Mismatched OT load - impact on frequency response?

[pdf64]

I'm working on re-assigning an unwanted TVM PA KT88 amp for bass.
It puts out close to 100W but unfortunately only down to ~50Hz, below that the OT must be saturating, and the waveform goes horrid.
Plus the OT secondary is 16 ohms only.
So I was thinking of replacing the OT; however, planning ahead, in case 2xKT88 is insufficient, I was thinking of getting one that would suit 4xKT88.
Then, with 2xKT88, connect twice the specified load impedance, to get a suitable reflected impedance back to the primary.
But will that mess up the low frequency capability? If so will it matter, as it's only running on half power?
I seem to remember reading a post here that raised the issue, eg mismatch one way and the high frequency spec gets hit, mismatch the other way and it's the low frequency.
Anyone care to comment on the subject / my idea?
The amp has a VB+ of about 600V, the current OT is about 5k and runs the KT88 in UL mode.
I'd like it to be able to cope with a 5 string, down to 29Hz Bb.
Also, a supplier for a ~2k5 : 4/8 ohm 200W OT?
Thanks - Pete

[Gingertube]

The power supply will not cope with 4 x KT88.

Just get a Hammond 1650RA (5,000 Ohms Raa 100 Watt rated down to 30Hz) to suit the 2 x KT88. This is from their HIFI tranny range rather than the Guitar Amp tranny range.

Hammond HiFi and Guitar Amp traany catalog here:
Hammond Mfrg. - Electronic Transformers - AUDIO Index

Cheers,
Ian

[Steve Conner]

As you lower the load impedance, the frequency range of the OT shifts downwards. So buying a 2k5 OT and running it at 5k is the wrong thing to do, it should be the other way around. Have you tried an 8 ohm load on the existing OT? (maybe with an upgrade to KT120s )

Having said that I think Gingertube's plan is the best, just buy a new OT. Although I will say, distortion from an undersized OT might actually make the amp sound better in a band context. Most bass cabinets can't handle the fundamental of the low B on a 5-string anyway, so you might as well generate some harmonics that will be audible.

[pdf64]

I'm still considering the options for this; it's working nicely as a bass amp, with a 5F6A based pre-amp and LTP, still with the original 5k:16ohm UL OT and KT88 pair.
I'm looking to get a 1650RA OT eventually, as per Ian's suggestion, if only for the more useful output impedance options, but am hoping that the additional iron may be beneficial too (existing OT only about 8 lbs compared to the 1650RA's 12 lbs).
The power output is sufficiant but Steve planted the idea of KT120 and I'm tempted to experiment with them.
The PT doesn't get overly warm in use, so hopefully it could accommodate the additional heater and B+ current.
The problem as I see it is the 47k grid leaks / low impedance bias supply that the KT120 require.
The LTP seems to manage sufficient signal output Vp-p into 100k and the bias winding is 70Vac so 100Vdc is available, but I'd appreciate some guidance for a design to buffer the outputs of the bias control pots to create something that's able to sink excess grid current from the KT120s.
I'm avoiding voltage regulator type circuits, as it seems a good idea for the bias supply to track changes in line voltage, so as to keep bias and B+ in equilibrium.
Pete

[Steve Conner]

Sounds like a job for MOSFET followers? You could use the bias winding with a voltage doubler circuit to create +/-100V dual rails, and run the MOSFETs between these.

[pdf64]

Yes, that could be a good way forward. In it's original incarnation as a PA amp, the bias winding also provided a split supply for transistor pre-amp channels.
Would the arrangement / devices Ian used in his Baby Huey EL84 Amp - Baby Huey - Page 61 - diyAudio cope with +/-100V? I'm not looking to push the power tubes into AB2.
ZVN0545A 450V 700mW MOSFETS and MPSA43 300V 625mW NPN current sources
Pete

[Gingertube]

Yes, that could be a good way forward. In it's original incarnation as a PA amp, the bias winding also provided a split supply for transistor pre-amp channels.
Would the arrangement / devices Ian used in his Baby Huey EL84 Amp - Baby Huey - Page 61 - diyAudio cope with +/-100V? I'm not looking to push the power tubes into AB2.
ZVN0545A 450V 700mW MOSFETS and MPSA43 300V 625mW NPN current sources
Pete

Setting the rails for current source loaded source follower drive of the output trubes.

