Jun 4, 2017

VT25, VT25A Push Pull rejuvenation part 2

On the bench for surgery.

This major changes just leave transformers, choke and the 4 pin sockets on the chassis.
I also kept the hum and push pull balance potentiometers but removed the electrolytics (replaced by Philips 021 low ESR) and made a new ground line from solid 1mm2 (17 AWG) silvered OFC wire.

Amp being almost completely stripped first thing to do was to change the 5687's socket.
Originally I used a phenolic Chinch one but after several tube tests the contacts became loose and the way I built my amp gave me no possibility for an easy replacement.
It is important to select good sockets to prevent any unwanted noise. When troubleshooting an old tube equipment, this is the first thing to check.

Today I use excellent Russian military ceramic ones with heavily silvered and super tight contacts. So tight that I can lift the amp when removing the tube! Not to be confused with cheap Russian's manufacture with thin clad metal frequently found on the web. These are the best I have on hand with old Schurter silver or gold plated ones (West German made in the 60's).

Supply choice

I had to make a decision, C or L input filter ?
Setting the amp for AB1 operation will favor choke input filter as there is some current variation depending on how deep I modulate the amplifier. In this way the choke acts like a constant current device, reason why it was widely used in class B amplifiers.

On the other hand I have noticed better dynamic with SRPP when using a CLC filter.
I finally opted for the choke input filter mainly because its inductance helps the current to change very little during the AC cycle thus providing an almost perfect DC to the whole circuit (E130 amplifier). Moreover this kind of filter is less stressing for the transformer and the rectifier and permits the use of a quite large capacitor behind choke without sacrificing the network time constant. In that way I can calculate a well filtered supply with just two cells (push pull configuration have a very good PSRR). Minimalism and efficiency.

It reminds the great electronics of the past with just two small capacitors in the main supply and no hum at all. These guys knew their job.

One important point when using a choke at input is a «starting current» through circuit. Below this minimum current the choke acts like a resistor only.
The minimum amount of Henries depends upon the total resistance in series with rectifier Rs and the internal resistance of the circuit Ri, which is the voltage to current ratio behind choke (Ohm's law).

                       Lmin ≥ ( Rs + Ri ) / 6πf                 where    f = supply frequency

Usually Rs is small compared to Ri and can be neglected so we can use a simplified formula.

                        Lmin ≥ Ri / 940 for 50Hz    and     Lmin ≥ Ri / 1130 for 60Hz

In this case, with an estimated current of 30mA (VT25A 2x10mA + 5687 10mA) @ 440V, the minimum would be ~14666/940 or 15,6 Henries. My choke is 10H / 95 ohm, to work properly I need to pump 15mA more and the simplest way is a bleeder resistor.
The resistor value will be 440V/15mA or 29,3K, closest standard value 30K and 6,7W dissipation. Need to use a 30W one and it will be hot, thus have to be cleverly located away of any heat sensitive part like electrolytic capacitor.

Two filtering cells command a quite large smoothing capacitor. For example a 200µF C1 capacitor will provide a very low impedance path to the 100/120Hz AC while keeping a time constant below 20 milliseconds for fast recovery and good transients. The ripple on C1 will be 2,64 10-1 Vpp or 9,3 10-2 Vrms (6J5 line preamp for calculation). A second cell with only 8/12µF will floor the ripple to negligible value to properly feed the SRPP.

30W bleeder resistor

Push pull balancing and hum potentiometers. Shunt resistor straight from choke to the point where all grounds will return.

PSU schematic

more to come

May 24, 2017

High End Munich 2017

Had the opportunity to get a professional pass for the MOC and I spent 3 days of pure enjoyment listening to some of the best audio gears an amateur could dream for.

Also had the chance and pleasure to meet Thomas Mayer and listen to his great sounding electronics. This usually happens once in a lifetime and it was a very rewarding.

Had a talk with Martin Brenner of Vinylista who brings interesting solutions for better LP's transcription. I was especially interested in the new Tenuto bronze turntable mat for my 301.

Took some pictures of what was the most relevant at this fair for the modest amateur and DIYer's I am.

Silbatone huge speakers system

GIP Speakers

Beautiful Western Electric reproductions.
Mr Koji Kikawa gave me a general products catalogue and I was surprised to see the number of speakers you can get from them.

Thomas Mayer great sounding electronics with ELROG tubes

I personally greatly appreciated the vinyl preamplifier coupled to a Garrard 301 on Vinylista plinth and Thomas Schick tonearm.

