Another Vari-mu - interstage thread...

[Big Bear]

Hi All,

New to this site but have been following for a number of years.

After a number of years building other peoples designs, I have decided to build my own vari-mu. Not for commercial release but for friends who own studios and the like. I do not mix / record / engineer any more but have always enjoyed building equipment so I thought a really good vari-mu would be the flavour of the day.

I have read all the threads here on vari-mu designs using interstages, a lot of very useful comments here by experienced members!

There are clearly a number of different options to pursue. I would prefer to stick with my original idea of 2 gain stages, but searching around there is a lot of good argument for more! I also certainly happy to use SS where required as I have enjoyed a lot of success with transistors in valve circuits.

I am not sure of the ethos of this site, but am very happy to start a thread with this build and share the outcomes. I am not as experienced as most here it seems but love the DIY community, and maybe this might serve as a good build thread for others to try or experiment with.

I am currently drawing up an initial diagram and have come up with my first hurdle which I hope someone can help me with;

I use Lundahl transformers a lot in my HIFI builds and plan on using the LL1660 as an interstage for this vari-mu. Does anyone have any experience with these? I currently only have the LL1660 20mA and do not own any that are not gapped for DC current.

Would the lower inductance of the primaries of the gapped LL1660 be an issue for this application? The LL1660 20mA has a primary inductance of 100H, where the LL1660PP has an inductance of 290H.

Thanks for any comments, looking forward to progressing this design!

Cheers

Bear

 

[Big Bear]

The attached image is a very rough initial idea of what I would like to do, please excuse the lack of components and component values or any little mistakes.

The idea behind the low impedance buffer is similar to the TAB U73 and a few other designs out there, some of which I have seen on this site. One question I have is does it matter where this buffer is? I have read a few members here suggest that they use a JFET buffer after the Release cap.

I may have missed the point of the low impedance circuit here.....? Would someone be able to expand on this please?

If this were a preamp I would prefer to use a current source in the cathodes of the input stage, is this possible to do with the vari-mu circuit where you want to monitor current through the cathode?

I will actually like to have an external side chain input, and sum this with the internal source pre rectification. This may need it's own amp / buffer. The practical advantages of this seem good enough to try out though!

Thanks for any comments. Learning a lot here.

Cheers

Bear

 

[Big Bear]

And the missing grounds !

 

[Big Bear]

If this were a preamp I would prefer to use a ccs in the cathodes of the input stage, is this possible to do with the vari-mu circuit where you want to monitor current through the cathode?

I guess I could monitor the current across a resistor from the HT as per the U73, and have the ccs in the cathodes still.

 

[abbey road d enfer]

If this were a preamp I would prefer to use a current source in the cathodes of the input stage, is this possible to do with the vari-mu circuit where you want to monitor current through the cathode?

The notion of a ccs is paradoxical in a vari-mu; gain varies because current is not constant.

 

[abbey road d enfer]

Would the lower inductance of the primaries of the gapped LL1660 be an issue for this application? The LL1660 20mA has a primary inductance of 100H, where the LL1660PP has an inductance of 290H.

It depends very much on the type of tube you want to use, configuration (duet, quartet, octet...), what performance you expect (in particular how much GR).
Hypothetically, using a pair of 6386 for 20dB GR, you may run them at 150V, 16mA each (Gm=5mS, Rp=3500ohms), down to 1-2mA (Gm=0.5mS, Rp=20k). You want the reactance of each half-primary to be equal to the highest Rp at the lowest frequency. For 20Hz, it computes at about 160H.
Now your gapped Lundahl xfmr, the rated inductance is with 20mA DC? If yes the value with no DC (or equal DC in each half) is definitely higher, probably about 150H, which means only 40H per half anyway, not enough.
Now you could load the secondary, that would extend the LF response somewhat, but then the gain and more importantly the max level under GR would suffer.

 

[Big Bear]

If this were a preamp I would prefer to use a current source in the cathodes of the input stage, is this possible to do with the vari-mu circuit where you want to monitor current through the cathode?

The notion of a ccs is paradoxical in a vari-mu; gain varies because current is not constant.

Hi, I appreciate that but I have seen active loads and current sources in vari-mu designs before. Maybe I am mistaken but is this 660 design not implimenting a ccs through cathode resistors?

http://www.silentarts.de/DIY/PM660/PM-schematic.pdf

 

[Big Bear]

Would the lower inductance of the primaries of the gapped LL1660 be an issue for this application? The LL1660 20mA has a primary inductance of 100H, where the LL1660PP has an inductance of 290H.

It depends very much on the type of tube you want to use, configuration (duet, quartet, octet...), what performance you expect (in particular how much GR).
Hypothetically, using a pair of 6386 for 20dB GR, you may run them at 150V, 16mA each (Gm=5mS, Rp=3500ohms), down to 1-2mA (Gm=0.5mS, Rp=20k). You want the reactance of each half-primary to be equal to the highest Rp at the lowest frequency. For 20Hz, it computes at about 160H.
Now your gapped Lundahl xfmr, the rated inductance is with 20mA DC? If yes the value with no DC (or equal DC in each half) is definitely higher, probably about 150H, which means only 40H per half anyway, not enough.
Now you could load the secondary, that would extend the LF response somewhat, but then the gain and more importantly the max level under GR would suffer.

Thanks for that, I thought that they would not be high enough but it's really helpful to see it explained.

I do plan on using a pair of 6AQ8s as I really like the sound, but am open to suggestions. Would happily move to a quad of GR valves if needed.

Are you able to comment on the low impedance SC buffer question?

Thanks for you help

 

[PRR]

> gain varies because current is not constant.

True.

A variable current source seems logical and modern. My sims say that is *not* what you want. At high levels, the cathodes must be held down by low-resistance; else the overdrive waveform is grossly distorted. <100r in the cathodes gives mild distortion.

> this 660 design not implimenting a ccs through cathode resistors?

No.

Most things that "improve" amplifiers *degrade* vary-Mu stages. Many ways different and some ways opposite of "conventional (amplifier-oriented) wisdom".

I have real doubts about a single stage after the vary-Mu. The vM's maximum undistorted output will be quite small in deep gain reduction. Yes, that is a high-gain tube; so high that its output resistance is rather large for transformer coupling.

(OTOH, a true Fairchild 660 is a single stage both vary-Mu and output booster. Massively parallel instead of building up in stages. Pick your path.)

 

[abbey road d enfer]

Maybe I am mistaken but is this 660 design not implimenting a ccs through cathode resistors?

http://www.silentarts.de/DIY/PM660/PM-schematic.pdf

It's not a ccs. It's a fixed-voltage source. Emtter-follower.

 

[abbey road d enfer]

One question I have is does it matter where this buffer is? I have read a few members here suggest that they use a JFET buffer after the Release cap.

This has been debated elsewhere.
The release resistor cannot be much higher than 200-300k because it is the grid leak for the GR tubes. If you want 10seconds release, teh capacitor must be about 5uF. The SC amp must be capable of charging the cap quickly. Typically 30-50V in a few milliseconds. If you want it quick, like 2msec, you need 75mA from the SC amp, with a nice low impedance, much less than 1kohm.
Now you could imagine reducing the cap to 1uF, increasing the release resistor to 1Meg, and need only 15mA from the SC amp. Then you may put a buffer between the release cap and the grids (via xfmr secondary). It can be done with a high-voltage FET. You will need a negative supply (about -100V). It can be done easily today; that was not the case when the 660/670 were designed.

 

[Big Bear]

Maybe I am mistaken but is this 660 design not implimenting a ccs through cathode resistors?

http://www.silentarts.de/DIY/PM660/PM-schematic.pdf

It's not a ccs. It's a fixed-voltage source. Emtter-follower.

That makes a lot more sense, thanks!

 

[Big Bear]

I found the Vari-mu META last night which has a vast amount of info on it, unfortunately a lot of the diagrams and links and now unavailable. So I will do my best to not ask questions that are covered there already.

