Trilogy H1 electrostatic amp

[spritzer]

Now that this amp has been released and we finally have an internal pic, it is time for a dedicated thread.  Here is some info about the launch event:

http://hifilounge.tumblr.com/post/162193518708/trilogy-h1-energiser-for-stax-earspeakers-launch

So 5000UKP... and nothing else which is relevant at all.  The usual HF retards claiming this was the second coming and all that BS but here are the internals...enjoy:

Not the best quality but good enough to know what is going on.  In before Jim and yup...this looks a lot like an SRX.  Some Russian dual triodes for the input and those look a lot like 6N6pi's to me.  Two plate resistors as well. 

So over all, a pretty horrible way to spend 5000UKP... 

[Ledzepp]

So 5000 pounds is about USD6500. About the same as the T8000.

[kevin gilmore]

There is only one pair of coupling caps, so it pretty much has to be Srx or Srx like

so the caps are 400v, so voltages roughly the same as t8000

 

Edited August 14, 2017 by kevin gilmore

[mypasswordis]

Couldn't the input tubes be cascoded like the T2 instead of the Srx's LTP into a common cathode with the plate/cathode resistor doing double duty?

[kevin gilmore]

Not enough voltage gain that way

 

[JimL]

1 hour ago, mypasswordis said:

Couldn't the input tubes be cascoded like the T2 instead of the Srx's LTP into a common cathode with the plate/cathode resistor doing double duty?

The SRX input  IS a cascoded LTP, AKA Hedge circuit, with cross-coupling per D.R. Birt's article "Self-Balancing Push-Pull Circuits" June 1960 in Wireless World to improve balance for a single-ended input.  The T2 also has cascoded LTP/diff amp input but CCS tail and plate loads but w/o cross-coupling.

 

If it's a variation on the SRX, so much for "...a completely clean sheet, creating an all-valve, zero feedback, fully differential design with a direct coupled output stage."  A limitation of the SRX input circuit is its high output impedance, which into the Miller capacitance of the output stage rolls off the open loop frequency response above 11 kHz, requiring feedback to flatten the response..

 

And, plate resistor outputs is so last century.  

Edited August 14, 2017 by JimL
clarification

[mypasswordis]

The definition of a cascode is a common emitter into a common base (or the tube equivalent), please tell me where you find that in the SRX.

4 minutes ago, JimL said:

The SRX input  IS a cascoded LTP, AKA Hedge circuit, with cross-coupling per D.R. Birt's article "Self-Balancing Push-Pull Circuits" June 1960 in Wireless World to improve balance for a single-ended input.  The T2 also has cascoded LTP/diff amp input but CCS tail and plate loads but w/o cross-coupling.

 

srx schem.pdf

[guzziguy]

????

Edited August 15, 2017 by guzziguy
How did this get here?

[JimL]

First, ignore the cross-coupling for a minute.  If you fix the voltage of the grid of the upper tubes, you'll see that you have a cascode circuit with a common cathode for the lower tubes and a "grounded grid" for the upper tubes.  Now, with the cross-coupling, what you have is the plate signal from the left lower tube going to both the cathode of left upper tube and the grid of the right upper tube, and vice-versa.

 

Remember that plate signals from the two lower tubes are out of phase.  So what the upper tubes see is double the voltage signal between the cathode and grid that they would without the cross-coupling.  It's still operating as a cascode, just that the voltage signal to the upper tube effectively doubles because the grid and cathode signals are out of phase.  If we analyze the upper tube in isolation, it looks like it has double the gm (change in current out for change in voltage in) because the same grid signal now results in twice the variation in current due to there also being an anti phase cathode signal.  This results in twice the voltage gain for the upper tube.  However, for the lower tube,  as it is feeding both upper tubes its current signal is split for the same voltage input, so its gm is halved.  The result is that the overall gain is about the same as a cascode without the cross-coupling.

 

So if there is no change in the voltage gain, what is the point of the cross-coupling?   Well, if there is a difference between the output signals of the lower tubes due to a single ended input, then the cross-coupling tends to balance it out because the output of the lower tubes feeds both the upper tubes.

