Rerurn to Romy the Cat's Site

Melquiades Amplifier
Topic: The Melquiades bass channels: the coupling capacitor.

Page 1 of 1 (12 items)

Posted by Romy the Cat on 08-13-2007

I was contemplating the 6 Channel Super Melquiades conversion – it will quite a revision – and I am slowly inclining to make it not 6 Channel but 7 channels…Anyhow, the post will not be about the needs of extra 7th channel but rather about one of the revisions that I am contemplating.

The bass channel of super Melquiades uses the Milq’s driver tube and a full 6C33C. The output transformer is from 7.5Hz (at full power!) to ~900Hz (-1db) and the channel is low-passed at 60Hz, first order. Currently the bass channels stages are coupled with 4 Electrocube 950B capacitors, 2uF each that makes it 8uF. The Electrocube 950B are from my judgment are the most neutral-sounding caps among many that I have tried, however, I always when I listening the caps did it in context of full-range channels. Will the 950 Cubes hold the ground if I make the epicenter of my attention juts the low frequency response?

In the Milq I have ~200V on the 6E6P’ plate and ~-100V on the 6C33C grid. It would make that I would need >350V-400V coupling capacitor. The 8uF that I use forms ~2Hz high-pass filter. I do not need so low but still I would like to stay with 5-10uF capacitor. So, does anyone has in your mind any other cap to propose to me?

Frankly speaking, I would like to confess a heresy that it bubbling up in my mine for a quite while: I would like to use in there electrolytic capacitor, the regular plain-vanilla 10uF/450V Nichicon, or perhaps something similar, none-polarized. I have a feeling that for LF applications the electrolytic are at great advantages, have no conformation thought. Does anyone can comment on it?

Rgs, Romy the Cat

Posted by Paul S on 08-13-2007
I have also been thinking of subbing electrolytics for the huge - up to 100 uF - caps in my speaker level crossovers, just to hear it for myself.  Although everyone says to use polys in the signal path, etc., I just wonder about all that mass, if nothing else, and what an alternative would sound like.  Elna also makes a new "computer-grade" hot-rod that looks nominally interesting, but I am not sure of the availability of 450V or even 250V in the sizes I would use.  They can't be any slower than the Solens...

Paul S

Posted by Romy the Cat on 08-13-2007

 Paul S wrote:
I have also been thinking of subbing electrolytics for the huge - up to 100 uF - caps in my speaker level crossovers, just to hear it for myself.  Although everyone says to use polys in the signal path, etc., I just wonder about all that mass, if nothing else, and what an alternative would sound like.  Elna also makes a new "computer-grade" hot-rod that looks nominally interesting, but I am not sure of the availability of 450V or even 250V in the sizes I would use.  They can't be any slower than the Solens...


if your speaker use large cap values then you would hardly fine anything better then get riding of those caps and run active basing: for bass applications it sounds extremely good at speakers level. The Solens that people love to use as the “non-electrolytic caps” in the crossover are notoriously bad caps.

In my case, where I do not need a huge value of capacitance I wonder if any other caps I might use, or perhaps to imply a different type of DB blocking (do not offer a transformer). I think it will lead to some experimentations….

Rgs, Romy the Cat

Posted by Paul S on 08-13-2007
Thanks for the link, Romy.

Very interesting.

I need to run the numbers for 150 Hz, both high and low pass, 2nd order.

Cement type resistor?  Mills have gone to the dogs...

Variable inductance OK (it might be handy...)?

Have you tried steel core for the big ones?

Who winds your coils?  I bought a load of those "ribbon" type air cores years ago, have them all over, and I'm thinking they sound flat, too...

I would still try the electrolytics...


Posted by cv on 08-13-2007
Here's an idea - uses a resistor in series with a current source on the negative supply to drop the voltage to the 6c33.
Not sure how happy (sonically) the 6C33 will be with a 50k resistor on its grid. It would be huge problem at HF and cause rolloff , not an issue with a bass amp, but it's always good to provide a low impedance path to ground for the grid in high gm valves.

If a problem, might need to insert a cathode follower buffer along with all the issues that entails...

Still, some food for thought. I'll dig up another scheme that uses a VR tube instead.

Not a mllion miles from the Newton bias....


Posted by cv on 08-14-2007

If you study the above, you'll note that the topology involves some compromise, namely, a trade-off between a lower series resistor (for better drive of the following valve) versus a lower current draw through the CCS . This draw has to come through the anode resistor of the driver valve, so you either increase the HT, reduce the anode load resistor, or replace the anode load with some kind of active loading.

Oh - a 12AT7 is shown above cos that was already setup in the circuit package I used. Feel free to picture a 6e5p there if it helps.

