srx revisited

[PretentiousFood]

JimL, you're absolutely right- thanks for the comments. The measurements were taken with bias applied to the grid, and a resistor (Rs) between source and ground. The output impedance really isn't that of the CCS.

 

I redid the measurements with the device connected as a current source (which I think is the same thing as you were describing, nikongod?) and indeed measured very close to 300K for 28mA or so. Much better.

 

 

Ignore Vg, the CCS was connected as a two-terminal circuit. This was tested with a 100R gate stoper. I repeated the measurement with a 1k gate stopper, and the impedance predictably dropped by a factor of 10.

 

I'd be curious to see if we can't do some sort of hybrid ring-of-two CCS that benefits from the high impedance of DMOS and the extra feedback.

[JimL]

luvdunhill:

No, I haven't.  Jung's measurements indicate that this provides an extra 10 dB improvement at 1 kHz and below, which is about a 3-fold increase in dynamic impedance.  I actually built my current sources as mini hard wired circuits, so the extra complexity, although small, gets to be a bit much with resistors and capacitors hanging out in space - it's already looking a bit like a free-form solder sculpture if you know what I mean

 

I don't think the increased impedance would make any difference in the current tail sinks for the differential amps.  It might make a difference in the output current loads since the improvement is mostly at low frequencies, where the headphone impedance is the highest.  However, even with the standard cascode 10M90S/DN2540 combination, nearly 90% of the signal current, and two thirds of the signal power is going to the headphones at 20 Hz, and by 100 Hz 99% of the current and 90% of the signal power is going to the headphones, so I don't know it it will make an audible improvement.

[congo5]

  Hi; I built this amp, boards and case. And while it looks homemade, it works so well that I want to know why?

What am I comparing too?......  KGSSHV, Megatron ,KGST,Egmont.

 

 

  other than the excellent details JimL posted in this thread this is the best information I have found:

 

  Spritzer said:

 

I've been giving this some thought and perhaps the best candidate for rework would be the ESX circuit. Three stages, mostly DC coupled and arguably, one of the finest tube circuits I've ever seen. The money we save on the various caps can be spent on a CCS for the output and even for the front end as well. With a C- supply there will be no offset issues and with a PSU of say +/-300V it will be a very capable amp. This is naturally a Stax circuit so I see no better way for the Mafia to show it's gratitude for 50+ years of great products but by reworking one of their designs. Here is the basic design:

http://www.bonavolta.ch/hobby/en/audio/hdgsta.htm

The second is what parts to use. ECC81 and ECC99 are the obvious choices and the resistors can be what ever that fits but the CCS is a bit of a challenge. I'd really like isolated parts to make assembly easier (no high voltage pads etc, required) but I'm coming up empty. 2SA1968 is an obvious choice but there is very little stock left and I'm not dipping into my stock to supply a group build. Any ideas for a suitable part?

and later said:

 

 

I'm not too keen to use the 6SN7 at this level even it it is a GTA/B.  I ran my "Stúfur" design with GTB's at 600V in grounded cathode and it wasn't ideal.  That was a cheap "volksamp" attempt so a small compromise wasn't a problem.  That said I did have the SRX pcb design open yesterday so I might build some.  I need to make it about half the size and have the tube sockets face the right way. 

 I am hoping some of the veteran builders will do this amp this year,

 

Maybe I got lucky with the 1950's Sylvania green label SN7GTA's / and not so Lucky buying  tubes for the KGST.

Either that or with all the CCS's this amp is just really good.

 

  thanks

[spritzer]

I've been keeping this circuit in the back of my mind and it could work to use ECL82's with it.  That is a true 600V tube so triode strap it and use the built in triode for the second stage. 

 

I'm going to do something similar for a mini megatron to match what I did with the Octave. 

[JimL]

Thanks for your comments, pongo5.

 

Spritzer:

Actually the 6SN7GTA/B is a pretty good tube for this amp.  The RCA tube manual specs are:

 

Max plate voltage:  450 volt

 

Max peak plate voltage:  1500 volts   NOTE!!!

