Reply by Lamont Cranston●November 10, 20222022-11-10
On Thursday, November 10, 2022 at 3:11:28 PM UTC-6, erichp...@hotmail.com wrote:
> > Seems to be a problem with the amp from the data book.
> > I did the same experiment shown above but with the databook amp.
> > Instead of it loading the LC, the Q more than doubles, instead of 1,000 it's 2500!
> > What's the fix for that? :-)
> > Mikek
> Perhaps some of all that bootstrapping is getting back into the LC tuned
> circuit - you have recreated the old-time Q-multiplier. Or the input
> levels are so high the buffer is overdriven?
>
> Padding down the input with a capacitive attentuator like Kleijer did
> could be a good way to reduce interaction. Also I guess at resonance the
> voltage is pretty high so overdriving is a real risk.
>
> piglet
I'll try a tiny input capacitor. If I create a 10x capacitive divider, I'll need a 10X amp to get back to
a gain of almost 1.
I put a series 0.3pf cap in series with the input. Same as the Kleijer amp.
Turns out that is a 10x was close it's 9.8 to 1, the output drops by a factor 9.8.
The input capacitance is 0.45pf and the input resistance measures 180MΩ.
This compares to the 200MΩ of the Kleijer amp, a little surprising sense the databook amp
has bootstrapping and the Kleijer amp doesn't.
I need to get some or make some smaller inductors so I can do these tests at 1, 10, 20, and 30MHz,
to see if or how much the input impedance drops.
Thanks, Mikek
Reply by piglet●November 10, 20222022-11-10
On 10/11/2022 5:53 pm, Lamont Cranston wrote:
> On Thursday, November 10, 2022 at 5:59:56 AM UTC-6, Lamont Cranston wrote:
>> Years ago, I built this Kleijer High input impedance amp.
>> http://www.crystal-radio.eu/fetamp/enfetamp.htm
>> The input impedance test:
>> I put an inductor on my Q meter and resonated it at 1MHz.
>> Tuning cap 151.1pf, Xc = 1,053Ω, Q = 1068
>> Rp of the LC is ,Rp = Q x Xc, so, 1,068 x 1,053 =1,124,604Ω =Rp.
>> Then I put the Kleijer amp input in parallel with the tuning capacitor.
>> I had to reduce the Q meter tuning capacitor by 1.1pf to get back to resonance.
>> The new Q with the Kleijer amp attached dropped to 1,062. So Rp= 1062 x 1053 = 1,118,286Ω
>> To get 1,118,286 Ω I need to parallel the original (unloaded) 1,124,604 Ω with 200,000,000 Ω.
>> i.e. 1,124,604Ω // 200,000,000Ω = 1,118,286Ω
>> So, I think the input R of the amp is 200MΩ.
>> Does this sound like a correct method?
>> Does the input capacitance matter, (as a load) if I'm resonating it out.
>> Mikek
>> P.S. Kleijer uses an air input cap, I used a tiny dot of roger 5880 pcb as my capacitor,
>> it would be interesting to make mine an air cap and see if there is a difference.
>
> Seems to be a problem with the amp from the data book.
> I did the same experimant shown above but with the databook amp.
> Instead of it loading the LC, the Q more than doubles, instead of 1,000 it's 2500!
> What's the fix for that? :-)
> Mikek
Perhaps some of all that bootstrapping is getting back into the LC tuned
circuit - you have recreated the old-time Q-multiplier. Or the input
levels are so high the buffer is overdriven?
Padding down the input with a capacitive attentuator like Kleijer did
could be a good way to reduce interaction. Also I guess at resonance the
volatge is pretty high so overdriving is a real risk.
piglet
Reply by Lamont Cranston●November 10, 20222022-11-10
On Thursday, November 10, 2022 at 5:59:56 AM UTC-6, Lamont Cranston wrote:
> Years ago, I built this Kleijer High input impedance amp.
> http://www.crystal-radio.eu/fetamp/enfetamp.htm
> The input impedance test:
> I put an inductor on my Q meter and resonated it at 1MHz.
