Matched power amp tubes

Hi all...

What would the difference be between matched and unmatched power amp tubes.

Some people building guitar amps claim that there is no difference for the amp while some people (hiFi-people) claim it is very important to have matched tubes in the output stage.

All insights are valuable


Best regards

/John

Current matching side to side in a push-pull transformer defines low frequency response and distortion. 

Thank you sir

So... mismatch increase distortion? And what about the low frequency response? ...do we lose low frequency?


Best regards

/John

right mismatch = more distortion and less low frequency. 

So... if I may ask. How big would that difference be before they are considered to be mismatched? I know that it might be difficult to describe perhaps but just to get an idea?

best regards

/John

Too many variables.  If you're a punk rock guitarist or Angus Young you may want it totally mismatched and in meltdown.  Angus can afford new output transformers daily. 

Thank you sir

Somewhat how I have thought about it as well. I just thought that I would like to understand it better from actual figures.


Best regards

/John

johnheath said:

Thank you sir

Somewhat how I have thought about it as well. I just thought that I would like to understand it better from actual figures.


Best regards

/John

Click to expand...

There are several aspects to matching tubes; the most important are gain matching and quiescent current matching.
Most amps have only one bias control, the bias voltage being applied to both sides of the push-pull stage. If the tubes are not matched for quiescent, there will be a difference between one side and the other, so the resulting electromagnetic force in  the xfmr's core will not be null, which causes assymmetry in the transfer between primary and secondary; that creates 2nd order distortion, but also, since the core is magnetized, the magnetic permittivity is considerably reduced. The effect is that the apparent inductance of the xfmr is much lower than expected, which reduces the low-frequency response.
As to gain matching, if the tubes are not matched, one side will be less responsive than the other, again creating assymetry in the response, and 2nd-harmonic distortion.
The relation between "un-match" and distortion and LF response depends on the actual transformer's construction and quite hard to predict, though not impossible. Some adventurous members of the LTspice group have actually done simulations, based on Chang's method for evaluating core saturation effects.

And as well as tube imbalance, it  can happen that the halves of a pp output transformer can often be 'unmatched' to some degree

So, should we balance for 'equal' currents in the output transformer primary coil halves ?  or  should it be the ac voltage we try to equalise.  (at some nominal frequency of excitation)

I generally try each method and measure the thd  ... first at the primary, then at the secondary.  I do find that individual bias adjust for the pp tube pairs is better than not.

Usually one finds a place where one is happy .... whether it be lower thd or higher 

abbey road d enfer said:

There are several aspects to matching tubes; the most important are gain matching and quiescent current matching.
Most amps have only one bias control, the bias voltage being applied to both sides of the push-pull stage. If the tubes are not matched for quiescent, there will be a difference between one side and the other, so the resulting electromagnetic force in  the xfmr's core will not be null, which causes assymmetry in the transfer between primary and secondary; that creates 2nd order distortion, but also, since the core is magnetized, the magnetic permittivity is considerably reduced. The effect is that the apparent inductance of the xfmr is much lower than expected, which reduces the low-frequency response.
As to gain matching, if the tubes are not matched, one side will be less responsive than the other, again creating assymetry in the response, and 2nd-harmonic distortion.
The relation between "un-match" and distortion and LF response depends on the actual transformer's construction and quite hard to predict, though not impossible. Some adventurous members of the LTspice group have actually done simulations, based on Chang's method for evaluating core saturation effects.

Click to expand...

Thank you sir

As always very informative... this was the kind of information I needed.