Edited August 15, 2017 by JimL
clarification

[mypasswordis]

I just spent awhile trying to find the article, can't find it so clearly it is a very popular and well-regarded article in textbooks and circuits today.  This is fun, I love playing devil's advocate even though it is clearly not a cascode. What do you think the open loop gain and output impedance of the front end of the SRX is, disregarding and with Miller capacitance? I'm going to put it into spice and do a comparison with a real cascode.

Edited August 15, 2017 by mypasswordis

[JimL]

The first place I saw this circuit was in an electrostatic amp design by Joe Curcio using 6DJ8 tubes, which was published in the very first issue of Glass Audio (Vol 1, no. 0), 1988.  That's where I found the reference to the Birt article.

 

I had to go to the MIT library to find a copy of the article on microfilm.  I have it in a pdf file so if you PM me, I can send you a copy.  Here is a quote from the article (part II):

"From a discussion of the general principles in last month's article, the requirements for the first stage of a practical amplifier are now fairly clear.  We have seen that for good balance and low push-push [as opposed to push-pull] gain, [tubes] V1 and V2 should have a high gm and large anode impedance......

 

A better plan is to substitute cascode stages for V1 and V2.  It is generally possible to achieve a higher gain in this way, and a screen grid supply is not required.

 

The grids of the upper triodes of the cascode pair require ideally to be at a constant potential relative to cathode.  This is not a difficult problem, as we may decouple the grids to cathode, and make the grid feed resistor large.

 

Alternatively, a cross coupling arrangement can be used as shown in figure 9 [this is the cross coupling the SRX uses].  The operation of this circuit is relatively interesting.  When a push-pull signal is applied, it can be seen that the drive to the upper triodes of the cascode pair is applied to both the grid and cathode, in anti phase.  As far as the cathode circuit is concerned, this turns out to be equivalent to doubling the  gm of the upper valve, and therefore the cathode impedance is halved and the voltage gain to this point is halved.  However, the grid-to-cathode voltage of each upper triode is the same as it would  be in a conventional cascode amplifier, and the overall gain is similar."

 

My comments and clarifications are in [ ].

 

The article then goes on to plot the V-I characteristics of the cross-coupled circuit using 12AX7 tubes, which, surprise, look just like a standard cascode, i.e. pentode-like.  You may not like to call this a cascode because it doesn't keep the grid of the upper tube constant, but considering that cascode was originally coined as a contraction of CASCaded triODE, that the connections of the two tubes are the same as a standard cascode (plate of lower tube to cathode of upper tube), the same current runs through both tubes, and that the V-I curves of this circuit are the same as a standard cascode circuit, I don't know what else you could call it.  Just don't let those two extra lines confuse you.

 

Note that the SRX input circuit is NOT an LTP into a common cathode because of the connection between the PLATE of the lower triode to the CATHODE of the upper triode ON THE SAME SIDE.  The common-cathode-like connection of the PLATE of the lower triode to the GRID of the upper triode is on THE OPPOSITE SIDE.  This is the cross-connection.  Take it out, substitute a fixed voltage to the grid of the upper triodes and you have the Hedge circuit, which is a cascode differential pair (balanced) or cascode LTP (single ended).

 

Given that the gain of the cross-coupled circuit is the same as the standard cascode, I calculated the gain as about 41-42 dB, and an output impedance per side of close to 147 kilohm, using the estimated rp and gain for the low current used in the SRX circuit. That's with my 250 kilohm plate resistors rather than the original 300k.  If you use the standard published 12AT7 parameters, you'll get a slightly different answer.

 

The input capacitance of the input circuit is low because of the cascode connection.  The issue is the high output impedance of the input stage into the output tube Miller capacitance.  With the 6SN7GTA/B, the Miller capacitance is about 84 pf, and the circuit rolls off above 11-12 kHz.  There is about 14 dB feedback at low frequencies, and so the calculated closed loop gain rolls off above about 53 kHz, which is what I measured.

 

"

Edited August 15, 2017 by JimL

[JimL]

So, what is the circuit board at the bottom of the box?  Regulated PS?