> Tuning cap 151.1pf, Xc = 1,053Ω, Q = 1068
> Rp of the LC is ,Rp = Q x Xc, so, 1,068 x 1,053 =1,124,604Ω =Rp.
> Then I put the Kleijer amp input in parallel with the tuning capacitor.
> I had to reduce the Q meter tuning capacitor by 1.1pf to get back to resonance.
> The new Q with the Kleijer amp attached dropped to 1,062. So Rp= 1062 x 1053 = 1,118,286Ω
> To get 1,118,286 Ω I need to parallel the original (unloaded) 1,124,604 Ω with 200,000,000 Ω.
> i.e. 1,124,604Ω // 200,000,000Ω = 1,118,286Ω
> So, I think the input R of the amp is 200MΩ.
> Does this sound like a correct method?
> Does the input capacitance matter, (as a load) if I'm resonating it out.
> Mikek
> P.S. Kleijer uses an air input cap, I used a tiny dot of roger 5880 pcb as my capacitor,
> it would be interesting to make mine an air cap and see if there is a difference.
Seems to be a problem with the amp from the data book.
I did the same experimant shown above but with the databook amp.
Instead of it loading the LC, the Q more than doubles, instead of 1,000 it's 2500!
What's the fix for that? :-)
Mikek
Reply by Lamont Cranston●November 10, 20222022-11-10
Years ago, I built this Kleijer High input impedance amp.
http://www.crystal-radio.eu/fetamp/enfetamp.htm
The input impedance test:
I put an inductor on my Q meter and resonated it at 1MHz.
Tuning cap 151.1pf, Xc = 1,053Ω, Q = 1068
Rp of the LC is ,Rp = Q x Xc, so, 1,068 x 1,053 =1,124,604Ω =Rp.
Then I put the Kleijer amp input in parallel with the tuning capacitor.
I had to reduce the Q meter tuning capacitor by 1.1pf to get back to resonance.
The new Q with the Kleijer amp attached dropped to 1,062. So Rp= 1062 x 1053 = 1,118,286Ω
To get 1,118,286 Ω I need to parallel the original (unloaded) 1,124,604 Ω with 200,000,000 Ω.
i.e. 1,124,604Ω // 200,000,000Ω = 1,118,286Ω
So, I think the input R of the amp is 200MΩ.
Does this sound like a correct method?
Does the input capacitance matter, (as a load) if I'm resonating it out.
Mikek
P.S. Kleijer uses an air input cap, I used a tiny dot of roger 5880 pcb as my capacitor,
it would be interesting to make mine an air cap and see if there is a difference.
Reply by Lamont Cranston●November 9, 20222022-11-09
On Wednesday, November 9, 2022 at 12:50:54 PM UTC-6, Lamont Cranston wrote:
> On Wednesday, November 9, 2022 at 8:58:44 AM UTC-6, erichp...@hotmail.com wrote:
>
> > 30MHz is your max frequency then lead length inductance of THT
> >components is unlikely to be a big problem.
> Don't forget the dielectric constant of pcb substrate is 4-5 times that
> of air (and lossy).
> Yes, and that is part of removing the ground plane, although it had a slight level peak at 29MHz and
> was up a little at 30MHz. That calmed down when I add the underside ground plane.
>
> The thread (on an IO Group) that got me started, was asking about building a Q meter with a range of
> 1.5MHz to 150Mhz. I don't care about that, I don't know that I will ever go above 10MHz, but increasing
> the frequency response is fun, interesting and for me educational.
> >
> > The BF256 is similar to 2SK192 and none of the parts in the new circuit
> > are special.
> Not, that I understand a lot about FET characteristics, but the 2SK192A has a minimally lower Crss and there is no spec for Ciss on the BF256C
> but the 2sk192A spec is 3.5pf, so I'm thinking it might also be a bit lower than the 256C. The Forward Transfer Admittance is higher on the 2sk192A
> and ! can't compare the power gain because they are measured at widely different frequencies.
> So, for those reasons I chose to go with the specified FET. It's either right or it's wrong and an expense I didn't need!