 

Max plate dissipation:  5 watts/plate, 7.5 watts both plates combined.

 

Amplification factor:  20

 

Note that these tubes were used in a lot of early TVs for vertical oscillators and deflection oscillators, same kind of duty as 6S4A tubes were designed for. - hence the peak plate voltage rating.  When I was debugging my SRX the tubes would sometimes be sitting there for a few minutes with 600+ volts from cathode to plate - no harm done.  They tested exactly the same on a tube tester afterwards as before.

 

 

By comparison:

6S4A

Max plate voltage:  550 volt

 

Max peak plate voltage:  2000 volts

 

Max plate dissipation:  8.5 watts

 

amplification factor:  16.5

 

 

ECC99:

Max plate voltage:  400 volts

 

Max peak plate voltage:  none specified

 

Max plate dissipation:  5 watts, dissipation for both plates combined not specified.

 

Amplification factor:  22

 

EL34

Max plate voltage:  800 volts (Mullard specifies 600 volts when triode strapped)

 

Max plate dissipation:  25 watts

 

Amplification factor (triode strapped):  10.8

 

 

The 6SN7GTA/B is at least as good a tube as the ECC99 if not better. - higher plate voltage, equal power dissipation, specified peak plate voltage of 1500 volt which is more than it will ever see in the SRX circuit.  Doesn't match the 6SA4 (great choice by KG) in power but it's not far off in voltage specs, I've found matching between tube sections to be excellent (generally less than 5-10 volts difference in circuit) and in the SRX I'm running it at about 4.5 watts total dissipation - 325 volts and 7 mA/plate.

 

Doesn't sound like much, but with the cascoded current sources, driving the current loads to clipping requires less than 5 MICROamps, so 99.9% of the standing current is available to drive the headphones.  This makes a major difference compared to using resistor loads (blechhh!) where the headphone (high impedance) is constantly fighting a losing battle versus the load resistor (low impedance by comparison) for the signal current.

 

The EL34 used in the ESX variant has a lot more power, but because of its lower amplification factor when triode strapped, plus its higher Miller capacitance, the closed loop frequency response barely makes it out to 20 kHz before rolling off.  And that's with current sources, if you use load resistors it starts to roll off earlier.

 

With 6SN7GTA/B tubes due to the higher amplification factor and slightly less Miller capacitance, the open loop response rolls off above 11 kHz, and the measured closed loop frequency response is flat to 20 kHz and r-3 dB at 46 kHz - that's at 100 volts RMS output into a 100 pf dummy load.

 

The 6S4A does very well here, slightly lower amplification factor but less Miller capacitance, the open loop response rolls off at about 20 kHz (calculated).

 

The issue with frequency response occurs because the input stage is a cascode, which has the disadvantage, as a driver stage, of having a high output impedance.  In combination with the Miller capacitance of the output stage, this results in the open loop response of the circuit rolling off at higher frequencies.  For example, with the 6SN7GTA/B it is -3dB at approx 11-12 kHz.   With the EL34 (I measured the grid to plate capacitance as about 10 pf it rolls off at about 8 kHz.  With the ECC99 it rolls off around 9 kHz.  The circuit doesn't have a lot of excess gain to begin with.  With the 6SN7 there is about 14 dB feedback (5-fold) at lower frequencies, rolling off to no feedback at around 50 kHz.  With the EL34 the feedback is around 8 dB (2.5-fold) at lower frequencies.  If you do stupid stuff like Mikhail did with substituting tubes for the input section you may wind up with no feedback and the circuit running open loop.

 

So if you're wondering whether I thought about tube choice a bit - yes, I did.

 

If you use ECL82s with the pentodes strapped as triodes you need to see if the amplification factor is high enough and the Miller capacitance low enough to make a good combination.  

 

I've just resubmitted my article on this to AudioXpress - apparently they didn't get my previous submission - where I discuss this and other things in infinite gory detail.

