From: none@nospam.org
"flipper" wrote in message
news:6ju2e6tkj9vjqb4mdvcjqf9o77cn0tbuta@4ax.com...
> On Mon, 15 Nov 2010 03:31:56 -0800 (PST), Patrick Turner
> wrote:
>
>>On Nov 14, 11:28 pm, flipper wrote:
>>> Well, RF is outside my realm but during various roamings I recall
>>> seeing a couple of things that might give you something to at least
>>> search for.
>>>
>>> The first is I seem to recall comments about "pre-distortion" used to
>>> compensate the Gilbert cell. Sorry that I can't recall more on that
>>> topic.
>>>
>>> The other is the Gilbert cell being very sensitive to component
>>> mismatch. That makes sense as we all know a 'diff amp' is and with
>>> the
>>> Gilbert being one feeding the emitters of two others one can only
>>> imagine the problem being worse, especially with discretes.
>>>
>>> Have you tried matching the transistors and some emitter
>>> degeneration?
>>
>>Yes indeed I tried adding in emitter resistors to the top 4 bjts and
>>this worsened all operation, and raised the carrier distortion at 1/2
>>max Vo to about 4%, and the signal looked like a class B amp with 3H
>>crossover THD, YUK.
>
> Emitter degeneration lowers gain, of course.
>
> Sorry, can't help much but I suspect it's discrete mismatch since I
> gather IC versions do better than you're seeing.
>
>
>>I could not get the circuit to give a clean reduction of envelope
>>shape to a straight line at the positive and negative maximum envelope
>>swings.
>>
>>Anyway, I understand the Gilbert cell now, but I won't be using it as
>>an alternative to the tube signal gene with AM.
>>
>>I then replaced the 6BX6 pentode with 6EJ7 in my SE AM modulator which
>>has a tuned RF transformer circuit for anode load.
>>The tuned anode coil is wrapped with a much smaller secondary coil
>>which is grounded one end and feeds aa 2k0 R divider load to vary the
>>output 0dB to -30dB in six switch positionslevel at the output
>>terminal. At -30dB Rout = 50 ohms and the low signal is strong enough
>>to drive all AM radios with say 5pF coupling to the antenna terminal.
>>
>>The 6EJ7 only very slightly reduced the envelope THD, but showed it
>>could produce an over modulated signal with less AF drive.
>>I also tried a 6EH7 which is the variable µ or "remote cut off"
>>version of the 6EJ7, but envelope THD was much worse.
>>
>>The SE pentode modulator was checked out for any faults in DC supplies
>>and basic condition, the first time in 13 years. No problems.
>>
>>The pentode is set up with B+ = 250V and Eg2 = 225V via 33k and 0.1uF
>>to 0V.
>>
>>The cathode has Rk = 1k8, and grid is biased with 1M to 0V. The RF
>>signal comes from a 12AT7 CF buffer after the 12AT7 Hartley oscillator
>>with a single tuned grid LC. When the RF ranges are is displayed on a
>>CRO, you cannot see any form of THD so I guess THD < 0.5%
>>Levels are about 7Vrms. Easy peasy for 1/2 f a 12AT7.
>>
>>The RF buffer driver has 0.02uF then 15k to pentode grid. Then I have
>>a 22k from pentode grid to a 0.1uF which couples to a 12AU7 CF which
>>buffers the audio input signal. The RF level before the modulator
>>remains fixed and determined by the oscillator level which does not
>>change much.
>>
>>So in effect, I have RF between 360kHz and 1.75MHz in two ranges and
>>simply mixed with an audio signal in a linear resistance network.
>>
>>But the pentode is set up to have no audio gain at its anode because
>>its load at AF is an RF coil of only 190uH, with tuning C max below
>>1,000pF.
>>At the the anode the LC is tuned to the same F as the oscillator
>>frequency, and the parallel LC has maybe 10k tuned impedance, so at RF
>>the pentode produces plenty of RF gain, but not excessive gain, and
>>the gain is slightly loaded by the secondary load of 2k0 resistance
>>divider load.
>>The LC RF transformer functions to keep any AF out of the anode output
>>but meanwhile, the high RF gain is possible because the Rk is bypassed
>>with 0.0033uF which has low enough Z at the range especially at high
>>RF, so that the slight roll off in RF level at HF is slightly
>>compensated.
>>
>>The 0.0033uF is a high Z at below 20kHz so that there is an audio
>>signal just below the grid input audio signal appearing at the
>>cathode. Thus the 6EJ7 functions as a CF at AF. The grid input signal
>>is thus generating a fairly linear change of voltage across the Rk and
>>therefore a fairly linear change of anode current. There is also a
>>change of AF anode current due to the Vk to Vg2 change but its
>>effect is minor, and can be neglected.
>>
>>The change of anode current at AF causes a corresponding change of
>>current in the tuned circuit and a fairly linear envelope shape. To
>>check the linearity, I measured more carefully before abandoning my
>>simple little modulator.
>>
>>This time I used my low THD signal < 0.01% from my 1kHz oscillator.
>>I could say that the 12AU7 buffer input to the modulator would
>>generate maybe 0.2% THD at the 7Vrms grid drive needed.
>>
>>I set the CRO to monitor the peaks of the un-modulated carrier at the
>>centre horizontal gratical. and adjusted the modulation and CRO level
>>so that the maximum peak RF voltage with full modulation was twice the
>>peak RF voltage with no modulation. The only problem I could see was
>>the slight inability of the modulation to reduce to zero on wave
>>trough minimums.
>>Considering the positive or top side of the modulated carrier wave, So
>>I could get a +ve envelope Vswing of 2.0 times the positive swing of
>>the unmodutaed carrier, while the negative swing side was 1.90 x
>>unmodulated carrier.
>>
>>The negative going envelope measured almost identical, and no worse.
>>From RDH, THD % = 100 x ( 0.5 [ 2.0 - 1.9 ] ) / ( 2.0 + 1.9 ) = 100
>>x (0.5 x 0.1) / 3.9 = 1.28%.
>>
>>The THD rapidly reduced away from the maximum modulation. The Dn at
>>1.3% can be seen as a slight flat on the peaks of either +ve or -ve
>>sine wave shapes.
>>
>>In a receiver with a simple diode + C + R detector of any kind, there
>>is a ripple voltage across the C, just like you get in a 1/2 wave PSU
>>rectifier. For the detector to function to filter the RF out and yet
>>allow the AF voltage to rise and fall linearly, the C must be just the
>>right value to give a time constant so that large high frequency AF
>>signals don't suffer slew distortion.
>>In my own detectors I use in most radios I use a CF buffer at the last
>>IF secondary and the cathode drives a germanium diode to charge a
>>220pF with 1M R which has 50Vdc idle voltage across it. This detector
[continued in next message]
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