I was hoping that I could transfer a lot of HIFI amplifier logic into this design but if this is not wise then maybe should look at other options. I do not really want to build an 8+ valve compressor with similar numbers of transformers, otherwise I could just build the poor mans 660. So SS side chains etc would certainly be attractive to me, it is a shame that the 660 with SS side chain in the META is not available as that would be good to read!

If one stage after the GR stage is not adequate then I am happy to add another stage. The benefits of an interstage between GR stage and make up gain stage are clear so I will stick with that. But is it necessary to then IT couple an extra gain stage? Or would a simple RC couple be fine here?

So for example - 6AQ8 - IT - 5965 - RC - 5965 - OP

I do already have the LL1689, that are 9:1 so would prefer to add an extra stage rather than parrallell two valves up.

( and this is where I turn you all off ) I did think that a DHT output stage would be really interesting to try, and maybe a 4P1L PP stage would be ridiculous but also sound brilliant. Might be an option at some point...

Thanks for all comments!

 

[joaquins]

I think you could avoid output TX... not recomended because of obvious reasons... The input TX I belive you need it there to have floating to put the CV there.

Maybe an option would be to use grounded input and a voltage control to catode because the gain is set with voltage between grid and catode. So changing voltage in catode instead of grid could be an option I guess, some guru here to make this clear. I don't know what happens with anode in this case...

I've seen a vari-mu with a single 12AU7 and a SS output stage, TX at input and intestage. Maybe a SS input with bias as CV may work too, but I think you are looking for all tube signal path. In this case I have PRR varimu all tube signalpath.

If you use RC filter at input to Thumb will be seen by feeding device, that could be a problem for some, not for others. In fact I guess PRR's this will happen.

Again with SS, I guess it's not imposible to avoid intestage with a floating SS stage that feeds directly the output TX. 100V floating could be made with a couple of not so high voltage BJT. Then the cathode CV I said before and the only TX is the output that is the less problematic because it's working on low-Z

JS

 

[DaveP]

Don't want to go off topic but this seems the best place at the moment.

I have wondered for a long time why an interstage is used in Vari-Mu designs, I got stuck on balance reasons at first.  I finally found out when researching the WE 1126 compressor which uses inductances across the grid resistors of the following tubes.  The write up article from Tele-Tech magazine August 1947
http://www.americanradiohistory.com/Tele_Tech_Master_Page_Guide.htm

This says that the coils are dumps to ground for very low frequencies....thumps.  The same must apply to an interstage which dumps to B+ an AC ground.  For this to be effective, the inductance must be high enough to form a low reactance pathway.  This also has implications for the following grid resistor which is usually on the low side 100~150k.  The article also explains why there are massive 1M grid stoppers on the output tubes.
best
DaveP

 

[Big Bear]

Hi JS,

Thanks for your response. The SS side chain in PRRs compressor looks ideal and exactly the sort of thing I would like to try out. I was of the opinion for a long time that all valve is all better, but am really enjoying the benefits of solid state devices and am happy to use them where needed. Generally I would prefer all valve signal path but am happy to use source followers and transistors in CCSs etc.

I can't find any Primary Z / secondary Z details for the side chain transformer on these designs, do you know what is commonly used?

Again, rough but here is another possible option.

 

[Big Bear]

Don't want to go off topic but this seems the best place at the moment.

I have wondered for a long time why an interstage is used in Vari-Mu designs, I got stuck on balance reasons at first.  I finally found out when researching the WE 1126 compressor which uses inductances across the grid resistors of the following tubes.  The write up article from Tele-Tech magazine August 1947
http://www.americanradiohistory.com/Tele_Tech_Master_Page_Guide.htm

This says that the coils are dumps to ground for very low frequencies....thumps.  The same must apply to an interstage which dumps to B+ an AC ground.  For this to be effective, the inductance must be high enough to form a low reactance pathway.  This also has implications for the following grid resistor which is usually on the low side 100~150k.  The article also explains why there are massive 1M grid stoppers on the output tubes.
best
DaveP

Thanks so much for that link, I have been looking for academic or commercial papers on this subject for a long time and it can be quite tricky to find. Will be very helpful for my studies!

 

[joaquins]

I'm thinking in build something like this one, but with DOA at output...

For my discrete mixer I don't think a DOA would harm... Also I don't see the point to avoid SS stages in a world that is SS... It's nice a tube gain stage, but to avoid OP TX you could use a SS output to get low Z... same for input and you could use a more easy to get IP TX and work at lower Z...

So  SS-buffer/TX/6AQ8/interstage/5965/RC/SS-buffer  for you... for me, SS/TX/12AU7/TX/SS. If I could use a current mode TX for interstage would be nice for me... I can wind my own output TX that would work for the input with a SSbuffer before it... I would like to use something I can made for interstage. I like to do as much as I can and learn in the process and if it ends in something that sound nice I really happy, if not, I keep modding and always is time to get a nice TX to replace mine.

JS

 

[abbey road d enfer]

I have wondered for a long time why an interstage is used in Vari-Mu designs, I got stuck on balance reasons at first.  I finally found out when researching the WE 1126 compressor which uses inductances across the grid resistors of the following tubes.  The write up article from Tele-Tech magazine August 1947
http://www.americanradiohistory.com/Tele_Tech_Master_Page_Guide.htm

This says that the coils are dumps to ground for very low frequencies....thumps.  The same must apply to an interstage which dumps to B+ an AC ground.  For this to be effective, the inductance must be high enough to form a low reactance pathway.  This also has implications for the following grid resistor which is usually on the low side 100~150k.  The article also explains why there are massive 1M grid stoppers on the output tubes.
best
DaveP

Mind, there is a HUGE difference between using inductors and a xfmr.
In the former case, the usual RC circuit is replaced by an RCL one, which creates a 2nd-order HPF instead of a 1st-order one. Indeed it reduces thumps, but in a measured amount.
Let's say in both cases you want 50Hz -3dB LF point (remember it's FM broadcast, 50Hz-15kHz BW). In the RC case, you can do 0.15uF/220k. In the RLC case, the cap goes down to 0.1uF and the inductor must be a whopping 8000 H!
Then you have to make a significant change. This RLC circuit is going to be the governing element in the overall LF response of the unit. Some calculations later, you end up with C=40n (0.04uF), L=600H and R=100k, for a -3dB point @30Hz and -0.1dB at 50Hz.

Now let's submit this to the rigors of a pulse with an attack time of 1ms. That's the pulse that comes from the side-chain/rectifier, to the grids of the transconductance tubes and amplified by them, and then hits the second-stage grids.
In the 1st case (5Hz RC) the resulting pulse creates a huge DC perturbation for about 150millisecondes, in particular with a large negative valley that may very well put the second-stage tube in cut-off; talk about a huge click in the speakers.
Second-case (hypothetic 5Hz RLC), the perturbation would be worse, about 200msec long with a positive peak following the negative valley.
Third case (30Hz RLC), the first peak lasts only about 6msec, the negative valley comes at 13 msec and the perturbation is gone at 20msec.
Indeed that's a serious improvement, considering that psycho-acoustical masking intervenes in the 20 msec following a transient.
Now you can ask why don't we choose an RC at a higher frequency and make it the dominant zero in the response? Well, in order to have a nice 50Hz response -0.5dB), we need a 16Hz -3dB point. I that case, the peak lasts for about 20msec, which seems about right...but in fact there's a negative valley that lasts for about 100msec that's gonna play tricks with the operating point and gain of the second-stage.

OTOH, none of this occurs with a xfmr, because the disturbance voltage is cancelled out by the differential nature of the xfmr.
Considering the design constraints that are put on the grid coils (high inductance, low stray capacitance, reduced saturation, shielding,...), I dont think a pair of them is a significant saving over a xfmr.

 

[abbey road d enfer]

This also has implications for the following grid resistor which is usually on the low side 100~150k.

The grid resistors are not needed anymore for galvanic path of the grids to gnd, but they are part of the RLC filter and define its Q.

The article also explains why there are massive 1M grid stoppers on the output tubes.

That's not exactly what I've seen. They explain the role of the small value res R25 & R26 for preventing grid-current toslap back in the previous stage, though.