[kevin gilmore]

does not look regulated to me.  at least not the high voltage, just a diode bridge and a pair of caps

low voltage looks like 3 terminal regulators

[mypasswordis]

14 hours ago, JimL said:

Given that the gain of the cross-coupled circuit is the same as the standard cascode, I calculated the gain as about 41-42 dB, and an output impedance per side of close to 147 kilohm, using the estimated rp and gain for the low current used in the SRX circuit. That's with my 250 kilohm plate resistors rather than the original 300k.  If you use the standard published 12AT7 parameters, you'll get a slightly different answer.

 

The input capacitance of the input circuit is low because of the cascode connection.  The issue is the high output impedance of the input stage into the output tube Miller capacitance.  With the 6SN7GTA/B, the Miller capacitance is about 84 pf, and the circuit rolls off above 11-12 kHz.  There is about 14 dB feedback at low frequencies, and so the calculated closed loop gain rolls off above about 53 kHz, which is what I measured.

 

"

You're not getting the point. Yes, everything can be "true" if you assume it to be true. You're simply assuming it's a cascode and then using a cascode calculator to get all your numbers. 

I found the schematic you're referring to in the Birt article you sent (thanks), and it's a totally different circuit from the SRX. This actually is a cascode, with some interesting voltage divider grid biasing action. Notice how it is a common cathode into a common base. I will try to simulate this in spice as well when I get the chance.

birt schem.pdf

[JimL]

Whatever, personally I think they're just variations on a theme.  But, run your simulations with the SRX circuit, the Birt circuit, and the SRX circuit without the cross-coupling, but with fixed bias on the upper tube grid, and see what you get.

Edited August 15, 2017 by JimL

[spritzer]

4 hours ago, kevin gilmore said:

does not look regulated to me.  at least not the high voltage, just a diode bridge and a pair of caps

low voltage looks like 3 terminal regulators

I agree, no HV regulation there and I found another pic taken at the same time:

A few things spring to mind, see the red and black wires coming off the input board?  Has to be the input wiring for the amplifier channels so they must just be on those small side boards.  Also, they used the same twisted wire for the output wiring so the left channel has a bit more capacitance to drive.  Pure engineering excellence here... 

I dug through the Russian tubes out there and the outputs sure look like 6N6pi and the inputs pretty much have to be 6N2pi which is the noval version of the 6SL7.  Also the small PCB with seemingly few parts sure points to the SRX.  This company also doesn't do SMD so that isn't an option to hide parts. 

[JimL]

Just what the world needs.  A variation of a 40 year old design, with passive power supply - for 5000 pounds??!!!?!?

[spritzer]

Well 50 years old but yeah...   I kinda want to make a clone of this but better done and sell if for a lot less. 

[mypasswordis]

I think I see zener strings and the transistors on the tiny heatsinks might be pass FETs with protection diodes on the bottom

[kevin gilmore]

Unlikely, they would have to be much bigger heatsinks, the output tubes are probably good for 9 watts total each

 

[JimL]

If spritzer is correct that the output tubes are 6N6P, the data sheet says their maximum plate dissipation is 8 watts total for both anodes, max static plate voltage is 450 volts.  Pretty similar to 6SN7GTA/B at 7.5 watts total and max static plate voltage of 450 volts.  I note that the most recent ECC99 data sheet lists plate dissipation per side as 3.5 watts (down from 5 watts in the original data sheets) and max static plate voltage of 400 volts.

Edited August 17, 2017 by JimL

[spritzer]

There is certainly a lot of diodes on the center board but yeah, those heatsinks are tiny. 

They pretty much have to be 6N6pi's as nothing else would fit. 

[Ali-Pacha]

Looks like Jim / mafia's work on re-updating SRX circuit did give some ideas to Schiit and Trilogy. But it also looks like both companies are not really into estat thing.

Ali

[JimL]

Well, it does appear that Schiit, Kingsound and Trilogy have all taken their designs from Stax circuits developed in the 1970s, without paying attention to anything done since.  A pity, really.

[Ali-Pacha]

Stax doesn't evolve that much neither BTW

Ali