Kondo setup

And during my roaming I found a well recorded vinyl by Audio Note

Merci Philippe et Laurent.

May 17, 2017

VT25, VT25A Push Pull rejuvenation part 1

This amp have been my music companion for more than two decades and the only one I kept among the few I built until I came across the Philips MC1/60 triode.
It had all the qualities expected from the 10 family of triodes and I especially enjoyed the VT25 Visseaux in this configuration for its rich texture and refined sound.

Both amps remained for a while in my system but listening after listening it became evident that the big triode SE was better in terms of speed and impactness. The VT25 amp finally went to the attic as I have no room for two amps.

I never totally gave up the idea of a better VT25 amplifier but was not in the necessity to ask myself "what makes the difference between these two amps ?" Moreover reflection time being always beneficial to find solutions to a given problem, even unconsciously, I let things slowly growing.

During the past year the few electronic devices I built made me realize the essential contribution of the PSU in the final result, and naturally came to me this question : Is my VT25 amp lack of punch a supply issue ?
So, I put it back on the bench for major modifications including revisited supply, better implementation, different bias point....and new driver.

In my very first prototype the bias point was quite in the middle of the load line and the amp worked almost A1, current flowing all the time at 22mA. This is not the best for triodes push pull operation, neither for power efficiency (even if I am not really concerned with 104dB speakers) nor distortion and I decided to move AB1 with a setting current in the 8/10mA range. The biasing resistor was just increased to a 3,5 Kohm value. On the new amp the resistors are 3K, 1% sorted, for a 10/12mA flowing current. For some years now I use the vintage Sprague Koolohm non inductive resistors or Kiwame carbon in tubes cathodes with good results.

I also totally reconsidered my work which was not the most pertinent in term of good implementation: long path from the decoupling caps to the active devices, potentiometer to set input level (excellent for unwanted hum and noise), multiple wiring points (the best for ground loops), poor filament filtering and bad supply time constant. Despite these negative points the amp was performing very decently and I can expect some clear improvements with the right modifications.

Prior to change the power supply and the amp being completely stripped, I modified the driver on one amp. I wanted to take advantage of the very good results of the SRPP stage in the MC1/60 amp and performed some blind test to see (...hear) which driver was the best. The 5687 transformer loaded SRPP is the winner, no discussion. Excellent image and tempo, deeper sound stage, extended low end (no current through IT primary means wider bandwidth). Better linearity and lower distortion are this setup assets when coupled to the Tango NC14. Furthermore the good PSRR of this circuit will help to simplify HV supply. A good point and first step to an improved amplifier.

5687 µ & ρ

5687 SRPP resistor, gain and current calculation

5687 DC load line. Tube will work in a more linear region of characteristics than the previous stage

One minor drawback is a more demanding stage. I need to provide a 2Vrms input voltage when 0,9 where sufficient with the former one to swing the amp full power. However I can switch to the E182CC / 7044 for higher gain if necessary.

Good 5687 and E182CC chosen from my stock for this test

Amplifier revisited.

A 5µF coupling cap will insure wide frequency response with a -6dB cut off @ 6Hz for a 5 Kohm AC load but like for a parafeed load the coupling cap is dependent of the transformer primary inductance and is calculated large enough to avoid resonance at an audible frequency. In this case, considering a 50H primary inductance for the NC14, the resonance frequency f0 = 1 / 2π √LC is about 10Hz and will give some bass boost to the circuit. Interesting with the VT25 family of tubes that usually lacks of solid bass extension.

To be continued

Mar 8, 2017

E80CC SRPP Line preamplifier Part 3

Power supply

Unlike the 6J5 line preamp I used a CLC filtered supply. I usually prefer the LC for its very smooth ripple behind choke but listening tests demonstrated better dynamic with SRPP. A good calculation of ripple rejection and time constant helps to achieve a fast recovery supply with a minimum of cells. To minimize intermodulation the supply is splited in two after the second capacitor

Ripple calculation

On C1 the ripple is estimated by formula     VC1~ = 10 I/C     where I is the current through circuit (in mA, here 7mA) and C the filtering capacitor (in µF).
VC1~ about 2.2Vpp or 0.77VRMS with a time constant of about 7 millisecond.

