On 2019-04-08 19:42, bill.sloman@ieee.org wrote:> On Tuesday, April 9, 2019 at 6:12:30 AM UTC+10, John Larkin wrote: >> On Mon, 8 Apr 2019 12:52:35 -0700 (PDT), sroberts6328@gmail.com >> wrote: >> >>> Funny, When I go to these VHF and Above conferences Agilent or >>> Keysight usually have a sales engineer and a pair of 200,000 VNAs >>> there for a weekend... Just for the test range and the noise >>> figure competition. >>> >>> Not exactly amateur, as most of the people who do microwave ham >>> radio > then 10 Ghz do work with microwave design at the day >>> job. >>> >>> Steve >> >> In this thread, we were kicking around ideas for working around >> the obvious dumb-FR4 losses. Like suspended substrate, FR4-kapton >> rigid-flex, things like that. Using an expensive laminate was >> already the not-preferred option.Exactly.> > Not strictly accurate. Joerg just wanted to know if he could get away > with FR4 because it is absolutely the cheapest and easiest option. > > A more expensive laminate might well be the cheapest option - looking > at the problem as a whole - but Joerg doesn't like spending any more > than he absolutely has to. >I remember the days when people said that under no circumstances could phenolic be used for UHF and higher. I still have examples to the contrary in the garage. In a similar way I have used low cost DC cables for pulse-echo at several GHz and it works well. Lossy, yes, but it's only the echoes we were after. What is key in most such cases isn't to achieve close to ideal RF performance but one that is merely good enough in terms of margins. -- Regards, Joerg http://www.analogconsultants.com/
77GHz Radar with FR4 board?
Started by ●April 1, 2019
Reply by ●April 17, 20192019-04-17
Reply by ●April 17, 20192019-04-17
On 2019-04-13 09:35, John Larkin wrote:> On Sat, 13 Apr 2019 09:01:38 -0700 (PDT), sroberts6328@gmail.com > wrote: > >> Jeorg is the one looking for board, I have enough Duroid for a life time of experiments. >> >> Steve > > I used to use duroid for prototypes, before I got good at Dremeling > FR4. You can score the duroid with an x-acto knife and then peel up > strips of copper. But adhesion is poor, so pads tend to lift. >If you do this milling out of FR-4 a lot one of the Carbide 3D routers could be really useful: https://carbide3d.com/shapeoko/ It would allow you to mill down very close to the top copper and maintain a consistently low remaining-layer thickness. -- Regards, Joerg http://www.analogconsultants.com/
Reply by ●April 17, 20192019-04-17
On Wed, 17 Apr 2019 07:47:15 -0700, Joerg <news@analogconsultants.com> wrote:>On 2019-04-08 19:42, bill.sloman@ieee.org wrote: >> On Tuesday, April 9, 2019 at 6:12:30 AM UTC+10, John Larkin wrote: >>> On Mon, 8 Apr 2019 12:52:35 -0700 (PDT), sroberts6328@gmail.com >>> wrote: >>> >>>> Funny, When I go to these VHF and Above conferences Agilent or >>>> Keysight usually have a sales engineer and a pair of 200,000 VNAs >>>> there for a weekend... Just for the test range and the noise >>>> figure competition. >>>> >>>> Not exactly amateur, as most of the people who do microwave ham >>>> radio > then 10 Ghz do work with microwave design at the day >>>> job. >>>> >>>> Steve >>> >>> In this thread, we were kicking around ideas for working around >>> the obvious dumb-FR4 losses. Like suspended substrate, FR4-kapton >>> rigid-flex, things like that. Using an expensive laminate was >>> already the not-preferred option. > > >Exactly. > >> >> Not strictly accurate. Joerg just wanted to know if he could get away >> with FR4 because it is absolutely the cheapest and easiest option. >> >> A more expensive laminate might well be the cheapest option - looking >> at the problem as a whole - but Joerg doesn't like spending any more >> than he absolutely has to. >> > >I remember the days when people said that under no circumstances could >phenolic be used for UHF and higher. I still have examples to the >contrary in the garage. In a similar way I have used low cost DC cables >for pulse-echo at several GHz and it works well. Lossy, yes, but it's >only the echoes we were after. > >What is key in most such cases isn't to achieve close to ideal RF >performance but one that is merely good enough in terms of margins.I'm using FR4 at 40 ps edges. Just keep the traces short. FR4 would be terrible to use for a high-Q planar filter, and maybe for a narrowband antenna. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●April 17, 20192019-04-17
