Rohde und Schwarz Begrenzungsverstärker ABR / U23

[gyraf]

Falk,

I would aim for first getting rid of all traces of hum, that part is possible.

From there it would be a matter of IRL evaluation what noise distance is really needed. Depends on quite a few factors, including how much of the noise gets modulated by GR, what it's sweet spot is, etc.. I'm not holding out on you, it's just something that can't be quantified in absentia.


Oscar,

Unfortunately no info at all about the transformers or their parts.. And your inbox is full..

Jakob E.

 

[Falk]

Hello,


@Jakob
thank you so much for giving me this motivating critical view.

I just turned my two power transformers (one for the 4A  of initial filament current and one for the 250V and 12V) 90° and the hum is almost gone. I could also try to isolate the Input TX and Interstage TX + cables since they are sensitive when I touch them.

The good news is: Noise is at -88dB FS now.  Thats almost fine for my mastering use. I am not doing any music that would need a dynamic range like this.

I attach a photo of the noise spectrum.
The 50 Hz is from mains TX the 160Hz comes from the heater TX. 

Apparently my two ground planes aren't a problem. How do you gentleman layout pcbs? Would you recommend to have one ground plane on the bottom or do you fill everything apart from the traces with a ground plane?

Best regards,
Falk

 

[Michael Tibes]

I've been following this thread loosely. In case you haven't seen it yet: I did some more digging now and came across this company http://vacuvox.com/. They build replicas with some enhancements. They linked a review of their older unit here http://vacuvox.com/wp-content/uploads/2018/10/Vacuvox-U23M-Promo.pdf and the measured noise spectrum there looks very much like yours...

Nevertheless I believe it should be possible to get rid of the hum.

Michael (also from Berlin  )

 

[opacheco]

Oscar,

Unfortunately no info at all about the transformers or their parts.. And your inbox is full..

Jakob E.

Jacob,
That's a really shame!....It will be a great help have some measurements for real transformers used in the U23 from a real unit!! but this is more difficult than find a original unit  in good shape!

I will check my inbox thanks!(you have my email address if will be necesary at any time!)

Thanks a apprecite your help and time so much!
opacheco

 

[emrr]

Two thoughts.

The IT with separate resistive loading of each coil is interesting.  I've only seen that in European designs, never an American one I can think of. 

An input pad like that is common when the unit is used in a bridging scenario with a an unknown number of other parallel amps.  Besides establishing bridging Z, it also gives isolation.  Allows use of a more common step up transformer, and can be removed if bridging isn't needed, as opposed to amps with dedicated bridging transformers that only work in that service. 

 

[opacheco]

Two thoughts.

The IT with separate resistive loading of each coil is interesting.  I've only seen that in European designs, never an American one I can think of. 

An input pad like that is common when the unit is used in a bridging scenario with a an unknown number of other parallel amps.  Besides establishing bridging Z, it also gives isolation.  Allows use of a more common step up transformer, and can be removed if bridging isn't needed, as opposed to amps with dedicated bridging transformers that only work in that service.

EmRR could you talk us about the Bridging Input Section type used in this unit please? What it does exactly? How this works?

Thanks in advance!
opacheco.

 

[Falk]

Hello everybody.

@Michael Tibes, thanks for the Link it was a great motivation to work on the SNR performance of the Unit.
Attached is a screen shot of the noise spectrum. The dB scale equals a dB(V) reading. So my unit is a good bit quiter in the low end than the unit from the comparison above. But its also noisier in the highs freq range.

The noisefloor of the unit in my mastering setup is -88dB(FS) RMS now.  Thats completely fine for me. I will shield some more cables and I am sure it will improve the performance.

Also I ordered a custom power transformers. This will be another advancement too.

 

[Falk]

The machine is almost working as I wish for it to be. I corrected the frequency response today. Before and after in the plot attached

 

[echoplex]

Hey Falk,

I'm impressed with what you have reached in the meantime.
-88dB(FS) RMS is pretty amazing and should suffice for even mastering applications.
And then the frequency graph you posted.. it's pretty amazing as well.

I rarely feel so hyped about a project on GDIY as I do with this one.
Please keep us updated and please remain so motivated.
This is really the most interesting project 2020 and I would love to see it proceeding.
Also eager to hear some samples and to see the PCBs

cheers

 

[gyraf]

bumpin' for support and interest.. you can't leave us hanging like this

 

[Falk]

Good evening gentleman,

I havent had time during the past weeks to go on with the project. I am also waiting for parts to arrive.
I am not satisfied with the SNR yet, so I redesigned the whole pcb and psu.

I keep you posted as soon as there are news.

@echoplex and @gyraf - Thanks for your support and motivation.

Falk

 

[Falk]

Hello Gentleman,

the new pcbs were dispatched today. The custom power tx will be dispatched tomorrow, mouser has already delivered the parts. Next week I will assembly the revision 1 board and see how things go.

What I did in revision 1 is to divide the PSU from the signal boards. Physically and regarding ground connections. That has two reasons:

1. I had the impression that the noise floor [-88dB(FS)] arises from having the same ground plane for psu and signal.
2. If the noise floor is still bad, I can put the PSU in a single case.

I also had some flaws on the first pcbs:
1. a wrong routing in the side chain
2. a ground plane on the top of the PCB and another on the bottom layer. That caused to have literally tons of ground loops.
3. also I decided to have EF83 as compression tube. The first boards could wether have EF83(audio remote cutoff) or EF89(radio remote cutoff). Both were great and sonically did not make too much of a difference.

I also played a lot with my spice model and got it fixed to be totally exact with the specs from the Rhode and Schwarz data sheet from 1954. This was a very tiring but also valuable process of deeply understanding the interaction between the positive diode bias in the side chain and the threshold control. This outcome will be the starting point for the calibration process. It allows for having the unit at a maximum ratio of a bit more than 4:1.