On C2 the ripple is calculated by the voltage divider formula    Vout = Vin Z2/Z2+Z1    where ZL1 is 37.6K @ 100 hz and ZC2 50 ohm at the same frequency.
VC2~ is about 1.05 mVRMS with a time constant of 24 millisecond. Thanks to the 60H choke that smooths out the ripple with great efficiency. A 20H one would have been enough but I had this one on hand.

The same applies on C3 with R = 2.7K and ZC3 133 ohm. It leaves a 0.05mVRMS ripple which represent a -130dB attenuation at a time constant of 33 millisecond.


Good sounding Philips/RTC EZ81 rectifier along with old Tango Hirata choke and some Siemens MP/JS and F&T capacitors are the parts of choice for this power supply.

Tube choice and tests

Not a long or tedious quest for different brands of tubes to be tested, just two makers for this fine tube; Philips (can bear Valvo, TFK, Siemens and so on names) and Tungsram. I attentively auditioned these two competitors and the result is disconcerting. In this setup the Philips that bears a reputation of excellence is not the winner of the test. The musical rendering appears fuzzy despite some great qualities in terms of tone and speed. It gives a blurry image that leaves the observer perplexed. On the other hand, the Tungsram brings music to a level of clarity and enjoyment seldom heard. It has incredible precision and gives a density to the sound that makes it perfect on any kind of material. Very detailed, not to say clinical sound but on the good side of absolute neutrality. You like it or not but it never leaves indifferent and if you are after audio perfection you get very close to it. Furthermore this tube gives an almost holographic image of the sound stage. Great, great tube !
As expected distortion is very low. I get 0.9% @ 1KHz and 2VRMS output, mainly second order. Noise floor is also very low and hard to measure on my FFT analyzer, thanks to the very good CMRR (or PSRR, it's the same) of this circuit. In facts we have the same power supply ripple rejection qualities than in parafeed setup and some will consider the SRPP a parafeed cousin with an active load in place of the cumbersome anode choke.

SRPP line preamp completed

...in its vintage blue gray and black look. This one is now étude numéro 3, the E140 amplifier was étude numéro 2 and 6J5 line preamp étude numéro 1

same connecting organization than the 6J5 line preamp. Outputs are doubled by Lemo 0 coax outlets and this feature will certainly disapear in future works. The little male plug is a nightmare to solder considering the 0.6 mm central pin !

Last minute update...

I made a transformer connection mod. Although I did not notice any sound change, it is intellectually more interesting not to have the C1 electrolytic decoupling cap in the signal path.

some CD's and LP's I appreciated a lot, and there are many others ...

Feb 8, 2017

E80CC SRPP Line preamplifier Part 2

SRPP calculation

The heart of the SRPP is the 3 resistors used in this circuit (the upper one is generally omitted but it is bad practice if you want a perfectly balanced push pull). Easy to calculate they just depend upon tube parameters and AC load.
With a fully decoupled cathode all have the same value which is :

                                               R = Rload + ρ / µ -1,5

Current through tubes can be determined by the following formula

                                          Io = 1/2 [ Vht / 2 ρ + R ( µ +1,5 ) ]             Ohm's law...

Gain calculation is way more complicated and is an extrapolation of the simple triode gain mathematical relation

                  G = µ Rload [ ρ + R ( µ +1 ) ] / ( ρ + R )2 + Rload [ 2 ρ + R ( µ +1 ) ]

An Excel file makes a very handy tool to calculate all these parameters. Just has to be filled with tubes characteristics and AC load.

I also use it to determine the DC load RaV1. Drawing this load line on tube curves helps a lot to check linearity and allows to precisely set the working point (NB: This load to not take in consideration voltage drop in the 2 upper resistors, it is necessary to add it to anode voltage found on curves to have the exact Vak value).
This load is Io formula's denominator

                                                 RaV1 = 2 ρ + R ( µ +1,5 )

It is almost like a 47K loaded single tube and data sheet shows that we can expect very low distortion. In the case of a classic anode follower we have 4.1% mainly second order at 50Vrms!. In this preamp we deal with smaller voltages and the even harmonics will be greatly canceled by the SRPP configuration. Furthermore the 1.3K value has been selected using a HP 3561A dynamic analyzer to get the best distortion figure.
The SRPP operation will be very close to the Philips values, with a 57K load we have a 3.07 mA current for a 4V cathode voltage and a calculated gain of 18,11.

