Long Tail Pair

[mr2racer]

Hey Anyone?

I have a pair of Dynaco Mark VI monoblocks. I received a long tail pair schematic from Bill Hardy at Vintage Amplifier Restoration. His schematic used 12AU7/12AU7 for the preamp/phase splitter. At low listening volumes the amps sound fine. But as I raise the volume they start to distort prematurely. I think this may be because I also changed the outputs from what I think are easy to drive 8417's to KT88's. I base this on the data sheets that tell me VG1 for 8417 is 15 while VG1 for either KT88 or 6550 is 75.

I apologize for my lame tube math ability. But if the mu of the 12AU7 is 17 with a 2 volt input does that give me 34 volts of drive? (I checked Morgan Jones and his description of 'drive' is two pages.) From an old thread I found this
Long-Tail Pair Design

I have three questions;

Am I right in assuming that the LTP I have is insufficient to drive 4 KT88's?

If I need more voltage gain could I substitute a 6922 for the gain stage? Or maybe a 12AT7?

Or should I use a 6922/6922 cascode front end like Joe Curcio's design for the Mark III?

 

[mr2racer]

I almost forgot. The KT88's are biased correctly. B+ for the outputs is around 550 volts

 

[HollowState]

At low listening volumes the amps sound fine. But as I raise the volume they start to distort prematurely. I think this may be because I also changed the outputs from what I think are easy to drive 8417's to KT88's. I base this on the data sheets that tell me VG1 for 8417 is 15 while VG1 for either KT88 or 6550 is 75.

Are you sure the 6550's are biased correctly? Bias means the negative voltage on the control grid used to control the tube quiescent current. With 550v on the plate and screens, you need about -35v in the G1, ± a couple of volts. Not -75. What you describe sounds like the 6550's are nearly cut off and the increasing signal is cutting them off even more. You want about 60mA of quiescent current for P-P AB1 operation. Insufficient signal drive voltage would result in less overall output power, not distortion.

 

[aroS3]

my calculations show that with -75 at G1 you are bias them around 27 mA ...which is far too low.... a better approach could be running them around 55 mA which is around -65 Volts at G1, but be careful not to exceed the max tube dissipation

 

[DF96]

Gain of LTP phase splitter is a little under mu/2, not mu. If you use a low mu valve like 12AU7 then you need a CCS in the tail otherwise you will get imbalance which will increase output stage distortion.

 

[mr2racer]

Sorry, I may not have been clear enough. I'm talking about the input voltage to the power tubes, the drive from the preamp section. How much voltage does the front end need to put out to properly drive the KT88s? The design was for 8417s and I have KT88s in the amp.

The outputs are biased at 50ma.

 

[mr2racer]

DF96, If gain is less than 8.5 volts is that enough to drive KT88s?

 

[DF96]

I don't know. The KT88 has about half the gain of an 8417, so it could need twice the voltage swing. A driver which is adequate for 8417 may be inadequate for KT88. If I made a change of output valve like that I would expect to have to recalculate the driver performance in detail, and maybe then redesign it.

As a rough guide, the peak signal voltage at the output grid needs to be equal to the grid bias.

 

[roline]

I designed a PCB with ECC99 for the LTP and I decided on a CCS for the tail to be sure to be able to drive the KT88 outputs. I wanted low output impedance to have sufficient drive for hungry tubes. They can dissipate 4 watts so can handle 28K plate resistors with ease. If you use a CCS, you can match the plate resistors.
For tonal influence, I wanted to use a 6SN7 for the input stage driving it. So many flavors to choose from......
Later research found a similar design from triode dick. etc
Triode Dick's Page
I decided on the Russian 6N6P version of ECC99 due to being a couple buck$ a piece.
If the outputs are biased at -55V for 50ma, then to get near full swing the drive would need to be say +-35 volts or so to pinch them off.... so thats about a 70Vpp swing from the LTP..

 

[wicked1]

I think this is related to this post...
How do you calculate the power (current) out of a LTP? I'm having bad luck finding answers in the Jones book, too.

 

[Bandersnatch]

my calculations show that with -75 at G1 you are bias them around 27 mA ...which is far too low.... a better approach could be running them around 55 mA which is around -65 Volts at G1, but be careful not to exceed the max tube dissipation

A 6550 or KT88 with 550V on its g2 will need a substantial amount of bias. The question should be, what current is the -75V delivering? I do agree that a 55-60 mA/tube idle current is desired...I don't think any of the 6550 or KT88 data sheets give enough information to determine bias voltage that accurately. Since you think so, do plese enlighten me as to your methods; share the wealth if you'd be so kind.
cheers,
Douglas

 

[mr2racer]

Bandersnatch,
This is what I was referencing on Duncan's page;

AB1 U/L 40% 553 -75.0 2x50 4500 100 2.0

I didn't mean -75 volts bias. This spec is for voltage on grid 1. If I'm reading Morgan correctly the - sign is for inverting not negstivr voltage. I thought that was the spec for the input voltage. Sorry

 

[Bandersnatch]

No, you've had it right the first time. 50 mA/tube idle with g2 and plate at 553 volts. 100 watts out at 2% distortion with a -75V bias on g1.

On LTP gain it is about mu/2 each plate to ground, it is mu plate to plate( this is one grid driven, the other grounded).
cheers,
Douglas

 

[mr2racer]

Hey Guys,
Is the drive voltage simply a function of gain? Or does a greater voltage swing give better quality sound? Like more dynamic range?