The -ve rail, the "Rule of Thumb" is that you need 3 x the bias voltage. For KT88 that probably means around -180V
The+ve rail, the source will swing up to 0V (or possibly 1 or 2 volts above if pushing into Class AB2). The thing which can make mosfets sound crappy is their reverse capacitance which is dependent on the drain to source voltage and which can therfore be modulated by the audio signal (the full audio signal is on the source). If you look at the reverse capacitance graph (Crss) on a mosfet datasheet you will see that the curve levels off and there is very little change in reverse capacitance for drain to source voltages above 20V. That mean you want to have a drain voltage of at least say +25V so as to keep 20V across the mosfet even on positive signal peaks.
Going higher in voltage just adds dissipation to the mosfet for little benefit.

You would want to run the mosfets at a minimum of say 3mA, with the source suitting at the output tube bias voltage (-60) and the drain at say +25V then 0.003 x (60 + 25) = 255mW dissipation in the ZVN0545A. The easiest power supply design is going to end up +90, -180 (from a 70V winding). The ZVN0545A would then start to look a bit stressed and I would recommend going to a 1 to 2 watt rated mosfet. Select the mosfet for low gate capacitance and low reverse capacitance which means using the smallest, lowest power device you can get away with, shoving an big but common IRF820 or similar in there is NOT the right thing to do.

For some reason I have not quite fathomed yet the +ve rail on the drain is fairly critical for good pace rhythm and attack (possibly it is a phase accuracy thing). Make sure you have good bypass caps at this node, if using an electrolytic you should parallel it with a poly cap, say a 470nF 100V polypropylene.

On the ring of 2 current source mosfet loads, use the MPSA42 for voltage withstand on the top, if worried about power dissipation then go to a MJE340. The performance of the current source is largely dictated by the lower device which has only a volt or 2 across it and so voltage rating is not a problem, you want lots of beta, a BC547C or similar would be the way to go.

Using a current source loaded source follower like this as the driver means you have just the output tube grid stop in the path from mosfet source to the output tube. Bias is applied to the mosfet gate. This means that you have eliminated the usual guitar amp max Rg1 abuse and that means that you can safely run the output tubes much hotter. For KT88 in this arrangement I would have no hesitation in biasing them at 35 Watts dissipation.

Hope I've answerd your questions.

Cheers,
Ian

[Steve Conner]

The only thing I'd add is, if you're using a low voltage for the positive rail, Zener protection for the MOSFET gates is vital. When the source hits the rail, the PI will keep driving the gate more positive, but the source can't follow it any more, so the gate will be overvolted.

The PI output is normally limited by power tube grid current. The MOSFET followers take that limit away. The next limit is the MOSFETs running out of Vds. If you increase the drain voltage then the next limit is the power tube grid stopper resistors burning out, or even the grids themselves.

[jazbo8]

Found an interesting graph showing the output load's effect on harmonic distortion and output power. The change in harmonic distortion will have an impact on the frequency response as previously mentioned.

[pdf64]

So for the sake of experimentation I tried a pair of KT120, still with the old LTP driver and 220k grid leaks; my thinking was that for a quick trial and with reasonablely low dissipation (~30mA cathode current), to appraise whether it was worth messing with the MOSFET buffers.
However, with KT88 the power output was about 90W into 16 ohms, with KT120 it's lower, about 80W; the impression from the scope is that one of the KT120 seems to be clipping earlier than the other (one of the LTP plate resistors has a variable resistor in series, so that ac signal balance can be achieved but tweaking this
With an 8ohm load, so bringing the p-p impedance down to 2k5, the output power with KT120 was even lower, about 70W.
Is it most likely that I've got a weak KT120, or is something else going on?
Pete

[Steve Conner]

Swap the tubes, does the early clipping follow the tube? If so you may well have a weak one.

[pdf64]

Thanks Steve, I suspected the KT120 but the clipping stayed on the 'top' half of the wave when swapped.
Another 75V on the supply to the LTP sorted it out - the KT120 certainly take some driving, I'm running the LTP at 450V.
It's putting out over 90W with the KT120 now; 1650RA OT, B+ ~580V static, 550V max sine.
Pete

[jazbo8]

Thanks Steve, I suspected the KT120 but the clipping stayed on the 'top' half of the wave when swapped.
Another 75V on the supply to the LTP sorted it out - the KT120 certainly take some driving, I'm running the LTP at 450V.
It's putting out over 90W with the KT120 now; 1650RA OT, B+ ~580V static, 550V max sine.
Pete

Could you please post the latest schematic? Even though the cause of your problem turned out to be PI related and not from load impedance mis-match, but, I just realized that the chart I posted above was for SE operation only. For PP operation, please refer to the chart below: