale and re-do the numbers in the old German font like the last one.  I did it in Paint but not after it crashed my computer several times because of lack of memory ???  I was doing it at the highest resolution so I guess it kept saving it after every stroke, making loads of copies, anyway I saved it after every 5 strokes and that seemed to cure the problem.

I have used Prit Stick to fix new scales before but its not very smooth.  This time I rubbed down the old scale letters with a fine emery paper and smeared it with PVA then rubbing it with a finger to get an even coat, I then placed the new scale on and it sticks without any glue transparency showing through.

Best
DaveP

 

[DaveP]

The final transformer has arrived and they are now all in place.



There are just a few components left to fit around the side chain as shown by the build schematic.



After that, it's only the front panel left to do.

Best
DaveP

 

[abbey road d enfer]

Now for some weird stuff!



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%

This is not weird. It's completely predictable from the shape of the flux lines.  At a distance and in the same plane, flux lines are parallel to the winding's axis, so one has to make sure that the axis of the receiving coil is orthogonal.

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!

that's because your shield is parallel to the flux lines, when it should be at right angle. In you picture, figure out the flux lines and you will see they jump around your shield.

But moving it near the middle finds a null point and the hum drops dramatically to 0.5@1mV/div =0.5mVp2p, that's a hum reduction to 6.4%

You have to remeber that the shield redirects the flux lines, so it is very likely to find a position where some of the flux lines are redirected in a way that cancels.
You also need to take into account the fact that mounting a xfmr on a chassis alters the flux lines pattern, so the ideal shield position for an open-air xfmr will be different than when the xfmr is mounted on the chassis (that is if the chassis is made out of magnetic material).

 

[abbey road d enfer]

Transformer screen test
with a 1mm thick Aluminium sheet it was 3.4

Al being an amagnetic but conductive material, it is not capable of significantly deflecting magnetic flux lines; the difference you measure owes greatly to the electrostatic shielding. There may be a tiny effect on flux lines due to Foucault effect (eddy currents), though.

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.

Indeed, a sandwich arrangement is better; not only because of the simple mathematical addition, but because the first shield rearranges the flux lines in a more orderly manner, making the second more efficient.

 

[DaveP]

Thank you for your explanations Abbey, better late than never

I do understand the magnetic issues, it was just the null point that was weird and you have explained that and the sandwich arrangement very well.

While you are here, do you know why the original circuit has a 100k resistor in parallel with the C/F Cap?



This is the only part of the circuit I don't get.

Best
DaveP

 

[abbey road d enfer]

While you are here, do you know why the original circuit has a 100k resistor in parallel with the C/F Cap?



This is the only part of the circuit I don't get.

Best
DaveP

At first I thought it would serve to damp the VLF resonance of the xfmr/cap combo, but it just doesn't click; the value is too high.
I can only guess and suggest some conjectures:
a) Having the magnetic core premagnetized in one direction makes it more prompt to respond to peaks of a certain polarity. Such a scheme would not be incongruous in an AM broadcast situation, where positive peaks are prohibited.
b) The small DC current (0.8mA) in the primary reduces the nominal inductance of the xfmr, making it less responsive to low-level LF. That may reduce LF modulation distortion.
c) There's a mistake in the schemo
d) The designer thought it would improve operation, but it does not, however it is so imperceptible he forgot to remove it
There may be a couple other stupid explanations, and there may be a good one, but I can't figure out what.
PRR may have  a better opinion...

 

[DaveP]

At first I thought it would serve to damp the VLF resonance of the xfmr/cap combo, but it just doesn't click; the value is too high.
I can only guess and suggest some conjectures:
a) Having the magnetic core premagnetized in one direction makes it more prompt to respond to peaks of a certain polarity. Such a scheme would not be incongruous in an AM broadcast situation, where positive peaks are prohibited.
b) The small DC current (0.8mA) in the primary reduces the nominal inductance of the xfmr, making it less responsive to low-level LF. That may reduce LF modulation distortion.
c) There's a mistake in the schemo
d) The designer thought it would improve operation, but it does not, however it is so imperceptible he forgot to remove it
There may be a couple other stupid explanations, and there may be a good one, but I can't figure out what.
PRR may have  a better opinion...

Thanks Abbey,

I think you may have cracked it with reducing the LF response.  Remembering that the original schematic included LF cut in the audio path, (FR 40-15kHz) it means that they balanced the side-chain with an equal cut.  There is no feedback path with a feed-forward design so balance is important.

This solves my final question because there is no LF (or HF) cut in my version, so I will not be including the 100k in my circuit, but I will have to reduce the 30k to 22k to compensate for the lack of a parallel 100k.  Great! ;D

Thank you very much.

Best
DaveP

 

[DaveP]

The main amp is now finished and all the cables and wires in place to connect to the front panel.



Things got quite crowded in there towards the end, but there were enough tags spare to fit everything in.

Front panel next.
best
DaveP

 

[thomasdf]

Awesome!