This ratio is what the original data sheet states. It says: The unit has a gain of -1dB (when input is +6dB; output +5dB) When signal rises for 10dB (input +16dB) the output signal is 7.5dB. Since the overall gain is -1dB the output actually rises 2.5dB what equals a ratio of 4:1.

I will integrate the following ratios: 1.25: 1; 1.5:1; 2:1; 3:1; 4:1

Any comments are highly appreciated.
Best regards,
Falk

 

[gyraf]

Very good job getting the simulation under control - however, don't be alarmed when this does not reflect upon reality (!). The important thing is that it shows you the expected interactions in the circuit.

Getting the ratios perfect may or may not be important for the subjective qualities of this unit, don't sweat it until you have a working unit at hand AND have a chance to evaluate in company with known-good ears. Do you have any mastering studios at hand where you are*? These sorts of people are often very good at determining the real impact of details like that - and it's my impression that a vast majority of them are genuinely interested in first-hand experiences of what is possible in electronic design. This is also the situation where you will want to explore the ranges of attack- and release timings. You can usually instigate genuine interest in perfecting your particular project by letting them have the prototype or an early production model at cost..

For tube selection - what are the reasons to stray away from the original EAF42's? I thought them to still be plentiful enough in the wild?

-88dB noise floor is not at all bad UNLESS it is hum products. A real-life evaluation will again tell you more than a thousand measurements.

Physically separating the PSU is a good idea, but please don't put it in an external box. Connecting high DC voltages is hard. If needed, perhaps go with bolting the "external" PSU enclosure to the back of the unit so cabling can be permanent.

/Jakob E.

*if not, perhaps I can set you up with someone, depending on where you are?

 

[Falk]

Hello Jacob,

thank you for another valuable bunch of hints on how to move on.

Getting the ratios perfect may or may not be important for the subjective qualities of this unit, don't sweat it until you have a working unit at hand AND have a chance to evaluate in company with known-good ears. Do you have any mastering studios at hand where you are*? These sorts of people are often very good at determining the real impact of details like that - and it's my impression that a vast majority of them are genuinely interested in first-hand experiences of what is possible in electronic design. This is also the situation where you will want to explore the ranges of attack- and release timings. You can usually instigate genuine interest in perfecting your particular project by letting them have the prototype or an early production model at cost..
I am located in the middle of Berlin and as soon as I have two running units I will hit the road to ring the bell at some downtown mastering studios to find company to get these machines on spot.

For tube selection - what are the reasons to stray away from the original EAF42's? I thought them to still be plentiful enough in the wild?
On the latter revisions of the U23 (actually the U23 Type ABR) Rhode und Schwarz themselves used the EF83 instead of the EAF42. Since I never planned to build a perfect clone of a historical piece of gear I simply went with the EF83 because of the much more easily available socket and tubes.

-88dB noise floor is not at all bad UNLESS it is hum products. A real-life evaluation will again tell you more than a thousand measurements.
Indeed its hum products and I think in the very best case the noise floor can go down to about -100dB(FS)
Thank you for the idea with attaching the PSU to the back. I will see if that is needed. The PSU is pretty big because of the regulator tube (108c1) and a big inductance of 7H.

*if not, perhaps I can set you up with someone, depending on where you are?
I appreciate this offer and would get back to this once the devices are up and running.

I will post more about the progress next week.
Best regards,
Falk

 

[jans]

Hi Falk,
I am also very much interested in your project. Hope you succeed!
Greetings from Amsterdam,
Jan

 

[Falk]

Good Morning,

@Jans thank you for your interest. I will keep everybody posted here.

What I achieved until now is: I skipped the input pad so the machine has a total gain of 20dB and highest possible SNR in my configuration. Maybe it would be possible to further step up internal levels but I sounds pretty damn cool like it is. The highest threshold is at +18dB(U).

I am still waiting for the power toroidal to arrive and am hoping that the nasty hum at 50Hz is going to get lower. Redesigning the PCB made the build less sensitive to noise and lowered the overall noise a bit but not tremendously. By redesigning the pcb I gained loads of space in the case which is super helpful for finding  alternate positions for all the iron.

I am figuring out some cool ratio settings with a pot to implement discrete values later. I like the attack and release time as is. Whats impessive about this unit is - It nails the rms level on spot but leaves most of the transients pass. Its truely a leveling amplifier.

I attach a picture from the bench.
Thank you everybody.
Falk

 

[Erik2345]

Nice work!

 

[gyraf]

Really good work there.

Have you tried re-orienting your audio transformers, as to get them to pick up less of your PSU hum? turning in one of the three possible axis's could perhaps get you further away from the hum?

/Jakob E.

 

[Falk]

Thank you Jakob, thank you Erik.

Unfortunately the new toroidal are noisy like crazy BUT! i had that mu-metal sheet flying around and wrapped it around the power tx - it was the cure. Noise spectrum at unity gain in my mastering chain is attached. While I am writing this I am listening music through the unit and it sounds ace.

I will make a break until next week and record some drums at the studio. Next week I will make a plan how to continue. Now that everything works, I need to build a second unit to see if I can reproduce whats been achieved.

Once the second unit works. I will finalize the front panel design. Still a good mile to go. Thanks everyone for hanging in here. Its great motivation to keep on working.

The 100Hz noise peak comes from the power supply a guess. Its unrelated to the transformers and their positions. Its always there and steady. Any idea on this?

Have a nice evening.
Falk

PS! The dB Scale is dB(FS) with FS = +16dB(U)

 

[volker]

> 100Hz noise

You could try more B+ filtering, just put in another bigger cap with alligator clips to see if it helps.