Also, the CCS is showing -7mA cos I couldn't be bothered to flip the symbol around and type in +7mA...


Posted by Romy the Cat on 08-14-2007

Yes, Chris, thanks.

I understand how it works, the same Newton Bias only on the other side of the driver’s tube. Dima have proposed something similar in past and last nigh we discussed with him what you suggested. He was planning to post his commentaries about it; I hope he will do it later on.

Besides all technical justifications and operation warnings that might or might not be applicable for this way of coupling (I will leave it for Dima to elaborate) there is one concern that I would bring in and that would be most likely missed from Dima’s view.

When I applied to Milq’s out tube the fixed bias voltage not as it done in Melquiades currently but from a negative gas tube (appropriately, via a separate cap and resistor) then the result was completely different then applying the gas dumper to the first tube. The voltage sourcing from the gas tube applied to the grid completely sucked out life from 6C33C and made the 6C33C to sound like I play basketball in a gym with the cotton floors… I have no explanation to this phenomena….

Still, I do not say that your Chris idea should not be tried on the 6C33C but it should be trued sonically-cautiously….

Rgs, Romy

Posted by cv on 08-14-2007
Absolutely, yes - I've never tried this scheme and I am not confident about the sonics. Given the demands of a 6C33's grid - and this "on paper" - I feel the results might be better with a cathode follower after the levle shifter. I any case, would love to hear Dima's thoughts on the matter.


Posted by Romy the Cat on 08-14-2007

In fact am not only planning to try an electrolytic capacitor as the LF Milq’s coupling  (if I found one with low linkage capacitance) but I would also planning to try the “my proprietary” cooked capacitors. I have to admit that it sounds very stupid and my discovery of that effect was stupid but it the result does “sound” very nice…

It was 7-8 years ago when I built an external box for my Bidat DAC’s power supply:

The power supply has a dozen or so electrolytic capacitors, notoriously bad by default and while I was taking the transformer outside of the DAC I decided to change the filtering caps to some better audio-quality caps. I had a few sets of “better” caps, which I changed but I did not like the results and I decided to put the old caps back.

When I did it I accidentally mixed up polarity in each and single cap that I put to that power supply – the markings were opposite to the conventional and I was quite sloppy. In a few seconds after I turned the DAC up I heard the typical capacitors cracking and felt some tiny smoke.  I turned the unit off as saw that the caps near-exploded, become larger and got in near oval shape. I do not know what triggered me to do it but I reversed the polarity of the caps and turned the DAC of. The unit went off, the filter still was working. How bit my surprise was when I learned that the DAC’s lower bass become insultingly better then the bass that I was accustom to get from Bidat?

I figured out that I most likely re-polarized the caps. I soldered out and trashed the cooked “puffed” capacitors and put in the set of the better unpolarized caps. How big my surprise was what the best unpolarized caps were not able to hold ground (after they break-in) in reference to the “cooked” caps. I remember that I was diving in a garbage container, trying to fish out my trash-bag with my pulled off cooked capacitors. I was able to recover the “cooked” caps and even as now my best Bidat has those ugly “cooked” capacitors. I never changed my DAC since then and never had even motivations to do so. I never also have any reason to complain about bass problem with my digital.

So, in  few weeks when I will be listening electrolytic cap as a coupling in my Super Milq bass channel I am consider to pre-cook a few caps. It might be a little danger to use the “cooked” caps at high voltage but from a different perspective… I never turned my Bidat off and it has been running with the cooked caps for over 7 years ….

Rgs, the Cat

Posted by Romy the Cat on 08-14-2007

 cv wrote:
…..and I am not confident about the sonics….

What might be very interesting also is to learn what DC coupling in power amps would do generally in term of Sound.

The first time I thought about it was a few years back when Vladimir Lamm published in Russian magazine his observations about the negative influence of direct-coupling on bass. It’s very hard to pay attention to what Vladimir say as he is very conditioned and each his word is just an array of unfortunate deceptions that he sprays in order sell his equipment. Still, the concept is very interesting. Over the years I was paying attention to a few power amps that were direct-coupled and I might say that all those asps did have that OTL-type, “easy come and easy go”- type of bass, even all of then were use with bass-disabled speakers.

Surely, it is impossible to generalize anything based upon a few “un-methodological” implementations and I would not question negatively the direct coupling. I am sure the different implantations, different topologies and different power supplies would lead to the different direct-coupled results but the conceptual question that Lamm raised in that article was very interesting: is any fundamental damage to LF in a direct coupling between a driver and output stage.