Line preamp schematic

Nothing special except an unusual input voltage divider. Gives some headroom on high level sources. Three positions, 0 -3 -6 dB permit a very fine gain control in association with the T pad. I used the excellent GRAYHILL 44 series rotary switches for both input/output selector and input attenuator. The quality is outstanding.
Like in the 6J5 line preamplifier just a few parts. Resistors are Holco H2, Philips (not Vishay) MRS25 and coupling capacitors ITT PMT/2R.

These are to my humble opinion (and to my friends ears...) some of the best caps for the price along with AEROVOX V161 and ERO MKC 1860 / KP1832. They give a very neutral and accurate restitution without spending hundreds on exotic parts for a somewhat slight improvement. I made tests with well-known pure copper or silver film and foil ones. There was an improvement in transparency or finesse on certain parts of the musical message but each time to the detriment of the overall tonal balance .....went back to the ITT's.

Looking inside...

Can't be a more compact wiring

Next step: power supply, fully assembled unit and tests

Jan 26, 2017

E80CC SRPP Line preamplifier Part 1

This project was initiated by a request of a line preamp with selectable input sensitivity to match sources with different voltage outputs. This preamp should also have a low output impedance to drive long connecting cables without any loss.
With such goals in mind it became evident that a simple 6J5 preamp would not fit these requirements.

1- The use of two gain control stacked together would greatly impact music transcription.
2- The gain of this preamp would have been too low to accommodate an input voltage divider up to -6dB.

A high gain stage with a voltage divider at input and a volume attenuator at output (which is the best way to control the gain from my humble opinion) was the way to go.

For that purpose there are a lot of good circuits with good linearity and gain, cascode, µ follower, totem pole, push pull. Push pull circuit, seldom used in preamplifier, is very interesting because it cancels all even harmonics. With such a circuit you can expect very low distortion using triodes but a phase splitter (transformer or tube) add complexity to the circuit, increases final cost and is hard to implement in a standard chassis.

One circuit shares this low distortion, high gain capability without the necessity of a splitter stage, the SRPP.

I have been working with for a long time and I know how it works (well), its limitations and drawbacks (very few when correctly understood and used).
Some would argue that it have a special sonic signature, especially on voices, with a kind of emphasis that could be appealing but do not reflects reality.
Believe me, when correctly used and adjusted (a FFT analyzer is of great help) it can be very neutral without being clinical or cold with huge dynamic and stunning resolution and accuracy on complex music.

SRPP (very) short history and use

Originally intended for video use (1943) it was widely used to drive low impedance capacitive loads with high gain and wide bandwidth.

SRPP as it should be set up.

As indicated by its name, the SRPP is a push pull even if not evident at first glance.
I won't bother you with any mathematical demonstration because it is not the main topic and there are very good publications that help to understand this circuit, Audio Xpress The Optimized SRPP and Tube Cad SRPP deconstructed.
Never forget it works like a push pull only when loaded and gain, distortion, bandwidth greatly depend upon this load. Moreover its linearity is a direct relation with the flowing current which itself depends upon load.
The good thing is if you have a well defined load that won't vary with the amplifier connected behind, you can fine tune the circuit to get the lowest distortion playing with only 3 resistors. Reason why I choose to load my SRPP with transformer.

Finding a good one was easy because I had some Tamura on hand.
The TKS20 used in this preamp is a 600/10K line input transformer but can be reversed used as there is no current flowing through primary.
Tamura are among the best transformers I had to work with. Unfortunately most are discontinued and prices skyrocket on the second hand market.

Previously I said that the best way to control gain was the use of an attenuator behind the transformer. I was lucky enough to have some Daven 600 ohm dual T pad, perfect for the purpose.
These pieces of engineering are unbelievably well made with plain silver contacts and sorted resistors for perfect tracking. They were built to last a lifetime. The only disadvantage is a 6dB insertion loss that have to be taken in account when calculating overall gain.
These attenuators are very hard to find today, especially NOS, and it is interesting to build its own T or bridged T attenuator. See Pots & Attenuators tutorial.

Tube choice

To reach my goals I needed a tube that gives a minimum overall gain of 4 (attenuator will divide it by 2). This means a SRPP gain of 16,30 (transformer voltage ratio seen by tubes is 4,08) and a tube with a µ of 20/30, a ρ of 7/12 K to correctly match the 10K dynamic (AC) load.
Double triodes like 12BH7A,13D3,E80CC would be good candidates. In this study it is stated that V1 tubes are identicals.
I finally used the E80CC because I had excellent results in the E140 amplifier and it is easy to source compared to the British 13D3. I made some simulations with the 12BH7A but gain was too low.

More to come...