On Wed, 17 Apr 2019 07:54:17 -0700, Joerg <news@analogconsultants.com> wrote:>On 2019-04-13 09:35, John Larkin wrote: >> On Sat, 13 Apr 2019 09:01:38 -0700 (PDT), sroberts6328@gmail.com >> wrote: >> >>> Jeorg is the one looking for board, I have enough Duroid for a life time of experiments. >>> >>> Steve >> >> I used to use duroid for prototypes, before I got good at Dremeling >> FR4. You can score the duroid with an x-acto knife and then peel up >> strips of copper. But adhesion is poor, so pads tend to lift. >> > >If you do this milling out of FR-4 a lot one of the Carbide 3D routers >could be really useful: > >https://carbide3d.com/shapeoko/ > >It would allow you to mill down very close to the top copper and >maintain a consistently low remaining-layer thickness.We have two milling machines, a big old Bridgeport and a fancy n/c Tormach. We can optionally chuck a high-speed router in either one, so we could shave out FR4 from below. Chucking a small regular end mill would probably be good enough. We can do that for you but it will cost you beer. PCB houses have told me that they don't mind routing from below. I wanted to reduce the capacitance and capacitive tempco of some big layer 1 pads. FR4 pcb capacitance has a ghastly tempco, like +900 PPM/K. My guys refurbed the ancient Bridgeport and the last step was to paint it. I wanted to make it purple, but they insisted on old boring grey. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●April 17, 20192019-04-17
onsdag den 17. april 2019 kl. 16.54.15 UTC+2 skrev Joerg:> On 2019-04-13 09:35, John Larkin wrote: > > On Sat, 13 Apr 2019 09:01:38 -0700 (PDT), sroberts6328@gmail.com > > wrote: > > > >> Jeorg is the one looking for board, I have enough Duroid for a life time of experiments. > >> > >> Steve > > > > I used to use duroid for prototypes, before I got good at Dremeling > > FR4. You can score the duroid with an x-acto knife and then peel up > > strips of copper. But adhesion is poor, so pads tend to lift. > > > > If you do this milling out of FR-4 a lot one of the Carbide 3D routers > could be really useful: > > https://carbide3d.com/shapeoko/ > > It would allow you to mill down very close to the top copper and > maintain a consistently low remaining-layer thickness.the shapeoko is beltdrive and a trimrouter, for small and accurate something like https://carbide3d.com/nomad/ this is more suited
Reply by ●April 18, 20192019-04-18
On 2019-04-17 08:20, John Larkin wrote:> On Wed, 17 Apr 2019 07:47:15 -0700, Joerg <news@analogconsultants.com> > wrote: > >> On 2019-04-08 19:42, bill.sloman@ieee.org wrote:[...]>>> Not strictly accurate. Joerg just wanted to know if he could get away >>> with FR4 because it is absolutely the cheapest and easiest option. >>> >>> A more expensive laminate might well be the cheapest option - looking >>> at the problem as a whole - but Joerg doesn't like spending any more >>> than he absolutely has to. >>> >> >> I remember the days when people said that under no circumstances could >> phenolic be used for UHF and higher. I still have examples to the >> contrary in the garage. In a similar way I have used low cost DC cables >> for pulse-echo at several GHz and it works well. Lossy, yes, but it's >> only the echoes we were after. >> >> What is key in most such cases isn't to achieve close to ideal RF >> performance but one that is merely good enough in terms of margins. > > I'm using FR4 at 40 ps edges. Just keep the traces short. >The fastest I have used FR-4 for so far was at 100psec and there it worked well. Nothing was longer than an inch though. No milling was involved. It got interesting when trying to measure resolution, dispersion and all that. Not having a 20GHz sampling scope required McGyver style methods such as flattened paperclips on thin transmission line. Paperclips have almost as many uses as duct tape.> FR4 would be terrible to use for a high-Q planar filter, and maybe for > a narrowband antenna. >The application would be wideband, several GHz BW, and traces less than 1". Losses are ok, dispersion not so much. -- Regards, Joerg http://www.analogconsultants.com/
Reply by ●April 18, 20192019-04-18
