Rohde und Schwarz Begrenzungsverstärker ABR / U23

[Winston OBoogie]

One thing that occurred to me earlier is a solution (modification) to help a little with potentially too low an inductance in that interstage.

Of course strapping the input valves as triodes would lower the source impedance to the interstage and relax the inductance requirements, but say you wanted to mostly keep the pentode characteristics.
Well, an ultra-linear connection would get the anode impedance down a bit, as well as lower any distortion.

If the main anode voltage is, say, 265V, a pair of simple resistive dividers from the EF83 anodes to ground.  Lets pick 165K on top and 100V on bottom for ease (scaling them up would lessen any extraneous anode load though).  At the mid point, feed that into the grid or gate of a triode or Mosfet and tie the triode/Mosfet's cathode/source to the EF83's G2 with a bit of current care of a 33K to ground.  Triode/Mosfet anode/drain is tied to actual B+.

This would be about a 37% ultra-linear feedback and would lower the anode impedance.
There is precedent for this in a compressor, albeit using cascodes rather than pentodes so it's not totally off the wall I don't think.  Values need to be adjusted and a cap possibly added, but hopefully someone gets the point.

Anyway, just a thought I had.

 

[Heikki]

Triode or pentode with screen feedback might introduce new problems. With pentode assuming the negative feedback lowers the plate resistance significantly. With triode or screen feedback pentode the changing plate resistance from low to high impedance needs to be considered. In gain reduction how much will the changing plate resistance affect frequency response? Is the sidechain sufficient for triode or screen feedback pentode? How much more control voltage will be needed and can the sidechain provide it?

I think lowering the plate resistors to 1.5k was the smart move. Gain goes down in the the gain reduction stage with 1.5k resistors, but it doesn't really matter because there's plenty of gain after it.

 

[Winston OBoogie]

Thanks Heikki.  It was more a thinking out loud for future experiment than suggesting  Falk do this on his current unit. 
I take your point about changing anode impedance, although it's something that happens with 90% of the variable mutual conductance compressors we discuss on here, and to a greater extent than I'm suggesting from a 30+ % of ultra-linear fb.

Edit:  Side chain voltage -  looks like, with 100V on G2, then -20V is the maximum before it poops out anyway.  Probably not a compressor for deep gain reductions whichever way it's cooked.  OK for mastering etc of course. 

I checked my scribbles from messing with the Edcor in this type of use - I had a 10K across the full primary and a zobel on the secondary. 


Cheers 
   




 






 

 

[Falk]

Good Evening.

Sorry for being so awfully less responsive but the studio work keeps me busy. Thank you so much for making this thread flourish. There will definitely be some compensation for all this sooner or later. I will need some more time testing but the last mastering sessions felt great and the customers didnt feel too much distortion anymore.

@Volker.. totally missed your message. Regarding the mu metal housings I had a tough one. I got a quote for about 400€ for ONE housing from the Netherlands. I have another for 10 housings for 900€ from France and a third for 10 housings for 600€ from England. Honestly thats too expensive for a DIY project which was aimed to build the compressor as cheap as possible. I will keep myself busy on finding a solution for this.

Aside of that: Input TX is 10k:10k, Interstage is 10k:10k, Output is 600r:600r, Sidechain is 600r:2k4
The transformers are the biggest compromise in this build but for the price they do a great job. This build was never aimed to become a boutique compressor rather than a unit that works well for what it is. It has some weak spots but its too early for me to evaluate if they really bother me or if they become the weak spots that every good friendship brings.

Best regards,
Falk

 

[gyraf]

You may not NEED mu-metal casings at all, unless you are physically limited to a certain amount of space in your build.  Remember, the main benefit of mumetal shielding is that it has some 6-10 times the magnetic softness (or magnetic conductivity) of standard soft iron

..which in the case of shielding only means that you have shielding efficiency of 6-10 times the same amount of iron. This off course paramount important for tight fits like Danner-casettes with integral power transformers - not necessarily for us

..which again means: Just put in 10-times the thickness of cheap soft iron and you're golden IF you have space for this

I have had really good results just getting some 8-10mm iron sheet lasercut to profile fitting the transformer, then stacking a handful of these with a top and bottom "lid" of same material. Yes, gets heavy - but do we care?

For smaller transformers; in building supply stores you can get off-the-shelf soft iron pipes in short lengths, threaded and with fitting end caps. Yes, you probably can't ship this solution internationally for obvious reasons  but it works really nice as well..


/Jakob E.

 

[Heikki]

Falk your compressor might have less noise and hum than a certain compressor that costs 7500€

http://vacuvox.com/wp-content/uploads/2018/10/Vacuvox-U23M-Promo.pdf

If I am not mistaken Diagramm 1 is the noise spectrum. Biggest hum spike at -70 dBu is not spectacular for a compressor that has only few dB of gain. It's probably inaudible with modern pop music, but if I were to buy a compressor that cost more than any car I've ever owned I would want it to be dead quiet.