> > Didn't you say before you measured input impedance by inserting variable
> > high resistances between input jack and FET gate and finding values that
> > gave 3dB drop? I recall the figure 30kohm from an early post?
> I did, and I think that was at 1MHz. Although later I tested again and with the 30KΩ the output was lower,
> I don't know if my generator was set to a higher frequency, but, I let that go until I have another amp
> to compare it to. I also don't know what strays I'm adding with my input series resistor.
> I like the Q meter, because it has a + and - 3pf tuning cap graduated in 10ths of a pf.
> A 0.2pf change on a high Q inductor is clearly seen, so I figure its a good way to find the input capacitance.
> With a with a high Q inductor 1250Q at 1MHz, a 50 point of Q drop is 45MΩ additional load resistance, I
> should easily see any drops. I'll need to setup my Q meter and verify that. Seems high, but that's what I calculate,
> although that does not include the losses in the gate capacitance and strays..
> Thanks, Mikek
Reply by Lamont Cranston●November 9, 20222022-11-09
On Wednesday, November 9, 2022 at 8:58:44 AM UTC-6, erichp...@hotmail.com wrote:
> 30MHz is your max frequency then lead length inductance of THT
>components is unlikely to be a big problem.
Don't forget the dielectric constant of pcb substrate is 4-5 times that
of air (and lossy).
Yes, and that is part of removing the ground plane, although it had a slight level peak at 29MHz and
was up a little at 30MHz. That calmed down when I add the underside ground plane.
The thread (on an IO Group) that got me started, was asking about building a Q meter with a range of
1.5MHz to 150Mhz. I don't care about that, I don't know that I will ever go above 10MHz, but increasing
the frequency response is fun, interesting and for me educational.
>
> The BF256 is similar to 2SK192 and none of the parts in the new circuit
> are special.
Not, that I understand a lot about FET characteristics, but the 2SK192A has a minimally lower Crss and there is no spec for Ciss on the BF256C
but the 2sk192A spec is 3.5pf, so I'm thinking it might also be a bit lower than the 256C. The Forward Transfer Admittance is higher on the 2sk192A
and ! can't compare the power gain because they are measured at widely different frequencies.
So, for those reasons I chose to go with the specified FET. It's either right or it's wrong and an expense I didn't need!
> Didn't you say before you measured input impedance by inserting variable
> high resistances between input jack and FET gate and finding values that
> gave 3dB drop? I recall the figure 30kohm from an early post?
I did, and I think that was at 1MHz. Although later I tested again and with the 30KΩ the output was lower,
I don't know if my generator was set to a higher frequency, but, I let that go until I have another amp
to compare it to. I also don't know what strays I'm adding with my input series resistor.
I like the Q meter, because it has a + and - 3pf tuning cap graduated in 10ths of a pf.
A 0.2pf change on a high Q inductor is clearly seen, so I figure its a good way to find the input capacitance.
With a with a high Q inducto,r 1250Q at 1MHz, a 50 point of Q drop is 45MΩ additional load resistance, I
should easily see any drops. I'll need to setup my Q meter and verify that. Seems high, but that's what I calculate,
although that does not include the losses in the gate capacitance and strays..
Thanks, Mikek
Reply by piglet●November 9, 20222022-11-09
On 09/11/2022 11:02 am, Lamont Cranston wrote:
>
>> In the real world performance will be dominated by physical layout
>> strays.
>
> There is the rub, proper layout, on iteration 3, I followed a couple of the ideas given here,
> removing the ground plane and getting the parts raised above the board, although
> without the ground plane I don't think raising the parts mattered. I suspect removing the ground plane
> from the enter circuit was,
>
> not as important as for the high impedance FET area?
>
> I used long leads to keep the parts physically separated, but I know that leads to inductive strays,
> so...
> l 'll be reading up on how to mitigate pcb strays before my next build.
> If anyone has a favorite site on the subject of stray mitigation, please post.
>
> Is there any advantage using smd components?
>
> I thought there was, but the last build did pretty well with leaded components.
>
> Is it worth using a pcb software program, so I could make thinner tracks to help minimize strays?
>
> >If you build it please let us know.