Edited April 4, 2015 by JimL

[spritzer]

My problem with the 6SN7 comes from not finding any evidence that the tube was actually redesigned to fit those parameters.  Sure they altered the filament a bit so you can series wire them in TV's but other than that they seem to be bone stock 6SN7's.  Call me a cynic but it smells of "sure we can push to that" as we have millions of these tubes in stock. 

 

The ECC99 spec sheet is very lacking but it's a sort of a hybrid between the 6n6pi and the 12BH7A.  No need for peak voltage in those as it's only applicable in TV use.  It's a worthless measurement for our use. 

[JimL]

Well, the plates on a 6SN7GTA/B are larger than those of the early non-A/B types, and about the same size as the ECC99.  Remember when the GTA/B were designed in pre-transistor days - if they couldn't meet the published specs we would have had a lot of consumer TVs blowing up and the tube company would have gone out of business.  Now, later there were higher voltage and power tubes for color TVs such as the 6BL7, etc., but color TVs needed the higher voltage and power.  No indication as far as I know that the GTA/B were in any way lacking for TV applications at the time they were in widespread use.  

 

Of course in the old days consumer equipment was designed conservatively for long life and reliability, but still, designers probably designed to, say 70-80% of max voltage and 65-70 of max power, which is what I did for the SRX revision.  I grew up in the 50s when B&W TVs were very common and there were no issues with TVs blowing up, period.  Also remember that later 6SN7s probably used the bigger plates just for convenience sake - no reason to keep the older lower power design in production when the A/B had the same specs.  You need to compare the plates on a pre-WWII 6SN7 to a 1960s GTA/B.  The 1950s GTs probably used the same plates as the GTA/B for simplicity of manufacturing - in fact, on Tubes Asylum one poster stated he saw a carton of 1960s TungSol 6SN7s of identical construction with some tubes labelled GTB and others labeled GT.

 

The peak plate voltage is simply an indication that the design will take that kind of voltage w/o failing, sparking, etc.  For stat amps where the highest cathode to plate voltage they will see is less than 700 volts (for a +/- 350 volt supply) this means that there shouldn't be any issues in terms of the tubes being damaged by too high a voltage.  So when someone says the 6NS7 is a 450 volt tube running at 700 volts - NO!  It's a tube capable of withstanding 1500 volts running up to 700 volts.  And in the SRX circuit it's sitting at about 300-350 volts at rest - perfectly fine.

 

I know you like to build lots of stuff - I suggest you build this with 6SN7GTAs, then if it sounds lousy or blows up, come back to me and we'll look into why. 

Edited April 4, 2015 by JimL

[JimL]

I would add one other comment - whether or not you see or read of any design changes between the GT and GTA/B versions, remember that when the 6SN7GTA and B were designed in 1950 and 1954 there were no transistors, just tubes.  They were designed and built by RCA, GE, Sylvania, all the big manufacturers, and used in all sorts of electronic instruments, industrial electronics, military equipment, etc.  Engineers design using manufacturer specifications.  Do you REALLY think that if the tubes were under designed, unreliable and didn't meet its specs hat it would have become as popular as it did?  Sorry, and no offense, but  cynicism or no cynicism that makes NO FUCKING SENSE!!  And I mean that in the nicest way. 

[luvdunhill]

Well-said

[spritzer]

That post shows a huge lack of knowledge of how the tube industry operated back in the "good old days".  Manufacturing tolerances were crap to say the least and everything was done to keep costs down.  There is also a huge difference between just about works and works comfortably. 

 

Let's also look at what was happening in the early 50's, that's when the improved novals are being phased in so they had to find some use for their obsolete octal tubes.  A manufacturer "inventing" new ways to use an existing product when the market place shifts, perish the thought... that would never happen. 