 

[gyraf]

well spotted..!  ;D

 

[JMan]

Falk your compressor might have less noise and hum than a certain compressor that costs 7500€

We’ll be waiting patiently to hear the report of the angry and belligerent phone call you receive  ;D

 

[Heikki]

We’ll be waiting patiently to hear the report of the angry and belligerent phone call you receive  ;D

I doubt he’ll find my phone number anywhere but if he does I won’t mind. He has already harassed pro audio manufacturer and studio owners, he should probably stop at that and not start harassing hobbyists.

 

[Falk]

Hello,

so again distortion issues. Today I was working on some electronic music again. Deep 909 kickdrum sounding distorted when passing through the unit. It distorts in amp mode and distorts even more when compressing. The waveform looks totally fine on the oscilloscope but its too much distortion. I compared it with my Stamchild which is dead clean at the same level. So what can it be? The low frequency distortion is not too much dependent on the level. Even if I lower the input 6dB it sounds the same.

I found out, that the wave form after the 6sn7 in the side chain (originally ecc40) looks distorted. Since its rectified afterwards it should not be too bad but what do you think this could cause audible distortion at any point?

I attach some spice simulation of my side chain circuitry as soon as the upload folder got emptied.
Thank you!

EDIT:
1)Can someone tell me what C4 in the original schematic is needed for? I remember that I had problems with it in the beginning and left it out at some point.

2)And again regarding the low frequency distortion. Can it be core saturation of the interstage (plate voltage) or the input tx (control voltage)?

3)Or can it be that a cathode voltage of 3.5V at the EF83 causes the distortion? I though by raising the cathode voltage I would gain some more clean headroom. Was that right?

 

[Heikki]

C4 is the screen bypass cap. If it is left out it will reduce gain of Rö3.

High low frequency distortion when not compressing could be caused by the transformers. Are there any specs how large low frequency signals the transformers can handle and how much DC current can the interstage transformer handle? If the plate currents of RÖ1 and RÖ2 are not in perfect balance it can cause distortion if the interstage transformer isn't designed to handle dc currents.

Edit: some more thoughts.

Signals large enough to clip the input of sidechain amp can cause distortion.

If your output transformer is 1:1, driving low impedance loads will cause distortion.

Are your plate chokes hammond 156C. That's one source for distortion if they are.

Not taking B+ to interstage transformer center tap and putting a cap in series with the primary might be worth a try. Not having DC current flow in the transformer primary makes life a lot easier.

 

[Falk]

Not taking B+ to interstage transformer center tap and putting a cap in series with the primary might be worth a try. Not having DC current flow in the transformer primary makes life a lot easier.

Hi Heiki,

thank you for your thoughts. I appreciate it a lot. I fed audio signal to the different stages. After the Interstage everything sounds fine. But as soon as I feed audio to the Input of the interstage distortion appears.

Do you mean to have a cap on each side between plate and primary inputs? Or once cap between center tap and B+?

EDIT:

I tried both, still the same amount of distortion. I also tried a transformer from Manfred Deppe in the instage position with regular wiring and the distortion occurred alongside with less low end.

But! I fed a symmetrical audio signal to the inputs of the instage and the sound is clean. So it appears it must be the balancing of the input tubes that cause the distortion. I will check for symmetry of B+ at the plate of the input tubes and will try to get a pair of matched tubes. Unfortunately I dont have a tube tester here.

Thanks a lot!

 

[Heikki]

The upload folder is full. PM me your email address and I'll email you a drawing of what I mean.

Edit:
My suggestion was to connect the primary of the interstage transformer like the primary of T2 in this schematic
https://web.archive.org/web/20180221103320im_/http://www.michaelkingston.fi/files/M670-schematic-Kingston-web.png

 

[Falk]

Good morning,

literally spent the whole day yesterday with trying to get things solved unfortunately with no results but some knowledge gained.

I fed symmetrical external audio via a 10u caps to each side the b+ of the mu tubes et voila I had super clean audio on the output which tells me that the distortions happens on the first stage. This rises some questions about the origin of the distortion that I would appreciate your thoughts about.
Image: https://www.dropbox.com/s/pzz4psn2b4wp4p2/IMG_9450.jpg?dl=0

1) Lowering the plate resistor from 12k5 to 1k5 also lowers the maximum swing of the anode voltage, right?
May this be a reason for distortion to occur?


2) When I lower the input by about 12dB distortion seems to disappear. This is btw. the same amount of gain I lost on the interstage when lowering the plate resistors.

3) I compensated for the gain loss by changing the input pad to a 1k - 2k - 1k where it was 2k5 - 320 - 2k5 before. So yep I might be feeding too much voltage to the grids. But I can only change this when having more gain in the first stage. This is because of the overall noise that would be amplified in later stages leading to new troubles

What I think of now is:

4) Changing the interstage to a 1:1 (15k) Edcor XSM (now its WSM) which has more Henries as some forum research revealed.
-> https://groupdiy.com/index.php?topic=55122.0
This would allow me to have a larger plate resistor. This giving me more gain allowing for a stiffer pad on the input and more voltage swing on the anode.

