How to choose the coupled capacitors for kHz signals?

[saulbit]

I am designing a circuit working from 9kHz~30MHz, including a few attenuators, two SPDTs and one amplifier, which all need some DC bias. I use some 10u 0805 MLCC capciators for coupled capciators in the signal transmitting microstrip. However, I found that even the amplifier could work normally from 9kHz to 30MHz, the throughout gain is varying from 0dB to 12dB as from 9kHz to 30MHz. So I test every step with open-ended coaxial cable, getting that the insertion loss of coupled capacitors causing the gain subtracted. So I use 10u 0805 capacitors, still having terrible loss. With larger value of capacitors, the loss becomes smaller. Now it comes to my question: the 0805 case MLCC capcacitors have a most value of 10u, so what effort should I make? Pursuing larger value capacitors or just make 10u capacitors parrelleled? Or would somebody tell me another way? Thanks.

 

[jiripolivka]

If I understand correctly, you have cascaded components having individual responses "flat" over 9 kHz to 30 MHz, and for coupling you use electrolytic 10 uF capacitors, all in "microstrip" meaning a 50-Ohm line.

To resolve your slant-response problem, I would recommend to test one or up to three those SMD capacitors connected in series (or cascaded) from your signal source to a load. If their response is flat over the full range, then the problem is elsewhere. The estimated reactance of one 10 uF capacitor at 10 kHz is ~160 Ohms, so it will attenuate well in a 50-Ohm line.
The coupling reactance should be rather < 10 Ohms at 9-10 kHZ, so you should use a higher capacitance value ( 16 x 10 uF capacitors in parallel), or, why can't you try to use a wideband transformer instead? Minicircuits has 50-Ohm 1:1 transformers covering 10 kHz to 125 MHz, model FTB 1-6. You can try to build such transformer on a ferrite core, too.

The easier way may be to use not 50-Ohm line but say 1 kOhm; this would call for < 1 uF coupling capacitors. As amplifiers there are good fast operational amplifiers available, also for other functions like switching and to make active filters.

 

[FvM]

The estimated reactance of one 10 uF capacitor at 10 kHz is ~160 Ohms

Seems, that there is an error in your calculation. Zc of 10 uF is 1.8 ohm at 9 kHz. It's causing 0.005 dB loss per 10 uF/50 ohm highpass. In other words, the reported problem isn't understandable.

 

[biff44]

there are 100 uf tantalums in that size. Why are you liiting youself to 0805 size?

 

[jiripolivka]

Thank you for correction! Still I would advise to test individual components of your cascaded system as I suggested, to find the weak point (s). Then you can correct it. And using transformers instead capacitors has the advantage that the DC bias-tee is included in a transformer. Maybe your DC circuit choke is too small at 9 kHz..

 

[Syncopator]

The inductive reactance of 10µF capacitors, at those frequencies, is likely to be very high. Have you thought of that?

 

[jiripolivka]

I think those SMD capacitors should be tantalum. I would suggest connecting in parallel with the 10 uF cap also a ceramic 0.1 uF capacitor.
The requester did not give other details of his setup. I recommend using a signal generator and a detector or RF mV meter to test responses of each part of the cascaded system. When a poor response is detected this way, it can be corrected.

---------- Post added at 23:53 ---------- Previous post was at 23:49 ----------

I have used fast opamps like MAX4286 or LT 6230-10, also operating from ~1 kHz to > 30 MHz, and never had any problem with a flat gain. I did not need coupling capacitors of 10 uF, my coupling capacitors were 0.1 uF, SMD ceramic. And my devices had 50-Ohm input/outputs, no problems.

 

[BigBoss]

The capacitor might be cracked.Normally it's diificult to physically observe..

 

[FvM]

The inductive reactance of 10µF capacitors, at those frequencies, is likely to be very high. Have you thought of that?

Did you notice, that the discussion is about SMD MLCC? The series inductance can be expected around 1 nH.

 

[biff44]

and his problem is at the LOW end of the frequency band, not at 30 mhz.

 

[saulbit]

Then what about your noise figure of the operation amplifier system( I mean the op circuit)? I have planned to use the OP( operation amplifier), but I am afraid of the 1/f noise from the OP at 9kHz. It might be terrible. The capacitor is used to couple the ac signal or in other words, to isolate diffierent DC biases such as the attentuator and the SPDT, which all have different DC biases. If I don't use the capacitor, the DC biases would be effected each other, especially for the SPDTs.

 

[FvM]

I am afraid of the 1/f noise from the OP at 9kHz. It might be terrible.

I don't see the question related to the thread topic, but there use to be datasheets.

P.S.: For the intended frequency range, it's easy to find an OP that has no noticable 1/f noise.