On Friday, April 19, 2019 at 12:58:17 AM UTC+10, Joerg wrote:> On 2019-04-17 08:20, John Larkin wrote: > > On Wed, 17 Apr 2019 07:47:15 -0700, Joerg <news@analogconsultants.com> > > wrote: > > > >> On 2019-04-08 19:42, bill.sloman@ieee.org wrote: > > [...] > > > >>> Not strictly accurate. Joerg just wanted to know if he could get away > >>> with FR4 because it is absolutely the cheapest and easiest option. > >>> > >>> A more expensive laminate might well be the cheapest option - looking > >>> at the problem as a whole - but Joerg doesn't like spending any more > >>> than he absolutely has to. > >>> > >> > >> I remember the days when people said that under no circumstances could > >> phenolic be used for UHF and higher. I still have examples to the > >> contrary in the garage. In a similar way I have used low cost DC cables > >> for pulse-echo at several GHz and it works well. Lossy, yes, but it's > >> only the echoes we were after. > >> > >> What is key in most such cases isn't to achieve close to ideal RF > >> performance but one that is merely good enough in terms of margins. > > > > I'm using FR4 at 40 ps edges. Just keep the traces short. > > The fastest I have used FR-4 for so far was at 100psec and there it > worked well. Nothing was longer than an inch though. No milling was > involved. > > It got interesting when trying to measure resolution, dispersion and all > that. Not having a 20GHz sampling scope required McGyver style methods > such as flattened paperclips on thin transmission line. Paperclips have > almost as many uses as duct tape. > > > FR4 would be terrible to use for a high-Q planar filter, and maybe for > > a narrowband antenna. > > > > The application would be wideband, several GHz BW, and traces less than > 1". Losses are ok, dispersion not so much.FR4 is dispersive, and so is microstrip (even with an ideal substrate and FR4 is lumps of glass fibre bonded with epoxy). Stripline is apparently not dispersive - if you bury the trace between two ground planes you've got strip-line. Difficult to get a characteristic impedance above 50R of course. -- Bill Sloman, Sydney
Reply by ●April 18, 20192019-04-18
On Thu, 18 Apr 2019 07:58:23 -0700, Joerg <news@analogconsultants.com> wrote:>On 2019-04-17 08:20, John Larkin wrote: >> On Wed, 17 Apr 2019 07:47:15 -0700, Joerg <news@analogconsultants.com> >> wrote: >> >>> On 2019-04-08 19:42, bill.sloman@ieee.org wrote: > >[...] > > >>>> Not strictly accurate. Joerg just wanted to know if he could get away >>>> with FR4 because it is absolutely the cheapest and easiest option. >>>> >>>> A more expensive laminate might well be the cheapest option - looking >>>> at the problem as a whole - but Joerg doesn't like spending any more >>>> than he absolutely has to. >>>> >>> >>> I remember the days when people said that under no circumstances could >>> phenolic be used for UHF and higher. I still have examples to the >>> contrary in the garage. In a similar way I have used low cost DC cables >>> for pulse-echo at several GHz and it works well. Lossy, yes, but it's >>> only the echoes we were after. >>> >>> What is key in most such cases isn't to achieve close to ideal RF >>> performance but one that is merely good enough in terms of margins. >> >> I'm using FR4 at 40 ps edges. Just keep the traces short. >> > >The fastest I have used FR-4 for so far was at 100psec and there it >worked well. Nothing was longer than an inch though. No milling was >involved. > >It got interesting when trying to measure resolution, dispersion and all >that. Not having a 20GHz sampling scope required McGyver style methods >such as flattened paperclips on thin transmission line. Paperclips have >almost as many uses as duct tape. > > >> FR4 would be terrible to use for a high-Q planar filter, and maybe for >> a narrowband antenna. >> > >The application would be wideband, several GHz BW, and traces less than >1". Losses are ok, dispersion not so much.You can get an old 11801 and an SD24 TDR head for a grand or so. It's a great tool. I have hundereds of TDR test photos. We'll sometimes add a couple SMAs and a test trace pattern to a production board, to see how we are doing. Here's one. https://www.dropbox.com/s/qsoqodfdrb80ln0/Z404_PCB.JPG?dl=0 https://www.dropbox.com/s/cxs5ad7vj2o1d52/Z404_traces.jpg?dl=0 https://www.dropbox.com/s/qsoqodfdrb80ln0/Z404_PCB.JPG?dl=0 https://www.dropbox.com/s/unyea2ahjruyvvf/Z404_test_bottom.JPG?dl=0 https://www.dropbox.com/s/u6yaue3bi3c6nad/Z404_TDR_From_J1.JPG?dl=0 In that case, I was curious about the behavior of a tight meander line, especially preshoot, so I added the test traces to a board we were about to order. Preshoot was small and we got all the impedances close on all the layers. I should have used edge launch connectors. There's an ugly transition when the coax from the scope hits the right-angle SMA on the board. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●April 18, 20192019-04-18