[Earspeakers]

All of this is generalities. In manufacturing roughly there's a sliding scale, with Satellite being the most rigorous (can't fix it after launch). Next is critical infrastructure (e.g. nuclear power plants), defense and scientific instrumentation. Then is general infrastructure (cellular, power system, etc). From there it goes into consumer which is hit or miss depending on the application and industry. I don't know the tube manufacturing explicitly but I'd be surprised if it wasn't similar. Many lines being like that, but surely the JAN products for A&D were done under Military contract as they've always been. Trust me, those contracts are explicit as to the performance. 

 

On par though Spritz is probably right, especially with the 6SN family which was general purpose. Plus these tubes are all long past their expiration date.  

Edited April 5, 2015 by Earspeakers

[JimL]

So there's a difference between how industry operated then and how it works now?  Keeping costs down has ALWAYS been a part of it.  But my limited experience with random 6SN7GTAs from the 1950s is that in the SRX circuit the plate voltages between two sections of the same tube are within 5-10 volts of each other and with Sylvania JAN 6SN7GTA tubes from the 1980s is that plate voltages are within 15 volts of each other.  So are tolerances between two transistors nowadays better?  Gimme a break! Here is a quote from Robert Tomer in "Betting the Most out of Vacuum Tubes" published in 1960, "It has been correctly stated by competent authorities in the field of industrial management that it is cheaper to make a quality (uniform) product, than it is to make a poor quality (non-uniform) product....Rejects add enormously to cost, and therefore, the fewer the rejects, the lower the manufacturing costs. ...no manufacturer can afford to produce a product that doesn't meet good quality standards."  That was a common belief at the time.

 

As to your comments about lousy quality control in tubes, you're chronology is off.  In the late 60s and 70s when tubes were considered obsolete and the big manufacturers were using the old machines to produce tubes, and/or selling the machinery to Asia, etc., yes, quality deteriorated. But in the 50s when tubes were THE electronic devices quality control was still good.  That's why a lot of tube buffs prefer NOS tubes from pre-war to the 1950s and wary 60s to tubes made later.

 

The point is that regardless of tolerances the tubes fulfill the basic specs for voltage, current and power rating.  Companies don't make money if their stuff breaks down frequently when used within spec.  That's basic engineering regardless of era.  And regardless of what you think of RCA, GE, etc., the 50s was when the Japanese sent their people over here to learn how to do quality control.  And even though novels were new, they were still designing circuits with 6SN7GTA/B in the 1950s - witness B&W TVs.  

 

BTW, I made a mistake saying that the pre-war plates were smaller - did that from memory but I looked at some pre-war 6SN7s and some 6SN7GTAs and they are about the same size.  However, when I compared a Sylvania 6sn7GT with a late 6SN7GTB, although the plates are basically the same size, the GTB had a doubly long "tail" where the plate was stapled together compared with the earlier tube.  That tail does give some extra cooling capacity.

 

To Earspeaker's point, NOS tubes generally don't go bad unless they've developed a vacuum leak.  People are using NOS tubes from before WWII which still meet specifications and work perfectly.  In fact until recently tube audio geeks uniformly recommended old tubes from the 50s and before as sounding better and lasting longer than new tubes.

 

So here's the bottom line:

Even if every 6SN7GTA/B out there just barely meets its specs, in the revised SRX with +/- 350 volt PS the standing voltage is 335 volts cathode-to-plate voltage which is less than 75% of maximum, and 7 mA standing current, so the combined power dissipation is 63% of maximum.  If you're worried about it, use JAN (Joint Army Navy) military tubes that have been tested to meet specs.  But even with commercial tubes, run as conservatively as they are in this circuit, the only way they would blow up is if they were defective in the first place.

Edited April 6, 2015 by JimL

[luvdunhill]

Perhaps if Rudistor hadn't used it in an 'stat amp your analysis might be accepted.

I think it's fine, enjoy!

[JimL]

Good point.  When someone makes a crap amp it poisons the well.  But when properly used with current sources it's a good tube within its limits.  Not as powerful as a 6S4A, not to mention a triode strapped EL34, but the very efficient current sources really help make up for the relative lack of power.

