Is it real or is it a Spice oddity?

[esox]

I have a question about a circuit behavior I see in LTspice.

I know the AnalogRat is just going to tell me again to go build it and try it and I'll get to that as soon as I've finished unearthing my workbench. Meanwhile, I have LTspice.

The circuit is an NPN CE amplifier with an n-jfet as a cascode (Borbely style, sort of). With Drain and Emitter resistances equal and gain varied by changing a series RC in parallel with the emitter R there is an obnoxious gain peak way out at 50-60MHz. I can't get rid of it except by rolling off the cascode effect at higher frequencies. I've tried hacking about with small L, C, R in various spots but since I don't really understand the cause it's still there.

It shows up with various transistors but here's one example,
bjt 2N4401 @ ~2mA
fet 2N5486
R = 4K
bias R 86K, 140K

Is it just an artifact of Spice or is there something real that I'll see when I get around to the workbench?

Mike

 

[Kit]

It´s hard to tell what your circuit looks like from your description and i don´t even know what LTspice is, but if you try and post a schematic ill do my best to help you.

Look on the bright side, 50-60 Mhz is inaudible. :grinis that a word?)

 

[esox]

MR Packrat scolded me about the lack of a schematic on his way out the door...

It's pretty basic..
http://i78.photobucket.com/albums/j90/ElmontTSmith/grief.gif

MIke

 

[bcarso]

Probably real. The bootstrapping of the gate via the emitter connection results in a negative input impedance term at high frequencies.

Although---since you are driving it with a voltage source it wouldn't so much be the input effect as something associated with the same bootstrapped connection affecting the signal coupled to the output, probably. In any event such things are common to bootstrapping connections at high frequencies, nice though they are for reducing input C at lower freqs.

I'll throw it into my sim when I get a chance. In the meantime, try running the gate to a fixed bias and see what happens.

EDIT: Hmmm. I guess the main effect is Ccb coupling HF signals into the source. It gets worse with a non-bootstrapped FET, although the response is broader.

EDIT 2: Using a really low Ccb transistor for the NPN (MPSH34) in sim a pF to ground from the output resistor tames it effectively, even with nonzero source Z's. It is likely that you would get this due to strays in a typical layout, but then you will have other C's and L's to contend with as well.

EDIT 3: With the original 4401 it takes about 1.5pF.

 

[AMZ-FX]

R6 and C3 seem pointless... what happens when you take them out?

regards, Jack

 

[esox]

bcarso,

Thanks, nice fix. I'll put this composite back into my diff pair and see if the same fix applies. I'm sure a real life layout will be - different. I'll feed it some fast edges and look for ringing.

AMX-FX,

I just change R6(0..inf) to shift the midband gain around. The gain peak seemed largely unaffected by midband gain.

Mike

 

[bcarso]

This circuit should be built with a very tight layout to minimize trace length inductance. At least the connections among the transistors should be very short. I recommend the Twin Industries ground-plane-one-side pad-per-hole material, or equivalent, being careful not to short things to the plane.

Also, have some close power supply bypassing (100nF ceramics or films from +24 to ground) to keep the local power supply impedance low.

You may also want a little R-C lowpass at the input (couple hundred ohms in series, couple nanofarads to ground) to limit input signal bandwidth.

Are you wanting to use this as a concertina-type "phase splitter"?

 

[esox]

I started out with a transformerless mic pre idea based on a differential pair with composite transistors made from 4x2N4403@250uA each, 2SJ74 cascode and BC550C as a CFP. Much like the mic pre on Mr Elliot's site + a cascode. Although I was planning on a cascoded current source instead of just resistors.

The high freq gain peak was disconcerting so I started reducing things to the simplest circuit that would exhibit the problem. Prior to your fix, all of the hacks I was trying were relative to the emitter of the bipolar.

This morning I tried your fix in my diff amp ckt with good results. It took a lot more capacitance ( ~50p ) but could that be due to the multiple input tranistors and a higher Gm in the 2SJ74?

My plan for the mic pre was to use local feedback only. The first stage is this composite monster, the 2nd stage is a 2SK170w/2SK246 cascode diff pair and the outputs are Class A bipolar CFP+current source drivers. Each diff stage has a gain switch. The first stage is intended to provide the lion's share of the gain.

I have no idea how it will sound but with a fix for the peak I'm a step closer to matching parts and soldering together a prototype. I have plenty of that pad-per-hole proto board and lots of copper tape. I was definitely planning on BW limiting the input with an RC and probably a couple of ferrite beads at the XLR.

Mike

 

[bcarso]

[quote author="esox"]I started out with a transformerless mic pre idea based on a differential pair with composite transistors made from 4x2N4403@250uA each, 2SJ74 cascode and BC550C as a CFP. Much like the mic pre on Mr Elliot's site + a cascode. Although I was planning on a cascoded current source instead of just resistors.

The high freq gain peak was disconcerting so I started reducing things to the simplest circuit that would exhibit the problem. Prior to your fix, all of the hacks I was trying were relative to the emitter of the bipolar.

This morning I tried your fix in my diff amp ckt with good results. It took a lot more capacitance ( ~50p ) but could that be due to the multiple input tranistors and a higher Gm in the 2SJ74?

My plan for the mic pre was to use local feedback only. The first stage is this composite monster, the 2nd stage is a 2SK170w/2SK246 cascode diff pair and the outputs are Class A bipolar CFP+current source drivers. Each diff stage has a gain switch. The first stage is intended to provide the lion's share of the gain.

I have no idea how it will sound but with a fix for the peak I'm a step closer to matching parts and soldering together a prototype. I have plenty of that pad-per-hole proto board and lots of copper tape. I was definitely planning on BW limiting the input with an RC and probably a couple of ferrite beads at the XLR.

Mike[/quote]

Surprised that there is quite that much disparity between the two circuits, although SJ74's are very high capacitance devices indeed, whereas the 5486 is the very-low-C process 50 part.

Although 4403's are great parts, especially the On Semi (previously Moto) ones, you might want to consider the Toshiba 2SA1316's, which have an astonishingly low rbb' along with very high beta, or if you invert the circuit the NPN complement the 2SC3329, which allows you then to use the higher-performance N-channel JFETs like the 2SK170 as your cascode device(s).

 

[aurt]

Spice Oddity. Ow.

<groans>

The mutant child of David Bowie and Geri Halliwell?

 

[bcarso]

[quote author="aurt"]Spice Oddity. Ow.

<groans>

The mutant child of David Bowie and Geri Halliwell?[/quote]

Damn, I just got that.

And if you have static and shock hazard problems, you may be The Man Who Failed to Earth?