>> piglet
>
> I did order the transistors and FETs last night, so there is a possibility,
> but other projects first.
>
> I can think of only one way to test the input impedance, that is with my Q meter.
> Set up a LC at resonance and then add the high input impedance circuit across the
> tuning capacitor. Then read out the change of the tuning capacitor and how
> much the Q drops, then do the math on the Q change.
>
> With the input current so low, is there another way?
> Thanks, Mikek
If 30MHz is your max frequency then lead length inductance of THT
components is unlikely to be a big problem. Your goal of Hi-Z means low
stray C and capacitance is proportional to area divided by distance. THT
parts will get you distance between nodes whereas SMD parts will reduce
node area. Usually SMD wins but with care you could get THT to work.
Don't forget the dielectric constant of pcb substrate is 4-5 times that
of air (and lossy).
The BF256 is similar to 2SK192 and none of the parts in the new circuit
are special.
Didn't you say before you measured input impedance by inserting variable
high resistances between input jack and FET gate and finding values that
gave 3dB drop? I recall the figure 30kohm from an early post?
The Q meter method should work too and since that appears to be your end
application that method might be the gold standard even though tedious?
piglet
Reply by Lamont Cranston●November 9, 20222022-11-09
> In the real world performance will be dominated by physical layout
> strays.
There is the rub, proper layout, on iteration 3, I followed a couple of the ideas given here,
removing the ground plane and getting the parts raised above the board, although
without the ground plane I don't think raising the parts mattered. I suspect removing the ground plane
from the enter circuit was,
not as important as for the high impedance FET area?
I used long leads to keep the parts physically separated, but I know that leads to inductive strays,
so...
l 'll be reading up on how to mitigate pcb strays before my next build.
If anyone has a favorite site on the subject of stray mitigation, please post.
Is there any advantage using smd components?
I thought there was, but the last build did pretty well with leaded components.
Is it worth using a pcb software program, so I could make thinner tracks to help minimize strays?
>If you build it please let us know.
> piglet
I did order the transistors and FETs last night, so there is a possibility,
but other projects first.
I can think of only one way to test the input impedance, that is with my Q meter.
Set up a LC at resonance and then add the high input impedance circuit across the
tuning capacitor. Then read out the change of the tuning capacitor and how
much the Q drops, then do the math on the Q change.
With the input current so low, is there another way?
Thanks, Mikek
Reply by Jeroen Belleman●November 9, 20222022-11-09
It's very similar to your previous circuit. Again there is the JFET
follower T1, followed by an NPN follower T4. T2 is a current source,
so that the gain of follower T1 will be very close to unity. There will
be virtually no AC voltage across R2, nor across R1, multiplying the
apparent resistance of the latter a lot.
C2 applies the output signal to follower T3 which bootstraps the
drain of T1, reducing the effect of Cgd of that JFET.
In conclusion, it should work fine.
Jeroen Belleman
Reply by piglet●November 9, 20222022-11-09
On 09/11/2022 7:53 am, piglet wrote:
> On 08/11/2022 10:05 pm, Lamont Cranston wrote:
>> Anyone care to look at this Chinese design for a high input impedance
>> amp,
>> and give me their thoughts about it?
>> Two FETs, two transistors, a little different than I have been
>> looking at.
>> https://www.dropbox.com/s/ypvmq9ushs4d2cf/High%20input%20impedance%20amp%20Chinese%20Crystal%20Radio%20Forum.jpg?dl=0
>>
>>
>> Thanks, Mikek
>
> It is fundamentally the same as the one you did yesterday. T3 and C2
> bootstrap away the FET drain capacitance, the other circuit didn't
> bother with the T3 follower using merely an RC. Not sure if the added
> complexity is worthwhile or risks more instability. Behavior with large
> signals may be different but at first glance can't say which is better
> or worse, need further thought!
>
> In the real world performance will be dominated by physical layout
> strays. If you build it please let us know.
>
> piglet
Followup #2 - both give roughly similar results - if you can afford the
higher supply voltage of the simpler circuit use that, if you want to
operate from 9V use the more complex circuit?
piglet