Best regards

/John

In the manual for the quad II power amp it is stated that  performance specs will be met with tubes that are matched to within 10% .This is a little bit vague as they dont say what properties need matching . My guess is the topology and the amount of feedback would probably be the determining factor in how much unbalance a particular amp can handle . One interesting thing about the Quad amp is resistance wise the two primary half winds vary by more than 100 ohms but then the tertiary feedback in the cathodes helps reduce the difference  . I did one time run a pair of KT88's and a GZ37 rectifier in one of my quads . It does draw a bit more current ,which the mains transformer can safely handle as long as your not powering radio tuners or preamps from the main amp supply ,It sounded great and gave a chunk more headroom than the KT66's ,the mains transformer did of course run a bit hotter too .The abillity to be able to vary (by a small amount)the drive voltages to the powertube grids and balance the static current across the transformer will improve the situation further ,but again cancellation wont ever be perfect at all frequencies and signal levels . Even tubes that are matched initially may well drift apart ,theres a certain amount  of burn-in  from new before things settle down . I think for guitar amps the extra fiddly controls for accurately setting primary currents and drive voltages are just  a worry for the player thats needless. If you want something a bit more special for home hifi  the Williamson amplifiers driver/output stage provide the controls to statically and dynamically balance . As far as the end user is concerned ,you'd at least need to be confident pokeing around measuring high voltages with a multimeter  for it to be useful.

alexc said:

And as well as tube imbalance, it  can happen that the halves of a pp output transformer can often be 'unmatched' to some degree

So, should we balance for 'equal' currents in the output transformer primary coil halves ?  or  should it be the ac voltage we try to equalise.  (at some nominal frequency of excitation)

I generally try each method and measure the thd  ... first at the primary, then at the secondary.  I do find that individual bias adjust for the pp tube pairs is better than not.

Usually one finds a place where one is happy .... whether it be lower thd or higher 

Click to expand...

Thank you sir

I have not measured OT's myself and I guess that small differences exists as well in different tubes of the same model and brand.

I also guess that in a guitar amp with a lot of gain it is less sensitive than in a HiFi amp? Considering THD and like that is.


Best regards

/John

Tubetec said:

In the manual for the quad II power amp it is stated that  performance specs will be met with tubes that are matched to within 10% .This is a little bit vague as they dont say what properties need matching . My guess is the topology and the amount of feedback would probably be the determining factor in how much unbalance a particular amp can handle . One interesting thing about the Quad amp is resistance wise the two primary half winds vary by more than 100 ohms but then the tertiary feedback in the cathodes helps reduce the difference  . I did one time run a pair of KT88's and a GZ37 rectifier in one of my quads . It does draw a bit more current ,which the mains transformer can safely handle as long as your not powering radio tuners or preamps from the main amp supply ,It sounded great and gave a chunk more headroom than the KT66's ,the mains transformer did of course run a bit hotter too .The abillity to be able to vary (by a small amount)the drive voltages to the powertube grids and balance the static current across the transformer will improve the situation further ,but again cancellation wont ever be perfect at all frequencies and signal levels . Even tubes that are matched initially may well drift apart ,theres a certain amount  of burn-in  from new before things settle down . I think for guitar amps the extra fiddly controls for accurately setting primary currents and drive voltages are just  a worry for the player thats needless. If you want something a bit more special for home hifi  the Williamson amplifiers driver/output stage provide the controls to statically and dynamically balance . As far as the end user is concerned ,you'd at least need to be confident pokeing around measuring high voltages with a multimeter  for it to be useful.

Click to expand...

Thank you sir

Yes, I know that almost nothing is fully accurate and therefore it is hard to match or balance everything might be tricky in a general sense. I have looked at the williamson amp and have seen the trimpots. I have my eyes on the Dynamo ST70 as a reference for my up-coming HiFi build... my biggest concern is to chose transformers... not what I need but rather how much I shall spend on them

I think the Dynamo is a fairly simple circuit that has remained pretty much unchanged over the years except the PSU that has undergone some changes in their upgraded "reissue". I guess many of these old amps sound good as they are... the major difference might be that people of today prefer to use their I-pads and telephones to plug them into a D/A converter and then into a power amp? This might need some tweaking on the input impedances i guess?


Best regards

/John

alexc said:

And as well as tube imbalance, it  can happen that the halves of a pp output transformer can often be 'unmatched' to some degree

Click to expand...