I do not know it would be possible to model it “honestly” in context of Melquiades as in Milq the driver stage biased freakishly and it yields “different” sound according to totality unknown reasons. From a different perspective in Melquiades the grid’s current are very idiosyncratically dumped and it might open an opportunities to take advantage to direct couplings and he low-level current waves do not bounce from it’s bias power supply. Go figure…

BTW, it looks like Dima will in a few days will eventually give up and build his first Melquiades… Perhaps he will experiment with all of it…..

Rgs, The caT

Posted by deemon on 08-14-2007
Hi all !

I thought about this resistor bias idea that Chris proposed , and found some some problems here . One of them discovered Chris himself - it is a big serial resistor in the grid chain , and it can cause not only HF rolloff ( it isn't a big problem in a bass channel ) but a problem with grid current . Big resistor can cause a signal distortion when we apply a big signal to the tube and signal peaks goes closer to zero voltage . But of course we can use a cathode follower after this bias resistor ....
Another issue is an additional current that this bias circuit sucks from the input tube plate . We need to decrease the plate load resistance in order to keep the same plate current , so we must loose some gain here . 
And the third problem is the most dangerous one - it's DC instability ! When you have two tubes DC coupled , you will get all variations of the first tube operation point being amplified by the second tube , it's bad ...... it might be not so bad if we have a resistors in the cathode chains of the tubes , but in our circuit we have both cathodes directly connected to the ground , and it means big DC amplification factor of the amp . You see - if we have only 3V DC variation on the first tube plate , we'll have 120 ma variation of the output tube current ! It's too bad of course ....  
By the way , Chris , what do you think about your current source ? How do you want to implement it ? I have some idea how to use it and solve this instability problem . It seems to me that the best current source is a pentode . It has very big plate resistance , and it seems to be a good solution . But we can make this current source VARIABLE , applying some voltage on the pentode control ( or maybe screen ) grid , and make this voltage depending of the mean current of the output tube . It will be a kind of automatic bias , or servo system . For example , if the plate voltage of the first tube increase - it will cause increased output tube current , after it the voltage on our pentode grid will increase too , so our current source will produce MORE current , bias will go to negative and the output tube current will return to its normal value . Maybe we'll design a regulation circuit that won't be too complicated . But how this servo system will affect the sound ? I don't know ......

Best regards

Posted by cv on 08-15-2007

Hi Dima,
Thanks for your comments; please forgive the chaos of what follows as I have to rush!

Agreed on all counts - the dc bias point would need adjustment and monitoring; I supposse it depends on how stable the electricity is from the wall; another story elsewhere on the site...

Regarding the CCS, the best engineered I've seen is Gary Pimm's stuff - he's made many careful measurements of various configurations and ensured that the CCS maintain their impedance at HF, and are stable. There are some details here:

(thankfully, as his orioginal pages are no longer active)

Simpler alternatives include a simple bipolar - basically a voltage reference is applied to the base, giving Vref-0.6V at the emitter; this maintains a constant current through the emitter. Performance is improved by cascading.

That said, I do like your idea of using a pentode. The idea I had was to use a pentode or something like a triode connected D3a, put a solid-state CCS in the cathode, and bias up the grid with a battery to give the CCS in the cathode enought headroom.
This would multiplies up the impedance of the CCS and also shields it from a large voltage swing (down by the mu of the triode in the latter case). Ie we use the tube as the upper element of a cascode, to fix the voltage across the CCS.
 Why do I feel this is important?

Well, say we decide to go down the route of buffering the level shifter with a cathode follower, and so use a large 100K resistor in the upper leg of our level shifter in order not to draw too much current through the 6e5P anode load.

If we have a poor CCS with impedance varying from 10M to 20M with signal, then the "gain" of the shifter will vary from 0.99 to 0.995. This is about 0.5% of nasty "non-linear" solid state distortion.

To fix this, we need either a very high impedance CCS, or one where the impedance is constant with respect to level. Given sufficiently high CCS impedance, we could shunt it with a 1M resistor. This would give an attenuation of roughly 0.9 but take out a lot of the variability (if present) in the CCS.

It may be that a well designed pentode CCS is good enough - especially if you bias up the grid so you can use a higher cathode resistor (the dynamic impedance is then the gain x cathode resistance + plate impedance). My conjecture is that we could improve the performance by replacing the cathode resistor with a solid state CCS, which would be shielded from large voltage swings by the pentode; less nasty non-linear solid state artifacts. The remaining impedance would be the anode-grid capacitance (via the bias battery to ground)

Of course, how this all relates to sound is another matter! I do feel that solid state is best used when the intrinsic device non-linearities are minimised by fixing current or voltage etc.

Again sorry this is rushed, let me know if anything is unclear.

Page 1 of 1 (12 items)