[JoaMat]

 ..... And while it looks homemade, it works so well that I want to know why?

 

I really like the homemade look and I can’t see any reason why it shouldn’t work well.

 

Is that Kevin Gilmore's layout you have used? Isn’t the tubes sticking out on solder side?

 

More pictures, please.

[kevin gilmore]

that is definitely my layout.

 

wish i could make boards at home, i could waste even more time

and make the house smell of ferric cloride.

[JimL]

Confirmed, pongo5 told me he used KG's board files.  I believe he also used the KGST PS board.  However, the circuit is simple enough that it can be wired point-to-point, which is what I did

[congo5]

I really like the homemade look and I can’t see any reason why it shouldn’t work well.

 

Is that Kevin Gilmore's layout you have used? Isn’t the tubes sticking out on solder side?

 

More pictures, please.

Yes Kevin's files and His kgbhultraminipsV4 PS,  The tubes are out the back like on Megatron.

pic's on other site as I can't figure out how to do it here.

Made the boards on a DIY Microcarve CNC (dust and noise but no ferric cloride)

and yes it is a fun way to waste a bunch of time>

Thanks to Kevin for handing out the files.

 

  Question:   the SRX sounds much better than my KGST, which was built using same parts as my KGSSHV.

while the KGST has more gain than SRX, it is also harsh and the bass is not as good.

no complaints with KGSSHV

 

  poor choice of tubes?

[nopants]

What are you trying to say, that the KGST is worse because it doesn't use 6SN7s? I don't really understand the comparison.

[congo5]

What are you trying to say, that the KGST is worse because it doesn't use 6SN7s? I don't really understand the comparison.

I'm saying that I did something wrong,,,  It should sound more like the others.

 

I didn't get the right tubes for it?  (good quality 6s4a's)

[nopants]

Sorry I read this right after the previous exchanges.

 

What tubes are you using, and have you tried other quads? I assume you aren't having offset/balance issues

[JoaMat]

/

Edited November 11, 2015 by JoaMat

[JoaMat]

JimL,

 

I'm wondering about the way you are handling the adjustment of the offset on srx revisited. I suppose you choose the current sink method for some good reason? Do you mind telling?

[JimL]

Sure.

So the current loads are there for obvious reasons as spritzer has often posted about over the years, and I discuss in my new post on output stage requirements elsewhere on the DIY section. The current sink is there to improve the balance of the output stage (it does the same for the input stage which is useful because I'm running it single-ended on the input side) and also, in combination with the current loads, completely isolates the output tubes from the power supply and from the other channel. This improves channel separation, and improves the differential operation of the outputs. All the output tubes see is the signal from the input stage, the headphones on the output side, and the other tube section in the channel via the cathode current sink enforcing balance.  This results in optimum function of the output tubes (modest, ain't I?  )

However, it is impossible to build output loads and sinks that match each other down to the last electron. So instead of trying to match them, I deliberately set the current sink to soak up about 3 mA more current than the two current loads combined put out.

To absorb this extra 3 mA, I put in the resistor string that runs between B+ of +325 volts and the current sink, which is sitting at about -310 volts. Varying the resistance of the string varies the cathode-to-plate current, allowing the plate voltage to be set close to ground. Now, this circuit does not allow balancing the plate voltages independently, so that the tube sections have to be well matched. In practice, with NOS tubes from the 1950s I've been able to get the output voltages within +/- 3 volts of ground with a couple volts drift, and with tubes from the 1980s I've been able to get the output voltages with +/- 6 volts of ground, which is more than good enough.

Edited April 9, 2015 by JimL

[nopants]

I can't contribute much to this thread, but I was I was reading about the use of 12SN7GTA's over the 6BX7 because of better matching between sections (Morgan Jones, Daughter of the Beast). I assume 12SN7* is a valid option if you're willing to tweak the filaments.

Edited April 7, 2015 by nopants