MPPC front end with SAV-551+ pHEMT bootstrap

So I'm doing this multi-pixel photon counter front end for a SEM cathodoluminescence detector. The proto uses a Hamamatsu S13362-3050DG cooled MPPC running at a gain of about 1.7E6, so each detection event produces

1.7E6 e = 0.3 pC,

with asymmetrical rise/fall times of about 5/30 ns. To overcome the 320 pF capacitance and maintain a 40-MHz bandwidth with reasonable noise performance, I built a bootstrap from one of the newish Mini-Circuits SAV-551+ pHEMTs on fine pitch blob board. It promptly oscillated at 5 GHz, but my trusty TDS 694C can see that.

I often scout that sort of problem using a 4-GHz FET probe with a 1 kV disc ceramic cap stuck in where the probe tip should go, and no ground connection--you just lay it someplace nearby and it picks up whatever's going. (The 1 kV caps have leads of the right diameter for the probe.)

Once I got the oscillation stopped, the noise floor in a 60-MHz bandwidth was 900 uV RMS after applying a gain of 8.2. The TIA is an ADA4899-1 with a 1k feedback resistor, so the input-referred current noise is

i_N = 900uV/8.2/1000/sqrt(60 MHz) = 14 pA/sqrt(Hz).

The Johnson noise of the 1k feedback resistor is 4.0 pA/sqrt(Hz).

Without the bootstrap, the noise would be dominated by the e_N C contribution due to the MPPC capacitance differentiating the voltage noise of the op amp:

i_NC = e_Namp * (2*pi*f*320 pF),

and the RMS definite integral from 0 to f_0 is

I_NC = 2*pi * 320 pF * e_Namp * sqrt(integral (f**2)) 0, f_0

The integral term is sqrt(f_0**3/3), so the total e_N*C noise is

I_NC = 2*pi * 320 pF * 1 nV/sqrt(Hz)/sqrt(3) * f_0**1.5

The ADA4899's 1-Hz input noise is about a nanovolt, so without the bootstrap it would be approximately

i_NC ~ 6.3 * 1E-9 * 320E-12 * 60e6**1.5/sqrt(3) = 540 pA/sqrt(Hz).

In order to get this down to the 14 pA level, the voltage noise of the pHEMT would have to be down around 26 pV/sqrt(Hz), which is at least a factor of five too low, even with the most optimistic assumptions. Double-checking underway. ;)

In any event, single-photon events are easily visible, and the dynamic range is pretty good, limited by the number of MPPC pixels (3600) and their recovery time of ~20 ns--assuming uniform illumination, sensitivity will drop by half at

N ~ #pixels/t_rec

or

2.8 nA primary photocurrent. That works out to almost 50 mA after multiplication, which would turn the chip to lava, so we aren't going to be limited that way.

Fun stuff.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs
Loading thread data ...

 S13362-3050DG cooled MPPC running at a gain of about 1.7E6, so each

I recently built a 600 MHz CCRO Colpitts oscillator with a SAV551, and it oscillated at some astronomical frequency above 600. Prudently, I allowed for a gate resistor, and it needed it. The resistor does probably contribute to phase noise. Maybe I should stick to bipolars.

Did you lay out a board or kluge up something? I did both fet and bipolar versions of my oscillator, and a bunch of other stuff, on one board from PCBWAY.

PCBWAY is great but their solder masks are sloppy, not good for BGAs.

formatting link

I do have a little DC data on the SAV551 thing. It's impressive.

Rds-on * Cd-off = 0.8 ps. Almost as good as a relay!

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

I usually use a Murata BLM18BB010 or 005 bead (10 or 5 ohms @ 100 MHz respectively). They peak at some much higher value out at 3-4 GHz where the problems usually lurk.

I built it on a tiny piece of blob board, BoardWorX SMT-400. Photo at .

The bootstrap is at bottom right, the ADA4899 is at bottom left (just to the right of the TO-92 voltage regulator) and the DIP at the top is an LT1260 triple CFA that I had in the drawer.

Never tried one, so I don't know. We use them for everything, though we may go back to Pentalogix for assembly due to the duty--it's a drag having to pay duty on the whole finished assembly.

Yeah, not great. I haven't seen one that bad so far.

The noise plots hint at a very low 1/f corner, too. On P. 7 of the datasheet there's a plot of Fmin (minimum noise figure) vs. frequency that shows a value below 0.1 dB at 500 MHz, and a table on P. 8 showing that it's 0.084 dB with an optimal gamma of 0.38 at 24 degrees phase.

0.084 dB is a fractional added power of 10**0.0084 - 1 ~ 0.0084 * ln(10) = 1.9%.

Ignoring the phase and loading, that means that our optimal noise match is

Ropt = 50*(1+0.38)/(1-0.38) = 111 ohms

and the noise voltage is about

sqrt(0.019*111/60.4) = 190 pV/sqrt(Hz).

That's quite dramatically better than the Avago parts, which were about

300 pV. So maybe that 55 pV number is vaguely conceivable.
--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

As I estimated 190 pV/sqrt(Hz) at 500 MHz in my other post, this is actually a plausible number near baseband--and one that's _SEVEN_POINT_SIX_DECIBELS_ quieter than the ATF38143! Now I really have to make an amp and measure it.

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Aggg! That looks like a junkyard after a nuclear war!

Somehow a noise figure below 1 dB sounds impossible to me. 0.1 dB is an equivalent temperature of 7K.

Reply to
John Larkin

The ADA4899 is the solder blob, with two blue (wire wrap) wires and a tiny piece of coax coming from it?

George H.

Reply to
George Herold

It was prettier before I started debugging the 5-GHz oscillation. But it's in a nice die cast box with a clear coat.

2-D electron gas devices (which is what a pHEMT is) can have very very low scattering. I'll spend part of tomorrow building an amp and measuring its actual noise.

The Early voltage is pretty good too--about 16 volts, and weirdly better at lower V_DS. So you can get some actual voltage gain out of them.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Yup. That's for the power pad. The feedback network (1 k || 3 pF) is sitting right on pins 1 and 2.

BTW the wire cutters arrived, very nice. Thanks!

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Sounds like a weird slope somewhere?

Different thing entirely, but the actual it-dies drain voltage is about 11. They are probably OK to run at 5 or 6.

Reply to
John Larkin

It looks like some other leakage mechanism turning on at around 2 V D-S. Maybe a region with different doping or something like that.

Good to know, thanks.

If this thing is really that good, I'm going to have a lot of fun with it. Hopefully its 1/f corner isn't as horrible as the SKY65050's (30 MHz!). I doubt it'll get the Gerhard Seal of Approval at low baseband, of course, but maybe he can wire up 10,000 chopamps in parallel. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Lots of gain. They go from about off to on pretty hard in a few tenths of a volt.

This is from my phemt NOR gate idea:

formatting link

Voltage gain must be around 50 in the best part.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

On a sunny day (Fri, 09 Aug 2019 12:29:35 -0700) it happened John Larkin wrote in :

The preamp in my Octagon LNB is the NE3512S02 N-CHANNEL HJ-FET. Seems EOL, but mysteriously is in my now 1 year old LNB. That FET has a noise of 0.35 dB at 2 GHz, and 0.4 dB at 10 GHz (where the LNB works). It is supposed to be one of the best LNBs around. ebay still has those from many sellers:

formatting link
Maybe China is making clone FETs?

Reply to
Jan Panteltje

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.