Does this sound logical to you gentleman? Any advice is highly appreciated.
(ALERT! Tube newbie might talk nonsense )

Btw. is there anyone with a tube tester in Berlin?

Thank you,
Falk

Edit: One more questions and some math:

Z = impedance
XL = inductive reactance
R = resistance

When considering the following:
Z = sqrt(R^2 + X^2)

X = sqrt (Z^2 - R^2)

for high impedance transformers R is much smaller than Z so we can almost ignore R.
In this case we can consider Z = X

From this point:

L = XL / ( 2 × π × f ) where XL is the inductive reactance given with 15k
L (60Hz) = 39H

L = XL / ( 2 × π × f ) where XL is the inductive reactance given with 10k
L (60Hz) = 26H

The thing is: How I came up with lowering the plate resistor was due to my simulation which uses 8H. I am not sure if I ever tried the original values actually. When I assume 20H (still a bit conservative) the simulation works great with original values. Maybe I really should not question what the engineers invented without computers. That makes me feel even more stupid.

 

[Falk]

I tried the original value resistors today with no good results. The search continues.

Have a good start of the week.
Falk

Edit: Some more hours of studying transformers and listening later. I noticed that one component of the distortion I notice is that the sub bass becomes longer. So there is some low frequency resonance. I was checking all the LCR combinations in the compressor for their resonance frequency and found out that the LCR group on the output has a resonance around 46Hz what is exactly the annoying thing I noticed with the 909 kick some posts before. I will try to dampen this resonant circuit and see what happens. If I get this thing solved I am sure its not far to get those units to work really great.

 

[Falk]

After a lot of testing and trying I finally got the units to play perfectly. The resonant behavior is gone and the bass is really tight now, sitting right in place. Also the overall THD is much lower now.

0.5% THD below 150 Hz
0.4% THD above 150 Hz

under full compression THD is around

up to 1.2% THD below 150 Hz
0.4% THD above 150 Hz

The units sound even better now. Although they have gained a bit more hiss now. I changed all the transformers in the audio path. Using 15k 1:1 Edcore XSM as input tx and interstage tx and a 10k 1:1 WSM as output tx.

I will test the machines again for some time and then do a final revision of the pcbs. I am updating the pcb with everything I find out and it really got better doing so.

I attach a plot of the noise. The db scale equals a dB(u) reading. RMS noise is around -70dB(u). I am really wondering why the hiss got so prominent since I swapped the input tx. Does anyone have an idea on this?

Edit: The hiss disappears as soon as I disconnect the input XLR.

Thank you,
Falk

 

[mjrippe]

I attach a plot of the noise. The db scale equals a dB(u) reading. RMS noise is around -70dB(u). I am really wondering why the hiss got so prominent since I swapped the input tx. Does anyone have an idea on this?

Edit: The hiss disappears as soon as I disconnect the input XLR.

Thank you,
Falk

That's not an XLR, that's a snake  ;D  Sorry, could not resist a bad audio joke.

So if all the xfmrs are 1:1 I don't see why attaching an input would add hiss that is not coming from the source.  Unless the way it is terminated causes a rising response?

 

[Falk]

Its terminated as this. C3 + C4 are NOT installed. Source is a THAT 1646 line driver with 50 ohms output impedance. As long as the units are not inserted the source is dead quiet.

 

[Winston OBoogie]

Do the replacement transformers have significantly more dcr?  That would imply more inductance which is good, but might explain the higher hiss.

I'm convinced that, even though you're running these into a bridging input -  the output transformer should really be a 4:1 or 8:1 rather than the 1:1 you have in there but, if it's working for ya... 

Edit:  It's possible I'm not understanding the hiss issue but, if it's such that the U23 with nothing plugged in is quieter than when your 50 ohm THAT driver chip is connected then:

What happens if you double the values of the U pad so, 2 X 2K series, and 1 X 4K shunt.

As is, with nothing connected there is a 2K primary termination reflected to the secondary.  As soon as your 50 ohm source is connected then you have 50R and 2 X 1K all in series, which are then all in parallel with the 2K primary term. so, a 1K source impedance total.

Doubling up on values would get you  2K source impedance with the THAT driver connected which might be closer to what the transformer and/or valves prefer. 
Just thinking out loud and rambling...
Good work  anyway Falk   

 

[Falk]

Hi Winston, thank you for your thoughts. I will give it a try later this week. I will also try 5:1 on the output then. I should have enough overall gain to go with 5:1 and have some 3 to 5dB left for gain makeup. We are really getting close to an amazing compressor. With the latest changes it handles low frequencies better as my Stamchild. I ll keep you posted. Speak soon.

Edit: New parts will only be delivered early June. Until then there will be no news. Aside of this - having a 600ohms impedance before the units instead of  a 50 ohms output cured the noise issues.