That is true.

So, should we balance for 'equal' currents in the output transformer primary coil halves ?  or  should it be the ac voltage we try to equalise.  (at some nominal frequency of excitation)

Click to expand...

Actually, quiescent currents should be individually biased for zero induction in the core, which is hard to measure without a magnetometer, then the tube's gain and xfmr unbalance should be adjusted by altering the drive level of one side. This is not a common feature, although some tweak the values of the phase-inverter anode resistors to achieve that.
Balancing quiescent current improves LF response AND distortion, balancing drive also improves distortion.

a good balance can also reduce hum,

most P-P output transformers are designed to handle a bit of mismatch, 

as far as unequal wave halves, the OPT takes care of most of that problem,

tubes sound good at a certain idle current,  a mismatch means that one or more tubes will be away from the sweet spot,

and tube life will obviously be shorter for the tube that has a higher idle current,

the turns count of the primary will determine how much DC current the core will see due to mismatch, an 800 turn primary with 5 ma mismatch will see 800 x .005 = 4 amp-turns of unbalanced DC,

a 4000 turn primary will see 4000 x .005 = 20 amp turns of unbalanced DC, a difference of 20/4= 5 times the DC flux.

With the usual cathode coupled phase invertor found in guitar amps its easy to include a pot between the ht and the two anode resistors of the phase invertor  and so have a means to adjust the relative gains of each side of the push pull .Trying a Williamson biasing arrangement with a vintage cathode biased fender 30 watt style amp might be interesting too. One other nice thing about having balancing adjustments  is that cancelation of psu noise tends to be better than where you have preset values .

Thank you guys

All in all I understand that it is better to have the tubes matched för better LF response and lower distortion... Even though you use matched tubes some degree of "mismatch" can be in the transformer itself making it hard to balance everything without knowing the mismatch in the transformer?

Using trimpots, like in the Williamson amp, could make up for some mismatch in both tubes and/or in the transformer?

Best regards

/John

In hifi amps, for 'balanced' or push-pull stages, it is common to see 'balance' adjustment pots - it can be used in any of the plate, cathodes and even screens (for pentode).

You can see this done for example, in some of the 'vari-mu' classis ..  the ua175, the rca ba25  etc.

It definately decreases thd in a stage, at least in some scenarios  ...  'static' or quiescent conditions for sure ...  maybe also under some actual operating conditions too!

So the question is how much of a good thing do we need ?

Surely, good hifi performance is benefitted by a balance adjust in the finals'. Same thing really for the driver ...

And to for the first stage..    So ...  it is done, to have say 3 different balances   

I have found it pretty easy to get quite tight balances under quiescent conditions  at each of the stages.

The first stage and driver stage can each improve THD+N performance something like a couple of tenths of a % - quite significant. The finals' balance can improve approaching 0.5% of even 1% in some cases.

Adjusting across stages can also do noise cancellation ie. 'debalancing' the first stage can improve the end-end THD+N.

A nice benefit of adjusting balances at different points in a circuit, is that you can often get a 'sweet spot' in terms of representative spectra    ...  balancing can tailor to some degree the profile of THD achieved  end-end.

Balancing takes time and has additional parts complexity .. and there is a diminishing rate of return!

One reason why 'vari-mu' limiters can be expensive to build and maintain - the balance of each stage is very important to achieve useable THD+N.

....

My el34 monoblocks use a 'plate balance' in the 6SN7 driver and individual 'bias pots' for balance at the grids of the EL34.

With careful adjustment, I could get the 1KHz THD+N at 0.13% at 30Wrms into 8ohm loads using 12dB global nfb.

At the end of the day, one can get pretty crazy about 'balance'  ..  thence going to 'constant current sources', long tailed pairs and bipolar HV supply rails  ...  to further force improved balance across the phases of  stages ....   