On 2019-04-18 08:21, bill.sloman@ieee.org wrote:> On Friday, April 19, 2019 at 12:58:17 AM UTC+10, Joerg wrote: >> On 2019-04-17 08:20, John Larkin wrote: >>> On Wed, 17 Apr 2019 07:47:15 -0700, Joerg >>> <news@analogconsultants.com> wrote: >>> >>>> On 2019-04-08 19:42, bill.sloman@ieee.org wrote: >> >> [...] >> >> >>>>> Not strictly accurate. Joerg just wanted to know if he could >>>>> get away with FR4 because it is absolutely the cheapest and >>>>> easiest option. >>>>> >>>>> A more expensive laminate might well be the cheapest option - >>>>> looking at the problem as a whole - but Joerg doesn't like >>>>> spending any more than he absolutely has to. >>>>> >>>> >>>> I remember the days when people said that under no >>>> circumstances could phenolic be used for UHF and higher. I >>>> still have examples to the contrary in the garage. In a similar >>>> way I have used low cost DC cables for pulse-echo at several >>>> GHz and it works well. Lossy, yes, but it's only the echoes we >>>> were after. >>>> >>>> What is key in most such cases isn't to achieve close to ideal >>>> RF performance but one that is merely good enough in terms of >>>> margins. >>> >>> I'm using FR4 at 40 ps edges. Just keep the traces short. >> >> The fastest I have used FR-4 for so far was at 100psec and there >> it worked well. Nothing was longer than an inch though. No milling >> was involved. >> >> It got interesting when trying to measure resolution, dispersion >> and all that. Not having a 20GHz sampling scope required McGyver >> style methods such as flattened paperclips on thin transmission >> line. Paperclips have almost as many uses as duct tape. >> >>> FR4 would be terrible to use for a high-Q planar filter, and >>> maybe for a narrowband antenna. >>> >> >> The application would be wideband, several GHz BW, and traces less >> than 1". Losses are ok, dispersion not so much. > > FR4 is dispersive, and so is microstrip (even with an ideal substrate > and FR4 is lumps of glass fibre bonded with epoxy). > > Stripline is apparently not dispersive - if you bury the trace > between two ground planes you've got strip-line. Difficult to get a > characteristic impedance above 50R of course. >We put it on the top layer due to other structures needed. At around 5GHz that wasn't a problem but this one will be >10x higher. With plunge milling it might work. -- Regards, Joerg http://www.analogconsultants.com/
Reply by ●April 18, 20192019-04-18
On 2019-04-18 08:22, John Larkin wrote:> On Thu, 18 Apr 2019 07:58:23 -0700, Joerg <news@analogconsultants.com> > wrote: > >> On 2019-04-17 08:20, John Larkin wrote:[...]>>> FR4 would be terrible to use for a high-Q planar filter, and maybe for >>> a narrowband antenna. >>> >> >> The application would be wideband, several GHz BW, and traces less than >> 1". Losses are ok, dispersion not so much. > > You can get an old 11801 and an SD24 TDR head for a grand or so. It's > a great tool. >Sure, but now that I am gradually retiring I don't want to increase the boat anchor clutter in my lab, rather get rid of some of them. Sez SWMBO. She has got a point there. If we ever want to move it will be a smaller place with a LOT less yard maintenance. Ideally the yard should consist of a rock and a cactus. One cactus. Also, many of the older gizmos now come in USB pods. My spectrum analyzer plus tracking generator which can double as a network analyzer are the size of three cigarette packs each. Plus they go to 4400GHz while the huge HP3577 can't go past 200MHz (it is better in dynamic range though). Recently I had a situation where my big old Dolch logic analyzer didn't have the needed serial decode functions. Bought an 8-channel pod for a little over $10, the size of two matchboxes, does all the decoding I could ever want, found the bug immediately.> I have hundereds of TDR test photos. We'll sometimes add a couple SMAs > and a test trace pattern to a production board, to see how we are > doing. > > Here's one. > > https://www.dropbox.com/s/qsoqodfdrb80ln0/Z404_PCB.JPG?dl=0 > > https://www.dropbox.com/s/cxs5ad7vj2o1d52/Z404_traces.jpg?dl=0 > > https://www.dropbox.com/s/qsoqodfdrb80ln0/Z404_PCB.JPG?dl=0 > > https://www.dropbox.com/s/unyea2ahjruyvvf/Z404_test_bottom.JPG?dl=0 > > https://www.dropbox.com/s/u6yaue3bi3c6nad/Z404_TDR_From_J1.JPG?dl=0 > > In that case, I was curious about the behavior of a tight meander > line, especially preshoot, so I added the test traces to a board we > were about to order. Preshoot was small and we got all the impedances > close on all the layers. >That's impressive with such a tight meander. Normally you'd think capacitive coupling would spoil things.> I should have used edge launch connectors. There's an ugly transition > when the coax from the scope hits the right-angle SMA on the board. >Or make a slit and ease in coax straight from the side of the board. I like to avoid connectors where I can. -- Regards, Joerg http://www.analogconsultants.com/