Judicious use of negative feedback(s) probably has more influence over performance than stage balance ..  and the real fun is combining several stages of nfb ...  like in the classic limiters like the ua175, rca ba25 etc.

...

The other thing to note ..... getting perfect balance at 'quiescent' operating conditions doesn't necessarily translate to similar figures in 'dynamic' operating conditions.    ie.  more comprehensive testing is usually required to get to the core of things!

..

The last pair of finals I purchased were a pair of modern Tung Sol 7581a (super 6L6 type) ... 'matched' by the vendor .....  at the quite affordable end of the price range  ....  in my SE stereo 'spud' amp  ... running +465V , 53mA each, the balance from Left,Right channels  was +/- 0.5mA  ! 

So out of the box, burned in a few hours and with no further adjustment, the finals were within 1mA of each other (quiescent and warmed-up).  Of course, I have seen differences of up to several .. up to 5mA  .. among unmatched 6L6s.
...

Something to note is of course, the measurement of parameters used in balancing adjustments  -  you can measure voltages easily enough, but  most  current measurments are derived from voltage measurements ..  through some resistance ...  which can be often be 5  .. 10% tolerance power resistors ...  such as are used in tube amps.

You can put an ammeter inline with a circuit, and that is very accurate, but usually .. currents are measured indirectly as voltages and so can be a few % uncertainty  ... 'real' balancing (high absolute acurracy) of voltages and currents in tubes can be very tricky business  ...

For that reason, it is useful to get 'as close as you can' with voltage/current measurements, and then do 'final tweaks' while watching the spectrum with THD+N readings  .....  at different points throughout the circuit.

_

johnheath said:

All in all I understand that it is better to have the tubes matched för better LF response and lower distortion...

Click to expand...

That is matched for quiescent, because it is essential that both sides of the primary receive the same amount of magnetic force, resulting in cancellation of the induction. As transformers are pretty well matched DC-wise (it just takes making sure the number of turns is identical and the winding are similarly wound), it often results in making the quiescent current identical on both sides. However, tubes matched for quiescent do not necessarily match dynamically, which results in gain assymetry. Same for transformers, because flux lines can differ from one side to the other.

Even though you use matched tubes some degree of "mismatch" can be in the transformer itself making it hard to balance everything without knowing the mismatch in the transformer?

Click to expand...

Evaluating the xfmr's assymetry is a tedious task; converting the results in a compensation figure is prone to error.

Using trimpots, like in the Williamson amp, could make up for some mismatch in both tubes and/or in the transformer?

Click to expand...

Exactly. It's the whole gain that needs to be "symmetrized", that means compensating the total unbalance resulting from all the factors: gain difference between the two triodes in the LTP phase-inverter, tolerance in the anode resitors of the PI, gain difference between the output tubes, xfmr assymetry,...
Typically a single symetry adjustment will provide the most significant optimization. As Alex mentioned, there is a sharp law of diminishing returns.

Best regards

/John
[/quote]

While on the subject of matching power tubes ...  and their transformers ..

The Dynaco output transformers are sometimes said to be 'more' tolerant of 'power tube imbalance'  ..  at least by some modern vendors ...

So what would it be, in the design/manufacture of the dynaco, that improves this property ?

....

As I understand the dynaco, it differs from most in that it uses (more) interleaving with (more) complex interconnections.

ie. more interleaved layers with paralleling of reverse-phased sections (and what not). Possibly better 'balanced' halves in the primary ? (maybe even less balanced!)

Possibly some other fancy features  in the Dynaco traffos as compared to ..  say .. my standard Edcor (larger cxpp series).

The cores and other basic properties are pretty similar -  I don't think any exotic materials are used in either ..

And the big Edcors measure pretty well - I get around -0.2dB at 20Hz to 20KHz and -3dB at 55KHz in my el34 monoblocs ..  basically with ruler flat magnitude and phase response;

Maybe the square wave/transient response is less refined in the edcor than in the dynaco - I may have to get hold of one and do some measurements!